d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
e97ba859b3
freebsd-dev/sys/dev/mpt/mpt.c

2539 lines
68 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
* Generic routines for LSI Fusion adapters.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
* FreeBSD Version.
*
* Copyright (c) 2000, 2001 by Greg Ansley
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice immediately at the beginning of the file, without modification,
* this list of conditions, and the following disclaimer.
* 2. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
*/
/*-
* Copyright (c) 2002, 2006 by Matthew Jacob
* All rights reserved.
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
*
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce at minimum a disclaimer
* substantially similar to the "NO WARRANTY" disclaimer below
* ("Disclaimer") and any redistribution must be conditioned upon including
* a substantially similar Disclaimer requirement for further binary
* redistribution.
* 3. Neither the names of the above listed copyright holders nor the names
* of any contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
*
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF THE COPYRIGHT
* OWNER OR CONTRIBUTOR IS ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
*
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
* Support from Chris Ellsworth in order to make SAS adapters work
* is gratefully acknowledged.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
*/
Role a microrev of the MPI Library in preparation for target mode work. Make my portions of the license clearer. Thank Chris Ellsworth for his support in getting a bunch of this done.
2006-02-25 07:45:54 +00:00
/*-
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* Copyright (c) 2004, Avid Technology, Inc. and its contributors.
* Copyright (c) 2005, WHEEL Sp. z o.o.
* Copyright (c) 2004, 2005 Justin T. Gibbs
* All rights reserved.
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
*
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce at minimum a disclaimer
* substantially similar to the "NO WARRANTY" disclaimer below
* ("Disclaimer") and any redistribution must be conditioned upon including
* a substantially similar Disclaimer requirement for further binary
* redistribution.
Correct attribution in clause three to address the correct copyright holders. The license that was approved for my changes to this driver originally came from LSI, but the changes to the driver core are not owned by LSI. MFC: 1 day
2005-08-03 14:08:41 +00:00
* 3. Neither the names of the above listed copyright holders nor the names
* of any contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
*
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF THE COPYRIGHT
* OWNER OR CONTRIBUTOR IS ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
*/
Fix where my automated script blew the SCM ID format conversion.
2004-08-21 17:44:57 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
#include <dev/mpt/mpt.h>
#include <dev/mpt/mpt_cam.h> /* XXX For static handler registration */
#include <dev/mpt/mpt_raid.h> /* XXX For static handler registration */
#include <dev/mpt/mpilib/mpi.h>
#include <dev/mpt/mpilib/mpi_ioc.h>
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#include <dev/mpt/mpilib/mpi_fc.h>
#include <dev/mpt/mpilib/mpi_targ.h>
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
#include <sys/sysctl.h>
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
#define MPT_MAX_TRYS 3
#define MPT_MAX_WAIT 300000
static int maxwait_ack = 0;
static int maxwait_int = 0;
static int maxwait_state = 0;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
TAILQ_HEAD(, mpt_softc) mpt_tailq = TAILQ_HEAD_INITIALIZER(mpt_tailq);
mpt_reply_handler_t *mpt_reply_handlers[MPT_NUM_REPLY_HANDLERS];
static mpt_reply_handler_t mpt_default_reply_handler;
static mpt_reply_handler_t mpt_config_reply_handler;
static mpt_reply_handler_t mpt_handshake_reply_handler;
static mpt_reply_handler_t mpt_event_reply_handler;
static void mpt_send_event_ack(struct mpt_softc *mpt, request_t *ack_req,
MSG_EVENT_NOTIFY_REPLY *msg, uint32_t context);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
static int mpt_send_event_request(struct mpt_softc *mpt, int onoff);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static int mpt_soft_reset(struct mpt_softc *mpt);
static void mpt_hard_reset(struct mpt_softc *mpt);
static int mpt_configure_ioc(struct mpt_softc *mpt);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
static int mpt_enable_ioc(struct mpt_softc *mpt, int);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/************************* Personality Module Support *************************/
/*
* We include one extra entry that is guaranteed to be NULL
* to simplify our itterator.
*/
static struct mpt_personality *mpt_personalities[MPT_MAX_PERSONALITIES + 1];
static __inline struct mpt_personality*
mpt_pers_find(struct mpt_softc *, u_int);
static __inline struct mpt_personality*
mpt_pers_find_reverse(struct mpt_softc *, u_int);
static __inline struct mpt_personality *
mpt_pers_find(struct mpt_softc *mpt, u_int start_at)
{
KASSERT(start_at <= MPT_MAX_PERSONALITIES,
("mpt_pers_find: starting position out of range\n"));
while (start_at < MPT_MAX_PERSONALITIES
&& (mpt->mpt_pers_mask & (0x1 << start_at)) == 0) {
start_at++;
}
return (mpt_personalities[start_at]);
}
/*
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
* Used infrequently, so no need to optimize like a forward
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* traversal where we use the MAX+1 is guaranteed to be NULL
* trick.
*/
static __inline struct mpt_personality *
mpt_pers_find_reverse(struct mpt_softc *mpt, u_int start_at)
{
while (start_at < MPT_MAX_PERSONALITIES
&& (mpt->mpt_pers_mask & (0x1 << start_at)) == 0) {
start_at--;
}
if (start_at < MPT_MAX_PERSONALITIES)
return (mpt_personalities[start_at]);
return (NULL);
}
#define MPT_PERS_FOREACH(mpt, pers) \
for (pers = mpt_pers_find(mpt, /*start_at*/0); \
pers != NULL; \
pers = mpt_pers_find(mpt, /*start_at*/pers->id+1))
#define MPT_PERS_FOREACH_REVERSE(mpt, pers) \
for (pers = mpt_pers_find_reverse(mpt, MPT_MAX_PERSONALITIES-1);\
pers != NULL; \
pers = mpt_pers_find_reverse(mpt, /*start_at*/pers->id-1))
static mpt_load_handler_t mpt_stdload;
static mpt_probe_handler_t mpt_stdprobe;
static mpt_attach_handler_t mpt_stdattach;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
static mpt_enable_handler_t mpt_stdenable;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static mpt_event_handler_t mpt_stdevent;
static mpt_reset_handler_t mpt_stdreset;
static mpt_shutdown_handler_t mpt_stdshutdown;
static mpt_detach_handler_t mpt_stddetach;
static mpt_unload_handler_t mpt_stdunload;
static struct mpt_personality mpt_default_personality =
{
.load = mpt_stdload,
.probe = mpt_stdprobe,
.attach = mpt_stdattach,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
.enable = mpt_stdenable,
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
.event = mpt_stdevent,
.reset = mpt_stdreset,
.shutdown = mpt_stdshutdown,
.detach = mpt_stddetach,
.unload = mpt_stdunload
};
static mpt_load_handler_t mpt_core_load;
static mpt_attach_handler_t mpt_core_attach;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
static mpt_enable_handler_t mpt_core_enable;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static mpt_reset_handler_t mpt_core_ioc_reset;
static mpt_event_handler_t mpt_core_event;
static mpt_shutdown_handler_t mpt_core_shutdown;
static mpt_shutdown_handler_t mpt_core_detach;
static mpt_unload_handler_t mpt_core_unload;
static struct mpt_personality mpt_core_personality =
{
.name = "mpt_core",
.load = mpt_core_load,
.attach = mpt_core_attach,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
.enable = mpt_core_enable,
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
.event = mpt_core_event,
.reset = mpt_core_ioc_reset,
.shutdown = mpt_core_shutdown,
.detach = mpt_core_detach,
.unload = mpt_core_unload,
};
/*
* Manual declaration so that DECLARE_MPT_PERSONALITY doesn't need
* ordering information. We want the core to always register FIRST.
* other modules are set to SI_ORDER_SECOND.
*/
static moduledata_t mpt_core_mod = {
"mpt_core", mpt_modevent, &mpt_core_personality
};
DECLARE_MODULE(mpt_core, mpt_core_mod, SI_SUB_DRIVERS, SI_ORDER_FIRST);
MODULE_VERSION(mpt_core, 1);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#define MPT_PERS_ATTACHED(pers, mpt) ((mpt)->mpt_pers_mask & (0x1 << pers->id))
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int
mpt_modevent(module_t mod, int type, void *data)
{
struct mpt_personality *pers;
int error;
pers = (struct mpt_personality *)data;
error = 0;
switch (type) {
case MOD_LOAD:
{
mpt_load_handler_t **def_handler;
mpt_load_handler_t **pers_handler;
int i;
for (i = 0; i < MPT_MAX_PERSONALITIES; i++) {
if (mpt_personalities[i] == NULL)
break;
}
if (i >= MPT_MAX_PERSONALITIES) {
error = ENOMEM;
break;
}
pers->id = i;
mpt_personalities[i] = pers;
/* Install standard/noop handlers for any NULL entries. */
def_handler = MPT_PERS_FIRST_HANDLER(&mpt_default_personality);
pers_handler = MPT_PERS_FIRST_HANDLER(pers);
while (pers_handler <= MPT_PERS_LAST_HANDLER(pers)) {
if (*pers_handler == NULL)
*pers_handler = *def_handler;
pers_handler++;
def_handler++;
}
error = (pers->load(pers));
if (error != 0)
mpt_personalities[i] = NULL;
break;
}
case MOD_SHUTDOWN:
break;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#if __FreeBSD_version >= 500000
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
case MOD_QUIESCE:
break;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#endif
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
case MOD_UNLOAD:
error = pers->unload(pers);
mpt_personalities[pers->id] = NULL;
break;
default:
error = EINVAL;
break;
}
return (error);
}
int
mpt_stdload(struct mpt_personality *pers)
{
/* Load is always successfull. */
return (0);
}
int
mpt_stdprobe(struct mpt_softc *mpt)
{
/* Probe is always successfull. */
return (0);
}
int
mpt_stdattach(struct mpt_softc *mpt)
{
/* Attach is always successfull. */
return (0);
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
int
mpt_stdenable(struct mpt_softc *mpt)
{
/* Enable is always successfull. */
return (0);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt_stdevent(struct mpt_softc *mpt, request_t *req, MSG_EVENT_NOTIFY_REPLY *msg)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "mpt_stdevent: 0x%x\n", msg->Event & 0xFF);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/* Event was not for us. */
return (0);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
void
mpt_stdreset(struct mpt_softc *mpt, int type)
{
}
void
mpt_stdshutdown(struct mpt_softc *mpt)
{
}
void
mpt_stddetach(struct mpt_softc *mpt)
{
}
int
mpt_stdunload(struct mpt_personality *pers)
{
/* Unload is always successfull. */
return (0);
}
/******************************* Bus DMA Support ******************************/
void
mpt_map_rquest(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct mpt_map_info *map_info;
map_info = (struct mpt_map_info *)arg;
map_info->error = error;
map_info->phys = segs->ds_addr;
}
/**************************** Reply/Event Handling ****************************/
int
mpt_register_handler(struct mpt_softc *mpt, mpt_handler_type type,
mpt_handler_t handler, uint32_t *phandler_id)
{
switch (type) {
case MPT_HANDLER_REPLY:
{
u_int cbi;
u_int free_cbi;
if (phandler_id == NULL)
return (EINVAL);
free_cbi = MPT_HANDLER_ID_NONE;
for (cbi = 0; cbi < MPT_NUM_REPLY_HANDLERS; cbi++) {
/*
* If the same handler is registered multiple
* times, don't error out. Just return the
* index of the original registration.
*/
if (mpt_reply_handlers[cbi] == handler.reply_handler) {
*phandler_id = MPT_CBI_TO_HID(cbi);
return (0);
}
/*
* Fill from the front in the hope that
* all registered handlers consume only a
* single cache line.
*
* We don't break on the first empty slot so
* that the full table is checked to see if
* this handler was previously registered.
*/
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (free_cbi == MPT_HANDLER_ID_NONE &&
(mpt_reply_handlers[cbi]
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
== mpt_default_reply_handler))
free_cbi = cbi;
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (free_cbi == MPT_HANDLER_ID_NONE) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (ENOMEM);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_reply_handlers[free_cbi] = handler.reply_handler;
*phandler_id = MPT_CBI_TO_HID(free_cbi);
break;
}
default:
mpt_prt(mpt, "mpt_register_handler unknown type %d\n", type);
return (EINVAL);
}
return (0);
}
int
mpt_deregister_handler(struct mpt_softc *mpt, mpt_handler_type type,
mpt_handler_t handler, uint32_t handler_id)
{
switch (type) {
case MPT_HANDLER_REPLY:
{
u_int cbi;
cbi = MPT_CBI(handler_id);
if (cbi >= MPT_NUM_REPLY_HANDLERS
|| mpt_reply_handlers[cbi] != handler.reply_handler)
return (ENOENT);
mpt_reply_handlers[cbi] = mpt_default_reply_handler;
break;
}
default:
mpt_prt(mpt, "mpt_deregister_handler unknown type %d\n", type);
return (EINVAL);
}
return (0);
}
static int
mpt_default_reply_handler(struct mpt_softc *mpt, request_t *req,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t reply_desc, MSG_DEFAULT_REPLY *reply_frame)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_prt(mpt,
"Default Handler Called: req=%p:%u reply_descriptor=%x frame=%p\n",
req, req->serno, reply_desc, reply_frame);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (reply_frame != NULL)
mpt_dump_reply_frame(mpt, reply_frame);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_prt(mpt, "Reply Frame Ignored\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (/*free_reply*/TRUE);
}
static int
mpt_config_reply_handler(struct mpt_softc *mpt, request_t *req,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t reply_desc, MSG_DEFAULT_REPLY *reply_frame)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
if (req != NULL) {
if (reply_frame != NULL) {
MSG_CONFIG *cfgp;
MSG_CONFIG_REPLY *reply;
cfgp = (MSG_CONFIG *)req->req_vbuf;
reply = (MSG_CONFIG_REPLY *)reply_frame;
req->IOCStatus = le16toh(reply_frame->IOCStatus);
bcopy(&reply->Header, &cfgp->Header,
sizeof(cfgp->Header));
}
req->state &= ~REQ_STATE_QUEUED;
req->state |= REQ_STATE_DONE;
TAILQ_REMOVE(&mpt->request_pending_list, req, links);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
if ((req->state & REQ_STATE_NEED_WAKEUP) != 0) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
wakeup(req);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
return (TRUE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
static int
mpt_handshake_reply_handler(struct mpt_softc *mpt, request_t *req,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t reply_desc, MSG_DEFAULT_REPLY *reply_frame)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
/* Nothing to be done. */
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
return (TRUE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
static int
mpt_event_reply_handler(struct mpt_softc *mpt, request_t *req,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t reply_desc, MSG_DEFAULT_REPLY *reply_frame)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
int free_reply;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
KASSERT(reply_frame != NULL, ("null reply in mpt_event_reply_handler"));
KASSERT(req != NULL, ("null request in mpt_event_reply_handler"));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
free_reply = TRUE;
switch (reply_frame->Function) {
case MPI_FUNCTION_EVENT_NOTIFICATION:
{
MSG_EVENT_NOTIFY_REPLY *msg;
struct mpt_personality *pers;
u_int handled;
handled = 0;
msg = (MSG_EVENT_NOTIFY_REPLY *)reply_frame;
MPT_PERS_FOREACH(mpt, pers)
handled += pers->event(mpt, req, msg);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
if (handled == 0 && mpt->mpt_pers_mask == 0) {
turn some WARNs for unknown events into INFOs
2006-03-04 21:46:34 +00:00
mpt_lprt(mpt, MPT_PRT_INFO,
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
"No Handlers For Any Event Notify Frames. "
"Event %#x (ACK %sequired).\n",
msg->Event, msg->AckRequired? "r" : "not r");
} else if (handled == 0) {
turn some WARNs for unknown events into INFOs
2006-03-04 21:46:34 +00:00
mpt_lprt(mpt, MPT_PRT_WARN,
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
"Unhandled Event Notify Frame. Event %#x "
"(ACK %sequired).\n",
msg->Event, msg->AckRequired? "r" : "not r");
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (msg->AckRequired) {
request_t *ack_req;
uint32_t context;
context = htole32(req->index|MPT_REPLY_HANDLER_EVENTS);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
ack_req = mpt_get_request(mpt, FALSE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (ack_req == NULL) {
struct mpt_evtf_record *evtf;
evtf = (struct mpt_evtf_record *)reply_frame;
evtf->context = context;
LIST_INSERT_HEAD(&mpt->ack_frames, evtf, links);
free_reply = FALSE;
break;
}
mpt_send_event_ack(mpt, ack_req, msg, context);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
/*
* Don't check for CONTINUATION_REPLY here
*/
return (free_reply);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
break;
}
case MPI_FUNCTION_PORT_ENABLE:
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG , "enable port reply\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
break;
case MPI_FUNCTION_EVENT_ACK:
break;
default:
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
mpt_prt(mpt, "unknown event function: %x\n",
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
reply_frame->Function);
break;
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
/*
* I'm not sure that this continuation stuff works as it should.
*
* I've had FC async events occur that free the frame up because
* the continuation bit isn't set, and then additional async events
* then occur using the same context. As you might imagine, this
* leads to Very Bad Thing.
*
* Let's just be safe for now and not free them up until we figure
* out what's actually happening here.
*/
#if 0
if ((reply_frame->MsgFlags & MPI_MSGFLAGS_CONTINUATION_REPLY) == 0) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
TAILQ_REMOVE(&mpt->request_pending_list, req, links);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
mpt_free_request(mpt, req);
mpt_prt(mpt, "event_reply %x for req %p:%u NOT a continuation",
reply_frame->Function, req, req->serno);
if (reply_frame->Function == MPI_FUNCTION_EVENT_NOTIFICATION) {
MSG_EVENT_NOTIFY_REPLY *msg =
(MSG_EVENT_NOTIFY_REPLY *)reply_frame;
mpt_prtc(mpt, " Event=0x%x AckReq=%d",
msg->Event, msg->AckRequired);
}
} else {
mpt_prt(mpt, "event_reply %x for %p:%u IS a continuation",
reply_frame->Function, req, req->serno);
if (reply_frame->Function == MPI_FUNCTION_EVENT_NOTIFICATION) {
MSG_EVENT_NOTIFY_REPLY *msg =
(MSG_EVENT_NOTIFY_REPLY *)reply_frame;
mpt_prtc(mpt, " Event=0x%x AckReq=%d",
msg->Event, msg->AckRequired);
}
mpt_prtc(mpt, "\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
#endif
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (free_reply);
}
/*
* Process an asynchronous event from the IOC.
*/
static int
mpt_core_event(struct mpt_softc *mpt, request_t *req,
MSG_EVENT_NOTIFY_REPLY *msg)
{
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "mpt_core_event: 0x%x\n",
msg->Event & 0xFF);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
switch(msg->Event & 0xFF) {
case MPI_EVENT_NONE:
break;
case MPI_EVENT_LOG_DATA:
{
int i;
/* Some error occured that LSI wants logged */
mpt_prt(mpt, "EvtLogData: IOCLogInfo: 0x%08x\n",
msg->IOCLogInfo);
mpt_prt(mpt, "\tEvtLogData: Event Data:");
for (i = 0; i < msg->EventDataLength; i++)
mpt_prtc(mpt, " %08x", msg->Data[i]);
mpt_prtc(mpt, "\n");
break;
}
case MPI_EVENT_EVENT_CHANGE:
/*
* This is just an acknowledgement
* of our mpt_send_event_request.
*/
break;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
case MPI_EVENT_SAS_DEVICE_STATUS_CHANGE:
break;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
default:
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
return (0);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
break;
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
return (1);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
static void
mpt_send_event_ack(struct mpt_softc *mpt, request_t *ack_req,
MSG_EVENT_NOTIFY_REPLY *msg, uint32_t context)
{
MSG_EVENT_ACK *ackp;
ackp = (MSG_EVENT_ACK *)ack_req->req_vbuf;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(ackp, 0, sizeof (*ackp));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
ackp->Function = MPI_FUNCTION_EVENT_ACK;
ackp->Event = msg->Event;
ackp->EventContext = msg->EventContext;
ackp->MsgContext = context;
mpt_check_doorbell(mpt);
mpt_send_cmd(mpt, ack_req);
}
/***************************** Interrupt Handling *****************************/
void
mpt_intr(void *arg)
{
struct mpt_softc *mpt;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t reply_desc;
int ntrips = 0;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt = (struct mpt_softc *)arg;
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG2, "enter mpt_intr\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
while ((reply_desc = mpt_pop_reply_queue(mpt)) != MPT_REPLY_EMPTY) {
request_t *req;
MSG_DEFAULT_REPLY *reply_frame;
uint32_t reply_baddr;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
uint32_t ctxt_idx;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
u_int cb_index;
u_int req_index;
int free_rf;
req = NULL;
reply_frame = NULL;
reply_baddr = 0;
if ((reply_desc & MPI_ADDRESS_REPLY_A_BIT) != 0) {
u_int offset;
/*
* Insure that the reply frame is coherent.
*/
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
reply_baddr = MPT_REPLY_BADDR(reply_desc);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
offset = reply_baddr - (mpt->reply_phys & 0xFFFFFFFF);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
bus_dmamap_sync_range(mpt->reply_dmat,
mpt->reply_dmap, offset, MPT_REPLY_SIZE,
BUS_DMASYNC_POSTREAD);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
reply_frame = MPT_REPLY_OTOV(mpt, offset);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
ctxt_idx = le32toh(reply_frame->MsgContext);
} else {
uint32_t type;
type = MPI_GET_CONTEXT_REPLY_TYPE(reply_desc);
ctxt_idx = reply_desc;
mpt_lprt(mpt, MPT_PRT_DEBUG1, "Context Reply: 0x%08x\n",
reply_desc);
switch (type) {
case MPI_CONTEXT_REPLY_TYPE_SCSI_INIT:
ctxt_idx &= MPI_CONTEXT_REPLY_CONTEXT_MASK;
break;
case MPI_CONTEXT_REPLY_TYPE_SCSI_TARGET:
ctxt_idx = GET_IO_INDEX(reply_desc);
if (mpt->tgt_cmd_ptrs == NULL) {
mpt_prt(mpt,
"mpt_intr: no target cmd ptrs\n");
reply_desc = MPT_REPLY_EMPTY;
break;
}
if (ctxt_idx >= mpt->tgt_cmds_allocated) {
mpt_prt(mpt,
"mpt_intr: bad tgt cmd ctxt %u\n",
ctxt_idx);
reply_desc = MPT_REPLY_EMPTY;
ntrips = 1000;
break;
}
req = mpt->tgt_cmd_ptrs[ctxt_idx];
if (req == NULL) {
mpt_prt(mpt, "no request backpointer "
"at index %u", ctxt_idx);
reply_desc = MPT_REPLY_EMPTY;
ntrips = 1000;
break;
}
/*
* Reformulate ctxt_idx to be just as if
* it were another type of context reply
* so the code below will find the request
* via indexing into the pool.
*/
ctxt_idx =
req->index | mpt->scsi_tgt_handler_id;
req = NULL;
break;
case MPI_CONTEXT_REPLY_TYPE_LAN:
mpt_prt(mpt, "LAN CONTEXT REPLY: 0x%08x\n",
reply_desc);
reply_desc = MPT_REPLY_EMPTY;
break;
default:
mpt_prt(mpt, "Context Reply 0x%08x?\n", type);
reply_desc = MPT_REPLY_EMPTY;
break;
}
if (reply_desc == MPT_REPLY_EMPTY) {
if (ntrips++ > 1000) {
break;
}
continue;
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
cb_index = MPT_CONTEXT_TO_CBI(ctxt_idx);
req_index = MPT_CONTEXT_TO_REQI(ctxt_idx);
if (req_index < MPT_MAX_REQUESTS(mpt)) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req = &mpt->request_pool[req_index];
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
} else {
mpt_prt(mpt, "WARN: mpt_intr index == %d (reply_desc =="
" 0x%x)\n", req_index, reply_desc);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
free_rf = mpt_reply_handlers[cb_index](mpt, req,
reply_desc, reply_frame);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
if (reply_frame != NULL && free_rf) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_free_reply(mpt, reply_baddr);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
/*
* If we got ourselves disabled, don't get stuck in a loop
*/
if (mpt->disabled) {
mpt_disable_ints(mpt);
break;
}
if (ntrips++ > 1000) {
break;
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG2, "exit mpt_intr\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
/******************************* Error Recovery *******************************/
void
mpt_complete_request_chain(struct mpt_softc *mpt, struct req_queue *chain,
u_int iocstatus)
{
MSG_DEFAULT_REPLY ioc_status_frame;
request_t *req;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(&ioc_status_frame, 0, sizeof(ioc_status_frame));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
ioc_status_frame.MsgLength = roundup2(sizeof(ioc_status_frame), 4);
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
ioc_status_frame.IOCStatus = iocstatus;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
while((req = TAILQ_FIRST(chain)) != NULL) {
MSG_REQUEST_HEADER *msg_hdr;
u_int cb_index;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
In mpt_complete_request_chain don't depend on somebody else to remove the request from the TAILQ.
2006-03-17 04:52:27 +00:00
TAILQ_REMOVE(chain, req, links);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
msg_hdr = (MSG_REQUEST_HEADER *)req->req_vbuf;
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
ioc_status_frame.Function = msg_hdr->Function;
ioc_status_frame.MsgContext = msg_hdr->MsgContext;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
cb_index = MPT_CONTEXT_TO_CBI(le32toh(msg_hdr->MsgContext));
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_reply_handlers[cb_index](mpt, req, msg_hdr->MsgContext,
&ioc_status_frame);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
}
/********************************* Diagnostics ********************************/
/*
* Perform a diagnostic dump of a reply frame.
*/
void
mpt_dump_reply_frame(struct mpt_softc *mpt, MSG_DEFAULT_REPLY *reply_frame)
{
mpt_prt(mpt, "Address Reply:\n");
mpt_print_reply(reply_frame);
}
/******************************* Doorbell Access ******************************/
static __inline uint32_t mpt_rd_db(struct mpt_softc *mpt);
static __inline uint32_t mpt_rd_intr(struct mpt_softc *mpt);
static __inline uint32_t
mpt_rd_db(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
return mpt_read(mpt, MPT_OFFSET_DOORBELL);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static __inline uint32_t
mpt_rd_intr(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
return mpt_read(mpt, MPT_OFFSET_INTR_STATUS);
}
/* Busy wait for a door bell to be read by IOC */
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_wait_db_ack(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int i;
for (i=0; i < MPT_MAX_WAIT; i++) {
if (!MPT_DB_IS_BUSY(mpt_rd_intr(mpt))) {
maxwait_ack = i > maxwait_ack ? i : maxwait_ack;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_OK);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
DELAY(200);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Busy wait for a door bell interrupt */
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_wait_db_int(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int i;
for (i=0; i < MPT_MAX_WAIT; i++) {
if (MPT_DB_INTR(mpt_rd_intr(mpt))) {
maxwait_int = i > maxwait_int ? i : maxwait_int;
return MPT_OK;
}
DELAY(100);
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Wait for IOC to transition to a give state */
void
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_check_doorbell(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
uint32_t db = mpt_rd_db(mpt);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if (MPT_STATE(db) != MPT_DB_STATE_RUNNING) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "Device not running\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
mpt_print_db(db);
}
}
/* Wait for IOC to transition to a give state */
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_wait_state(struct mpt_softc *mpt, enum DB_STATE_BITS state)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int i;
for (i = 0; i < MPT_MAX_WAIT; i++) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
uint32_t db = mpt_rd_db(mpt);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if (MPT_STATE(db) == state) {
maxwait_state = i > maxwait_state ? i : maxwait_state;
return (MPT_OK);
}
DELAY(100);
}
return (MPT_FAIL);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/************************* Intialization/Configuration ************************/
static int mpt_download_fw(struct mpt_softc *mpt);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* Issue the reset COMMAND to the IOC */
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static int
mpt_soft_reset(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "soft reset\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* Have to use hard reset if we are not in Running state */
if (MPT_STATE(mpt_rd_db(mpt)) != MPT_DB_STATE_RUNNING) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "soft reset failed: device not running\n");
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* If door bell is in use we don't have a chance of getting
* a word in since the IOC probably crashed in message
* processing. So don't waste our time.
*/
if (MPT_DB_IS_IN_USE(mpt_rd_db(mpt))) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "soft reset failed: doorbell wedged\n");
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Send the reset request to the IOC */
mpt_write(mpt, MPT_OFFSET_DOORBELL,
MPI_FUNCTION_IOC_MESSAGE_UNIT_RESET << MPI_DOORBELL_FUNCTION_SHIFT);
if (mpt_wait_db_ack(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "soft reset failed: ack timeout\n");
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Wait for the IOC to reload and come out of reset state */
if (mpt_wait_state(mpt, MPT_DB_STATE_READY) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "soft reset failed: device did not restart\n");
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
return (MPT_FAIL);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
return MPT_OK;
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static int
mpt_enable_diag_mode(struct mpt_softc *mpt)
{
int try;
try = 20;
while (--try) {
if ((mpt_read(mpt, MPT_OFFSET_DIAGNOSTIC) & MPI_DIAG_DRWE) != 0)
break;
/* Enable diagnostic registers */
mpt_write(mpt, MPT_OFFSET_SEQUENCE, 0xFF);
mpt_write(mpt, MPT_OFFSET_SEQUENCE, MPI_WRSEQ_1ST_KEY_VALUE);
mpt_write(mpt, MPT_OFFSET_SEQUENCE, MPI_WRSEQ_2ND_KEY_VALUE);
mpt_write(mpt, MPT_OFFSET_SEQUENCE, MPI_WRSEQ_3RD_KEY_VALUE);
mpt_write(mpt, MPT_OFFSET_SEQUENCE, MPI_WRSEQ_4TH_KEY_VALUE);
mpt_write(mpt, MPT_OFFSET_SEQUENCE, MPI_WRSEQ_5TH_KEY_VALUE);
DELAY(100000);
}
if (try == 0)
return (EIO);
return (0);
}
static void
mpt_disable_diag_mode(struct mpt_softc *mpt)
{
mpt_write(mpt, MPT_OFFSET_SEQUENCE, 0xFFFFFFFF);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* This is a magic diagnostic reset that resets all the ARM
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
* processors in the chip.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static void
mpt_hard_reset(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int error;
int wait;
uint32_t diagreg;
mpt_lprt(mpt, MPT_PRT_DEBUG, "hard reset\n");
error = mpt_enable_diag_mode(mpt);
if (error) {
mpt_prt(mpt, "WARNING - Could not enter diagnostic mode !\n");
mpt_prt(mpt, "Trying to reset anyway.\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
diagreg = mpt_read(mpt, MPT_OFFSET_DIAGNOSTIC);
/*
* This appears to be a workaround required for some
* firmware or hardware revs.
*/
mpt_write(mpt, MPT_OFFSET_DIAGNOSTIC, diagreg | MPI_DIAG_DISABLE_ARM);
DELAY(1000);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* Diag. port is now active so we can now hit the reset bit */
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_write(mpt, MPT_OFFSET_DIAGNOSTIC, diagreg | MPI_DIAG_RESET_ADAPTER);
/*
* Ensure that the reset has finished. We delay 1ms
* prior to reading the register to make sure the chip
* has sufficiently completed its reset to handle register
* accesses.
*/
wait = 5000;
do {
DELAY(1000);
diagreg = mpt_read(mpt, MPT_OFFSET_DIAGNOSTIC);
} while (--wait && (diagreg & MPI_DIAG_RESET_ADAPTER) == 0);
if (wait == 0) {
mpt_prt(mpt, "WARNING - Failed hard reset! "
"Trying to initialize anyway.\n");
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* If we have firmware to download, it must be loaded before
* the controller will become operational. Do so now.
*/
if (mpt->fw_image != NULL) {
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
error = mpt_download_fw(mpt);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (error) {
mpt_prt(mpt, "WARNING - Firmware Download Failed!\n");
mpt_prt(mpt, "Trying to initialize anyway.\n");
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* Reseting the controller should have disabled write
* access to the diagnostic registers, but disable
* manually to be sure.
*/
mpt_disable_diag_mode(mpt);
}
static void
mpt_core_ioc_reset(struct mpt_softc *mpt, int type)
{
/*
* Complete all pending requests with a status
* appropriate for an IOC reset.
*/
mpt_complete_request_chain(mpt, &mpt->request_pending_list,
MPI_IOCSTATUS_INVALID_STATE);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/*
* Reset the IOC when needed. Try software command first then if needed
* poke at the magic diagnostic reset. Note that a hard reset resets
* *both* IOCs on dual function chips (FC929 && LSI1030) as well as
* fouls up the PCI configuration registers.
*/
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_reset(struct mpt_softc *mpt, int reinit)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
struct mpt_personality *pers;
int ret;
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
int retry_cnt = 0;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
/*
* Try a soft reset. If that fails, get out the big hammer.
*/
again:
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if ((ret = mpt_soft_reset(mpt)) != MPT_OK) {
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
int cnt;
for (cnt = 0; cnt < 5; cnt++) {
/* Failed; do a hard reset */
mpt_hard_reset(mpt);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
/*
* Wait for the IOC to reload
* and come out of reset state
*/
ret = mpt_wait_state(mpt, MPT_DB_STATE_READY);
if (ret == MPT_OK) {
break;
}
/*
* Okay- try to check again...
*/
ret = mpt_wait_state(mpt, MPT_DB_STATE_READY);
if (ret == MPT_OK) {
break;
}
mpt_prt(mpt, "mpt_reset: failed hard reset (%d:%d)\n",
retry_cnt, cnt);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
if (retry_cnt == 0) {
/*
* Invoke reset handlers. We bump the reset count so
* that mpt_wait_req() understands that regardless of
* the specified wait condition, it should stop its wait.
*/
mpt->reset_cnt++;
MPT_PERS_FOREACH(mpt, pers)
pers->reset(mpt, ret);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
if (reinit) {
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
ret = mpt_enable_ioc(mpt, 1);
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
if (ret == MPT_OK) {
mpt_enable_ints(mpt);
}
}
if (ret != MPT_OK && retry_cnt++ < 2) {
goto again;
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return ret;
}
/* Return a command buffer to the free queue */
void
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_free_request(struct mpt_softc *mpt, request_t *req)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
request_t *nxt;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
struct mpt_evtf_record *record;
uint32_t reply_baddr;
Parameterize MPT_MAX_REQUESTS based upon device type (FC has Global Credits of 1024- Ultra4 256). Rename 'requests' tag to 'request_pool' for clarity. Make sure we do correct xpt_freeze_simq/CAM_RELEASE_SIMQ if we run out of chip resources. MFC after: 6 days
2002-09-23 19:41:10 +00:00
if (req == NULL || req != &mpt->request_pool[req->index]) {
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
panic("mpt_free_request bad req ptr\n");
return;
}
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
if ((nxt = req->chain) != NULL) {
req->chain = NULL;
mpt_free_request(mpt, nxt); /* NB: recursion */
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
KASSERT(req->state != REQ_STATE_FREE, ("freeing free request"));
KASSERT(!(req->state & REQ_STATE_LOCKED), ("freeing locked request"));
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
KASSERT(MPT_OWNED(mpt), ("mpt_free_request: mpt not locked\n"));
KASSERT(mpt_req_on_free_list(mpt, req) == 0,
("mpt_free_request: req %p:%u func %x already on freelist",
req, req->serno, ((MSG_REQUEST_HEADER *)req->req_vbuf)->Function));
KASSERT(mpt_req_on_pending_list(mpt, req) == 0,
("mpt_free_request: req %p:%u func %x on pending list",
req, req->serno, ((MSG_REQUEST_HEADER *)req->req_vbuf)->Function));
#ifdef INVARIANTS
mpt_req_not_spcl(mpt, req, "mpt_free_request", __LINE__);
#endif
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
req->ccb = NULL;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (LIST_EMPTY(&mpt->ack_frames)) {
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
/*
* Insert free ones at the tail
*/
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
req->serno = 0;
req->state = REQ_STATE_FREE;
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
#ifdef INVARIANTS
memset(req->req_vbuf, 0xff, sizeof (MSG_REQUEST_HEADER));
#endif
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
TAILQ_INSERT_TAIL(&mpt->request_free_list, req, links);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (mpt->getreqwaiter != 0) {
mpt->getreqwaiter = 0;
wakeup(&mpt->request_free_list);
}
return;
}
/*
* Process an ack frame deferred due to resource shortage.
*/
record = LIST_FIRST(&mpt->ack_frames);
LIST_REMOVE(record, links);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
req->state = REQ_STATE_ALLOCATED;
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt_assign_serno(mpt, req);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_event_ack(mpt, req, &record->reply, record->context);
reply_baddr = (uint32_t)((uint8_t *)record - mpt->reply)
+ (mpt->reply_phys & 0xFFFFFFFF);
mpt_free_reply(mpt, reply_baddr);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Get a command buffer from the free queue */
request_t *
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_get_request(struct mpt_softc *mpt, int sleep_ok)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
request_t *req;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
retry:
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
KASSERT(MPT_OWNED(mpt), ("mpt_get_request: mpt not locked\n"));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req = TAILQ_FIRST(&mpt->request_free_list);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if (req != NULL) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
KASSERT(req == &mpt->request_pool[req->index],
("mpt_get_request: corrupted request free list\n"));
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
KASSERT(req->state == REQ_STATE_FREE,
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
("req %p:%u not free on free list %x index %d function %x",
req, req->serno, req->state, req->index,
((MSG_REQUEST_HEADER *)req->req_vbuf)->Function));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
TAILQ_REMOVE(&mpt->request_free_list, req, links);
req->state = REQ_STATE_ALLOCATED;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
req->chain = NULL;
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt_assign_serno(mpt, req);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
} else if (sleep_ok != 0) {
mpt->getreqwaiter = 1;
mpt_sleep(mpt, &mpt->request_free_list, PUSER, "mptgreq", 0);
goto retry;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
return (req);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Pass the command to the IOC */
void
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_cmd(struct mpt_softc *mpt, request_t *req)
{
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
if (mpt->verbose > MPT_PRT_DEBUG2) {
Remove handrolled raw dump of a request from mpt_send_cmd and replace with a new debug function.
2006-05-26 05:41:14 +00:00
mpt_dump_request(mpt, req);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
bus_dmamap_sync(mpt->request_dmat, mpt->request_dmap,
Code cleanup: use mpt_prt instead of device_printf.
2002-09-24 21:33:43 +00:00
BUS_DMASYNC_PREWRITE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req->state |= REQ_STATE_QUEUED;
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
KASSERT(mpt_req_on_free_list(mpt, req) == 0,
("req %p:%u func %x on freelist list in mpt_send_cmd",
req, req->serno, ((MSG_REQUEST_HEADER *)req->req_vbuf)->Function));
KASSERT(mpt_req_on_pending_list(mpt, req) == 0,
("req %p:%u func %x already on pending list in mpt_send_cmd",
req, req->serno, ((MSG_REQUEST_HEADER *)req->req_vbuf)->Function));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
TAILQ_INSERT_HEAD(&mpt->request_pending_list, req, links);
mpt_write(mpt, MPT_OFFSET_REQUEST_Q, (uint32_t) req->req_pbuf);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/*
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* Wait for a request to complete.
*
* Inputs:
* mpt softc of controller executing request
* req request to wait for
* sleep_ok nonzero implies may sleep in this context
* time_ms timeout in ms. 0 implies no timeout.
*
* Return Values:
* 0 Request completed
* non-0 Timeout fired before request completion.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int
mpt_wait_req(struct mpt_softc *mpt, request_t *req,
mpt_req_state_t state, mpt_req_state_t mask,
int sleep_ok, int time_ms)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int error;
int timeout;
u_int saved_cnt;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* timeout is in ms. 0 indicates infinite wait.
* Convert to ticks or 500us units depending on
* our sleep mode.
*/
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (sleep_ok != 0) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
timeout = (time_ms * hz) / 1000;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
} else {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
timeout = time_ms * 2;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req->state |= REQ_STATE_NEED_WAKEUP;
mask &= ~REQ_STATE_NEED_WAKEUP;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
saved_cnt = mpt->reset_cnt;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
while ((req->state & mask) != state && mpt->reset_cnt == saved_cnt) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (sleep_ok != 0) {
error = mpt_sleep(mpt, req, PUSER, "mptreq", timeout);
if (error == EWOULDBLOCK) {
timeout = 0;
break;
}
} else {
if (time_ms != 0 && --timeout == 0) {
break;
}
DELAY(500);
mpt_intr(mpt);
}
}
req->state &= ~REQ_STATE_NEED_WAKEUP;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (mpt->reset_cnt != saved_cnt) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (EIO);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
if (time_ms && timeout <= 0) {
MSG_REQUEST_HEADER *msg_hdr = req->req_vbuf;
mpt_prt(mpt, "mpt_wait_req(%x) timed out\n", msg_hdr->Function);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (ETIMEDOUT);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (0);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/*
* Send a command to the IOC via the handshake register.
*
* Only done at initialization time and for certain unusual
* commands such as device/bus reset as specified by LSI.
*/
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_handshake_cmd(struct mpt_softc *mpt, size_t len, void *cmd)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int i;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
uint32_t data, *data32;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* Check condition of the IOC */
data = mpt_rd_db(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if ((MPT_STATE(data) != MPT_DB_STATE_READY
&& MPT_STATE(data) != MPT_DB_STATE_RUNNING
&& MPT_STATE(data) != MPT_DB_STATE_FAULT)
|| MPT_DB_IS_IN_USE(data)) {
mpt_prt(mpt, "handshake aborted - invalid doorbell state\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
mpt_print_db(data);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (EBUSY);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* We move things in 32 bit chunks */
len = (len + 3) >> 2;
data32 = cmd;
/* Clear any left over pending doorbell interupts */
if (MPT_DB_INTR(mpt_rd_intr(mpt)))
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
/*
* Tell the handshake reg. we are going to send a command
* and how long it is going to be.
*/
data = (MPI_FUNCTION_HANDSHAKE << MPI_DOORBELL_FUNCTION_SHIFT) |
(len << MPI_DOORBELL_ADD_DWORDS_SHIFT);
mpt_write(mpt, MPT_OFFSET_DOORBELL, data);
/* Wait for the chip to notice */
if (mpt_wait_db_int(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_send_handshake_cmd timeout1\n");
return (ETIMEDOUT);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Clear the interrupt */
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
if (mpt_wait_db_ack(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_send_handshake_cmd timeout2\n");
return (ETIMEDOUT);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Send the command */
for (i = 0; i < len; i++) {
mpt_write(mpt, MPT_OFFSET_DOORBELL, *data32++);
if (mpt_wait_db_ack(mpt) != MPT_OK) {
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
mpt_prt(mpt,
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
"mpt_send_handshake_cmd timeout! index = %d\n",
i);
return (ETIMEDOUT);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
}
return MPT_OK;
}
/* Get the response from the handshake register */
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_recv_handshake_reply(struct mpt_softc *mpt, size_t reply_len, void *reply)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int left, reply_left;
u_int16_t *data16;
MSG_DEFAULT_REPLY *hdr;
/* We move things out in 16 bit chunks */
reply_len >>= 1;
data16 = (u_int16_t *)reply;
hdr = (MSG_DEFAULT_REPLY *)reply;
/* Get first word */
if (mpt_wait_db_int(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_recv_handshake_cmd timeout1\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return ETIMEDOUT;
}
*data16++ = mpt_read(mpt, MPT_OFFSET_DOORBELL) & MPT_DB_DATA_MASK;
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
/* Get Second Word */
if (mpt_wait_db_int(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_recv_handshake_cmd timeout2\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return ETIMEDOUT;
}
*data16++ = mpt_read(mpt, MPT_OFFSET_DOORBELL) & MPT_DB_DATA_MASK;
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
/*
* With the second word, we can now look at the length.
* Warn about a reply that's too short (except for IOC FACTS REPLY)
*/
if ((reply_len >> 1) != hdr->MsgLength &&
(hdr->Function != MPI_FUNCTION_IOC_FACTS)){
#if __FreeBSD_version >= 500000
mpt_prt(mpt, "reply length does not match message length: "
Fix fat-fingered version define.
2006-04-01 19:49:55 +00:00
"got %x; expected %zx for function %x\n",
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
hdr->MsgLength << 2, reply_len << 1, hdr->Function);
#else
Code cleanup: use mpt_prt instead of device_printf.
2002-09-24 21:33:43 +00:00
mpt_prt(mpt, "reply length does not match message length: "
Fix fat-fingered version define.
2006-04-01 19:49:55 +00:00
"got %x; expected %x for function %x\n",
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
hdr->MsgLength << 2, reply_len << 1, hdr->Function);
#endif
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/* Get rest of the reply; but don't overflow the provided buffer */
left = (hdr->MsgLength << 1) - 2;
reply_left = reply_len - 2;
while (left--) {
u_int16_t datum;
if (mpt_wait_db_int(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_recv_handshake_cmd timeout3\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return ETIMEDOUT;
}
datum = mpt_read(mpt, MPT_OFFSET_DOORBELL);
if (reply_left-- > 0)
*data16++ = datum & MPT_DB_DATA_MASK;
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
}
/* One more wait & clear at the end */
if (mpt_wait_db_int(mpt) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_recv_handshake_cmd timeout4\n");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return ETIMEDOUT;
}
mpt_write(mpt, MPT_OFFSET_INTR_STATUS, 0);
if ((hdr->IOCStatus & MPI_IOCSTATUS_MASK) != MPI_IOCSTATUS_SUCCESS) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (mpt->verbose >= MPT_PRT_TRACE)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
mpt_print_reply(hdr);
return (MPT_FAIL | hdr->IOCStatus);
}
return (0);
}
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_get_iocfacts(struct mpt_softc *mpt, MSG_IOC_FACTS_REPLY *freplp)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
MSG_IOC_FACTS f_req;
int error;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(&f_req, 0, sizeof f_req);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
f_req.Function = MPI_FUNCTION_IOC_FACTS;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
f_req.MsgContext = htole32(MPT_REPLY_HANDLER_HANDSHAKE);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
error = mpt_send_handshake_cmd(mpt, sizeof f_req, &f_req);
if (error)
return(error);
error = mpt_recv_handshake_reply(mpt, sizeof (*freplp), freplp);
return (error);
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_get_portfacts(struct mpt_softc *mpt, MSG_PORT_FACTS_REPLY *freplp)
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
{
MSG_PORT_FACTS f_req;
int error;
/* XXX: Only getting PORT FACTS for Port 0 */
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
memset(&f_req, 0, sizeof f_req);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
f_req.Function = MPI_FUNCTION_PORT_FACTS;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
f_req.MsgContext = htole32(MPT_REPLY_HANDLER_HANDSHAKE);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
error = mpt_send_handshake_cmd(mpt, sizeof f_req, &f_req);
if (error)
return(error);
error = mpt_recv_handshake_reply(mpt, sizeof (*freplp), freplp);
return (error);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/*
* Send the initialization request. This is where we specify how many
* SCSI busses and how many devices per bus we wish to emulate.
* This is also the command that specifies the max size of the reply
* frames from the IOC that we will be allocating.
*/
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_ioc_init(struct mpt_softc *mpt, uint32_t who)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int error = 0;
MSG_IOC_INIT init;
MSG_IOC_INIT_REPLY reply;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(&init, 0, sizeof init);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
init.WhoInit = who;
init.Function = MPI_FUNCTION_IOC_INIT;
if (mpt->is_fc) {
init.MaxDevices = 255;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
} else if (mpt->is_sas) {
init.MaxDevices = mpt->mpt_max_devices;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
} else {
init.MaxDevices = 16;
}
init.MaxBuses = 1;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
init.MsgVersion = htole16(MPI_VERSION);
init.HeaderVersion = htole16(MPI_HEADER_VERSION);
init.ReplyFrameSize = htole16(MPT_REPLY_SIZE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
init.MsgContext = htole32(MPT_REPLY_HANDLER_HANDSHAKE);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if ((error = mpt_send_handshake_cmd(mpt, sizeof init, &init)) != 0) {
return(error);
}
error = mpt_recv_handshake_reply(mpt, sizeof reply, &reply);
return (error);
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
/*
* Utiltity routine to read configuration headers and pages
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int
mpt_issue_cfg_req(struct mpt_softc *mpt, request_t *req, u_int Action,
u_int PageVersion, u_int PageLength, u_int PageNumber,
u_int PageType, uint32_t PageAddress, bus_addr_t addr,
bus_size_t len, int sleep_ok, int timeout_ms)
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
{
MSG_CONFIG *cfgp;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
SGE_SIMPLE32 *se;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
cfgp = req->req_vbuf;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
memset(cfgp, 0, sizeof *cfgp);
cfgp->Action = Action;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
cfgp->Function = MPI_FUNCTION_CONFIG;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
cfgp->Header.PageVersion = PageVersion;
cfgp->Header.PageLength = PageLength;
cfgp->Header.PageNumber = PageNumber;
cfgp->Header.PageType = PageType;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
cfgp->PageAddress = PageAddress;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
se = (SGE_SIMPLE32 *)&cfgp->PageBufferSGE;
se->Address = addr;
MPI_pSGE_SET_LENGTH(se, len);
MPI_pSGE_SET_FLAGS(se, (MPI_SGE_FLAGS_SIMPLE_ELEMENT |
MPI_SGE_FLAGS_LAST_ELEMENT | MPI_SGE_FLAGS_END_OF_BUFFER |
MPI_SGE_FLAGS_END_OF_LIST |
((Action == MPI_CONFIG_ACTION_PAGE_WRITE_CURRENT
|| Action == MPI_CONFIG_ACTION_PAGE_WRITE_NVRAM)
? MPI_SGE_FLAGS_HOST_TO_IOC : MPI_SGE_FLAGS_IOC_TO_HOST)));
cfgp->MsgContext = htole32(req->index | MPT_REPLY_HANDLER_CONFIG);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
mpt_check_doorbell(mpt);
mpt_send_cmd(mpt, req);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (mpt_wait_req(mpt, req, REQ_STATE_DONE, REQ_STATE_DONE,
sleep_ok, timeout_ms));
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int
mpt_read_cfg_header(struct mpt_softc *mpt, int PageType, int PageNumber,
uint32_t PageAddress, CONFIG_PAGE_HEADER *rslt,
int sleep_ok, int timeout_ms)
{
request_t *req;
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
MSG_CONFIG *cfgp;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int error;
req = mpt_get_request(mpt, sleep_ok);
if (req == NULL) {
mpt_prt(mpt, "mpt_read_cfg_header: Get request failed!\n");
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
return (ENOMEM);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
error = mpt_issue_cfg_req(mpt, req, MPI_CONFIG_ACTION_PAGE_HEADER,
/*PageVersion*/0, /*PageLength*/0, PageNumber,
PageType, PageAddress, /*addr*/0, /*len*/0,
sleep_ok, timeout_ms);
if (error != 0) {
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
mpt_free_request(mpt, req);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "read_cfg_header timed out\n");
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
return (ETIMEDOUT);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
switch (req->IOCStatus & MPI_IOCSTATUS_MASK) {
case MPI_IOCSTATUS_SUCCESS:
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
cfgp = req->req_vbuf;
bcopy(&cfgp->Header, rslt, sizeof(*rslt));
error = 0;
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
break;
case MPI_IOCSTATUS_CONFIG_INVALID_PAGE:
mpt_lprt(mpt, MPT_PRT_DEBUG,
"Invalid Page Type %d Number %d Addr 0x%0x\n",
PageType, PageNumber, PageAddress);
error = EINVAL;
break;
default:
mpt_prt(mpt, "mpt_read_cfg_header: Config Info Status %x\n",
req->IOCStatus);
error = EIO;
break;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
mpt_free_request(mpt, req);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (error);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
Fix things so that: a) we don't believe what the board tells us all the time (if the BIOS hasn't run, port page 2 and port page 0 tend to be garbage) b) add the missing code to set parameters for the SPI cards. MFC after: 0 days
2002-09-01 23:07:09 +00:00
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_read_cfg_page(struct mpt_softc *mpt, int Action, uint32_t PageAddress,
CONFIG_PAGE_HEADER *hdr, size_t len, int sleep_ok,
int timeout_ms)
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
request_t *req;
int error;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req = mpt_get_request(mpt, sleep_ok);
if (req == NULL) {
mpt_prt(mpt, "mpt_read_cfg_page: Get request failed!\n");
return (-1);
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
error = mpt_issue_cfg_req(mpt, req, Action, hdr->PageVersion,
hdr->PageLength, hdr->PageNumber,
hdr->PageType & MPI_CONFIG_PAGETYPE_MASK,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
PageAddress, req->req_pbuf + MPT_RQSL(mpt),
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len, sleep_ok, timeout_ms);
if (error != 0) {
mpt_prt(mpt, "read_cfg_page(%d) timed out\n", Action);
return (-1);
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if ((req->IOCStatus & MPI_IOCSTATUS_MASK) != MPI_IOCSTATUS_SUCCESS) {
mpt_prt(mpt, "mpt_read_cfg_page: Config Info Status %x\n",
req->IOCStatus);
mpt_free_request(mpt, req);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
return (-1);
}
bus_dmamap_sync(mpt->request_dmat, mpt->request_dmap,
BUS_DMASYNC_POSTREAD);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
memcpy(hdr, ((uint8_t *)req->req_vbuf)+MPT_RQSL(mpt), len);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
mpt_free_request(mpt, req);
return (0);
}
Fix things so that: a) we don't believe what the board tells us all the time (if the BIOS hasn't run, port page 2 and port page 0 tend to be garbage) b) add the missing code to set parameters for the SPI cards. MFC after: 0 days
2002-09-01 23:07:09 +00:00
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_write_cfg_page(struct mpt_softc *mpt, int Action, uint32_t PageAddress,
CONFIG_PAGE_HEADER *hdr, size_t len, int sleep_ok,
int timeout_ms)
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
request_t *req;
u_int hdr_attr;
int error;
hdr_attr = hdr->PageType & MPI_CONFIG_PAGEATTR_MASK;
if (hdr_attr != MPI_CONFIG_PAGEATTR_CHANGEABLE &&
hdr_attr != MPI_CONFIG_PAGEATTR_PERSISTENT) {
mpt_prt(mpt, "page type 0x%x not changeable\n",
hdr->PageType & MPI_CONFIG_PAGETYPE_MASK);
return (-1);
}
hdr->PageType &= MPI_CONFIG_PAGETYPE_MASK,
req = mpt_get_request(mpt, sleep_ok);
if (req == NULL)
return (-1);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
memcpy(((caddr_t)req->req_vbuf)+MPT_RQSL(mpt), hdr, len);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/* Restore stripped out attributes */
hdr->PageType |= hdr_attr;
error = mpt_issue_cfg_req(mpt, req, Action, hdr->PageVersion,
hdr->PageLength, hdr->PageNumber,
hdr->PageType & MPI_CONFIG_PAGETYPE_MASK,
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
PageAddress, req->req_pbuf + MPT_RQSL(mpt),
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len, sleep_ok, timeout_ms);
if (error != 0) {
mpt_prt(mpt, "mpt_write_cfg_page timed out\n");
return (-1);
}
if ((req->IOCStatus & MPI_IOCSTATUS_MASK) != MPI_IOCSTATUS_SUCCESS) {
mpt_prt(mpt, "mpt_write_cfg_page: Config Info Status %x\n",
req->IOCStatus);
mpt_free_request(mpt, req);
return (-1);
}
mpt_free_request(mpt, req);
return (0);
}
/*
* Read IOC configuration information
*/
static int
mpt_read_config_info_ioc(struct mpt_softc *mpt)
{
CONFIG_PAGE_HEADER hdr;
struct mpt_raid_volume *mpt_raid;
int rv;
int i;
size_t len;
rv = mpt_read_cfg_header(mpt, MPI_CONFIG_PAGETYPE_IOC,
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
2, 0, &hdr, FALSE, 5000);
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
/*
* If it's an invalid page, so what? Not a supported function....
*/
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
if (rv == EINVAL) {
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
return (0);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
}
if (rv) {
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
return (rv);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
#if __FreeBSD_version >= 500000
Correct bad format args. Interesting- the lines that generated the error on sparc64 hadn't changed since the last checkin, pass LINT on other platforms and mpt doesn't work on sparc64 anyway and the tinderbox build didn't work for me in a cross build case on my main build machine (which runs RELENG_6). Sigh. Still need to try harder.
2006-03-26 07:16:17 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "IOC Page 2 Header: ver %x, len %zx, "
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
"num %x, type %x\n", hdr.PageVersion,
hdr.PageLength * sizeof(uint32_t),
hdr.PageNumber, hdr.PageType);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
#else
mpt_lprt(mpt, MPT_PRT_DEBUG, "IOC Page 2 Header: ver %x, len %z, "
"num %x, type %x\n", hdr.PageVersion,
hdr.PageLength * sizeof(uint32_t),
hdr.PageNumber, hdr.PageType);
#endif
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len = hdr.PageLength * sizeof(uint32_t);
Remove a couple of explicit memset(0) ops that were zeroing past the end of an allocation. This fixes the malloc 'use after free' panic on boot that many were seeing. It doesn't solve the problem of the allocations being cached and then written past their bounds later. That will take more work. Submitted by: kan
2005-10-08 05:16:45 +00:00
mpt->ioc_page2 = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
if (mpt->ioc_page2 == NULL) {
mpt_prt(mpt, "unable to allocate memory for IOC page 2\n");
mpt_raid_free_mem(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (ENOMEM);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
memcpy(&mpt->ioc_page2->Header, &hdr, sizeof(hdr));
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
rv = mpt_read_cur_cfg_page(mpt, 0,
&mpt->ioc_page2->Header, len, FALSE, 5000);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (rv) {
mpt_prt(mpt, "failed to read IOC Page 2\n");
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid_free_mem(mpt);
return (EIO);
}
if (mpt->ioc_page2->CapabilitiesFlags != 0) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
uint32_t mask;
mpt_prt(mpt, "Capabilities: (");
for (mask = 1; mask != 0; mask <<= 1) {
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
if ((mpt->ioc_page2->CapabilitiesFlags & mask) == 0) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
continue;
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
switch (mask) {
case MPI_IOCPAGE2_CAP_FLAGS_IS_SUPPORT:
mpt_prtc(mpt, " RAID-0");
break;
case MPI_IOCPAGE2_CAP_FLAGS_IME_SUPPORT:
mpt_prtc(mpt, " RAID-1E");
break;
case MPI_IOCPAGE2_CAP_FLAGS_IM_SUPPORT:
mpt_prtc(mpt, " RAID-1");
break;
case MPI_IOCPAGE2_CAP_FLAGS_SES_SUPPORT:
mpt_prtc(mpt, " SES");
break;
case MPI_IOCPAGE2_CAP_FLAGS_SAFTE_SUPPORT:
mpt_prtc(mpt, " SAFTE");
break;
case MPI_IOCPAGE2_CAP_FLAGS_CROSS_CHANNEL_SUPPORT:
mpt_prtc(mpt, " Multi-Channel-Arrays");
default:
break;
}
}
mpt_prtc(mpt, " )\n");
if ((mpt->ioc_page2->CapabilitiesFlags
& (MPI_IOCPAGE2_CAP_FLAGS_IS_SUPPORT
| MPI_IOCPAGE2_CAP_FLAGS_IME_SUPPORT
| MPI_IOCPAGE2_CAP_FLAGS_IM_SUPPORT)) != 0) {
mpt_prt(mpt, "%d Active Volume%s(%d Max)\n",
mpt->ioc_page2->NumActiveVolumes,
mpt->ioc_page2->NumActiveVolumes != 1
? "s " : " ",
mpt->ioc_page2->MaxVolumes);
mpt_prt(mpt, "%d Hidden Drive Member%s(%d Max)\n",
mpt->ioc_page2->NumActivePhysDisks,
mpt->ioc_page2->NumActivePhysDisks != 1
? "s " : " ",
mpt->ioc_page2->MaxPhysDisks);
}
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len = mpt->ioc_page2->MaxVolumes * sizeof(struct mpt_raid_volume);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt->raid_volumes = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (mpt->raid_volumes == NULL) {
mpt_prt(mpt, "Could not allocate RAID volume data\n");
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid_free_mem(mpt);
return (ENOMEM);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* Copy critical data out of ioc_page2 so that we can
* safely refresh the page without windows of unreliable
* data.
*/
mpt->raid_max_volumes = mpt->ioc_page2->MaxVolumes;
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
len = sizeof(*mpt->raid_volumes->config_page) +
(sizeof (RAID_VOL0_PHYS_DISK) * (mpt->ioc_page2->MaxPhysDisks - 1));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
for (i = 0; i < mpt->ioc_page2->MaxVolumes; i++) {
mpt_raid = &mpt->raid_volumes[i];
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid->config_page =
malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (mpt_raid->config_page == NULL) {
mpt_prt(mpt, "Could not allocate RAID page data\n");
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid_free_mem(mpt);
return (ENOMEM);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
}
mpt->raid_page0_len = len;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len = mpt->ioc_page2->MaxPhysDisks * sizeof(struct mpt_raid_disk);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt->raid_disks = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (mpt->raid_disks == NULL) {
mpt_prt(mpt, "Could not allocate RAID disk data\n");
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid_free_mem(mpt);
return (ENOMEM);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt->raid_max_disks = mpt->ioc_page2->MaxPhysDisks;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
/*
* Load page 3.
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
rv = mpt_read_cfg_header(mpt, MPI_CONFIG_PAGETYPE_IOC,
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
3, 0, &hdr, FALSE, 5000);
if (rv) {
mpt_raid_free_mem(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (EIO);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "IOC Page 3 Header: %x %x %x %x\n",
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
hdr.PageVersion, hdr.PageLength, hdr.PageNumber, hdr.PageType);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
len = hdr.PageLength * sizeof(uint32_t);
Remove a couple of explicit memset(0) ops that were zeroing past the end of an allocation. This fixes the malloc 'use after free' panic on boot that many were seeing. It doesn't solve the problem of the allocations being cached and then written past their bounds later. That will take more work. Submitted by: kan
2005-10-08 05:16:45 +00:00
mpt->ioc_page3 = malloc(len, M_DEVBUF, M_NOWAIT | M_ZERO);
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
if (mpt->ioc_page3 == NULL) {
mpt_prt(mpt, "unable to allocate memory for IOC page 3\n");
mpt_raid_free_mem(mpt);
return (ENOMEM);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
memcpy(&mpt->ioc_page3->Header, &hdr, sizeof(hdr));
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
rv = mpt_read_cur_cfg_page(mpt, 0,
&mpt->ioc_page3->Header, len, FALSE, 5000);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (rv) {
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
mpt_raid_free_mem(mpt);
return (EIO);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_raid_wakeup(mpt);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
return (0);
}
/*
* Enable IOC port
*/
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_port_enable(struct mpt_softc *mpt, int port)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
request_t *req;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
MSG_PORT_ENABLE *enable_req;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int error;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
req = mpt_get_request(mpt, /*sleep_ok*/FALSE);
if (req == NULL)
return (-1);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req = req->req_vbuf;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(enable_req, 0, MPT_RQSL(mpt));
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req->Function = MPI_FUNCTION_PORT_ENABLE;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
enable_req->MsgContext = htole32(req->index | MPT_REPLY_HANDLER_CONFIG);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req->PortNumber = port;
mpt_check_doorbell(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "enabling port %d\n", port);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_cmd(mpt, req);
error = mpt_wait_req(mpt, req, REQ_STATE_DONE, REQ_STATE_DONE,
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
FALSE, (mpt->is_sas || mpt->is_fc)? 30000 : 3000);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (error != 0) {
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_prt(mpt, "port %d enable timed out\n", port);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (-1);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
mpt_free_request(mpt, req);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "enabled port %d\n", port);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return (0);
}
/*
* Enable/Disable asynchronous event reporting.
*/
static int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_send_event_request(struct mpt_softc *mpt, int onoff)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
request_t *req;
MSG_EVENT_NOTIFY *enable_req;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
req = mpt_get_request(mpt, FALSE);
if (req == NULL) {
return (ENOMEM);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req = req->req_vbuf;
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
memset(enable_req, 0, sizeof *enable_req);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req->Function = MPI_FUNCTION_EVENT_NOTIFICATION;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
enable_req->MsgContext = htole32(req->index | MPT_REPLY_HANDLER_EVENTS);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
enable_req->Switch = onoff;
mpt_check_doorbell(mpt);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "%sabling async events\n",
onoff ? "en" : "dis");
/*
* Send the command off, but don't wait for it.
*/
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
mpt_send_cmd(mpt, req);
return (0);
}
/*
* Un-mask the interupts on the chip.
*/
void
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_enable_ints(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
/* Unmask every thing except door bell int */
mpt_write(mpt, MPT_OFFSET_INTR_MASK, MPT_INTR_DB_MASK);
}
/*
* Mask the interupts on the chip.
*/
void
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_disable_ints(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
/* Mask all interrupts */
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
mpt_write(mpt, MPT_OFFSET_INTR_MASK,
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
MPT_INTR_REPLY_MASK | MPT_INTR_DB_MASK);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static void
mpt_sysctl_attach(struct mpt_softc *mpt)
{
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#if __FreeBSD_version >= 500000
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(mpt->dev);
struct sysctl_oid *tree = device_get_sysctl_tree(mpt->dev);
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), OID_AUTO,
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
"debug", CTLFLAG_RW, &mpt->verbose, 0,
"Debugging/Verbose level");
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
#endif
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
int
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_attach(struct mpt_softc *mpt)
{
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
struct mpt_personality *pers;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int i;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
int error;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
for (i = 0; i < MPT_MAX_PERSONALITIES; i++) {
pers = mpt_personalities[i];
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (pers == NULL) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
continue;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if (pers->probe(mpt) == 0) {
error = pers->attach(mpt);
if (error != 0) {
mpt_detach(mpt);
return (error);
}
mpt->mpt_pers_mask |= (0x1 << pers->id);
pers->use_count++;
}
}
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
/*
* Now that we've attached everything, do the enable function
* for all of the personalities. This allows the personalities
* to do setups that are appropriate for them prior to enabling
* any ports.
*/
for (i = 0; i < MPT_MAX_PERSONALITIES; i++) {
pers = mpt_personalities[i];
if (pers != NULL && MPT_PERS_ATTACHED(pers, mpt) != 0) {
error = pers->enable(mpt);
if (error != 0) {
mpt_prt(mpt, "personality %s attached but would"
" not enable (%d)\n", pers->name, error);
mpt_detach(mpt);
return (error);
}
}
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (0);
}
int
mpt_shutdown(struct mpt_softc *mpt)
{
struct mpt_personality *pers;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
MPT_PERS_FOREACH_REVERSE(mpt, pers) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
pers->shutdown(mpt);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (0);
}
int
mpt_detach(struct mpt_softc *mpt)
{
struct mpt_personality *pers;
MPT_PERS_FOREACH_REVERSE(mpt, pers) {
pers->detach(mpt);
mpt->mpt_pers_mask &= ~(0x1 << pers->id);
pers->use_count--;
}
return (0);
}
int
mpt_core_load(struct mpt_personality *pers)
{
int i;
/*
* Setup core handlers and insert the default handler
* into all "empty slots".
*/
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
for (i = 0; i < MPT_NUM_REPLY_HANDLERS; i++) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_reply_handlers[i] = mpt_default_reply_handler;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_reply_handlers[MPT_CBI(MPT_REPLY_HANDLER_EVENTS)] =
mpt_event_reply_handler;
mpt_reply_handlers[MPT_CBI(MPT_REPLY_HANDLER_CONFIG)] =
mpt_config_reply_handler;
mpt_reply_handlers[MPT_CBI(MPT_REPLY_HANDLER_HANDSHAKE)] =
mpt_handshake_reply_handler;
return (0);
}
/*
* Initialize per-instance driver data and perform
* initial controller configuration.
*/
int
mpt_core_attach(struct mpt_softc *mpt)
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
{
int val;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
int error;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
LIST_INIT(&mpt->ack_frames);
/* Put all request buffers on the free list */
TAILQ_INIT(&mpt->request_pending_list);
TAILQ_INIT(&mpt->request_free_list);
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
TAILQ_INIT(&mpt->request_timeout_list);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
for (val = 0; val < MPT_MAX_REQUESTS(mpt); val++) {
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
request_t *req = &mpt->request_pool[val];
req->state = REQ_STATE_ALLOCATED;
mpt_free_request(mpt, req);
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
for (val = 0; val < MPT_MAX_LUNS; val++) {
STAILQ_INIT(&mpt->trt[val].atios);
STAILQ_INIT(&mpt->trt[val].inots);
}
STAILQ_INIT(&mpt->trt_wildcard.atios);
STAILQ_INIT(&mpt->trt_wildcard.inots);
mpt->scsi_tgt_handler_id = MPT_HANDLER_ID_NONE;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_sysctl_attach(mpt);
mpt_lprt(mpt, MPT_PRT_DEBUG, "doorbell req = %s\n",
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_ioc_diag(mpt_read(mpt, MPT_OFFSET_DOORBELL)));
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
error = mpt_configure_ioc(mpt);
return (error);
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
int
mpt_core_enable(struct mpt_softc *mpt)
{
/*
* We enter with the IOC enabled, but async events
* not enabled, ports not enabled and interrupts
* not enabled.
*/
/*
* Enable asynchronous event reporting- all personalities
* have attached so that they should be able to now field
* async events.
*/
mpt_send_event_request(mpt, 1);
/*
* Catch any pending interrupts
*
* This seems to be crucial- otherwise
* the portenable below times out.
*/
mpt_intr(mpt);
/*
* Enable Interrupts
*/
mpt_enable_ints(mpt);
/*
* Catch any pending interrupts
*
* This seems to be crucial- otherwise
* the portenable below times out.
*/
mpt_intr(mpt);
/*
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
* Enable the port.
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
*/
if (mpt_send_port_enable(mpt, 0) != MPT_OK) {
mpt_prt(mpt, "failed to enable port 0\n");
return (ENXIO);
}
return (0);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
void
mpt_core_shutdown(struct mpt_softc *mpt)
{
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_disable_ints(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
void
mpt_core_detach(struct mpt_softc *mpt)
{
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_disable_ints(mpt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
int
mpt_core_unload(struct mpt_personality *pers)
{
/* Unload is always successfull. */
return (0);
}
#define FW_UPLOAD_REQ_SIZE \
(sizeof(MSG_FW_UPLOAD) - sizeof(SGE_MPI_UNION) \
+ sizeof(FW_UPLOAD_TCSGE) + sizeof(SGE_SIMPLE32))
static int
mpt_upload_fw(struct mpt_softc *mpt)
{
uint8_t fw_req_buf[FW_UPLOAD_REQ_SIZE];
MSG_FW_UPLOAD_REPLY fw_reply;
MSG_FW_UPLOAD *fw_req;
FW_UPLOAD_TCSGE *tsge;
SGE_SIMPLE32 *sge;
uint32_t flags;
int error;
memset(&fw_req_buf, 0, sizeof(fw_req_buf));
fw_req = (MSG_FW_UPLOAD *)fw_req_buf;
fw_req->ImageType = MPI_FW_UPLOAD_ITYPE_FW_IOC_MEM;
fw_req->Function = MPI_FUNCTION_FW_UPLOAD;
fw_req->MsgContext = htole32(MPT_REPLY_HANDLER_HANDSHAKE);
tsge = (FW_UPLOAD_TCSGE *)&fw_req->SGL;
tsge->DetailsLength = 12;
tsge->Flags = MPI_SGE_FLAGS_TRANSACTION_ELEMENT;
tsge->ImageSize = htole32(mpt->fw_image_size);
sge = (SGE_SIMPLE32 *)(tsge + 1);
flags = (MPI_SGE_FLAGS_LAST_ELEMENT | MPI_SGE_FLAGS_END_OF_BUFFER
| MPI_SGE_FLAGS_END_OF_LIST | MPI_SGE_FLAGS_SIMPLE_ELEMENT
| MPI_SGE_FLAGS_32_BIT_ADDRESSING | MPI_SGE_FLAGS_IOC_TO_HOST);
flags <<= MPI_SGE_FLAGS_SHIFT;
sge->FlagsLength = htole32(flags | mpt->fw_image_size);
sge->Address = htole32(mpt->fw_phys);
error = mpt_send_handshake_cmd(mpt, sizeof(fw_req_buf), &fw_req_buf);
if (error)
return(error);
error = mpt_recv_handshake_reply(mpt, sizeof(fw_reply), &fw_reply);
return (error);
}
static void
mpt_diag_outsl(struct mpt_softc *mpt, uint32_t addr,
uint32_t *data, bus_size_t len)
{
uint32_t *data_end;
data_end = data + (roundup2(len, sizeof(uint32_t)) / 4);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
pci_enable_io(mpt->dev, SYS_RES_IOPORT);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_pio_write(mpt, MPT_OFFSET_DIAG_ADDR, addr);
while (data != data_end) {
mpt_pio_write(mpt, MPT_OFFSET_DIAG_DATA, *data);
data++;
}
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
pci_disable_io(mpt->dev, SYS_RES_IOPORT);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
static int
mpt_download_fw(struct mpt_softc *mpt)
{
MpiFwHeader_t *fw_hdr;
int error;
uint32_t ext_offset;
uint32_t data;
mpt_prt(mpt, "Downloading Firmware - Image Size %d\n",
mpt->fw_image_size);
error = mpt_enable_diag_mode(mpt);
if (error != 0) {
mpt_prt(mpt, "Could not enter diagnostic mode!\n");
return (EIO);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_write(mpt, MPT_OFFSET_DIAGNOSTIC,
MPI_DIAG_RW_ENABLE|MPI_DIAG_DISABLE_ARM);
fw_hdr = (MpiFwHeader_t *)mpt->fw_image;
mpt_diag_outsl(mpt, fw_hdr->LoadStartAddress, (uint32_t*)fw_hdr,
fw_hdr->ImageSize);
ext_offset = fw_hdr->NextImageHeaderOffset;
while (ext_offset != 0) {
MpiExtImageHeader_t *ext;
ext = (MpiExtImageHeader_t *)((uintptr_t)fw_hdr + ext_offset);
ext_offset = ext->NextImageHeaderOffset;
mpt_diag_outsl(mpt, ext->LoadStartAddress, (uint32_t*)ext,
ext->ImageSize);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
pci_enable_io(mpt->dev, SYS_RES_IOPORT);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/* Setup the address to jump to on reset. */
mpt_pio_write(mpt, MPT_OFFSET_DIAG_ADDR, fw_hdr->IopResetRegAddr);
mpt_pio_write(mpt, MPT_OFFSET_DIAG_DATA, fw_hdr->IopResetVectorValue);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/*
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
* The controller sets the "flash bad" status after attempting
* to auto-boot from flash. Clear the status so that the controller
* will continue the boot process with our newly installed firmware.
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_pio_write(mpt, MPT_OFFSET_DIAG_ADDR, MPT_DIAG_MEM_CFG_BASE);
data = mpt_pio_read(mpt, MPT_OFFSET_DIAG_DATA) | MPT_DIAG_MEM_CFG_BADFL;
mpt_pio_write(mpt, MPT_OFFSET_DIAG_ADDR, MPT_DIAG_MEM_CFG_BASE);
mpt_pio_write(mpt, MPT_OFFSET_DIAG_DATA, data);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
pci_disable_io(mpt->dev, SYS_RES_IOPORT);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* Re-enable the processor and clear the boot halt flag.
*/
data = mpt_read(mpt, MPT_OFFSET_DIAGNOSTIC);
data &= ~(MPI_DIAG_PREVENT_IOC_BOOT|MPI_DIAG_DISABLE_ARM);
mpt_write(mpt, MPT_OFFSET_DIAGNOSTIC, data);
mpt_disable_diag_mode(mpt);
return (0);
}
/*
* Allocate/Initialize data structures for the controller. Called
* once at instance startup.
*/
static int
mpt_configure_ioc(struct mpt_softc *mpt)
{
MSG_PORT_FACTS_REPLY pfp;
MSG_IOC_FACTS_REPLY facts;
int try;
int needreset;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
uint32_t max_chain_depth;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
needreset = 0;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
for (try = 0; try < MPT_MAX_TRYS; try++) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/*
* No need to reset if the IOC is already in the READY state.
*
* Force reset if initialization failed previously.
* Note that a hard_reset of the second channel of a '929
* will stop operation of the first channel. Hopefully, if the
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
* first channel is ok, the second will not require a hard
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
* reset.
*/
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (needreset || MPT_STATE(mpt_rd_db(mpt)) !=
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
MPT_DB_STATE_READY) {
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (mpt_reset(mpt, FALSE) != MPT_OK) {
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
continue;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
needreset = 0;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
if (mpt_get_iocfacts(mpt, &facts) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_get_iocfacts failed\n");
needreset = 1;
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
continue;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt->mpt_global_credits = le16toh(facts.GlobalCredits);
mpt->request_frame_size = le16toh(facts.RequestFrameSize);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt->ioc_facts_flags = facts.Flags;
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "MPI Version=%d.%d.%d.%d\n",
le16toh(facts.MsgVersion) >> 8,
le16toh(facts.MsgVersion) & 0xFF,
le16toh(facts.HeaderVersion) >> 8,
le16toh(facts.HeaderVersion) & 0xFF);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
/*
* Now that we know request frame size, we can calculate
* the actual (reasonable) segment limit for read/write I/O.
*
* This limit is constrained by:
*
* + The size of each area we allocate per command (and how
* many chain segments we can fit into it).
* + The total number of areas we've set up.
* + The actual chain depth the card will allow.
*
* The first area's segment count is limited by the I/O request
* at the head of it. We cannot allocate realistically more
* than MPT_MAX_REQUESTS areas. Therefore, to account for both
* conditions, we'll just start out with MPT_MAX_REQUESTS-2.
*
*/
max_chain_depth = facts.MaxChainDepth;
/* total number of request areas we (can) allocate */
mpt->max_seg_cnt = MPT_MAX_REQUESTS(mpt) - 2;
/* converted to the number of chain areas possible */
mpt->max_seg_cnt *= MPT_NRFM(mpt);
/* limited by the number of chain areas the card will support */
if (mpt->max_seg_cnt > max_chain_depth) {
mpt_lprt(mpt, MPT_PRT_DEBUG,
"chain depth limited to %u (from %u)\n",
max_chain_depth, mpt->max_seg_cnt);
mpt->max_seg_cnt = max_chain_depth;
}
/* converted to the number of simple sges in chain segments. */
mpt->max_seg_cnt *= (MPT_NSGL(mpt) - 1);
mpt_lprt(mpt, MPT_PRT_DEBUG,
"Maximum Segment Count: %u\n", mpt->max_seg_cnt);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG,
"MsgLength=%u IOCNumber = %d\n",
facts.MsgLength, facts.IOCNumber);
mpt_lprt(mpt, MPT_PRT_DEBUG,
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
"IOCFACTS: GlobalCredits=%d BlockSize=%u bytes "
"Request Frame Size %u bytes Max Chain Depth %u\n",
mpt->mpt_global_credits, facts.BlockSize,
mpt->request_frame_size << 2, max_chain_depth);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG,
"IOCFACTS: Num Ports %d, FWImageSize %d, "
"Flags=%#x\n", facts.NumberOfPorts,
le32toh(facts.FWImageSize), facts.Flags);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
if ((facts.Flags & MPI_IOCFACTS_FLAGS_FW_DOWNLOAD_BOOT) != 0) {
struct mpt_map_info mi;
int error;
/*
* In some configurations, the IOC's firmware is
* stored in a shared piece of system NVRAM that
* is only accessable via the BIOS. In this
* case, the firmware keeps a copy of firmware in
* RAM until the OS driver retrieves it. Once
* retrieved, we are responsible for re-downloading
* the firmware after any hard-reset.
*/
mpt->fw_image_size = le32toh(facts.FWImageSize);
error = mpt_dma_tag_create(mpt, mpt->parent_dmat,
/*alignment*/1, /*boundary*/0,
/*lowaddr*/BUS_SPACE_MAXADDR_32BIT,
/*highaddr*/BUS_SPACE_MAXADDR, /*filter*/NULL,
/*filterarg*/NULL, mpt->fw_image_size,
/*nsegments*/1, /*maxsegsz*/mpt->fw_image_size,
/*flags*/0, &mpt->fw_dmat);
if (error != 0) {
mpt_prt(mpt, "cannot create fw dma tag\n");
return (ENOMEM);
}
error = bus_dmamem_alloc(mpt->fw_dmat,
(void **)&mpt->fw_image, BUS_DMA_NOWAIT,
&mpt->fw_dmap);
if (error != 0) {
mpt_prt(mpt, "cannot allocate fw mem.\n");
bus_dma_tag_destroy(mpt->fw_dmat);
return (ENOMEM);
}
mi.mpt = mpt;
mi.error = 0;
bus_dmamap_load(mpt->fw_dmat, mpt->fw_dmap,
mpt->fw_image, mpt->fw_image_size, mpt_map_rquest,
&mi, 0);
mpt->fw_phys = mi.phys;
error = mpt_upload_fw(mpt);
if (error != 0) {
mpt_prt(mpt, "fw upload failed.\n");
bus_dmamap_unload(mpt->fw_dmat, mpt->fw_dmap);
bus_dmamem_free(mpt->fw_dmat, mpt->fw_image,
mpt->fw_dmap);
bus_dma_tag_destroy(mpt->fw_dmat);
mpt->fw_image = NULL;
return (EIO);
}
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
if (mpt_get_portfacts(mpt, &pfp) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_get_portfacts failed\n");
needreset = 1;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
continue;
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG,
"PORTFACTS: Type %x PFlags %x IID %d MaxDev %d\n",
pfp.PortType, pfp.ProtocolFlags, pfp.PortSCSIID,
pfp.MaxDevices);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt->mpt_port_type = pfp.PortType;
mpt->mpt_proto_flags = pfp.ProtocolFlags;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
if (pfp.PortType != MPI_PORTFACTS_PORTTYPE_SCSI &&
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
pfp.PortType != MPI_PORTFACTS_PORTTYPE_SAS &&
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
pfp.PortType != MPI_PORTFACTS_PORTTYPE_FC) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "Unsupported Port Type (%x)\n",
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
pfp.PortType);
return (ENXIO);
}
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt->mpt_max_tgtcmds = le16toh(pfp.MaxPostedCmdBuffers);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
if (pfp.PortType == MPI_PORTFACTS_PORTTYPE_FC) {
mpt->is_fc = 1;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt->is_sas = 0;
} else if (pfp.PortType == MPI_PORTFACTS_PORTTYPE_SAS) {
mpt->is_fc = 0;
mpt->is_sas = 1;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
} else {
mpt->is_fc = 0;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt->is_sas = 0;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
}
mpt->mpt_ini_id = pfp.PortSCSIID;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt->mpt_max_devices = pfp.MaxDevices;
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
/*
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
* Set our expected role with what this port supports.
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
*/
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt->role = MPT_ROLE_NONE;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (pfp.ProtocolFlags & MPI_PORTFACTS_PROTOCOL_INITIATOR) {
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt->role |= MPT_ROLE_INITIATOR;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
if (pfp.ProtocolFlags & MPI_PORTFACTS_PROTOCOL_TARGET) {
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
mpt->role |= MPT_ROLE_TARGET;
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
}
A large set of changes: + Add boatloads of KASSERTs and *really* check out more locking issues (to catch recursions when we actually go to real locking in CAM soon). The KASSERTs also caught lots of other issues like using commands that were put back on free lists, etc. + Target mode: role setting is derived directly from port capabilities. There is no need to set a role any more. Some target mode resources are allocated early on (ELS), but target command buffer allocation is deferred until the first lun enable. + Fix some breakages I introduced with target mode in that some commands are *repeating* commands. That is, the reply shows up but the command isn't really done (we don't free it). We still need to take it off the pending list because when we resubmit it, bad things then happen. + Fix more of the way that timed out commands and bus reset is done. The actual TMF response code was being ignored. + For SPI, honor BIOS settings. This doesn't quite fix the problems we've seen where we can't seem to (re)negotiate U320 on all drives but avoids it instead by letting us honor the BIOS settings. I'm sure this is not quite right and will have to change again soon.
2006-04-11 16:47:30 +00:00
if (mpt->role == MPT_ROLE_NONE) {
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_prt(mpt, "port does not support either target or "
"initiator role\n");
return (ENXIO);
}
if (mpt_enable_ioc(mpt, 0) != MPT_OK) {
mpt_prt(mpt, "unable to initialize IOC\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
return (ENXIO);
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
}
/*
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
* Read IOC configuration information.
Work in progress toward fixing IM checked in after having lost one set to a peninsula power failure last night. After this, I can see both submembers and the raid volumes again, but speed negotiation is still broken. Add a mpt_raid_free_mem function to centralize the resource reclaim and fixed a small memory leak. Remove restriction on number of targets for systems with IM enabled- you can have setups that have both IM volumes as well as other devices. Fix target id selection for passthru and nonpastrhu cases. Move complete command dumpt to MPT_PRT_DEBUG1 level so that just setting debug level gets mostly informative albeit less verbose dumping.
2006-05-27 17:26:57 +00:00
*
* We need this to determine whether or not we have certain
* settings for Integrated Mirroring (e.g.).
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
*/
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_read_config_info_ioc(mpt);
A chunk of cleanup, both stylistic and substantive. We now also read configuration information for the SCSI cards- this allows us to try and say what the speed settings now are. Start, but not yet complete, the process of reorgs && #defines so that we can backport to RELENG_4 pretty soon.
2002-08-20 23:04:08 +00:00
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
/* Everything worked */
break;
}
if (try >= MPT_MAX_TRYS) {
Code cleanup: use mpt_prt instead of device_printf.
2002-09-24 21:33:43 +00:00
mpt_prt(mpt, "failed to initialize IOC");
Add support for the LSI-Logic Fusion/MP architecture. This is an architecture that present a thing message passing interface to the OS. You can query as to how many ports and what kind are attached and enable them and so on. A less grand view is that this is just another way to package SCSI (SPI or FC) and FC-IP into a one-driver interface set. This driver support the following hardware: LSI FC909: Single channel, 1Gbps, Fibre Channel (FC-SCSI only) LSI FC929: Dual Channel, 1-2Gbps, Fibre Channel (FC-SCSI only) LSI 53c1020: Single Channel, Ultra4 (320M) (Untested) LSI 53c1030: Dual Channel, Ultra4 (320M) Currently it's in fair shape, but expect a lot of changes over the next few weeks as it stabilizes. Credits: The driver is mostly from some folks from Jeff Roberson's company- I've been slowly migrating it to broader support that I it came to me as. The hardware used in developing support came from: FC909: LSI-Logic, Advansys (now Connetix) FC929: LSI-Logic 53c1030: Antares Microsystems (they make a very fine board!) MFC after: 3 weeks
2002-08-11 23:34:20 +00:00
return (EIO);
}
return (0);
}
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
static int
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
mpt_enable_ioc(struct mpt_softc *mpt, int portenable)
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
{
uint32_t pptr;
int val;
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
if (mpt_send_ioc_init(mpt, MPI_WHOINIT_HOST_DRIVER) != MPT_OK) {
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
mpt_prt(mpt, "mpt_send_ioc_init failed\n");
return (EIO);
}
mpt_lprt(mpt, MPT_PRT_DEBUG, "mpt_send_ioc_init ok\n");
if (mpt_wait_state(mpt, MPT_DB_STATE_RUNNING) != MPT_OK) {
mpt_prt(mpt, "IOC failed to go to run state\n");
return (ENXIO);
}
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
mpt_lprt(mpt, MPT_PRT_DEBUG, "IOC now at RUNSTATE\n");
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
/*
* Give it reply buffers
*
* Do *not* exceed global credits.
*/
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
for (val = 0, pptr = mpt->reply_phys;
(pptr + MPT_REPLY_SIZE) < (mpt->reply_phys + PAGE_SIZE);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
pptr += MPT_REPLY_SIZE) {
mpt_free_reply(mpt, pptr);
if (++val == mpt->mpt_global_credits - 1)
break;
}
/*
Fix some of the previus changes 'better'. There's something strange going on with async events. They seem to be be treated differently for different Fusion implementations. Some will really tell you when it's okay to free the request that started them. Some won't. Very disconcerting. This is particularily bad when the chip (FC in this case) tells you in the reply that it's not a continuation reply, which means you can free the request that its associated with. However, if you do that, I've found that additional async event replies come back for that message context after you freed it. Very Bad Things Happen. Put in a reply register debounce. Warn about out of range context indices. Use more MPILIB defines where possible. Replace bzero with memset. Add tons more KASSERTS. Do a *lot* more request free list auditting and serial number usages. Get rid of the warning about the short IOC Facts Reply. Go back to 16 bits of context index. Do a lot more target state auditting as well. Make a tag out of not only the ioindex but the request index as well and worry less about keeping a full serial number.
2006-04-01 07:12:18 +00:00
* Enable the port if asked. This is only done if we're resetting
* the IOC after initial startup.
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
*/
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (portenable) {
/*
* Enable asynchronous event reporting
*/
mpt_send_event_request(mpt, 1);
Do initial cut of SAS HBA support. These controllers (106X) seem to support automatically both SATA and SAS drives. The async SAS event handling we catch but ignore at present (so automagic attach/detach isn't hooked up yet). Do 64 bit PCI support- we can now work on systems with > 4GB of memory. Do large transfer support- we now can support up to reported chain depth, or the length of our request area. We simply allocate additional request elements when we would run out of room for chain lists. Tested on Ultra320, FC and SAS controllers on AMD64 and i386 platforms. There were no RAID cards available for me to regression test. The error recovery for this driver still is pretty bad.
2006-02-11 01:35:29 +00:00
Some fairly major changes to this driver. A) Fibre Channel Target Mode support mostly works (SAS/SPI won't be too far behind). I'd say that this probably works just about as well as isp(4) does right now. Still, it and isp(4) and the whole target mode stack need a bit of tightening. B) The startup sequence has been changed so that after all attaches are done, a set of enable functions are called. The idea here is that the attaches do whatever needs to be done *prior* to a port being enabled and the enables do what need to be done for enabling stuff for a port after it's been enabled. This means that we also have events handled by their proper handlers as we start up. C) Conditional code that means that this driver goes back all the way to RELENG_4 in terms of support. D) Quite a lot of little nitty bug fixes- some discovered by doing RELENG_4 support. We've been living under Giant *waaaayyyyy* too long and it's made some of us (me) sloppy. E) Some shutdown hook stuff that makes sure we don't blow up during a reboot (like by the arrival of a new command from an initiator). There's been some testing and LINT checking, but not as complete as would be liked. Regression testing with Fusion RAID instances has not been possible. Caveat Emptor. Sponsored by: LSI-Logic.
2006-03-25 07:08:27 +00:00
if (mpt_send_port_enable(mpt, 0) != MPT_OK) {
mpt_prt(mpt, "failed to enable port 0\n");
return (ENXIO);
}
}
Fix mpt_reset to try mpt_hard_reset more than once, and to try mpt_soft_reset more than once. And to wait for MPT_DB_STATE_READY twice. I mean, this is crucial- give the IOC a chance to get ready. If mpt_reset is called to reinit things, and we succeed, make sure to re-enable interrupts. This is what has mostly led to system lockup after having to hard reset the chip. Also, if we think that interrupts aren't function in mpt_cam_timeout, for goodness sake, turn them on again. In read_cfg_header, return distinguishing errnos so the caller can decide what's an error. It's *not* an error to fail to read a RAID page from a non-RAID capable device like the FC929X. Some whitespace fixes (removing spaces from ends of lines).
2006-02-28 07:44:50 +00:00
return (MPT_OK);
Massive overhaul of MPT Fusion driver: o Add timeout error recovery (from a thread context to avoid the deferral of other critical interrupts). o Properly recover commands across controller reset events. o Update the driver to handle events and status codes that have been added to the MPI spec since the driver was originally written. o Make the driver more modular to improve maintainability and support dynamic "personality" registration (e.g. SCSI Initiator, RAID, SAS, FC, etc). o Shorten and simplify the common I/O path to improve driver performance. o Add RAID volume and RAID member state/settings reporting. o Add periodic volume resynchronization status reporting. o Add support for sysctl tunable resync rate, member write cache enable, and volume transaction queue depth. Sponsored by ---------------- Avid Technologies Inc: SCSI error recovery, driver re-organization, update of MPI library headers, portions of dynamic personality registration, and misc bug fixes. Wheel Open Technologies: RAID event notification, RAID member pass-thru support, firmware upload/download support, enhanced RAID resync speed, portions of dynamic personality registration, and misc bug fixes. Detailed Changes ================ mpt.c mpt_cam.c mpt_raid.c mpt_pci.c: o Add support for personality modules. Each module exports load, and unload module scope methods as well as probe, attach, event, reset, shutdown, and detach per-device instance methods mpt.c mpt.h mpt_pci.c: o The driver now associates a callback function (via an index) with every transaction submitted to the controller. This allows the main interrupt handler to absolve itself of any knowledge of individual transaction/response types by simply calling the callback function "registered" for the transaction. We use a callback index instead of a callback function pointer in each requests so we can properly handle responses (e.g. event notifications) that are not associated with a transaction. Personality modules dynamically register their callbacks with the driver core to receive the callback index to use for their handlers. o Move the interrupt handler into mpt.c. The ISR algorithm is bus transport and OS independent and thus had no reason to be in mpt_pci.c. o Simplify configuration message reply handling by copying reply frame data for the requester and storing completion status in the original request structure. o Add the mpt_complete_request_chain() helper method and use it to implement reset handlers that must abort transactions. o Keep track of all pending requests on the new requests_pending_list in the softc. o Add default handlers to mpt.c to handle generic event notifications and controller reset activities. The event handler code is largely the same as in the original driver. The reset handler is new and terminates any pending transactions with a status code indicating the controller needs to be re-initialized. o Add some endian support to the driver. A complete audit is still required for this driver to have any hope of operating in a big-endian environment. o Use inttypes.h and __inline. Come closer to being style(9) compliant. o Remove extraneous use of typedefs. o Convert request state from a strict enumeration to a series of flags. This allows us to, for example, tag transactions that have timed-out while retaining the state that the transaction is still in-flight on the controller. o Add mpt_wait_req() which allows a caller to poll or sleep for the completion of a request. Use this to simplify and factor code out from many initialization routines. We also use this to sleep for task management request completions in our CAM timeout handler. mpt.c: o Correct a bug in the event handler where request structures were freed even if the request reply was marked as a continuation reply. Continuation replies indicate that the controller still owns the request and freeing these replies prematurely corrupted controller state. o Implement firmware upload and download. On controllers that do not have dedicated NVRAM (as in the Sun v20/v40z), the firmware image is downloaded to the controller by the system BIOS. This image occupies precious controller RAM space until the host driver fetches the image, reducing the number of concurrent I/Os the controller can processes. The uploaded image is used to re-program the controller during hard reset events since the controller cannot fetch the firmware on its own. Implementing this feature allows much higher queue depths when RAID volumes are configured. o Changed configuration page accessors to allow threads to sleep rather than busy wait for completion. o Removed hard coded data transfer sizes from configuration page routines so that RAID configuration page processing is possible. mpt_reg.h: o Move controller register definitions into a separate file. mpt.h: o Re-arrange includes to allow inlined functions to be defined in mpt.h. o Add reply, event, and reset handler definitions. o Add softc fields for handling timeout and controller reset recovery. mpt_cam.c: o Move mpt_freebsd.c to mpt_cam.c. Move all core functionality, such as event handling, into mpt.c leaving only CAM SCSI support here. o Revamp completion handler to provide correct CAM status for all currently defined SCSI MPI message result codes. o Register event and reset handlers with the MPT core. Modify the event handler to notify CAM of bus reset events. The controller reset handler will abort any transactions that have timed out. All other pending CAM transactions are correctly aborted by the core driver's reset handler. o Allocate a single request up front to perform task management operations. This guarantees that we can always perform a TMF operation even when the controller is saturated with other operations. The single request also serves as a perfect mechanism of guaranteeing that only a single TMF is in flight at a time - something that is required according to the MPT Fusion documentation. o Add a helper function for issuing task management requests to the controller. This is used to abort individual requests or perform a bus reset. o Modify the CAM XPT_BUS_RESET ccb handler to wait for and properly handle the status of the bus reset task management frame used to reset the bus. The previous code assumed that the reset request would always succeed. o Add timeout recovery support. When a timeout occurs, the timed-out request is added to a queue to be processed by our recovery thread and the thread is woken up. The recovery thread processes timed-out command serially, attempting first to abort them and then falling back to a bus reset if an abort fails. o Add calls to mpt_reset() to reset the controller if any handshake command, bus reset attempt or abort attempt fails due to a timeout. o Export a secondary "bus" to CAM that exposes all volume drive members as pass-thru devices, allowing CAM to perform proper speed negotiation to hidden devices. o Add a CAM async event handler tracking the AC_FOUND_DEVICE event. Use this to trigger calls to set the per-volume queue depth once the volume is fully registered with CAM. This is required to avoid hitting firmware limits on volume queue depth. Exceeding the limit causes the firmware to hang. mpt_cam.h: o Add several helper functions for interfacing to CAM and performing timeout recovery. mpt_pci.c: o Disable interrupts on the controller before registering and enabling interrupt delivery to the OS. Otherwise we risk receiving interrupts before the driver is ready to receive them. o Make use of compatibility macros that allow the driver to be compiled under 4.x and 5.x. mpt_raid.c: o Add a per-controller instance RAID thread to perform settings changes and query status (minimizes CPU busy wait loops). o Use a shutdown handler to disable "Member Write Cache Enable" (MWCE) setting for RAID arrays set to enable MWCE During Rebuild. o Change reply handler function signature to allow handlers to defer the deletion of reply frames. Use this to allow the event reply handler to queue up events that need to be acked if no resources are available to immediately ack an event. Queued events are processed in mpt_free_request() where resources are freed. This avoids a panic on resource shortage. o Parse and print out RAID controller capabilities during driver probe. o Define, allocate, and maintain RAID data structures for volumes, hidden member physical disks and spare disks. o Add dynamic sysctls for per-instance setting of the log level, array resync rate, array member cache enable, and volume queue depth. mpt_debug.c: o Add mpt_lprt and mpt_lprtc for printing diagnostics conditioned on a particular log level to aid in tracking down driver issues. o Add mpt_decode_value() which parses the bits in an integer value based on a parsing table (mask, value, name string, tuples). mpilib/*: o Update mpi library header files to latest distribution from LSI. Submitted by: gibbs Approved by: re
2005-07-10 15:05:39 +00:00
}
Reference in New Issue Copy Permalink