d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
933fab9d73
freebsd-dev/sys/dev/ciss/ciss.c

4660 lines
133 KiB
C
Raw Normal View History

Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*-
* Copyright (c) 2001 Michael Smith
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
* Copyright (c) 2004 Paul Saab
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* All rights reserved.
*
* 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 the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* 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.
*
* $FreeBSD$
*/
/*
* Common Interface for SCSI-3 Support driver.
*
* CISS claims to provide a common interface between a generic SCSI
* transport and an intelligent host adapter.
*
* This driver supports CISS as defined in the document "CISS Command
* Interface for SCSI-3 Support Open Specification", Version 1.04,
* Valence Number 1, dated 20001127, produced by Compaq Computer
* Corporation. This document appears to be a hastily and somewhat
* arbitrarlily cut-down version of a larger (and probably even more
* chaotic and inconsistent) Compaq internal document. Various
* details were also gleaned from Compaq's "cciss" driver for Linux.
*
* We provide a shim layer between the CISS interface and CAM,
* offloading most of the queueing and being-a-disk chores onto CAM.
* Entry to the driver is via the PCI bus attachment (ciss_probe,
* ciss_attach, etc) and via the CAM interface (ciss_cam_action,
* ciss_cam_poll). The Compaq CISS adapters are, however, poor SCSI
* citizens and we have to fake up some responses to get reasonable
* behaviour out of them. In addition, the CISS command set is by no
* means adequate to support the functionality of a RAID controller,
* and thus the supported Compaq adapters utilise portions of the
* control protocol from earlier Compaq adapter families.
*
* Note that we only support the "simple" transport layer over PCI.
* This interface (ab)uses the I2O register set (specifically the post
* queues) to exchange commands with the adapter. Other interfaces
* are available, but we aren't supposed to know about them, and it is
* dubious whether they would provide major performance improvements
* except under extreme load.
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
*
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* Currently the only supported CISS adapters are the Compaq Smart
* Array 5* series (5300, 5i, 532). Even with only three adapters,
* Compaq still manage to have interface variations.
*
*
* Thanks must go to Fred Harris and Darryl DeVinney at Compaq, as
* well as Paul Saab at Yahoo! for their assistance in making this
* driver happen.
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
*
* More thanks must go to John Cagle at HP for the countless hours
* spent making this driver "work" with the MSA* series storage
* enclosures. Without his help (and nagging), this driver could not
* be used with these enclosures.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/stat.h>
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#include <sys/kthread.h>
#include <sys/queue.h>
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
#include <sys/sysctl.h>
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_periph.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt_sim.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
#include <machine/bus.h>
#include <machine/endian.h>
#include <machine/resource.h>
#include <sys/rman.h>
Prefer new location of pci include files (which have only been in the tree for two or more years now), except in a few places where there's code to be compatible with older versions of FreeBSD.
2003-08-22 05:54:52 +00:00
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
#include <dev/ciss/cissreg.h>
#include <dev/ciss/cissio.h>
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
#include <dev/ciss/cissvar.h>
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
Mark MALLOC_DEFINEs static that have no corresponding MALLOC_DECLAREs. This means that their use is restricted to a single C file.
2011-11-07 06:44:47 +00:00
static MALLOC_DEFINE(CISS_MALLOC_CLASS, "ciss_data",
"ciss internal data buffers");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* pci interface */
static int ciss_lookup(device_t dev);
static int ciss_probe(device_t dev);
static int ciss_attach(device_t dev);
static int ciss_detach(device_t dev);
static int ciss_shutdown(device_t dev);
/* (de)initialisation functions, control wrappers */
static int ciss_init_pci(struct ciss_softc *sc);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
static int ciss_setup_msix(struct ciss_softc *sc);
static int ciss_init_perf(struct ciss_softc *sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static int ciss_wait_adapter(struct ciss_softc *sc);
static int ciss_flush_adapter(struct ciss_softc *sc);
static int ciss_init_requests(struct ciss_softc *sc);
static void ciss_command_map_helper(void *arg, bus_dma_segment_t *segs,
int nseg, int error);
static int ciss_identify_adapter(struct ciss_softc *sc);
static int ciss_init_logical(struct ciss_softc *sc);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
static int ciss_init_physical(struct ciss_softc *sc);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
static int ciss_filter_physical(struct ciss_softc *sc, struct ciss_lun_report *cll);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static int ciss_identify_logical(struct ciss_softc *sc, struct ciss_ldrive *ld);
static int ciss_get_ldrive_status(struct ciss_softc *sc, struct ciss_ldrive *ld);
static int ciss_update_config(struct ciss_softc *sc);
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
static int ciss_accept_media(struct ciss_softc *sc, struct ciss_ldrive *ld);
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
static void ciss_init_sysctl(struct ciss_softc *sc);
static void ciss_soft_reset(struct ciss_softc *sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void ciss_free(struct ciss_softc *sc);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
static void ciss_spawn_notify_thread(struct ciss_softc *sc);
static void ciss_kill_notify_thread(struct ciss_softc *sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* request submission/completion */
static int ciss_start(struct ciss_request *cr);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
static void ciss_done(struct ciss_softc *sc, cr_qhead_t *qh);
static void ciss_perf_done(struct ciss_softc *sc, cr_qhead_t *qh);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void ciss_intr(void *arg);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
static void ciss_perf_intr(void *arg);
static void ciss_perf_msi_intr(void *arg);
static void ciss_complete(struct ciss_softc *sc, cr_qhead_t *qh);
static int _ciss_report_request(struct ciss_request *cr, int *command_status, int *scsi_status, const char *func);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static int ciss_synch_request(struct ciss_request *cr, int timeout);
static int ciss_poll_request(struct ciss_request *cr, int timeout);
static int ciss_wait_request(struct ciss_request *cr, int timeout);
Fix warning; line 1640: ciss_abort_request defined but not used.
2001-12-18 08:01:48 +00:00
#if 0
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static int ciss_abort_request(struct ciss_request *cr);
Fix warning; line 1640: ciss_abort_request defined but not used.
2001-12-18 08:01:48 +00:00
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* request queueing */
static int ciss_get_request(struct ciss_softc *sc, struct ciss_request **crp);
static void ciss_preen_command(struct ciss_request *cr);
static void ciss_release_request(struct ciss_request *cr);
/* request helpers */
static int ciss_get_bmic_request(struct ciss_softc *sc, struct ciss_request **crp,
int opcode, void **bufp, size_t bufsize);
static int ciss_user_command(struct ciss_softc *sc, IOCTL_Command_struct *ioc);
/* DMA map/unmap */
static int ciss_map_request(struct ciss_request *cr);
static void ciss_request_map_helper(void *arg, bus_dma_segment_t *segs,
int nseg, int error);
static void ciss_unmap_request(struct ciss_request *cr);
/* CAM interface */
static int ciss_cam_init(struct ciss_softc *sc);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
static void ciss_cam_rescan_target(struct ciss_softc *sc,
int bus, int target);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void ciss_cam_action(struct cam_sim *sim, union ccb *ccb);
static int ciss_cam_action_io(struct cam_sim *sim, struct ccb_scsiio *csio);
static int ciss_cam_emulate(struct ciss_softc *sc, struct ccb_scsiio *csio);
static void ciss_cam_poll(struct cam_sim *sim);
static void ciss_cam_complete(struct ciss_request *cr);
static void ciss_cam_complete_fixup(struct ciss_softc *sc, struct ccb_scsiio *csio);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
static struct cam_periph *ciss_find_periph(struct ciss_softc *sc,
int bus, int target);
static int ciss_name_device(struct ciss_softc *sc, int bus, int target);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* periodic status monitoring */
static void ciss_periodic(void *arg);
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
static void ciss_nop_complete(struct ciss_request *cr);
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
static void ciss_disable_adapter(struct ciss_softc *sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void ciss_notify_event(struct ciss_softc *sc);
static void ciss_notify_complete(struct ciss_request *cr);
static int ciss_notify_abort(struct ciss_softc *sc);
static int ciss_notify_abort_bmic(struct ciss_softc *sc);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
static void ciss_notify_hotplug(struct ciss_softc *sc, struct ciss_notify *cn);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void ciss_notify_logical(struct ciss_softc *sc, struct ciss_notify *cn);
static void ciss_notify_physical(struct ciss_softc *sc, struct ciss_notify *cn);
/* debugging output */
static void ciss_print_request(struct ciss_request *cr);
static void ciss_print_ldrive(struct ciss_softc *sc, struct ciss_ldrive *ld);
static const char *ciss_name_ldrive_status(int status);
static int ciss_decode_ldrive_status(int status);
static const char *ciss_name_ldrive_org(int org);
static const char *ciss_name_command_status(int status);
/*
* PCI bus interface.
*/
static device_method_t ciss_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, ciss_probe),
DEVMETHOD(device_attach, ciss_attach),
DEVMETHOD(device_detach, ciss_detach),
DEVMETHOD(device_shutdown, ciss_shutdown),
{ 0, 0 }
};
static driver_t ciss_pci_driver = {
"ciss",
ciss_methods,
sizeof(struct ciss_softc)
};
static devclass_t ciss_devclass;
DRIVER_MODULE(ciss, pci, ciss_pci_driver, ciss_devclass, 0, 0);
o ciss.ko depends on cam and pci. PR: kern/105989 Submitted by: nork MFC after: 1 month
2006-11-30 15:14:17 +00:00
MODULE_DEPEND(ciss, cam, 1, 1, 1);
MODULE_DEPEND(ciss, pci, 1, 1, 1);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Control device interface.
*/
static d_open_t ciss_open;
static d_close_t ciss_close;
static d_ioctl_t ciss_ioctl;
static struct cdevsw ciss_cdevsw = {
Device megapatch 4/6: Introduce d_version field in struct cdevsw, this must always be initialized to D_VERSION. Flip sense of D_NOGIANT flag to D_NEEDGIANT, this involves removing four D_NOGIANT flags and adding 145 D_NEEDGIANT flags.
2004-02-21 21:10:55 +00:00
.d_version = D_VERSION,
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
.d_flags = 0,
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_open = ciss_open,
.d_close = ciss_close,
.d_ioctl = ciss_ioctl,
.d_name = "ciss",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
};
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
/*
* This tunable can be set at boot time and controls whether physical devices
* that are marked hidden by the firmware should be exposed anyways.
*/
static unsigned int ciss_expose_hidden_physical = 0;
TUNABLE_INT("hw.ciss.expose_hidden_physical", &ciss_expose_hidden_physical);
Add `hw.ciss.nop_message_heartbeat' tunable (default disabled) for NOP-message polling in ciss_periodic(). Note that setting the tunable to non-zero can be workaround only for `ADAPTER HEARTBEAT FAILED' problem, and may freeze the system w/o the problem. Reviewed by: scottl Reported by: Attila Nagy MFC after: 3 days
2008-02-10 06:21:52 +00:00
static unsigned int ciss_nop_message_heartbeat = 0;
TUNABLE_INT("hw.ciss.nop_message_heartbeat", &ciss_nop_message_heartbeat);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
/*
* This tunable can force a particular transport to be used:
* <= 0 : use default
* 1 : force simple
* 2 : force performant
*/
static int ciss_force_transport = 0;
TUNABLE_INT("hw.ciss.force_transport", &ciss_force_transport);
/*
* This tunable can force a particular interrupt delivery method to be used:
* <= 0 : use default
* 1 : force INTx
* 2 : force MSIX
*/
static int ciss_force_interrupt = 0;
TUNABLE_INT("hw.ciss.force_interrupt", &ciss_force_interrupt);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* CISS adapters amazingly don't have a defined programming interface
* value. (One could say some very despairing things about PCI and
* people just not getting the general idea.) So we are forced to
* stick with matching against subvendor/subdevice, and thus have to
* be updated for every new CISS adapter that appears.
*/
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
#define CISS_BOARD_UNKNWON 0
#define CISS_BOARD_SA5 1
#define CISS_BOARD_SA5B 2
#define CISS_BOARD_NOMSI (1<<4)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static struct
{
u_int16_t subvendor;
u_int16_t subdevice;
int flags;
char *desc;
} ciss_vendor_data[] = {
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
{ 0x0e11, 0x4070, CISS_BOARD_SA5|CISS_BOARD_NOMSI, "Compaq Smart Array 5300" },
{ 0x0e11, 0x4080, CISS_BOARD_SA5B|CISS_BOARD_NOMSI, "Compaq Smart Array 5i" },
{ 0x0e11, 0x4082, CISS_BOARD_SA5B|CISS_BOARD_NOMSI, "Compaq Smart Array 532" },
{ 0x0e11, 0x4083, CISS_BOARD_SA5B|CISS_BOARD_NOMSI, "HP Smart Array 5312" },
Add PCID for 6i controller Submitted by: John Cagle <john.cagle@hp.com>
2003-11-07 03:01:48 +00:00
{ 0x0e11, 0x4091, CISS_BOARD_SA5, "HP Smart Array 6i" },
Add support for the HP Smart Array 6400 EM Change the interrupt mask for the Smart Array 6xxx controllers after discussions w/ HP.
2003-04-03 20:09:27 +00:00
{ 0x0e11, 0x409A, CISS_BOARD_SA5, "HP Smart Array 641" },
{ 0x0e11, 0x409B, CISS_BOARD_SA5, "HP Smart Array 642" },
{ 0x0e11, 0x409C, CISS_BOARD_SA5, "HP Smart Array 6400" },
{ 0x0e11, 0x409D, CISS_BOARD_SA5, "HP Smart Array 6400 EM" },
Update PCI ids to add the E200, E200i, P400, and P400i storage controllers. Remove the E400 since it is not a real product. Submitted by: HP
2005-11-11 16:45:33 +00:00
{ 0x103C, 0x3211, CISS_BOARD_SA5, "HP Smart Array E200i" },
{ 0x103C, 0x3212, CISS_BOARD_SA5, "HP Smart Array E200" },
{ 0x103C, 0x3213, CISS_BOARD_SA5, "HP Smart Array E200i" },
{ 0x103C, 0x3214, CISS_BOARD_SA5, "HP Smart Array E200i" },
{ 0x103C, 0x3215, CISS_BOARD_SA5, "HP Smart Array E200i" },
Add pci id's for the SmartArray 6422 and V100 controllers. Also add a whole bunch of pci id's for future controllers. Submitted by: John Cagle <first.last@hp.com>
2004-06-05 18:12:56 +00:00
{ 0x103C, 0x3220, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x3222, CISS_BOARD_SA5, "HP Smart Array" },
Add support for the P600 and name the E400.
2005-04-28 14:40:23 +00:00
{ 0x103C, 0x3223, CISS_BOARD_SA5, "HP Smart Array P800" },
Remove 6422, V100 and add the P600. Submitted by: John Cagle
2005-01-20 22:51:38 +00:00
{ 0x103C, 0x3225, CISS_BOARD_SA5, "HP Smart Array P600" },
Add pci id's for the SmartArray 6422 and V100 controllers. Also add a whole bunch of pci id's for future controllers. Submitted by: John Cagle <first.last@hp.com>
2004-06-05 18:12:56 +00:00
{ 0x103C, 0x3230, CISS_BOARD_SA5, "HP Smart Array" },
Update PCI ids to add the E200, E200i, P400, and P400i storage controllers. Remove the E400 since it is not a real product. Submitted by: HP
2005-11-11 16:45:33 +00:00
{ 0x103C, 0x3231, CISS_BOARD_SA5, "HP Smart Array" },
Add pci id's for the SmartArray 6422 and V100 controllers. Also add a whole bunch of pci id's for future controllers. Submitted by: John Cagle <first.last@hp.com>
2004-06-05 18:12:56 +00:00
{ 0x103C, 0x3232, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x3233, CISS_BOARD_SA5, "HP Smart Array" },
Update PCI ids to add the E200, E200i, P400, and P400i storage controllers. Remove the E400 since it is not a real product. Submitted by: HP
2005-11-11 16:45:33 +00:00
{ 0x103C, 0x3234, CISS_BOARD_SA5, "HP Smart Array P400" },
{ 0x103C, 0x3235, CISS_BOARD_SA5, "HP Smart Array P400i" },
Add pci id's for the SmartArray 6422 and V100 controllers. Also add a whole bunch of pci id's for future controllers. Submitted by: John Cagle <first.last@hp.com>
2004-06-05 18:12:56 +00:00
{ 0x103C, 0x3236, CISS_BOARD_SA5, "HP Smart Array" },
Updated PCI ID's from the vendor
2009-05-19 01:41:11 +00:00
{ 0x103C, 0x3237, CISS_BOARD_SA5, "HP Smart Array E500" },
Add pci id's for the SmartArray 6422 and V100 controllers. Also add a whole bunch of pci id's for future controllers. Submitted by: John Cagle <first.last@hp.com>
2004-06-05 18:12:56 +00:00
{ 0x103C, 0x3238, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x3239, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x323A, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x323B, CISS_BOARD_SA5, "HP Smart Array" },
{ 0x103C, 0x323C, CISS_BOARD_SA5, "HP Smart Array" },
Updated PCI ID's from the vendor
2009-05-19 01:41:11 +00:00
{ 0x103C, 0x323D, CISS_BOARD_SA5, "HP Smart Array P700m" },
Add support for the P212, P410, P410i, P411, and P812 HP Smart Array controllers. Submitted by: Scott Benesh at HP
2008-06-10 17:51:51 +00:00
{ 0x103C, 0x3241, CISS_BOARD_SA5, "HP Smart Array P212" },
{ 0x103C, 0x3243, CISS_BOARD_SA5, "HP Smart Array P410" },
{ 0x103C, 0x3245, CISS_BOARD_SA5, "HP Smart Array P410i" },
{ 0x103C, 0x3247, CISS_BOARD_SA5, "HP Smart Array P411" },
{ 0x103C, 0x3249, CISS_BOARD_SA5, "HP Smart Array P812" },
Updated PCI ID's from the vendor
2009-05-19 01:41:11 +00:00
{ 0x103C, 0x324A, CISS_BOARD_SA5, "HP Smart Array P712m" },
{ 0x103C, 0x324B, CISS_BOARD_SA5, "HP Smart Array" },
Update PCI IDs for ciss controllers that are supported by this driver. Submitted by: scott.benesh@hp.com Obtained from: Yahoo! Inc and HP MFC after: 1 week
2012-01-26 17:04:17 +00:00
{ 0x103C, 0x3350, CISS_BOARD_SA5, "HP Smart Array P222" },
{ 0x103C, 0x3351, CISS_BOARD_SA5, "HP Smart Array P420" },
{ 0x103C, 0x3352, CISS_BOARD_SA5, "HP Smart Array P421" },
{ 0x103C, 0x3353, CISS_BOARD_SA5, "HP Smart Array P822" },
{ 0x103C, 0x3354, CISS_BOARD_SA5, "HP Smart Array P420i" },
{ 0x103C, 0x3355, CISS_BOARD_SA5, "HP Smart Array P220i" },
{ 0x103C, 0x3356, CISS_BOARD_SA5, "HP Smart Array P721m" },
Replace (ab)uses of "NULL" where "0" is really meant.
2002-08-22 21:24:01 +00:00
{ 0, 0, 0, NULL }
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
};
/************************************************************************
* Find a match for the device in our list of known adapters.
*/
static int
ciss_lookup(device_t dev)
{
int i;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
for (i = 0; ciss_vendor_data[i].desc != NULL; i++)
if ((pci_get_subvendor(dev) == ciss_vendor_data[i].subvendor) &&
(pci_get_subdevice(dev) == ciss_vendor_data[i].subdevice)) {
return(i);
}
return(-1);
}
/************************************************************************
* Match a known CISS adapter.
*/
static int
ciss_probe(device_t dev)
{
int i;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
i = ciss_lookup(dev);
if (i != -1) {
device_set_desc(dev, ciss_vendor_data[i].desc);
Use BUS_PROBE_DEFAULT for pci probe return value
2005-03-05 18:30:12 +00:00
return(BUS_PROBE_DEFAULT);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
return(ENOENT);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Attach the driver to this adapter.
*/
static int
ciss_attach(device_t dev)
{
struct ciss_softc *sc;
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
int error;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
#ifdef CISS_DEBUG
/* print structure/union sizes */
debug_struct(ciss_command);
debug_struct(ciss_header);
debug_union(ciss_device_address);
debug_struct(ciss_cdb);
debug_struct(ciss_report_cdb);
debug_struct(ciss_notify_cdb);
debug_struct(ciss_notify);
debug_struct(ciss_message_cdb);
debug_struct(ciss_error_info_pointer);
debug_struct(ciss_error_info);
debug_struct(ciss_sg_entry);
debug_struct(ciss_config_table);
debug_struct(ciss_bmic_cdb);
debug_struct(ciss_bmic_id_ldrive);
debug_struct(ciss_bmic_id_lstatus);
debug_struct(ciss_bmic_id_table);
debug_struct(ciss_bmic_id_pdrive);
debug_struct(ciss_bmic_blink_pdrive);
debug_struct(ciss_bmic_flush_cache);
debug_const(CISS_MAX_REQUESTS);
debug_const(CISS_MAX_LOGICAL);
debug_const(CISS_INTERRUPT_COALESCE_DELAY);
debug_const(CISS_INTERRUPT_COALESCE_COUNT);
debug_const(CISS_COMMAND_ALLOC_SIZE);
debug_const(CISS_COMMAND_SG_LENGTH);
debug_type(cciss_pci_info_struct);
debug_type(cciss_coalint_struct);
debug_type(cciss_coalint_struct);
debug_type(NodeName_type);
debug_type(NodeName_type);
debug_type(Heartbeat_type);
debug_type(BusTypes_type);
debug_type(FirmwareVer_type);
debug_type(DriverVer_type);
debug_type(IOCTL_Command_struct);
#endif
sc = device_get_softc(dev);
sc->ciss_dev = dev;
Don't panic due to unlocking an unowned mutex if we fail during attach. PR: kern/139053 Reviewed by: scottl Approved by: ed (mentor) MFC after: 2 weeks
2009-12-17 17:46:08 +00:00
mtx_init(&sc->ciss_mtx, "cissmtx", NULL, MTX_DEF);
Initialize the callout structure earlier in attach before calling any routines that can fail since ciss_free() always tries to stop and drain the callout.
2010-04-28 18:49:45 +00:00
callout_init_mtx(&sc->ciss_periodic, &sc->ciss_mtx, 0);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Do PCI-specific init.
*/
if ((error = ciss_init_pci(sc)) != 0)
goto out;
/*
* Initialise driver queues.
*/
ciss_initq_free(sc);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_initq_notify(sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
/*
* Initalize device sysctls.
*/
ciss_init_sysctl(sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Initialise command/request pool.
*/
if ((error = ciss_init_requests(sc)) != 0)
goto out;
/*
* Get adapter information.
*/
if ((error = ciss_identify_adapter(sc)) != 0)
goto out;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
/*
* Find all the physical devices.
*/
if ((error = ciss_init_physical(sc)) != 0)
goto out;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Build our private table of logical devices.
*/
if ((error = ciss_init_logical(sc)) != 0)
goto out;
/*
* Enable interrupts so that the CAM scan can complete.
*/
CISS_TL_SIMPLE_ENABLE_INTERRUPTS(sc);
/*
* Initialise the CAM interface.
*/
if ((error = ciss_cam_init(sc)) != 0)
goto out;
/*
* Start the heartbeat routine and event chain.
*/
ciss_periodic(sc);
/*
* Create the control device.
*/
sc->ciss_dev_t = make_dev(&ciss_cdevsw, device_get_unit(sc->ciss_dev),
UID_ROOT, GID_OPERATOR, S_IRUSR | S_IWUSR,
"ciss%d", device_get_unit(sc->ciss_dev));
sc->ciss_dev_t->si_drv1 = sc;
/*
* The adapter is running; synchronous commands can now sleep
* waiting for an interrupt to signal completion.
*/
sc->ciss_flags |= CISS_FLAG_RUNNING;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_spawn_notify_thread(sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
error = 0;
out:
Don't panic due to unlocking an unowned mutex if we fail during attach. PR: kern/139053 Reviewed by: scottl Approved by: ed (mentor) MFC after: 2 weeks
2009-12-17 17:46:08 +00:00
if (error != 0) {
/* ciss_free() expects the mutex to be held */
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_free(sc);
Don't panic due to unlocking an unowned mutex if we fail during attach. PR: kern/139053 Reviewed by: scottl Approved by: ed (mentor) MFC after: 2 weeks
2009-12-17 17:46:08 +00:00
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(error);
}
/************************************************************************
* Detach the driver from this adapter.
*/
static int
ciss_detach(device_t dev)
{
struct ciss_softc *sc = device_get_softc(dev);
debug_called(1);
Don't allow the driver to be unloaded if the device node is open.
2004-04-14 19:45:07 +00:00
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
if (sc->ciss_flags & CISS_FLAG_CONTROL_OPEN) {
mtx_unlock(&sc->ciss_mtx);
Don't allow the driver to be unloaded if the device node is open.
2004-04-14 19:45:07 +00:00
return (EBUSY);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
}
Don't allow the driver to be unloaded if the device node is open.
2004-04-14 19:45:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* flush adapter cache */
ciss_flush_adapter(sc);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
/* release all resources. The mutex is released and freed here too. */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_free(sc);
return(0);
}
/************************************************************************
* Prepare adapter for system shutdown.
*/
static int
ciss_shutdown(device_t dev)
{
struct ciss_softc *sc = device_get_softc(dev);
debug_called(1);
Satisfy witness during shutdown
2007-06-05 05:03:13 +00:00
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* flush adapter cache */
ciss_flush_adapter(sc);
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
if (sc->ciss_soft_reset)
ciss_soft_reset(sc);
Satisfy witness during shutdown
2007-06-05 05:03:13 +00:00
mtx_unlock(&sc->ciss_mtx);
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
}
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
static void
ciss_init_sysctl(struct ciss_softc *sc)
{
SYSCTL_ADD_INT(device_get_sysctl_ctx(sc->ciss_dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(sc->ciss_dev)),
OID_AUTO, "soft_reset", CTLFLAG_RW, &sc->ciss_soft_reset, 0, "");
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Perform PCI-specific attachment actions.
*/
static int
ciss_init_pci(struct ciss_softc *sc)
{
uintptr_t cbase, csize, cofs;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
uint32_t method, supported_methods;
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
int error, sqmask, i;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
void *intr;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
/*
* Work out adapter type.
*/
i = ciss_lookup(sc->ciss_dev);
if (i < 0) {
ciss_printf(sc, "unknown adapter type\n");
return (ENXIO);
}
if (ciss_vendor_data[i].flags & CISS_BOARD_SA5) {
sqmask = CISS_TL_SIMPLE_INTR_OPQ_SA5;
} else if (ciss_vendor_data[i].flags & CISS_BOARD_SA5B) {
sqmask = CISS_TL_SIMPLE_INTR_OPQ_SA5B;
} else {
/*
* XXX Big hammer, masks/unmasks all possible interrupts. This should
* work on all hardware variants. Need to add code to handle the
* "controller crashed" interupt bit that this unmasks.
*/
sqmask = ~0;
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Allocate register window first (we need this to find the config
* struct).
*/
error = ENXIO;
sc->ciss_regs_rid = CISS_TL_SIMPLE_BAR_REGS;
if ((sc->ciss_regs_resource =
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
bus_alloc_resource_any(sc->ciss_dev, SYS_RES_MEMORY,
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
&sc->ciss_regs_rid, RF_ACTIVE)) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "can't allocate register window\n");
return(ENXIO);
}
sc->ciss_regs_bhandle = rman_get_bushandle(sc->ciss_regs_resource);
sc->ciss_regs_btag = rman_get_bustag(sc->ciss_regs_resource);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Find the BAR holding the config structure. If it's not the one
* we already mapped for registers, map it too.
*/
sc->ciss_cfg_rid = CISS_TL_SIMPLE_READ(sc, CISS_TL_SIMPLE_CFG_BAR) & 0xffff;
if (sc->ciss_cfg_rid != sc->ciss_regs_rid) {
if ((sc->ciss_cfg_resource =
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
bus_alloc_resource_any(sc->ciss_dev, SYS_RES_MEMORY,
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
&sc->ciss_cfg_rid, RF_ACTIVE)) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "can't allocate config window\n");
return(ENXIO);
}
cbase = (uintptr_t)rman_get_virtual(sc->ciss_cfg_resource);
csize = rman_get_end(sc->ciss_cfg_resource) -
rman_get_start(sc->ciss_cfg_resource) + 1;
} else {
cbase = (uintptr_t)rman_get_virtual(sc->ciss_regs_resource);
csize = rman_get_end(sc->ciss_regs_resource) -
rman_get_start(sc->ciss_regs_resource) + 1;
}
cofs = CISS_TL_SIMPLE_READ(sc, CISS_TL_SIMPLE_CFG_OFF);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Use the base/size/offset values we just calculated to
* sanity-check the config structure. If it's OK, point to it.
*/
if ((cofs + sizeof(struct ciss_config_table)) > csize) {
ciss_printf(sc, "config table outside window\n");
return(ENXIO);
}
sc->ciss_cfg = (struct ciss_config_table *)(cbase + cofs);
debug(1, "config struct at %p", sc->ciss_cfg);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
/*
* Calculate the number of request structures/commands we are
* going to provide for this adapter.
*/
sc->ciss_max_requests = min(CISS_MAX_REQUESTS, sc->ciss_cfg->max_outstanding_commands);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Validate the config structure. If we supported other transport
* methods, we could select amongst them at this point in time.
*/
if (strncmp(sc->ciss_cfg->signature, "CISS", 4)) {
ciss_printf(sc, "config signature mismatch (got '%c%c%c%c')\n",
sc->ciss_cfg->signature[0], sc->ciss_cfg->signature[1],
sc->ciss_cfg->signature[2], sc->ciss_cfg->signature[3]);
return(ENXIO);
}
/*
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
* Select the mode of operation, prefer Performant.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if (!(sc->ciss_cfg->supported_methods &
(CISS_TRANSPORT_METHOD_SIMPLE | CISS_TRANSPORT_METHOD_PERF))) {
ciss_printf(sc, "No supported transport layers: 0x%x\n",
sc->ciss_cfg->supported_methods);
}
switch (ciss_force_transport) {
case 1:
supported_methods = CISS_TRANSPORT_METHOD_SIMPLE;
break;
case 2:
supported_methods = CISS_TRANSPORT_METHOD_PERF;
break;
default:
supported_methods = sc->ciss_cfg->supported_methods;
break;
}
setup:
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
if ((supported_methods & CISS_TRANSPORT_METHOD_PERF) != 0) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
method = CISS_TRANSPORT_METHOD_PERF;
sc->ciss_perf = (struct ciss_perf_config *)(cbase + cofs +
sc->ciss_cfg->transport_offset);
if (ciss_init_perf(sc)) {
supported_methods &= ~method;
goto setup;
}
} else if (supported_methods & CISS_TRANSPORT_METHOD_SIMPLE) {
method = CISS_TRANSPORT_METHOD_SIMPLE;
} else {
ciss_printf(sc, "No supported transport methods: 0x%x\n",
sc->ciss_cfg->supported_methods);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(ENXIO);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
/*
* Tell it we're using the low 4GB of RAM. Set the default interrupt
* coalescing options.
*/
sc->ciss_cfg->requested_method = method;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc->ciss_cfg->command_physlimit = 0;
sc->ciss_cfg->interrupt_coalesce_delay = CISS_INTERRUPT_COALESCE_DELAY;
sc->ciss_cfg->interrupt_coalesce_count = CISS_INTERRUPT_COALESCE_COUNT;
Turn SCSI pre-fetch ON. This is mainly for 64XX and 64X based controllers and allows the controller to prefetch 1-2k on certain PCI memory reads to the host. The spec says this should only be used for IA32 based systems. Informed of feature by: John Cagle <first.last@hp.com>
2004-05-17 17:27:38 +00:00
#ifdef __i386__
sc->ciss_cfg->host_driver |= CISS_DRIVER_SCSI_PREFETCH;
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (ciss_update_config(sc)) {
ciss_printf(sc, "adapter refuses to accept config update (IDBR 0x%x)\n",
CISS_TL_SIMPLE_READ(sc, CISS_TL_SIMPLE_IDBR));
return(ENXIO);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if ((sc->ciss_cfg->active_method & method) == 0) {
supported_methods &= ~method;
if (supported_methods == 0) {
ciss_printf(sc, "adapter refuses to go into available transports "
"mode (0x%x, 0x%x)\n", supported_methods,
sc->ciss_cfg->active_method);
return(ENXIO);
} else
goto setup;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/*
* Wait for the adapter to come ready.
*/
if ((error = ciss_wait_adapter(sc)) != 0)
return(error);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
/* Prepare to possibly use MSIX and/or PERFORMANT interrupts. Normal
* interrupts have a rid of 0, this will be overridden if MSIX is used.
*/
sc->ciss_irq_rid[0] = 0;
if (method == CISS_TRANSPORT_METHOD_PERF) {
ciss_printf(sc, "PERFORMANT Transport\n");
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
if ((ciss_force_interrupt != 1) && (ciss_setup_msix(sc) == 0)) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
intr = ciss_perf_msi_intr;
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
} else {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
intr = ciss_perf_intr;
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
}
Make MSI and PERFORMANT interrupts work correctly. Only require the minimum number of MSIX interrupts that are needed, and don't strictly check for 4. Enable enough interrupt mask bits so that the controller will generate interrupts in PERFORMANT mode. This fixes the hang-on-boot issues that people were seeing with newer controllers.
2009-09-16 23:10:10 +00:00
/* XXX The docs say that the 0x01 bit is only for SAS controllers.
* Unfortunately, there is no good way to know if this is a SAS
* controller. Hopefully enabling this bit universally will work OK.
* It seems to work fine for SA6i controllers.
*/
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
sc->ciss_interrupt_mask = CISS_TL_PERF_INTR_OPQ | CISS_TL_PERF_INTR_MSI;
Make MSI and PERFORMANT interrupts work correctly. Only require the minimum number of MSIX interrupts that are needed, and don't strictly check for 4. Enable enough interrupt mask bits so that the controller will generate interrupts in PERFORMANT mode. This fixes the hang-on-boot issues that people were seeing with newer controllers.
2009-09-16 23:10:10 +00:00
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
} else {
ciss_printf(sc, "SIMPLE Transport\n");
/* MSIX doesn't seem to work in SIMPLE mode, only enable if it forced */
if (ciss_force_interrupt == 2)
/* If this fails, we automatically revert to INTx */
ciss_setup_msix(sc);
sc->ciss_perf = NULL;
intr = ciss_intr;
Work again to fix the interrupt masking problems. We now recognize that there are 3 different interrupt enable bits, 2 for different families of cards, and 1 for when MSI is used. Also apply a big hammer backstop for cards that aren't recognized. This should fix all of the interrupt issues at boot.
2008-08-29 01:23:16 +00:00
sc->ciss_interrupt_mask = sqmask;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Turn off interrupts before we go routing anything.
*/
CISS_TL_SIMPLE_DISABLE_INTERRUPTS(sc);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Allocate and set up our interrupt.
*/
if ((sc->ciss_irq_resource =
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
bus_alloc_resource_any(sc->ciss_dev, SYS_RES_IRQ, &sc->ciss_irq_rid[0],
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
RF_ACTIVE | RF_SHAREABLE)) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "can't allocate interrupt\n");
return(ENXIO);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss's interrupt handler was missing the INTR_ENTROPY flag.
2004-08-16 23:13:16 +00:00
if (bus_setup_intr(sc->ciss_dev, sc->ciss_irq_resource,
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
INTR_TYPE_CAM|INTR_MPSAFE, NULL, intr, sc,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
&sc->ciss_intr)) {
ciss_printf(sc, "can't set up interrupt\n");
return(ENXIO);
}
/*
* Allocate the parent bus DMA tag appropriate for our PCI
* interface.
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
*
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* Note that "simple" adapters can only address within a 32-bit
* span.
*/
Convert a number of drivers to obtaining their parent DMA tag from their PCI device attachment.
2012-03-12 08:03:51 +00:00
if (bus_dma_tag_create(bus_get_dma_tag(sc->ciss_dev),/* PCI parent */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
1, 0, /* alignment, boundary */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
BUS_SPACE_MAXADDR, /* lowaddr */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
BUS_SPACE_MAXSIZE_32BIT, /* maxsize */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
CISS_MAX_SG_ELEMENTS, /* nsegments */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
0, /* flags */
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
NULL, NULL, /* lockfunc, lockarg */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
&sc->ciss_parent_dmat)) {
ciss_printf(sc, "can't allocate parent DMA tag\n");
return(ENOMEM);
}
/*
* Create DMA tag for mapping buffers into adapter-addressable
* space.
*/
if (bus_dma_tag_create(sc->ciss_parent_dmat, /* parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
MAXBSIZE, CISS_MAX_SG_ELEMENTS, /* maxsize, nsegments */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
BUS_DMA_ALLOCNOW, /* flags */
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
busdma_lock_mutex, &sc->ciss_mtx, /* lockfunc, lockarg */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
&sc->ciss_buffer_dmat)) {
ciss_printf(sc, "can't allocate buffer DMA tag\n");
return(ENOMEM);
}
return(0);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
/************************************************************************
* Setup MSI/MSIX operation (Performant only)
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
* Four interrupts are available, but we only use 1 right now. If MSI-X
* isn't avaialble, try using MSI instead.
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
*/
static int
ciss_setup_msix(struct ciss_softc *sc)
{
int val, i;
/* Weed out devices that don't actually support MSI */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
i = ciss_lookup(sc->ciss_dev);
if (ciss_vendor_data[i].flags & CISS_BOARD_NOMSI)
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
return (EINVAL);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
/*
* Only need to use the minimum number of MSI vectors, as the driver
* doesn't support directed MSIX interrupts.
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
val = pci_msix_count(sc->ciss_dev);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
if (val < CISS_MSI_COUNT) {
val = pci_msi_count(sc->ciss_dev);
device_printf(sc->ciss_dev, "got %d MSI messages]\n", val);
if (val < CISS_MSI_COUNT)
return (EINVAL);
}
Make MSI and PERFORMANT interrupts work correctly. Only require the minimum number of MSIX interrupts that are needed, and don't strictly check for 4. Enable enough interrupt mask bits so that the controller will generate interrupts in PERFORMANT mode. This fixes the hang-on-boot issues that people were seeing with newer controllers.
2009-09-16 23:10:10 +00:00
val = MIN(val, CISS_MSI_COUNT);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
if (pci_alloc_msix(sc->ciss_dev, &val) != 0) {
if (pci_alloc_msi(sc->ciss_dev, &val) != 0)
return (EINVAL);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
sc->ciss_msi = val;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
if (bootverbose)
ciss_printf(sc, "Using %d MSIX interrupt%s\n", val,
(val != 1) ? "s" : "");
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
for (i = 0; i < val; i++)
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
sc->ciss_irq_rid[i] = i + 1;
return (0);
}
/************************************************************************
* Setup the Performant structures.
*/
static int
ciss_init_perf(struct ciss_softc *sc)
{
struct ciss_perf_config *pc = sc->ciss_perf;
int reply_size;
/*
* Create the DMA tag for the reply queue.
*/
reply_size = sizeof(uint64_t) * sc->ciss_max_requests;
if (bus_dma_tag_create(sc->ciss_parent_dmat, /* parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR_32BIT, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
reply_size, 1, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->ciss_reply_dmat)) {
ciss_printf(sc, "can't allocate reply DMA tag\n");
return(ENOMEM);
}
/*
* Allocate memory and make it available for DMA.
*/
if (bus_dmamem_alloc(sc->ciss_reply_dmat, (void **)&sc->ciss_reply,
BUS_DMA_NOWAIT, &sc->ciss_reply_map)) {
ciss_printf(sc, "can't allocate reply memory\n");
return(ENOMEM);
}
bus_dmamap_load(sc->ciss_reply_dmat, sc->ciss_reply_map, sc->ciss_reply,
reply_size, ciss_command_map_helper, &sc->ciss_reply_phys, 0);
bzero(sc->ciss_reply, reply_size);
sc->ciss_cycle = 0x1;
sc->ciss_rqidx = 0;
/*
* Preload the fetch table with common command sizes. This allows the
* hardware to not waste bus cycles for typical i/o commands, but also not
* tax the driver to be too exact in choosing sizes. The table is optimized
* for page-aligned i/o's, but since most i/o comes from the various pagers,
* it's a reasonable assumption to make.
*/
pc->fetch_count[CISS_SG_FETCH_NONE] = (sizeof(struct ciss_command) + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_1] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 1 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_2] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 2 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_4] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 4 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_8] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 8 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_16] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 16 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_32] =
(sizeof(struct ciss_command) + sizeof(struct ciss_sg_entry) * 32 + 15) / 16;
pc->fetch_count[CISS_SG_FETCH_MAX] = (CISS_COMMAND_ALLOC_SIZE + 15) / 16;
pc->rq_size = sc->ciss_max_requests; /* XXX less than the card supports? */
pc->rq_count = 1; /* XXX Hardcode for a single queue */
pc->rq_bank_hi = 0;
pc->rq_bank_lo = 0;
pc->rq[0].rq_addr_hi = 0x0;
pc->rq[0].rq_addr_lo = sc->ciss_reply_phys;
return(0);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Wait for the adapter to come ready.
*/
static int
ciss_wait_adapter(struct ciss_softc *sc)
{
int i;
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Wait for the adapter to come ready.
*/
if (!(sc->ciss_cfg->active_method & CISS_TRANSPORT_METHOD_READY)) {
ciss_printf(sc, "waiting for adapter to come ready...\n");
for (i = 0; !(sc->ciss_cfg->active_method & CISS_TRANSPORT_METHOD_READY); i++) {
DELAY(1000000); /* one second */
if (i > 30) {
ciss_printf(sc, "timed out waiting for adapter to come ready\n");
return(EIO);
}
}
}
return(0);
}
/************************************************************************
* Flush the adapter cache.
*/
static int
ciss_flush_adapter(struct ciss_softc *sc)
{
struct ciss_request *cr;
struct ciss_bmic_flush_cache *cbfc;
int error, command_status;
debug_called(1);
cr = NULL;
cbfc = NULL;
/*
* Build a BMIC request to flush the cache. We don't disable
* it, as we may be going to do more I/O (eg. we are emulating
* the Synchronise Cache command).
*/
if ((cbfc = malloc(sizeof(*cbfc), CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO)) == NULL) {
error = ENOMEM;
goto out;
}
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_FLUSH_CACHE,
(void **)&cbfc, sizeof(*cbfc))) != 0)
goto out;
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC FLUSH_CACHE command (%d)\n", error);
goto out;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS:
break;
default:
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
ciss_printf(sc, "error flushing cache (%s)\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
out:
if (cbfc != NULL)
free(cbfc, CISS_MALLOC_CLASS);
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
static void
ciss_soft_reset(struct ciss_softc *sc)
{
struct ciss_request *cr = NULL;
struct ciss_command *cc;
int i, error = 0;
for (i = 0; i < sc->ciss_max_logical_bus; i++) {
/* only reset proxy controllers */
if (sc->ciss_controllers[i].physical.bus == 0)
continue;
if ((error = ciss_get_request(sc, &cr)) != 0)
break;
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_SOFT_RESET,
NULL, 0)) != 0)
break;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Provide a way to soft reset a proxy controller such as an MSA20 or MSA500. This is useful if you need to reset one of the storage arrays on reboot.
2005-04-19 06:11:16 +00:00
cc->header.address = sc->ciss_controllers[i];
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0)
break;
ciss_release_request(cr);
}
if (error)
ciss_printf(sc, "error resetting controller (%d)\n", error);
if (cr != NULL)
ciss_release_request(cr);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Allocate memory for the adapter command structures, initialise
* the request structures.
*
* Note that the entire set of commands are allocated in a single
* contiguous slab.
*/
static int
ciss_init_requests(struct ciss_softc *sc)
{
struct ciss_request *cr;
int i;
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Quiet down ciss unless bootverbose is set.
2003-07-12 22:32:27 +00:00
if (bootverbose)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "using %d of %d available commands\n",
sc->ciss_max_requests, sc->ciss_cfg->max_outstanding_commands);
/*
* Create the DMA tag for commands.
*/
if (bus_dma_tag_create(sc->ciss_parent_dmat, /* parent */
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
32, 0, /* alignment, boundary */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
BUS_SPACE_MAXADDR_32BIT, /* lowaddr */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
CISS_COMMAND_ALLOC_SIZE *
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc->ciss_max_requests, 1, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
0, /* flags */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
NULL, NULL, /* lockfunc, lockarg */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
&sc->ciss_command_dmat)) {
ciss_printf(sc, "can't allocate command DMA tag\n");
return(ENOMEM);
}
/*
* Allocate memory and make it available for DMA.
*/
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
if (bus_dmamem_alloc(sc->ciss_command_dmat, (void **)&sc->ciss_command,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
BUS_DMA_NOWAIT, &sc->ciss_command_map)) {
ciss_printf(sc, "can't allocate command memory\n");
return(ENOMEM);
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
bus_dmamap_load(sc->ciss_command_dmat, sc->ciss_command_map,sc->ciss_command,
Give the correct size for the command map to busdma.
2003-06-26 04:01:34 +00:00
CISS_COMMAND_ALLOC_SIZE * sc->ciss_max_requests,
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_command_map_helper, &sc->ciss_command_phys, 0);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
bzero(sc->ciss_command, CISS_COMMAND_ALLOC_SIZE * sc->ciss_max_requests);
/*
* Set up the request and command structures, push requests onto
* the free queue.
*/
for (i = 1; i < sc->ciss_max_requests; i++) {
cr = &sc->ciss_request[i];
cr->cr_sc = sc;
cr->cr_tag = i;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cr->cr_cc = (struct ciss_command *)((uintptr_t)sc->ciss_command +
CISS_COMMAND_ALLOC_SIZE * i);
cr->cr_ccphys = sc->ciss_command_phys + CISS_COMMAND_ALLOC_SIZE * i;
Make ciss usable under PAE Approved by: re (scottl)
2003-05-21 07:17:06 +00:00
bus_dmamap_create(sc->ciss_buffer_dmat, 0, &cr->cr_datamap);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_enqueue_free(cr);
}
return(0);
}
static void
ciss_command_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
uint32_t *addr;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
addr = arg;
*addr = segs[0].ds_addr;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Identify the adapter, print some information about it.
*/
static int
ciss_identify_adapter(struct ciss_softc *sc)
{
struct ciss_request *cr;
int error, command_status;
debug_called(1);
cr = NULL;
/*
* Get a request, allocate storage for the adapter data.
*/
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_ID_CTLR,
(void **)&sc->ciss_id,
sizeof(*sc->ciss_id))) != 0)
goto out;
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC ID_CTLR command (%d)\n", error);
goto out;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS: /* buffer right size */
break;
case CISS_CMD_STATUS_DATA_UNDERRUN:
case CISS_CMD_STATUS_DATA_OVERRUN:
ciss_printf(sc, "data over/underrun reading adapter information\n");
default:
ciss_printf(sc, "error reading adapter information (%s)\n",
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
/* sanity-check reply */
if (!sc->ciss_id->big_map_supported) {
ciss_printf(sc, "adapter does not support BIG_MAP\n");
error = ENXIO;
goto out;
}
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
#if 0
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* XXX later revisions may not need this */
sc->ciss_flags |= CISS_FLAG_FAKE_SYNCH;
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* XXX only really required for old 5300 adapters? */
sc->ciss_flags |= CISS_FLAG_BMIC_ABORT;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* print information */
Quiet down ciss unless bootverbose is set.
2003-07-12 22:32:27 +00:00
if (bootverbose) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#if 0 /* XXX proxy volumes??? */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, " %d logical drive%s configured\n",
sc->ciss_id->configured_logical_drives,
(sc->ciss_id->configured_logical_drives == 1) ? "" : "s");
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, " firmware %4.4s\n", sc->ciss_id->running_firmware_revision);
ciss_printf(sc, " %d SCSI channels\n", sc->ciss_id->scsi_bus_count);
ciss_printf(sc, " signature '%.4s'\n", sc->ciss_cfg->signature);
ciss_printf(sc, " valence %d\n", sc->ciss_cfg->valence);
ciss_printf(sc, " supported I/O methods 0x%b\n",
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
sc->ciss_cfg->supported_methods,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
"\20\1READY\2simple\3performant\4MEMQ\n");
ciss_printf(sc, " active I/O method 0x%b\n",
sc->ciss_cfg->active_method, "\20\2simple\3performant\4MEMQ\n");
ciss_printf(sc, " 4G page base 0x%08x\n",
sc->ciss_cfg->command_physlimit);
ciss_printf(sc, " interrupt coalesce delay %dus\n",
sc->ciss_cfg->interrupt_coalesce_delay);
ciss_printf(sc, " interrupt coalesce count %d\n",
sc->ciss_cfg->interrupt_coalesce_count);
ciss_printf(sc, " max outstanding commands %d\n",
sc->ciss_cfg->max_outstanding_commands);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
ciss_printf(sc, " bus types 0x%b\n", sc->ciss_cfg->bus_types,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
"\20\1ultra2\2ultra3\10fibre1\11fibre2\n");
ciss_printf(sc, " server name '%.16s'\n", sc->ciss_cfg->server_name);
ciss_printf(sc, " heartbeat 0x%x\n", sc->ciss_cfg->heartbeat);
}
out:
if (error) {
if (sc->ciss_id != NULL) {
free(sc->ciss_id, CISS_MALLOC_CLASS);
sc->ciss_id = NULL;
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
/************************************************************************
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
* Helper routine for generating a list of logical and physical luns.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
static struct ciss_lun_report *
ciss_report_luns(struct ciss_softc *sc, int opcode, int nunits)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_report_cdb *crc;
struct ciss_lun_report *cll;
int command_status;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
int report_size;
int error = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
cr = NULL;
cll = NULL;
/*
* Get a request, allocate storage for the address list.
*/
if ((error = ciss_get_request(sc, &cr)) != 0)
goto out;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
report_size = sizeof(*cll) + nunits * sizeof(union ciss_device_address);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if ((cll = malloc(report_size, CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO)) == NULL) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_printf(sc, "can't allocate memory for lun report\n");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
error = ENOMEM;
goto out;
}
/*
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
* Build the Report Logical/Physical LUNs command.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cr->cr_data = cll;
cr->cr_length = report_size;
cr->cr_flags = CISS_REQ_DATAIN;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.address.physical.mode = CISS_HDR_ADDRESS_MODE_PERIPHERAL;
cc->header.address.physical.bus = 0;
cc->header.address.physical.target = 0;
cc->cdb.cdb_length = sizeof(*crc);
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_READ;
cc->cdb.timeout = 30; /* XXX better suggestions? */
crc = (struct ciss_report_cdb *)&(cc->cdb.cdb[0]);
bzero(crc, sizeof(*crc));
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
crc->opcode = opcode;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
crc->length = htonl(report_size); /* big-endian field */
cll->list_size = htonl(report_size - sizeof(*cll)); /* big-endian field */
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Submit the request and wait for it to complete. (timeout
* here should be much greater than above)
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_printf(sc, "error sending %d LUN command (%d)\n", opcode, error);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
goto out;
}
/*
* Check response. Note that data over/underrun is OK.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS: /* buffer right size */
case CISS_CMD_STATUS_DATA_UNDERRUN: /* buffer too large, not bad */
break;
case CISS_CMD_STATUS_DATA_OVERRUN:
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_printf(sc, "WARNING: more units than driver limit (%d)\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
CISS_MAX_LOGICAL);
break;
default:
ciss_printf(sc, "error detecting logical drive configuration (%s)\n",
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
ciss_release_request(cr);
cr = NULL;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
out:
if (cr != NULL)
ciss_release_request(cr);
if (error && cll != NULL) {
free(cll, CISS_MALLOC_CLASS);
cll = NULL;
}
return(cll);
}
/************************************************************************
* Find logical drives on the adapter.
*/
static int
ciss_init_logical(struct ciss_softc *sc)
{
struct ciss_lun_report *cll;
int error = 0, i, j;
int ndrives;
debug_called(1);
cll = ciss_report_luns(sc, CISS_OPCODE_REPORT_LOGICAL_LUNS,
CISS_MAX_LOGICAL);
if (cll == NULL) {
error = ENXIO;
goto out;
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* sanity-check reply */
ndrives = (ntohl(cll->list_size) / sizeof(union ciss_device_address));
Several changes to fix livelock under high load, introduced by r203489: - change the way in which command queue overflow is handled; - do not expose to CAM two command slots, used for driver's internal purposes; - allow driver to use up to 1024 command slots, instead of 256 before.
2010-03-03 17:58:41 +00:00
if ((ndrives < 0) || (ndrives > CISS_MAX_LOGICAL)) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "adapter claims to report absurd number of logical drives (%d > %d)\n",
ndrives, CISS_MAX_LOGICAL);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
error = ENXIO;
goto out;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/*
* Save logical drive information.
*/
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (bootverbose) {
ciss_printf(sc, "%d logical drive%s\n",
ndrives, (ndrives > 1 || ndrives == 0) ? "s" : "");
}
sc->ciss_logical =
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
malloc(sc->ciss_max_logical_bus * sizeof(struct ciss_ldrive *),
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_logical == NULL) {
error = ENXIO;
goto out;
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
for (i = 0; i <= sc->ciss_max_logical_bus; i++) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
sc->ciss_logical[i] =
malloc(CISS_MAX_LOGICAL * sizeof(struct ciss_ldrive),
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_logical[i] == NULL) {
error = ENXIO;
goto out;
}
for (j = 0; j < CISS_MAX_LOGICAL; j++)
sc->ciss_logical[i][j].cl_status = CISS_LD_NONEXISTENT;
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
for (i = 0; i < CISS_MAX_LOGICAL; i++) {
if (i < ndrives) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
struct ciss_ldrive *ld;
int bus, target;
bus = CISS_LUN_TO_BUS(cll->lun[i].logical.lun);
target = CISS_LUN_TO_TARGET(cll->lun[i].logical.lun);
ld = &sc->ciss_logical[bus][target];
ld->cl_address = cll->lun[i];
ld->cl_controller = &sc->ciss_controllers[bus];
if (ciss_identify_logical(sc, ld) != 0)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
continue;
/*
* If the drive has had media exchanged, we should bring it online.
*/
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (ld->cl_lstatus->media_exchanged)
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
ciss_accept_media(sc, ld);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
out:
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (cll != NULL)
free(cll, CISS_MALLOC_CLASS);
return(error);
}
static int
ciss_init_physical(struct ciss_softc *sc)
{
struct ciss_lun_report *cll;
int error = 0, i;
int nphys;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
int bus, target;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
debug_called(1);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
bus = 0;
target = 0;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cll = ciss_report_luns(sc, CISS_OPCODE_REPORT_PHYSICAL_LUNS,
CISS_MAX_PHYSICAL);
if (cll == NULL) {
error = ENXIO;
goto out;
}
nphys = (ntohl(cll->list_size) / sizeof(union ciss_device_address));
if (bootverbose) {
ciss_printf(sc, "%d physical device%s\n",
nphys, (nphys > 1 || nphys == 0) ? "s" : "");
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
/*
* Figure out the bus mapping.
* Logical buses include both the local logical bus for local arrays and
* proxy buses for remote arrays. Physical buses are numbered by the
* controller and represent physical buses that hold physical devices.
* We shift these bus numbers so that everything fits into a single flat
* numbering space for CAM. Logical buses occupy the first 32 CAM bus
* numbers, and the physical bus numbers are shifted to be above that.
* This results in the various driver arrays being indexed as follows:
*
* ciss_controllers[] - indexed by logical bus
* ciss_cam_sim[] - indexed by both logical and physical, with physical
* being shifted by 32.
* ciss_logical[][] - indexed by logical bus
* ciss_physical[][] - indexed by physical bus
*
* XXX This is getting more and more hackish. CISS really doesn't play
* well with a standard SCSI model; devices are addressed via magic
* cookies, not via b/t/l addresses. Since there is no way to store
* the cookie in the CAM device object, we have to keep these lookup
* tables handy so that the devices can be found quickly at the cost
* of wasting memory and having a convoluted lookup scheme. This
* driver should probably be converted to block interface.
*/
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
* If the L2 and L3 SCSI addresses are 0, this signifies a proxy
* controller. A proxy controller is another physical controller
* behind the primary PCI controller. We need to know about this
* so that BMIC commands can be properly targeted. There can be
* proxy controllers attached to a single PCI controller, so
* find the highest numbered one so the array can be properly
* sized.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
sc->ciss_max_logical_bus = 1;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
for (i = 0; i < nphys; i++) {
if (cll->lun[i].physical.extra_address == 0) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
bus = cll->lun[i].physical.bus;
sc->ciss_max_logical_bus = max(sc->ciss_max_logical_bus, bus) + 1;
} else {
bus = CISS_EXTRA_BUS2(cll->lun[i].physical.extra_address);
sc->ciss_max_physical_bus = max(sc->ciss_max_physical_bus, bus);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
}
}
sc->ciss_controllers =
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
malloc(sc->ciss_max_logical_bus * sizeof (union ciss_device_address),
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_controllers == NULL) {
ciss_printf(sc, "Could not allocate memory for controller map\n");
error = ENOMEM;
goto out;
}
/* setup a map of controller addresses */
for (i = 0; i < nphys; i++) {
if (cll->lun[i].physical.extra_address == 0) {
sc->ciss_controllers[cll->lun[i].physical.bus] = cll->lun[i];
}
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
sc->ciss_physical =
malloc(sc->ciss_max_physical_bus * sizeof(struct ciss_pdrive *),
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_physical == NULL) {
ciss_printf(sc, "Could not allocate memory for physical device map\n");
error = ENOMEM;
goto out;
}
for (i = 0; i < sc->ciss_max_physical_bus; i++) {
sc->ciss_physical[i] =
malloc(sizeof(struct ciss_pdrive) * CISS_MAX_PHYSTGT,
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_physical[i] == NULL) {
ciss_printf(sc, "Could not allocate memory for target map\n");
error = ENOMEM;
goto out;
}
}
ciss_filter_physical(sc, cll);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
out:
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (cll != NULL)
free(cll, CISS_MALLOC_CLASS);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(error);
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
static int
ciss_filter_physical(struct ciss_softc *sc, struct ciss_lun_report *cll)
{
u_int32_t ea;
int i, nphys;
int bus, target;
nphys = (ntohl(cll->list_size) / sizeof(union ciss_device_address));
for (i = 0; i < nphys; i++) {
if (cll->lun[i].physical.extra_address == 0)
continue;
/*
* Filter out devices that we don't want. Level 3 LUNs could
* probably be supported, but the docs don't give enough of a
* hint to know how.
*
* The mode field of the physical address is likely set to have
* hard disks masked out. Honor it unless the user has overridden
* us with the tunable. We also munge the inquiry data for these
* disks so that they only show up as passthrough devices. Keeping
* them visible in this fashion is useful for doing things like
* flashing firmware.
*/
ea = cll->lun[i].physical.extra_address;
if ((CISS_EXTRA_BUS3(ea) != 0) || (CISS_EXTRA_TARGET3(ea) != 0) ||
(CISS_EXTRA_MODE2(ea) == 0x3))
continue;
if ((ciss_expose_hidden_physical == 0) &&
(cll->lun[i].physical.mode == CISS_HDR_ADDRESS_MODE_MASK_PERIPHERAL))
continue;
/*
* Note: CISS firmware numbers physical busses starting at '1', not
* '0'. This numbering is internal to the firmware and is only
* used as a hint here.
*/
bus = CISS_EXTRA_BUS2(ea) - 1;
target = CISS_EXTRA_TARGET2(ea);
sc->ciss_physical[bus][target].cp_address = cll->lun[i];
sc->ciss_physical[bus][target].cp_online = 1;
}
return (0);
}
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
static int
ciss_inquiry_logical(struct ciss_softc *sc, struct ciss_ldrive *ld)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct scsi_inquiry *inq;
int error;
int command_status;
cr = NULL;
bzero(&ld->cl_geometry, sizeof(ld->cl_geometry));
if ((error = ciss_get_request(sc, &cr)) != 0)
goto out;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
cr->cr_data = &ld->cl_geometry;
cr->cr_length = sizeof(ld->cl_geometry);
cr->cr_flags = CISS_REQ_DATAIN;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cc->header.address = ld->cl_address;
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
cc->cdb.cdb_length = 6;
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_READ;
cc->cdb.timeout = 30;
inq = (struct scsi_inquiry *)&(cc->cdb.cdb[0]);
inq->opcode = INQUIRY;
inq->byte2 = SI_EVPD;
inq->page_code = CISS_VPD_LOGICAL_DRIVE_GEOMETRY;
Add the CAM Target Layer (CTL). CTL is a disk and processor device emulation subsystem originally written for Copan Systems under Linux starting in 2003. It has been shipping in Copan (now SGI) products since 2005. It was ported to FreeBSD in 2008, and thanks to an agreement between SGI (who acquired Copan's assets in 2010) and Spectra Logic in 2010, CTL is available under a BSD-style license. The intent behind the agreement was that Spectra would work to get CTL into the FreeBSD tree. Some CTL features: - Disk and processor device emulation. - Tagged queueing - SCSI task attribute support (ordered, head of queue, simple tags) - SCSI implicit command ordering support. (e.g. if a read follows a mode select, the read will be blocked until the mode select completes.) - Full task management support (abort, LUN reset, target reset, etc.) - Support for multiple ports - Support for multiple simultaneous initiators - Support for multiple simultaneous backing stores - Persistent reservation support - Mode sense/select support - Error injection support - High Availability support (1) - All I/O handled in-kernel, no userland context switch overhead. (1) HA Support is just an API stub, and needs much more to be fully functional. ctl.c: The core of CTL. Command handlers and processing, character driver, and HA support are here. ctl.h: Basic function declarations and data structures. ctl_backend.c, ctl_backend.h: The basic CTL backend API. ctl_backend_block.c, ctl_backend_block.h: The block and file backend. This allows for using a disk or a file as the backing store for a LUN. Multiple threads are started to do I/O to the backing device, primarily because the VFS API requires that to get any concurrency. ctl_backend_ramdisk.c: A "fake" ramdisk backend. It only allocates a small amount of memory to act as a source and sink for reads and writes from an initiator. Therefore it cannot be used for any real data, but it can be used to test for throughput. It can also be used to test initiators' support for extremely large LUNs. ctl_cmd_table.c: This is a table with all 256 possible SCSI opcodes, and command handler functions defined for supported opcodes. ctl_debug.h: Debugging support. ctl_error.c, ctl_error.h: CTL-specific wrappers around the CAM sense building functions. ctl_frontend.c, ctl_frontend.h: These files define the basic CTL frontend port API. ctl_frontend_cam_sim.c: This is a CTL frontend port that is also a CAM SIM. This frontend allows for using CTL without any target-capable hardware. So any LUNs you create in CTL are visible in CAM via this port. ctl_frontend_internal.c, ctl_frontend_internal.h: This is a frontend port written for Copan to do some system-specific tasks that required sending commands into CTL from inside the kernel. This isn't entirely relevant to FreeBSD in general, but can perhaps be repurposed. ctl_ha.h: This is a stubbed-out High Availability API. Much more is needed for full HA support. See the comments in the header and the description of what is needed in the README.ctl.txt file for more details. ctl_io.h: This defines most of the core CTL I/O structures. union ctl_io is conceptually very similar to CAM's union ccb. ctl_ioctl.h: This defines all ioctls available through the CTL character device, and the data structures needed for those ioctls. ctl_mem_pool.c, ctl_mem_pool.h: Generic memory pool implementation used by the internal frontend. ctl_private.h: Private data structres (e.g. CTL softc) and function prototypes. This also includes the SCSI vendor and product names used by CTL. ctl_scsi_all.c, ctl_scsi_all.h: CTL wrappers around CAM sense printing functions. ctl_ser_table.c: Command serialization table. This defines what happens when one type of command is followed by another type of command. ctl_util.c, ctl_util.h: CTL utility functions, primarily designed to be used from userland. See ctladm for the primary consumer of these functions. These include CDB building functions. scsi_ctl.c: CAM target peripheral driver and CTL frontend port. This is the path into CTL for commands from target-capable hardware/SIMs. README.ctl.txt: CTL code features, roadmap, to-do list. usr.sbin/Makefile: Add ctladm. ctladm/Makefile, ctladm/ctladm.8, ctladm/ctladm.c, ctladm/ctladm.h, ctladm/util.c: ctladm(8) is the CTL management utility. It fills a role similar to camcontrol(8). It allow configuring LUNs, issuing commands, injecting errors and various other control functions. usr.bin/Makefile: Add ctlstat. ctlstat/Makefile ctlstat/ctlstat.8, ctlstat/ctlstat.c: ctlstat(8) fills a role similar to iostat(8). It reports I/O statistics for CTL. sys/conf/files: Add CTL files. sys/conf/NOTES: Add device ctl. sys/cam/scsi_all.h: To conform to more recent specs, the inquiry CDB length field is now 2 bytes long. Add several mode page definitions for CTL. sys/cam/scsi_all.c: Handle the new 2 byte inquiry length. sys/dev/ciss/ciss.c, sys/dev/ata/atapi-cam.c, sys/cam/scsi/scsi_targ_bh.c, scsi_target/scsi_cmds.c, mlxcontrol/interface.c: Update for 2 byte inquiry length field. scsi_da.h: Add versions of the format and rigid disk pages that are in a more reasonable format for CTL. amd64/conf/GENERIC, i386/conf/GENERIC, ia64/conf/GENERIC, sparc64/conf/GENERIC: Add device ctl. i386/conf/PAE: The CTL frontend SIM at least does not compile cleanly on PAE. Sponsored by: Copan Systems, SGI and Spectra Logic MFC after: 1 month
2012-01-12 00:34:33 +00:00
scsi_ulto2b(sizeof(ld->cl_geometry), inq->length);
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error getting geometry (%d)\n", error);
goto out;
}
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS:
case CISS_CMD_STATUS_DATA_UNDERRUN:
break;
case CISS_CMD_STATUS_DATA_OVERRUN:
ciss_printf(sc, "WARNING: Data overrun\n");
break;
default:
ciss_printf(sc, "Error detecting logical drive geometry (%s)\n",
ciss_name_command_status(command_status));
break;
}
out:
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Identify a logical drive, initialise state related to it.
*/
static int
ciss_identify_logical(struct ciss_softc *sc, struct ciss_ldrive *ld)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_bmic_cdb *cbc;
int error, command_status;
debug_called(1);
cr = NULL;
/*
* Build a BMIC request to fetch the drive ID.
*/
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_ID_LDRIVE,
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
(void **)&ld->cl_ldrive,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sizeof(*ld->cl_ldrive))) != 0)
goto out;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cc->header.address = *ld->cl_controller; /* target controller */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cbc = (struct ciss_bmic_cdb *)&(cc->cdb.cdb[0]);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cbc->log_drive = CISS_LUN_TO_TARGET(ld->cl_address.logical.lun);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC LDRIVE command (%d)\n", error);
goto out;
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS: /* buffer right size */
break;
case CISS_CMD_STATUS_DATA_UNDERRUN:
case CISS_CMD_STATUS_DATA_OVERRUN:
ciss_printf(sc, "data over/underrun reading logical drive ID\n");
default:
ciss_printf(sc, "error reading logical drive ID (%s)\n",
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
ciss_release_request(cr);
cr = NULL;
/*
* Build a CISS BMIC command to get the logical drive status.
*/
if ((error = ciss_get_ldrive_status(sc, ld)) != 0)
goto out;
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
/*
* Get the logical drive geometry.
*/
if ((error = ciss_inquiry_logical(sc, ld)) != 0)
goto out;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Print the drive's basic characteristics.
*/
Quiet down ciss unless bootverbose is set.
2003-07-12 22:32:27 +00:00
if (bootverbose) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_printf(sc, "logical drive (b%dt%d): %s, %dMB ",
CISS_LUN_TO_BUS(ld->cl_address.logical.lun),
CISS_LUN_TO_TARGET(ld->cl_address.logical.lun),
Print the correct logical drive lun number returned by the controller.
2003-10-15 18:52:44 +00:00
ciss_name_ldrive_org(ld->cl_ldrive->fault_tolerance),
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
((ld->cl_ldrive->blocks_available / (1024 * 1024)) *
ld->cl_ldrive->block_size));
ciss_print_ldrive(sc, ld);
}
out:
if (error != 0) {
/* make the drive not-exist */
ld->cl_status = CISS_LD_NONEXISTENT;
if (ld->cl_ldrive != NULL) {
free(ld->cl_ldrive, CISS_MALLOC_CLASS);
ld->cl_ldrive = NULL;
}
if (ld->cl_lstatus != NULL) {
free(ld->cl_lstatus, CISS_MALLOC_CLASS);
ld->cl_lstatus = NULL;
}
}
if (cr != NULL)
ciss_release_request(cr);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(error);
}
/************************************************************************
* Get status for a logical drive.
*
* XXX should we also do this in response to Test Unit Ready?
*/
static int
ciss_get_ldrive_status(struct ciss_softc *sc, struct ciss_ldrive *ld)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_bmic_cdb *cbc;
int error, command_status;
/*
* Build a CISS BMIC command to get the logical drive status.
*/
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_ID_LSTATUS,
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
(void **)&ld->cl_lstatus,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sizeof(*ld->cl_lstatus))) != 0)
goto out;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cc->header.address = *ld->cl_controller; /* target controller */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cbc = (struct ciss_bmic_cdb *)&(cc->cdb.cdb[0]);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cbc->log_drive = CISS_LUN_TO_TARGET(ld->cl_address.logical.lun);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC LSTATUS command (%d)\n", error);
goto out;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS: /* buffer right size */
break;
case CISS_CMD_STATUS_DATA_UNDERRUN:
case CISS_CMD_STATUS_DATA_OVERRUN:
ciss_printf(sc, "data over/underrun reading logical drive status\n");
default:
ciss_printf(sc, "error reading logical drive status (%s)\n",
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
/*
* Set the drive's summary status based on the returned status.
*
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
* XXX testing shows that a failed JBOD drive comes back at next
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* boot in "queued for expansion" mode. WTF?
*/
ld->cl_status = ciss_decode_ldrive_status(ld->cl_lstatus->status);
out:
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
/************************************************************************
* Notify the adapter of a config update.
*/
static int
ciss_update_config(struct ciss_softc *sc)
{
int i;
debug_called(1);
CISS_TL_SIMPLE_WRITE(sc, CISS_TL_SIMPLE_IDBR, CISS_TL_SIMPLE_IDBR_CFG_TABLE);
for (i = 0; i < 1000; i++) {
if (!(CISS_TL_SIMPLE_READ(sc, CISS_TL_SIMPLE_IDBR) &
CISS_TL_SIMPLE_IDBR_CFG_TABLE)) {
return(0);
}
DELAY(1000);
}
return(1);
}
/************************************************************************
* Accept new media into a logical drive.
*
* XXX The drive has previously been offline; it would be good if we
* could make sure it's not open right now.
*/
static int
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
ciss_accept_media(struct ciss_softc *sc, struct ciss_ldrive *ld)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_bmic_cdb *cbc;
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
int command_status;
int error = 0, ldrive;
make this compile/work with CISS_DEBUG defined.
2004-04-26 19:28:08 +00:00
ldrive = CISS_LUN_TO_TARGET(ld->cl_address.logical.lun);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
debug(0, "bringing logical drive %d back online");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Build a CISS BMIC command to bring the drive back online.
*/
if ((error = ciss_get_bmic_request(sc, &cr, CISS_BMIC_ACCEPT_MEDIA,
NULL, 0)) != 0)
goto out;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cc->header.address = *ld->cl_controller; /* target controller */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cbc = (struct ciss_bmic_cdb *)&(cc->cdb.cdb[0]);
make this compile/work with CISS_DEBUG defined.
2004-04-26 19:28:08 +00:00
cbc->log_drive = ldrive;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
* Submit the request and wait for it to complete.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC ACCEPT MEDIA command (%d)\n", error);
goto out;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS: /* all OK */
/* we should get a logical drive status changed event here */
break;
default:
ciss_printf(cr->cr_sc, "error accepting media into failed logical drive (%s)\n",
ciss_name_command_status(command_status));
break;
}
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
out:
if (cr != NULL)
ciss_release_request(cr);
return(error);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Release adapter resources.
*/
static void
ciss_free(struct ciss_softc *sc)
{
Make ciss usable under PAE Approved by: re (scottl)
2003-05-21 07:17:06 +00:00
struct ciss_request *cr;
Free some previously leaked memory on module unload.
2006-04-19 19:56:10 +00:00
int i, j;
Make ciss usable under PAE Approved by: re (scottl)
2003-05-21 07:17:06 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
/* we're going away */
sc->ciss_flags |= CISS_FLAG_ABORTING;
/* terminate the periodic heartbeat routine */
Streamline locking in ciss_free()
2007-05-02 04:44:31 +00:00
callout_stop(&sc->ciss_periodic);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* cancel the Event Notify chain */
ciss_notify_abort(sc);
White space cleanup
2003-12-13 07:54:07 +00:00
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_kill_notify_thread(sc);
Streamline locking in ciss_free()
2007-05-02 04:44:31 +00:00
/* disconnect from CAM */
if (sc->ciss_cam_sim) {
for (i = 0; i < sc->ciss_max_logical_bus; i++) {
if (sc->ciss_cam_sim[i]) {
xpt_bus_deregister(cam_sim_path(sc->ciss_cam_sim[i]));
cam_sim_free(sc->ciss_cam_sim[i], 0);
}
}
for (i = CISS_PHYSICAL_BASE; i < sc->ciss_max_physical_bus +
CISS_PHYSICAL_BASE; i++) {
if (sc->ciss_cam_sim[i]) {
xpt_bus_deregister(cam_sim_path(sc->ciss_cam_sim[i]));
cam_sim_free(sc->ciss_cam_sim[i], 0);
}
}
free(sc->ciss_cam_sim, CISS_MALLOC_CLASS);
}
if (sc->ciss_cam_devq)
cam_simq_free(sc->ciss_cam_devq);
move the cleanup of the control device into ciss_free and add some ifdefs for the diffrent kthread_create API between -current and -stable
2004-04-18 02:39:01 +00:00
/* remove the control device */
Streamline locking in ciss_free()
2007-05-02 04:44:31 +00:00
mtx_unlock(&sc->ciss_mtx);
move the cleanup of the control device into ciss_free and add some ifdefs for the diffrent kthread_create API between -current and -stable
2004-04-18 02:39:01 +00:00
if (sc->ciss_dev_t != NULL)
destroy_dev(sc->ciss_dev_t);
Streamline locking in ciss_free()
2007-05-02 04:44:31 +00:00
/* Final cleanup of the callout. */
callout_drain(&sc->ciss_periodic);
mtx_destroy(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* free the controller data */
if (sc->ciss_id != NULL)
free(sc->ciss_id, CISS_MALLOC_CLASS);
/* release I/O resources */
if (sc->ciss_regs_resource != NULL)
bus_release_resource(sc->ciss_dev, SYS_RES_MEMORY,
sc->ciss_regs_rid, sc->ciss_regs_resource);
if (sc->ciss_cfg_resource != NULL)
bus_release_resource(sc->ciss_dev, SYS_RES_MEMORY,
sc->ciss_cfg_rid, sc->ciss_cfg_resource);
if (sc->ciss_intr != NULL)
bus_teardown_intr(sc->ciss_dev, sc->ciss_irq_resource, sc->ciss_intr);
if (sc->ciss_irq_resource != NULL)
bus_release_resource(sc->ciss_dev, SYS_RES_IRQ,
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
sc->ciss_irq_rid[0], sc->ciss_irq_resource);
if (sc->ciss_msi)
pci_release_msi(sc->ciss_dev);
Make ciss usable under PAE Approved by: re (scottl)
2003-05-21 07:17:06 +00:00
while ((cr = ciss_dequeue_free(sc)) != NULL)
bus_dmamap_destroy(sc->ciss_buffer_dmat, cr->cr_datamap);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (sc->ciss_buffer_dmat)
bus_dma_tag_destroy(sc->ciss_buffer_dmat);
/* destroy command memory and DMA tag */
if (sc->ciss_command != NULL) {
bus_dmamap_unload(sc->ciss_command_dmat, sc->ciss_command_map);
bus_dmamem_free(sc->ciss_command_dmat, sc->ciss_command, sc->ciss_command_map);
}
Make ciss usable under PAE Approved by: re (scottl)
2003-05-21 07:17:06 +00:00
if (sc->ciss_command_dmat)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
bus_dma_tag_destroy(sc->ciss_command_dmat);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if (sc->ciss_reply) {
bus_dmamap_unload(sc->ciss_reply_dmat, sc->ciss_reply_map);
bus_dmamem_free(sc->ciss_reply_dmat, sc->ciss_reply, sc->ciss_reply_map);
}
if (sc->ciss_reply_dmat)
bus_dma_tag_destroy(sc->ciss_reply_dmat);
/* destroy DMA tags */
if (sc->ciss_parent_dmat)
bus_dma_tag_destroy(sc->ciss_parent_dmat);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (sc->ciss_logical) {
Free some previously leaked memory on module unload.
2006-04-19 19:56:10 +00:00
for (i = 0; i <= sc->ciss_max_logical_bus; i++) {
for (j = 0; j < CISS_MAX_LOGICAL; j++) {
if (sc->ciss_logical[i][j].cl_ldrive)
free(sc->ciss_logical[i][j].cl_ldrive, CISS_MALLOC_CLASS);
if (sc->ciss_logical[i][j].cl_lstatus)
free(sc->ciss_logical[i][j].cl_lstatus, CISS_MALLOC_CLASS);
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
free(sc->ciss_logical[i], CISS_MALLOC_CLASS);
Free some previously leaked memory on module unload.
2006-04-19 19:56:10 +00:00
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
free(sc->ciss_logical, CISS_MALLOC_CLASS);
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
if (sc->ciss_physical) {
for (i = 0; i < sc->ciss_max_physical_bus; i++)
free(sc->ciss_physical[i], CISS_MALLOC_CLASS);
free(sc->ciss_physical, CISS_MALLOC_CLASS);
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (sc->ciss_controllers)
free(sc->ciss_controllers, CISS_MALLOC_CLASS);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Give a command to the adapter.
*
* Note that this uses the simple transport layer directly. If we
* want to add support for other layers, we'll need a switch of some
* sort.
*
* Note that the simple transport layer has no way of refusing a
* command; we only have as many request structures as the adapter
* supports commands, so we don't have to check (this presumes that
* the adapter can handle commands as fast as we throw them at it).
*/
static int
ciss_start(struct ciss_request *cr)
{
struct ciss_command *cc; /* XXX debugging only */
int error;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug(2, "post command %d tag %d ", cr->cr_tag, cc->header.host_tag);
/*
* Map the request's data.
*/
if ((error = ciss_map_request(cr)))
return(error);
#if 0
ciss_print_request(cr);
#endif
return(0);
}
/************************************************************************
* Fetch completed request(s) from the adapter, queue them for
* completion handling.
*
* Note that this uses the simple transport layer directly. If we
* want to add support for other layers, we'll need a switch of some
* sort.
*
* Note that the simple transport mechanism does not require any
* reentrancy protection; the OPQ read is atomic. If there is a
* chance of a race with something else that might move the request
* off the busy list, then we will have to lock against that
* (eg. timeouts, etc.)
*/
static void
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_done(struct ciss_softc *sc, cr_qhead_t *qh)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_request *cr;
struct ciss_command *cc;
u_int32_t tag, index;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(3);
/*
* Loop quickly taking requests from the adapter and moving them
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
* to the completed queue.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
for (;;) {
tag = CISS_TL_SIMPLE_FETCH_CMD(sc);
if (tag == CISS_TL_SIMPLE_OPQ_EMPTY)
break;
index = tag >> 2;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
debug(2, "completed command %d%s", index,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
(tag & CISS_HDR_HOST_TAG_ERROR) ? " with error" : "");
if (index >= sc->ciss_max_requests) {
ciss_printf(sc, "completed invalid request %d (0x%x)\n", index, tag);
continue;
}
cr = &(sc->ciss_request[index]);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.host_tag = tag; /* not updated by adapter */
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_enqueue_complete(cr, qh);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
}
static void
ciss_perf_done(struct ciss_softc *sc, cr_qhead_t *qh)
{
struct ciss_request *cr;
struct ciss_command *cc;
u_int32_t tag, index;
debug_called(3);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
* Loop quickly taking requests from the adapter and moving them
* to the completed queue.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
for (;;) {
tag = sc->ciss_reply[sc->ciss_rqidx];
if ((tag & CISS_CYCLE_MASK) != sc->ciss_cycle)
break;
index = tag >> 2;
debug(2, "completed command %d%s\n", index,
(tag & CISS_HDR_HOST_TAG_ERROR) ? " with error" : "");
if (index < sc->ciss_max_requests) {
cr = &(sc->ciss_request[index]);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cc->header.host_tag = tag; /* not updated by adapter */
ciss_enqueue_complete(cr, qh);
} else {
ciss_printf(sc, "completed invalid request %d (0x%x)\n", index, tag);
}
if (++sc->ciss_rqidx == sc->ciss_max_requests) {
sc->ciss_rqidx = 0;
sc->ciss_cycle ^= 1;
}
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Take an interrupt from the adapter.
*/
static void
ciss_intr(void *arg)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr_qhead_t qh;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
struct ciss_softc *sc = (struct ciss_softc *)arg;
/*
* The only interrupt we recognise indicates that there are
* entries in the outbound post queue.
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
STAILQ_INIT(&qh);
ciss_done(sc, &qh);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_complete(sc, &qh);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
static void
ciss_perf_intr(void *arg)
{
struct ciss_softc *sc = (struct ciss_softc *)arg;
/* Clear the interrupt and flush the bridges. Docs say that the flush
* needs to be done twice, which doesn't seem right.
*/
CISS_TL_PERF_CLEAR_INT(sc);
CISS_TL_PERF_FLUSH_INT(sc);
ciss_perf_msi_intr(sc);
}
static void
ciss_perf_msi_intr(void *arg)
{
cr_qhead_t qh;
struct ciss_softc *sc = (struct ciss_softc *)arg;
STAILQ_INIT(&qh);
ciss_perf_done(sc, &qh);
mtx_lock(&sc->ciss_mtx);
ciss_complete(sc, &qh);
mtx_unlock(&sc->ciss_mtx);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Process completed requests.
*
* Requests can be completed in three fashions:
*
* - by invoking a callback function (cr_complete is non-null)
* - by waking up a sleeper (cr_flags has CISS_REQ_SLEEP set)
* - by clearing the CISS_REQ_POLL flag in interrupt/timeout context
*/
static void
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_complete(struct ciss_softc *sc, cr_qhead_t *qh)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_request *cr;
debug_called(2);
/*
* Loop taking requests off the completed queue and performing
* completion processing on them.
*/
for (;;) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if ((cr = ciss_dequeue_complete(sc, qh)) == NULL)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
ciss_unmap_request(cr);
White space cleanup
2003-12-13 07:54:07 +00:00
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if ((cr->cr_flags & CISS_REQ_BUSY) == 0)
ciss_printf(sc, "WARNING: completing non-busy request\n");
cr->cr_flags &= ~CISS_REQ_BUSY;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* If the request has a callback, invoke it.
*/
if (cr->cr_complete != NULL) {
cr->cr_complete(cr);
continue;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* If someone is sleeping on this request, wake them up.
*/
if (cr->cr_flags & CISS_REQ_SLEEP) {
cr->cr_flags &= ~CISS_REQ_SLEEP;
wakeup(cr);
continue;
}
/*
* If someone is polling this request for completion, signal.
*/
if (cr->cr_flags & CISS_REQ_POLL) {
cr->cr_flags &= ~CISS_REQ_POLL;
continue;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Give up and throw the request back on the free queue. This
* should never happen; resources will probably be lost.
*/
ciss_printf(sc, "WARNING: completed command with no submitter\n");
ciss_enqueue_free(cr);
}
}
/************************************************************************
* Report on the completion status of a request, and pass back SCSI
* and command status values.
*/
static int
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
_ciss_report_request(struct ciss_request *cr, int *command_status, int *scsi_status, const char *func)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_command *cc;
struct ciss_error_info *ce;
debug_called(2);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ce = (struct ciss_error_info *)&(cc->sg[0]);
/*
* We don't consider data under/overrun an error for the Report
* Logical/Physical LUNs commands.
*/
if ((cc->header.host_tag & CISS_HDR_HOST_TAG_ERROR) &&
It seems ciss should ignore overrun and underrun on a SCSI INQUIRY command. This fixes some weird booting issues on newer versions of the firmware on the MSA20. Reported by: Philippe Pegon <Philippe dot Pegon at crc dot u-strasbg dot fr>
2005-12-16 06:50:55 +00:00
((ce->command_status == CISS_CMD_STATUS_DATA_OVERRUN) ||
(ce->command_status == CISS_CMD_STATUS_DATA_UNDERRUN)) &&
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
((cc->cdb.cdb[0] == CISS_OPCODE_REPORT_LOGICAL_LUNS) ||
It seems ciss should ignore overrun and underrun on a SCSI INQUIRY command. This fixes some weird booting issues on newer versions of the firmware on the MSA20. Reported by: Philippe Pegon <Philippe dot Pegon at crc dot u-strasbg dot fr>
2005-12-16 06:50:55 +00:00
(cc->cdb.cdb[0] == CISS_OPCODE_REPORT_PHYSICAL_LUNS) ||
(cc->cdb.cdb[0] == INQUIRY))) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.host_tag &= ~CISS_HDR_HOST_TAG_ERROR;
debug(2, "ignoring irrelevant under/overrun error");
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check the command's error bit, if clear, there's no status and
* everything is OK.
*/
if (!(cc->header.host_tag & CISS_HDR_HOST_TAG_ERROR)) {
if (scsi_status != NULL)
*scsi_status = SCSI_STATUS_OK;
if (command_status != NULL)
*command_status = CISS_CMD_STATUS_SUCCESS;
return(0);
} else {
if (command_status != NULL)
*command_status = ce->command_status;
if (scsi_status != NULL) {
if (ce->command_status == CISS_CMD_STATUS_TARGET_STATUS) {
*scsi_status = ce->scsi_status;
} else {
*scsi_status = -1;
}
}
if (bootverbose)
ciss_printf(cr->cr_sc, "command status 0x%x (%s) scsi status 0x%x\n",
ce->command_status, ciss_name_command_status(ce->command_status),
ce->scsi_status);
if (ce->command_status == CISS_CMD_STATUS_INVALID_COMMAND) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_printf(cr->cr_sc, "invalid command, offense size %d at %d, value 0x%x, function %s\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ce->additional_error_info.invalid_command.offense_size,
ce->additional_error_info.invalid_command.offense_offset,
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ce->additional_error_info.invalid_command.offense_value,
func);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#if 0
ciss_print_request(cr);
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(1);
}
/************************************************************************
* Issue a request and don't return until it's completed.
*
* Depending on adapter status, we may poll or sleep waiting for
* completion.
*/
static int
ciss_synch_request(struct ciss_request *cr, int timeout)
{
if (cr->cr_sc->ciss_flags & CISS_FLAG_RUNNING) {
return(ciss_wait_request(cr, timeout));
} else {
return(ciss_poll_request(cr, timeout));
}
}
/************************************************************************
* Issue a request and poll for completion.
*
* Timeout in milliseconds.
*/
static int
ciss_poll_request(struct ciss_request *cr, int timeout)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr_qhead_t qh;
struct ciss_softc *sc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
int error;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(2);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
STAILQ_INIT(&qh);
sc = cr->cr_sc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cr->cr_flags |= CISS_REQ_POLL;
if ((error = ciss_start(cr)) != 0)
return(error);
do {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if (sc->ciss_perf)
ciss_perf_done(sc, &qh);
else
ciss_done(sc, &qh);
ciss_complete(sc, &qh);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (!(cr->cr_flags & CISS_REQ_POLL))
return(0);
DELAY(1000);
} while (timeout-- >= 0);
return(EWOULDBLOCK);
}
/************************************************************************
* Issue a request and sleep waiting for completion.
*
* Timeout in milliseconds. Note that a spurious wakeup will reset
* the timeout.
*/
static int
ciss_wait_request(struct ciss_request *cr, int timeout)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
int error;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(2);
cr->cr_flags |= CISS_REQ_SLEEP;
if ((error = ciss_start(cr)) != 0)
return(error);
Do not catch signals when waiting for a request. This fixes a nasty race when issuing commands from userland.
2004-04-14 18:55:28 +00:00
while ((cr->cr_flags & CISS_REQ_SLEEP) && (error != EWOULDBLOCK)) {
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
error = msleep(cr, &cr->cr_sc->ciss_mtx, PRIBIO, "cissREQ", (timeout * hz) / 1000);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
return(error);
}
Fix warning; line 1640: ciss_abort_request defined but not used.
2001-12-18 08:01:48 +00:00
#if 0
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Abort a request. Note that a potential exists here to race the
* request being completed; the caller must deal with this.
*/
static int
ciss_abort_request(struct ciss_request *ar)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_message_cdb *cmc;
int error;
debug_called(1);
/* get a request */
if ((error = ciss_get_request(ar->cr_sc, &cr)) != 0)
return(error);
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
/* build the abort command */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.address.mode.mode = CISS_HDR_ADDRESS_MODE_PERIPHERAL; /* addressing? */
cc->header.address.physical.target = 0;
cc->header.address.physical.bus = 0;
cc->cdb.cdb_length = sizeof(*cmc);
cc->cdb.type = CISS_CDB_TYPE_MESSAGE;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_NONE;
cc->cdb.timeout = 30;
cmc = (struct ciss_message_cdb *)&(cc->cdb.cdb[0]);
cmc->opcode = CISS_OPCODE_MESSAGE_ABORT;
cmc->type = CISS_MESSAGE_ABORT_TASK;
cmc->abort_tag = ar->cr_tag; /* endianness?? */
/*
* Send the request and wait for a response. If we believe we
* aborted the request OK, clear the flag that indicates it's
* running.
*/
error = ciss_synch_request(cr, 35 * 1000);
if (!error)
error = ciss_report_request(cr, NULL, NULL);
ciss_release_request(cr);
return(error);
}
Fix warning; line 1640: ciss_abort_request defined but not used.
2001-12-18 08:01:48 +00:00
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Fetch and initialise a request
*/
static int
ciss_get_request(struct ciss_softc *sc, struct ciss_request **crp)
{
struct ciss_request *cr;
debug_called(2);
/*
* Get a request and clean it up.
*/
if ((cr = ciss_dequeue_free(sc)) == NULL)
return(ENOMEM);
cr->cr_data = NULL;
cr->cr_flags = 0;
cr->cr_complete = NULL;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
cr->cr_private = NULL;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr->cr_sg_tag = CISS_SG_MAX; /* Backstop to prevent accidents */
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_preen_command(cr);
*crp = cr;
return(0);
}
static void
ciss_preen_command(struct ciss_request *cr)
{
struct ciss_command *cc;
u_int32_t cmdphys;
/*
* Clean up the command structure.
*
* Note that we set up the error_info structure here, since the
* length can be overwritten by any command.
*/
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.sg_in_list = 0; /* kinda inefficient this way */
cc->header.sg_total = 0;
cc->header.host_tag = cr->cr_tag << 2;
cc->header.host_tag_zeroes = 0;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cmdphys = cr->cr_ccphys;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->error_info.error_info_address = cmdphys + sizeof(struct ciss_command);
cc->error_info.error_info_length = CISS_COMMAND_ALLOC_SIZE - sizeof(struct ciss_command);
}
/************************************************************************
* Release a request to the free list.
*/
static void
ciss_release_request(struct ciss_request *cr)
{
struct ciss_softc *sc;
debug_called(2);
sc = cr->cr_sc;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* release the request to the free queue */
ciss_requeue_free(cr);
}
/************************************************************************
* Allocate a request that will be used to send a BMIC command. Do some
* of the common setup here to avoid duplicating it everywhere else.
*/
static int
ciss_get_bmic_request(struct ciss_softc *sc, struct ciss_request **crp,
int opcode, void **bufp, size_t bufsize)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_bmic_cdb *cbc;
void *buf;
int error;
int dataout;
debug_called(2);
cr = NULL;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
buf = NULL;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Get a request.
*/
if ((error = ciss_get_request(sc, &cr)) != 0)
goto out;
/*
* Allocate data storage if requested, determine the data direction.
*/
dataout = 0;
if ((bufsize > 0) && (bufp != NULL)) {
if (*bufp == NULL) {
if ((buf = malloc(bufsize, CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO)) == NULL) {
error = ENOMEM;
goto out;
}
} else {
buf = *bufp;
dataout = 1; /* we are given a buffer, so we are writing */
}
}
/*
* Build a CISS BMIC command to get the logical drive ID.
*/
cr->cr_data = buf;
cr->cr_length = bufsize;
if (!dataout)
cr->cr_flags = CISS_REQ_DATAIN;
White space cleanup
2003-12-13 07:54:07 +00:00
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.address.physical.mode = CISS_HDR_ADDRESS_MODE_PERIPHERAL;
cc->header.address.physical.bus = 0;
cc->header.address.physical.target = 0;
cc->cdb.cdb_length = sizeof(*cbc);
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = dataout ? CISS_CDB_DIRECTION_WRITE : CISS_CDB_DIRECTION_READ;
cc->cdb.timeout = 0;
cbc = (struct ciss_bmic_cdb *)&(cc->cdb.cdb[0]);
bzero(cbc, sizeof(*cbc));
cbc->opcode = dataout ? CISS_ARRAY_CONTROLLER_WRITE : CISS_ARRAY_CONTROLLER_READ;
cbc->bmic_opcode = opcode;
cbc->size = htons((u_int16_t)bufsize);
out:
if (error) {
if (cr != NULL)
ciss_release_request(cr);
} else {
*crp = cr;
if ((bufp != NULL) && (*bufp == NULL) && (buf != NULL))
*bufp = buf;
}
return(error);
}
/************************************************************************
* Handle a command passed in from userspace.
*/
static int
ciss_user_command(struct ciss_softc *sc, IOCTL_Command_struct *ioc)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_error_info *ce;
Check the status of the user command and return the proper error to the user.
2003-07-12 22:37:03 +00:00
int error = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
cr = NULL;
/*
* Get a request.
*/
Spin until a request structure is available in the ioctl path.
2006-05-30 06:42:02 +00:00
while (ciss_get_request(sc, &cr) != 0)
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
msleep(sc, &sc->ciss_mtx, PPAUSE, "cissREQ", hz);
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Allocate an in-kernel databuffer if required, copy in user data.
*/
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cr->cr_length = ioc->buf_size;
if (ioc->buf_size > 0) {
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
if ((cr->cr_data = malloc(ioc->buf_size, CISS_MALLOC_CLASS, M_NOWAIT)) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
error = ENOMEM;
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
goto out_unlocked;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
if ((error = copyin(ioc->buf, cr->cr_data, ioc->buf_size))) {
debug(0, "copyin: bad data buffer %p/%d", ioc->buf, ioc->buf_size);
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
goto out_unlocked;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
}
/*
* Build the request based on the user command.
*/
bcopy(&ioc->LUN_info, &cc->header.address, sizeof(cc->header.address));
bcopy(&ioc->Request, &cc->cdb, sizeof(cc->cdb));
/* XXX anything else to populate here? */
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Run the command.
*/
if ((error = ciss_synch_request(cr, 60 * 1000))) {
debug(0, "request failed - %d", error);
goto out;
}
/*
Check the status of the user command and return the proper error to the user.
2003-07-12 22:37:03 +00:00
* Check to see if the command succeeded.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
ce = (struct ciss_error_info *)&(cc->sg[0]);
There's no need to call ciss_report_request in the passthru ioctl routine since the error will be reported back to the user buffer. This will quiet down the bootverbose case when using an ACU which does brute force discovery of the physical and logical devices.
2004-04-19 17:16:06 +00:00
if ((cc->header.host_tag & CISS_HDR_HOST_TAG_ERROR) == 0)
Check the status of the user command and return the proper error to the user.
2003-07-12 22:37:03 +00:00
bzero(ce, sizeof(*ce));
/*
* Copy the results back to the user.
*/
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
bcopy(ce, &ioc->error_info, sizeof(*ce));
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if ((ioc->buf_size > 0) &&
(error = copyout(cr->cr_data, ioc->buf, ioc->buf_size))) {
debug(0, "copyout: bad data buffer %p/%d", ioc->buf, ioc->buf_size);
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
goto out_unlocked;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/* done OK */
error = 0;
Fix locking around copyout() operations.
2009-09-16 23:17:22 +00:00
out_unlocked:
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
out:
if ((cr != NULL) && (cr->cr_data != NULL))
free(cr->cr_data, CISS_MALLOC_CLASS);
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
/************************************************************************
* Map a request into bus-visible space, initialise the scatter/gather
* list.
*/
static int
ciss_map_request(struct ciss_request *cr)
{
struct ciss_softc *sc;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
int error = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(2);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc = cr->cr_sc;
/* check that mapping is necessary */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
if (cr->cr_flags & CISS_REQ_MAPPED)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
cr->cr_flags |= CISS_REQ_MAPPED;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
bus_dmamap_sync(sc->ciss_command_dmat, sc->ciss_command_map,
BUS_DMASYNC_PREWRITE);
if (cr->cr_data != NULL) {
error = bus_dmamap_load(sc->ciss_buffer_dmat, cr->cr_datamap,
cr->cr_data, cr->cr_length,
ciss_request_map_helper, cr, 0);
if (error != 0)
return (error);
} else {
/*
* Post the command to the adapter.
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr->cr_sg_tag = CISS_SG_NONE;
cr->cr_flags |= CISS_REQ_BUSY;
if (sc->ciss_perf)
CISS_TL_PERF_POST_CMD(sc, cr);
else
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
CISS_TL_SIMPLE_POST_CMD(sc, cr->cr_ccphys);
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
}
static void
ciss_request_map_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct ciss_command *cc;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
struct ciss_request *cr;
struct ciss_softc *sc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
int i;
debug_called(2);
White space cleanup
2003-12-13 07:54:07 +00:00
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
cr = (struct ciss_request *)arg;
sc = cr->cr_sc;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
for (i = 0; i < nseg; i++) {
cc->sg[i].address = segs[i].ds_addr;
cc->sg[i].length = segs[i].ds_len;
cc->sg[i].extension = 0;
}
/* we leave the s/g table entirely within the command */
cc->header.sg_in_list = nseg;
cc->header.sg_total = nseg;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
if (cr->cr_flags & CISS_REQ_DATAIN)
bus_dmamap_sync(sc->ciss_buffer_dmat, cr->cr_datamap, BUS_DMASYNC_PREREAD);
if (cr->cr_flags & CISS_REQ_DATAOUT)
bus_dmamap_sync(sc->ciss_buffer_dmat, cr->cr_datamap, BUS_DMASYNC_PREWRITE);
Fix an apparently harmless typo. Approved by: re
2009-07-20 03:59:00 +00:00
if (nseg == 0)
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr->cr_sg_tag = CISS_SG_NONE;
else if (nseg == 1)
cr->cr_sg_tag = CISS_SG_1;
else if (nseg == 2)
cr->cr_sg_tag = CISS_SG_2;
else if (nseg <= 4)
cr->cr_sg_tag = CISS_SG_4;
else if (nseg <= 8)
cr->cr_sg_tag = CISS_SG_8;
else if (nseg <= 16)
cr->cr_sg_tag = CISS_SG_16;
else if (nseg <= 32)
cr->cr_sg_tag = CISS_SG_32;
else
cr->cr_sg_tag = CISS_SG_MAX;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
/*
* Post the command to the adapter.
*/
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr->cr_flags |= CISS_REQ_BUSY;
if (sc->ciss_perf)
CISS_TL_PERF_POST_CMD(sc, cr);
else
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
CISS_TL_SIMPLE_POST_CMD(sc, cr->cr_ccphys);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Unmap a request from bus-visible space.
*/
static void
ciss_unmap_request(struct ciss_request *cr)
{
struct ciss_softc *sc;
debug_called(2);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc = cr->cr_sc;
/* check that unmapping is necessary */
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
if ((cr->cr_flags & CISS_REQ_MAPPED) == 0)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return;
White space cleanup
2003-12-13 07:54:07 +00:00
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
bus_dmamap_sync(sc->ciss_command_dmat, sc->ciss_command_map,
BUS_DMASYNC_POSTWRITE);
if (cr->cr_data == NULL)
goto out;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (cr->cr_flags & CISS_REQ_DATAIN)
bus_dmamap_sync(sc->ciss_buffer_dmat, cr->cr_datamap, BUS_DMASYNC_POSTREAD);
if (cr->cr_flags & CISS_REQ_DATAOUT)
bus_dmamap_sync(sc->ciss_buffer_dmat, cr->cr_datamap, BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->ciss_buffer_dmat, cr->cr_datamap);
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
out:
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cr->cr_flags &= ~CISS_REQ_MAPPED;
}
/************************************************************************
* Attach the driver to CAM.
*
* We put all the logical drives on a single SCSI bus.
*/
static int
ciss_cam_init(struct ciss_softc *sc)
{
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
int i, maxbus;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
/*
* Allocate a devq. We can reuse this for the masked physical
* devices if we decide to export these as well.
*/
Several changes to fix livelock under high load, introduced by r203489: - change the way in which command queue overflow is handled; - do not expose to CAM two command slots, used for driver's internal purposes; - allow driver to use up to 1024 command slots, instead of 256 before.
2010-03-03 17:58:41 +00:00
if ((sc->ciss_cam_devq = cam_simq_alloc(sc->ciss_max_requests - 2)) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "can't allocate CAM SIM queue\n");
return(ENOMEM);
}
/*
* Create a SIM.
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
*
* This naturally wastes a bit of memory. The alternative is to allocate
* and register each bus as it is found, and then track them on a linked
* list. Unfortunately, the driver has a few places where it needs to
* look up the SIM based solely on bus number, and it's unclear whether
* a list traversal would work for these situations.
*/
maxbus = max(sc->ciss_max_logical_bus, sc->ciss_max_physical_bus +
CISS_PHYSICAL_BASE);
sc->ciss_cam_sim = malloc(maxbus * sizeof(struct cam_sim*),
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
CISS_MALLOC_CLASS, M_NOWAIT | M_ZERO);
if (sc->ciss_cam_sim == NULL) {
ciss_printf(sc, "can't allocate memory for controller SIM\n");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(ENOMEM);
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
for (i = 0; i < sc->ciss_max_logical_bus; i++) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if ((sc->ciss_cam_sim[i] = cam_sim_alloc(ciss_cam_action, ciss_cam_poll,
"ciss", sc,
device_get_unit(sc->ciss_dev),
Fix a mistake made in the MPSAFE commit that caused CAM to serialize requests to the controller.
2007-10-12 17:03:41 +00:00
&sc->ciss_mtx,
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
2,
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
sc->ciss_max_requests - 2,
sc->ciss_cam_devq)) == NULL) {
ciss_printf(sc, "can't allocate CAM SIM for controller %d\n", i);
return(ENOMEM);
}
/*
* Register bus with this SIM.
*/
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
if (i == 0 || sc->ciss_controllers[i].physical.bus != 0) {
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
if (xpt_bus_register(sc->ciss_cam_sim[i], sc->ciss_dev, i) != 0) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
ciss_printf(sc, "can't register SCSI bus %d\n", i);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
return (ENXIO);
}
}
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
}
for (i = CISS_PHYSICAL_BASE; i < sc->ciss_max_physical_bus +
CISS_PHYSICAL_BASE; i++) {
if ((sc->ciss_cam_sim[i] = cam_sim_alloc(ciss_cam_action, ciss_cam_poll,
"ciss", sc,
device_get_unit(sc->ciss_dev),
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
&sc->ciss_mtx, 1,
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
sc->ciss_max_requests - 2,
sc->ciss_cam_devq)) == NULL) {
ciss_printf(sc, "can't allocate CAM SIM for controller %d\n", i);
return (ENOMEM);
}
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
if (xpt_bus_register(sc->ciss_cam_sim[i], sc->ciss_dev, i) != 0) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_printf(sc, "can't register SCSI bus %d\n", i);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
return (ENXIO);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
}
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
return(0);
}
/************************************************************************
* Initiate a rescan of the 'logical devices' SIM
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
*/
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
static void
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_cam_rescan_target(struct ciss_softc *sc, int bus, int target)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
union ccb *ccb;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if ((ccb = xpt_alloc_ccb_nowait()) == NULL) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "rescan failed (can't allocate CCB)\n");
return;
}
White space cleanup
2003-12-13 07:54:07 +00:00
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if (xpt_create_path(&ccb->ccb_h.path, xpt_periph,
cam_sim_path(sc->ciss_cam_sim[bus]),
target, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "rescan failed (can't create path)\n");
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_free_ccb(ccb);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return;
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_rescan(ccb);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* scan is now in progress */
}
/************************************************************************
* Handle requests coming from CAM
*/
static void
ciss_cam_action(struct cam_sim *sim, union ccb *ccb)
{
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
struct ciss_softc *sc;
struct ccb_scsiio *csio;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
int bus, target;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
int physical;
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
sc = cam_sim_softc(sim);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
bus = cam_sim_bus(sim);
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
csio = (struct ccb_scsiio *)&ccb->csio;
target = csio->ccb_h.target_id;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
physical = CISS_IS_PHYSICAL(bus);
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
switch (ccb->ccb_h.func_code) {
/* perform SCSI I/O */
case XPT_SCSI_IO:
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
if (!ciss_cam_action_io(sim, csio))
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return;
break;
/* perform geometry calculations */
case XPT_CALC_GEOMETRY:
{
struct ccb_calc_geometry *ccg = &ccb->ccg;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
struct ciss_ldrive *ld;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug(1, "XPT_CALC_GEOMETRY %d:%d:%d", cam_sim_bus(sim), ccb->ccb_h.target_id, ccb->ccb_h.target_lun);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
ld = NULL;
if (!physical)
ld = &sc->ciss_logical[bus][target];
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
* Use the cached geometry settings unless the fault tolerance
* is invalid.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
if (physical || ld->cl_geometry.fault_tolerance == 0xFF) {
Properly get the drive geometry from the controller. This should fix booting off of volumes > 255GB.
2003-02-05 08:43:46 +00:00
u_int32_t secs_per_cylinder;
ccg->heads = 255;
ccg->secs_per_track = 32;
secs_per_cylinder = ccg->heads * ccg->secs_per_track;
ccg->cylinders = ccg->volume_size / secs_per_cylinder;
} else {
ccg->heads = ld->cl_geometry.heads;
ccg->secs_per_track = ld->cl_geometry.sectors;
ccg->cylinders = ntohs(ld->cl_geometry.cylinders);
}
ccb->ccb_h.status = CAM_REQ_CMP;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
}
/* handle path attribute inquiry */
case XPT_PATH_INQ:
{
struct ccb_pathinq *cpi = &ccb->cpi;
debug(1, "XPT_PATH_INQ %d:%d:%d", cam_sim_bus(sim), ccb->ccb_h.target_id, ccb->ccb_h.target_lun);
cpi->version_num = 1;
cpi->hba_inquiry = PI_TAG_ABLE; /* XXX is this correct? */
cpi->target_sprt = 0;
cpi->hba_misc = 0;
cpi->max_target = CISS_MAX_LOGICAL;
cpi->max_lun = 0; /* 'logical drive' channel only */
cpi->initiator_id = CISS_MAX_LOGICAL;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "msmith@freebsd.org", HBA_IDLEN);
strncpy(cpi->dev_name, cam_sim_name(sim), DEV_IDLEN);
cpi->unit_number = cam_sim_unit(sim);
cpi->bus_id = cam_sim_bus(sim);
cpi->base_transfer_speed = 132 * 1024; /* XXX what to set this to? */
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
cpi->transport = XPORT_SPI;
cpi->transport_version = 2;
cpi->protocol = PROTO_SCSI;
cpi->protocol_version = SCSI_REV_2;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cpi->maxio = (CISS_MAX_SG_ELEMENTS - 1) * PAGE_SIZE;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_GET_TRAN_SETTINGS:
{
struct ccb_trans_settings *cts = &ccb->cts;
int bus, target;
Tell CAM that CISS logical devices support tagged queueing. This fixes the low "max device openings" count that has led to poor performance in FreeBSD 7.0 and 7.1
2009-02-20 07:43:43 +00:00
struct ccb_trans_settings_spi *spi = &cts->xport_specific.spi;
struct ccb_trans_settings_scsi *scsi = &cts->proto_specific.scsi;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
bus = cam_sim_bus(sim);
target = cts->ccb_h.target_id;
debug(1, "XPT_GET_TRAN_SETTINGS %d:%d", bus, target);
/* disconnect always OK */
The first of 3 major steps to move the CAM layer forward to using the CAM_NEW_TRAN_CODE that has been in the tree for some years now. This first step consists solely of adding to or correcting CAM_NEW_TRAN_CODE pieces in the kernel source tree such that a both a GENERIC (at least on i386) and a LINT build with CAM_NEW_TRAN_CODE as an option will compile correctly and run (at least with some the h/w I have). After a short settle time, the other pieces (making CAM_NEW_TRAN_CODE the default and updating libcam and camcontrol) will be brought in. This will be an incompatible change in that the size of structures related to XPT_PATH_INQ and XPT_{GET,SET}_TRAN_SETTINGS change in both size and content. However, basic system operation and basic system utilities work well enough with this change. Reviewed by: freebsd-scsi and specific stakeholders
2006-10-31 05:53:29 +00:00
cts->protocol = PROTO_SCSI;
cts->protocol_version = SCSI_REV_2;
cts->transport = XPORT_SPI;
cts->transport_version = 2;
spi->valid = CTS_SPI_VALID_DISC;
spi->flags = CTS_SPI_FLAGS_DISC_ENB;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
Tell CAM that CISS logical devices support tagged queueing. This fixes the low "max device openings" count that has led to poor performance in FreeBSD 7.0 and 7.1
2009-02-20 07:43:43 +00:00
scsi->valid = CTS_SCSI_VALID_TQ;
scsi->flags = CTS_SCSI_FLAGS_TAG_ENB;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cts->ccb_h.status = CAM_REQ_CMP;
break;
}
default: /* we can't do this */
debug(1, "unspported func_code = 0x%x", ccb->ccb_h.func_code);
ccb->ccb_h.status = CAM_REQ_INVALID;
break;
}
xpt_done(ccb);
}
/************************************************************************
* Handle a CAM SCSI I/O request.
*/
static int
ciss_cam_action_io(struct cam_sim *sim, struct ccb_scsiio *csio)
{
struct ciss_softc *sc;
int bus, target;
struct ciss_request *cr;
struct ciss_command *cc;
int error;
sc = cam_sim_softc(sim);
bus = cam_sim_bus(sim);
target = csio->ccb_h.target_id;
debug(2, "XPT_SCSI_IO %d:%d:%d", bus, target, csio->ccb_h.target_lun);
/* check that the CDB pointer is not to a physical address */
if ((csio->ccb_h.flags & CAM_CDB_POINTER) && (csio->ccb_h.flags & CAM_CDB_PHYS)) {
debug(3, " CDB pointer is to physical address");
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
/* if there is data transfer, it must be to/from a virtual address */
if ((csio->ccb_h.flags & CAM_DIR_MASK) != CAM_DIR_NONE) {
if (csio->ccb_h.flags & CAM_DATA_PHYS) { /* we can't map it */
debug(3, " data pointer is to physical address");
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
if (csio->ccb_h.flags & CAM_SCATTER_VALID) { /* we want to do the s/g setup */
debug(3, " data has premature s/g setup");
csio->ccb_h.status = CAM_REQ_CMP_ERR;
}
}
/* abandon aborted ccbs or those that have failed validation */
if ((csio->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_INPROG) {
debug(3, "abandoning CCB due to abort/validation failure");
return(EINVAL);
}
/* handle emulation of some SCSI commands ourself */
if (ciss_cam_emulate(sc, csio))
return(0);
/*
* Get a request to manage this command. If we can't, return the
* ccb, freeze the queue and flag so that we unfreeze it when a
* request completes.
*/
if ((error = ciss_get_request(sc, &cr)) != 0) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
xpt_freeze_simq(sim, 1);
Several changes to fix livelock under high load, introduced by r203489: - change the way in which command queue overflow is handled; - do not expose to CAM two command slots, used for driver's internal purposes; - allow driver to use up to 1024 command slots, instead of 256 before.
2010-03-03 17:58:41 +00:00
sc->ciss_flags |= CISS_FLAG_BUSY;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
csio->ccb_h.status |= CAM_REQUEUE_REQ;
return(error);
}
/*
* Build the command.
*/
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cr->cr_data = csio->data_ptr;
cr->cr_length = csio->dxfer_len;
cr->cr_complete = ciss_cam_complete;
cr->cr_private = csio;
White space cleanup
2003-12-13 07:54:07 +00:00
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
/*
* Target the right logical volume.
*/
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
if (CISS_IS_PHYSICAL(bus))
cc->header.address =
sc->ciss_physical[CISS_CAM_TO_PBUS(bus)][target].cp_address;
else
cc->header.address =
sc->ciss_logical[bus][target].cl_address;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->cdb.cdb_length = csio->cdb_len;
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE; /* XXX ordered tags? */
if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_OUT) {
cr->cr_flags = CISS_REQ_DATAOUT;
cc->cdb.direction = CISS_CDB_DIRECTION_WRITE;
} else if ((csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN) {
cr->cr_flags = CISS_REQ_DATAIN;
cc->cdb.direction = CISS_CDB_DIRECTION_READ;
} else {
cr->cr_flags = 0;
cc->cdb.direction = CISS_CDB_DIRECTION_NONE;
}
cc->cdb.timeout = (csio->ccb_h.timeout / 1000) + 1;
if (csio->ccb_h.flags & CAM_CDB_POINTER) {
bcopy(csio->cdb_io.cdb_ptr, &cc->cdb.cdb[0], csio->cdb_len);
} else {
bcopy(csio->cdb_io.cdb_bytes, &cc->cdb.cdb[0], csio->cdb_len);
}
/*
* Submit the request to the adapter.
*
* Note that this may fail if we're unable to map the request (and
* if we ever learn a transport layer other than simple, may fail
* if the adapter rejects the command).
*/
if ((error = ciss_start(cr)) != 0) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
xpt_freeze_simq(sim, 1);
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_RELEASE_SIMQ;
Support EINPROGRESS and properly deal with 64bit physical addressing.
2003-08-12 17:55:53 +00:00
if (error == EINPROGRESS) {
error = 0;
} else {
csio->ccb_h.status |= CAM_REQUEUE_REQ;
ciss_release_request(cr);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(error);
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
}
/************************************************************************
* Emulate SCSI commands the adapter doesn't handle as we might like.
*/
static int
ciss_cam_emulate(struct ciss_softc *sc, struct ccb_scsiio *csio)
{
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
int bus, target;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
u_int8_t opcode;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
target = csio->ccb_h.target_id;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
bus = cam_sim_bus(xpt_path_sim(csio->ccb_h.path));
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
opcode = (csio->ccb_h.flags & CAM_CDB_POINTER) ?
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*(u_int8_t *)csio->cdb_io.cdb_ptr : csio->cdb_io.cdb_bytes[0];
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
if (CISS_IS_PHYSICAL(bus)) {
if (sc->ciss_physical[CISS_CAM_TO_PBUS(bus)][target].cp_online != 1) {
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_SEL_TIMEOUT;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
xpt_done((union ccb *)csio);
return(1);
} else
return(0);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
If a device is not ONLINE, do not attempt to submit any commands or I/O to the volume. This solves a panic when removing a disk in a RAID 0 volume. Reported by: le
2004-06-15 19:40:47 +00:00
* Handle requests for volumes that don't exist or are not online.
* A selection timeout is slightly better than an illegal request.
* Other errors might be better.
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
If a device is not ONLINE, do not attempt to submit any commands or I/O to the volume. This solves a panic when removing a disk in a RAID 0 volume. Reported by: le
2004-06-15 19:40:47 +00:00
if (sc->ciss_logical[bus][target].cl_status != CISS_LD_ONLINE) {
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_SEL_TIMEOUT;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
xpt_done((union ccb *)csio);
return(1);
}
/* if we have to fake Synchronise Cache */
if (sc->ciss_flags & CISS_FLAG_FAKE_SYNCH) {
/*
* If this is a Synchronise Cache command, typically issued when
* a device is closed, flush the adapter and complete now.
*/
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
if (((csio->ccb_h.flags & CAM_CDB_POINTER) ?
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*(u_int8_t *)csio->cdb_io.cdb_ptr : csio->cdb_io.cdb_bytes[0]) == SYNCHRONIZE_CACHE) {
ciss_flush_adapter(sc);
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_REQ_CMP;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
xpt_done((union ccb *)csio);
return(1);
}
}
return(0);
}
/************************************************************************
* Check for possibly-completed commands.
*/
static void
ciss_cam_poll(struct cam_sim *sim)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr_qhead_t qh;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
struct ciss_softc *sc = cam_sim_softc(sim);
debug_called(2);
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
STAILQ_INIT(&qh);
if (sc->ciss_perf)
ciss_perf_done(sc, &qh);
else
ciss_done(sc, &qh);
ciss_complete(sc, &qh);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Handle completion of a command - pass results back through the CCB
*/
static void
ciss_cam_complete(struct ciss_request *cr)
{
struct ciss_softc *sc;
struct ciss_command *cc;
struct ciss_error_info *ce;
struct ccb_scsiio *csio;
int scsi_status;
int command_status;
debug_called(2);
sc = cr->cr_sc;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ce = (struct ciss_error_info *)&(cc->sg[0]);
csio = (struct ccb_scsiio *)cr->cr_private;
/*
* Extract status values from request.
*/
ciss_report_request(cr, &command_status, &scsi_status);
csio->scsi_status = scsi_status;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Handle specific SCSI status values.
*/
switch(scsi_status) {
/* no status due to adapter error */
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
case -1:
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug(0, "adapter error");
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_REQ_CMP_ERR;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* no status due to command completed OK */
case SCSI_STATUS_OK: /* CISS_SCSI_STATUS_GOOD */
debug(2, "SCSI_STATUS_OK");
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_REQ_CMP;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
/* check condition, sense data included */
case SCSI_STATUS_CHECK_COND: /* CISS_SCSI_STATUS_CHECK_CONDITION */
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
debug(0, "SCSI_STATUS_CHECK_COND sense size %d resid %d\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ce->sense_length, ce->residual_count);
bzero(&csio->sense_data, SSD_FULL_SIZE);
bcopy(&ce->sense_info[0], &csio->sense_data, ce->sense_length);
Add descriptor sense support to CAM, and honor sense residuals properly in CAM. Desriptor sense is a new sense data format that originated in SPC-3. Among other things, it allows for an 8-byte info field, which is necessary to pass back block numbers larger than 4 bytes. This change adds a number of new functions to scsi_all.c (and therefore libcam) that abstract out most access to sense data. This includes a bump of CAM_VERSION, because the CCB ABI has changed. Userland programs that use the CAM pass(4) driver will need to be recompiled. camcontrol.c: Change uses of scsi_extract_sense() to use scsi_extract_sense_len(). Use scsi_get_sks() instead of accessing sense key specific data directly. scsi_modes: Update the control mode page to the latest version (SPC-4). scsi_cmds.c, scsi_target.c: Change references to struct scsi_sense_data to struct scsi_sense_data_fixed. This should be changed to allow the user to specify fixed or descriptor sense, and then use scsi_set_sense_data() to build the sense data. ps3cdrom.c: Use scsi_set_sense_data() instead of setting sense data manually. cam_periph.c: Use scsi_extract_sense_len() instead of using scsi_extract_sense() or accessing sense data directly. cam_ccb.h: Bump the CAM_VERSION from 0x15 to 0x16. The change of struct scsi_sense_data from 32 to 252 bytes changes the size of struct ccb_scsiio, but not the size of union ccb. So the version must be bumped to prevent structure mis-matches. scsi_all.h: Lots of updated SCSI sense data and other structures. Add function prototypes for the new sense data functions. Take out the inline implementation of scsi_extract_sense(). It is now too large to put in a header file. Add macros to calculate whether fields are present and filled in fixed and descriptor sense data scsi_all.c: In scsi_op_desc(), allow the user to pass in NULL inquiry data, and we'll assume a direct access device in that case. Changed the SCSI RESERVED sense key name and description to COMPLETED, as it is now defined in the spec. Change the error recovery action for a number of read errors to prevent lots of retries when the drive has said that the block isn't accessible. This speeds up reconstruction of the block by any RAID software running on top of the drive (e.g. ZFS). In scsi_sense_desc(), allow for invalid sense key numbers. This allows calling this routine without checking the input values first. Change scsi_error_action() to use scsi_extract_sense_len(), and handle things when invalid asc/ascq values are encountered. Add a new routine, scsi_desc_iterate(), that will call the supplied function for every descriptor in descriptor format sense data. Add scsi_set_sense_data(), and scsi_set_sense_data_va(), which build descriptor and fixed format sense data. They currently default to fixed format sense data. Add a number of scsi_get_*() functions, which get different types of sense data fields from either fixed or descriptor format sense data, if the data is present. Add a number of scsi_*_sbuf() functions, which print formatted versions of various sense data fields. These functions work for either fixed or descriptor sense. Add a number of scsi_sense_*_sbuf() functions, which have a standard calling interface and print the indicated field. These functions take descriptors only. Add scsi_sense_desc_sbuf(), which will print a formatted version of the given sense descriptor. Pull out a majority of the scsi_sense_sbuf() function and put it into scsi_sense_only_sbuf(). This allows callers that don't use struct ccb_scsiio to easily utilize the printing routines. Revamp that function to handle descriptor sense and use the new sense fetching and printing routines. Move scsi_extract_sense() into scsi_all.c, and implement it in terms of the new function, scsi_extract_sense_len(). The _len() version takes a length (which should be the sense length - residual) and can indicate which fields are present and valid in the sense data. Add a couple of new scsi_get_*() routines to get the sense key, asc, and ascq only. mly.c: Rename struct scsi_sense_data to struct scsi_sense_data_fixed. sbp_targ.c: Use the new sense fetching routines to get sense data instead of accessing it directly. sbp.c: Change the firewire/SCSI sense data transformation code to use struct scsi_sense_data_fixed instead of struct scsi_sense_data. This should be changed later to use scsi_set_sense_data(). ciss.c: Calculate the sense residual properly. Use scsi_get_sense_key() to fetch the sense key. mps_sas.c, mpt_cam.c: Set the sense residual properly. iir.c: Use scsi_set_sense_data() instead of building sense data by hand. iscsi_subr.c: Use scsi_extract_sense_len() instead of grabbing sense data directly. umass.c: Use scsi_set_sense_data() to build sense data. Grab the sense key using scsi_get_sense_key(). Calculate the sense residual properly. isp_freebsd.h: Use scsi_get_*() routines to grab asc, ascq, and sense key values. Calculate and set the sense residual. MFC after: 3 days Sponsored by: Spectra Logic Corporation
2011-10-03 20:32:55 +00:00
if (csio->sense_len > ce->sense_length)
csio->sense_resid = csio->sense_len - ce->sense_length;
else
csio->sense_resid = 0;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
csio->resid = ce->residual_count;
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_SCSI_STATUS_ERROR | CAM_AUTOSNS_VALID;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
#ifdef CISS_DEBUG
{
struct scsi_sense_data *sns = (struct scsi_sense_data *)&ce->sense_info[0];
Add descriptor sense support to CAM, and honor sense residuals properly in CAM. Desriptor sense is a new sense data format that originated in SPC-3. Among other things, it allows for an 8-byte info field, which is necessary to pass back block numbers larger than 4 bytes. This change adds a number of new functions to scsi_all.c (and therefore libcam) that abstract out most access to sense data. This includes a bump of CAM_VERSION, because the CCB ABI has changed. Userland programs that use the CAM pass(4) driver will need to be recompiled. camcontrol.c: Change uses of scsi_extract_sense() to use scsi_extract_sense_len(). Use scsi_get_sks() instead of accessing sense key specific data directly. scsi_modes: Update the control mode page to the latest version (SPC-4). scsi_cmds.c, scsi_target.c: Change references to struct scsi_sense_data to struct scsi_sense_data_fixed. This should be changed to allow the user to specify fixed or descriptor sense, and then use scsi_set_sense_data() to build the sense data. ps3cdrom.c: Use scsi_set_sense_data() instead of setting sense data manually. cam_periph.c: Use scsi_extract_sense_len() instead of using scsi_extract_sense() or accessing sense data directly. cam_ccb.h: Bump the CAM_VERSION from 0x15 to 0x16. The change of struct scsi_sense_data from 32 to 252 bytes changes the size of struct ccb_scsiio, but not the size of union ccb. So the version must be bumped to prevent structure mis-matches. scsi_all.h: Lots of updated SCSI sense data and other structures. Add function prototypes for the new sense data functions. Take out the inline implementation of scsi_extract_sense(). It is now too large to put in a header file. Add macros to calculate whether fields are present and filled in fixed and descriptor sense data scsi_all.c: In scsi_op_desc(), allow the user to pass in NULL inquiry data, and we'll assume a direct access device in that case. Changed the SCSI RESERVED sense key name and description to COMPLETED, as it is now defined in the spec. Change the error recovery action for a number of read errors to prevent lots of retries when the drive has said that the block isn't accessible. This speeds up reconstruction of the block by any RAID software running on top of the drive (e.g. ZFS). In scsi_sense_desc(), allow for invalid sense key numbers. This allows calling this routine without checking the input values first. Change scsi_error_action() to use scsi_extract_sense_len(), and handle things when invalid asc/ascq values are encountered. Add a new routine, scsi_desc_iterate(), that will call the supplied function for every descriptor in descriptor format sense data. Add scsi_set_sense_data(), and scsi_set_sense_data_va(), which build descriptor and fixed format sense data. They currently default to fixed format sense data. Add a number of scsi_get_*() functions, which get different types of sense data fields from either fixed or descriptor format sense data, if the data is present. Add a number of scsi_*_sbuf() functions, which print formatted versions of various sense data fields. These functions work for either fixed or descriptor sense. Add a number of scsi_sense_*_sbuf() functions, which have a standard calling interface and print the indicated field. These functions take descriptors only. Add scsi_sense_desc_sbuf(), which will print a formatted version of the given sense descriptor. Pull out a majority of the scsi_sense_sbuf() function and put it into scsi_sense_only_sbuf(). This allows callers that don't use struct ccb_scsiio to easily utilize the printing routines. Revamp that function to handle descriptor sense and use the new sense fetching and printing routines. Move scsi_extract_sense() into scsi_all.c, and implement it in terms of the new function, scsi_extract_sense_len(). The _len() version takes a length (which should be the sense length - residual) and can indicate which fields are present and valid in the sense data. Add a couple of new scsi_get_*() routines to get the sense key, asc, and ascq only. mly.c: Rename struct scsi_sense_data to struct scsi_sense_data_fixed. sbp_targ.c: Use the new sense fetching routines to get sense data instead of accessing it directly. sbp.c: Change the firewire/SCSI sense data transformation code to use struct scsi_sense_data_fixed instead of struct scsi_sense_data. This should be changed later to use scsi_set_sense_data(). ciss.c: Calculate the sense residual properly. Use scsi_get_sense_key() to fetch the sense key. mps_sas.c, mpt_cam.c: Set the sense residual properly. iir.c: Use scsi_set_sense_data() instead of building sense data by hand. iscsi_subr.c: Use scsi_extract_sense_len() instead of grabbing sense data directly. umass.c: Use scsi_set_sense_data() to build sense data. Grab the sense key using scsi_get_sense_key(). Calculate the sense residual properly. isp_freebsd.h: Use scsi_get_*() routines to grab asc, ascq, and sense key values. Calculate and set the sense residual. MFC after: 3 days Sponsored by: Spectra Logic Corporation
2011-10-03 20:32:55 +00:00
debug(0, "sense key %x", scsi_get_sense_key(sns, csio->sense_len -
csio->sense_resid, /*show_errors*/ 1));
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
case SCSI_STATUS_BUSY: /* CISS_SCSI_STATUS_BUSY */
debug(0, "SCSI_STATUS_BUSY");
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_SCSI_BUSY;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
default:
debug(0, "unknown status 0x%x", csio->scsi_status);
Return CAM_RELEASE_SIMQ flag only when it is needed, when SIM really was frozen before and should be released.
2010-02-04 14:40:08 +00:00
csio->ccb_h.status |= CAM_REQ_CMP_ERR;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
}
/* handle post-command fixup */
ciss_cam_complete_fixup(sc, csio);
ciss_release_request(cr);
Several changes to fix livelock under high load, introduced by r203489: - change the way in which command queue overflow is handled; - do not expose to CAM two command slots, used for driver's internal purposes; - allow driver to use up to 1024 command slots, instead of 256 before.
2010-03-03 17:58:41 +00:00
if (sc->ciss_flags & CISS_FLAG_BUSY) {
sc->ciss_flags &= ~CISS_FLAG_BUSY;
if (csio->ccb_h.status & CAM_RELEASE_SIMQ)
xpt_release_simq(xpt_path_sim(csio->ccb_h.path), 0);
else
csio->ccb_h.status |= CAM_RELEASE_SIMQ;
}
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
xpt_done((union ccb *)csio);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/********************************************************************************
* Fix up the result of some commands here.
*/
static void
ciss_cam_complete_fixup(struct ciss_softc *sc, struct ccb_scsiio *csio)
{
struct scsi_inquiry_data *inq;
struct ciss_ldrive *cl;
When hacking INQUIRY result, make sure that it is right INQUIRY and there is enough of result to hack.
2010-02-05 12:40:18 +00:00
uint8_t *cdb;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
int bus, target;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
When hacking INQUIRY result, make sure that it is right INQUIRY and there is enough of result to hack.
2010-02-05 12:40:18 +00:00
cdb = (csio->ccb_h.flags & CAM_CDB_POINTER) ?
(uint8_t *)csio->cdb_io.cdb_ptr : csio->cdb_io.cdb_bytes;
if (cdb[0] == INQUIRY &&
(cdb[1] & SI_EVPD) == 0 &&
(csio->ccb_h.flags & CAM_DIR_MASK) == CAM_DIR_IN &&
csio->dxfer_len >= SHORT_INQUIRY_LENGTH) {
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
inq = (struct scsi_inquiry_data *)csio->data_ptr;
target = csio->ccb_h.target_id;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
bus = cam_sim_bus(xpt_path_sim(csio->ccb_h.path));
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
/*
* Don't let hard drives be seen by the DA driver. They will still be
* attached by the PASS driver.
*/
if (CISS_IS_PHYSICAL(bus)) {
if (SID_TYPE(inq) == T_DIRECT)
inq->device = (inq->device & 0xe0) | T_NODEVICE;
return;
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
cl = &sc->ciss_logical[bus][target];
White space cleanup
2003-12-13 07:54:07 +00:00
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
padstr(inq->vendor, "COMPAQ", 8);
padstr(inq->product, ciss_name_ldrive_org(cl->cl_ldrive->fault_tolerance), 8);
padstr(inq->revision, ciss_name_ldrive_status(cl->cl_lstatus->status), 16);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
}
/********************************************************************************
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
* Find a peripheral attached at (target)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
*/
static struct cam_periph *
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_find_periph(struct ciss_softc *sc, int bus, int target)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct cam_periph *periph;
struct cam_path *path;
int status;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
status = xpt_create_path(&path, NULL, cam_sim_path(sc->ciss_cam_sim[bus]),
target, 0);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (status == CAM_REQ_CMP) {
periph = cam_periph_find(path, NULL);
xpt_free_path(path);
} else {
periph = NULL;
}
return(periph);
}
/********************************************************************************
* Name the device at (target)
*
* XXX is this strictly correct?
*/
Be consistent about "static" functions: if the function is marked static in its prototype, mark it static at the definition too. Inspired by: FlexeLint warning #512
2002-09-28 17:15:38 +00:00
static int
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_name_device(struct ciss_softc *sc, int bus, int target)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct cam_periph *periph;
Name non-physical devices.
2004-07-28 06:18:05 +00:00
if (CISS_IS_PHYSICAL(bus))
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
return (0);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if ((periph = ciss_find_periph(sc, bus, target)) != NULL) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
sprintf(sc->ciss_logical[bus][target].cl_name, "%s%d",
periph->periph_name, periph->unit_number);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
sc->ciss_logical[bus][target].cl_name[0] = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(ENOENT);
}
/************************************************************************
* Periodic status monitoring.
*/
static void
ciss_periodic(void *arg)
{
struct ciss_softc *sc;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
struct ciss_request *cr = NULL;
struct ciss_command *cc = NULL;
int error = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc = (struct ciss_softc *)arg;
/*
* Check the adapter heartbeat.
*/
if (sc->ciss_cfg->heartbeat == sc->ciss_heartbeat) {
sc->ciss_heart_attack++;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
debug(0, "adapter heart attack in progress 0x%x/%d",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc->ciss_heartbeat, sc->ciss_heart_attack);
if (sc->ciss_heart_attack == 3) {
ciss_printf(sc, "ADAPTER HEARTBEAT FAILED\n");
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
ciss_disable_adapter(sc);
return;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
} else {
sc->ciss_heartbeat = sc->ciss_cfg->heartbeat;
sc->ciss_heart_attack = 0;
debug(3, "new heartbeat 0x%x", sc->ciss_heartbeat);
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
/*
* Send the NOP message and wait for a response.
*/
Add `hw.ciss.nop_message_heartbeat' tunable (default disabled) for NOP-message polling in ciss_periodic(). Note that setting the tunable to non-zero can be workaround only for `ADAPTER HEARTBEAT FAILED' problem, and may freeze the system w/o the problem. Reviewed by: scottl Reported by: Attila Nagy MFC after: 3 days
2008-02-10 06:21:52 +00:00
if (ciss_nop_message_heartbeat != 0 && (error = ciss_get_request(sc, &cr)) == 0) {
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
cr->cr_complete = ciss_nop_complete;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
cc->cdb.cdb_length = 1;
cc->cdb.type = CISS_CDB_TYPE_MESSAGE;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_WRITE;
cc->cdb.timeout = 0;
cc->cdb.cdb[0] = CISS_OPCODE_MESSAGE_NOP;
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
if ((error = ciss_start(cr)) != 0) {
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
ciss_printf(sc, "SENDING NOP MESSAGE FAILED\n");
}
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* If the notify event request has died for some reason, or has
* not started yet, restart it.
*/
if (!(sc->ciss_flags & CISS_FLAG_NOTIFY_OK)) {
debug(0, "(re)starting Event Notify chain");
ciss_notify_event(sc);
}
/*
* Reschedule.
*/
Streamline locking in ciss_free()
2007-05-02 04:44:31 +00:00
callout_reset(&sc->ciss_periodic, CISS_HEARTBEAT_RATE * hz, ciss_periodic, sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
static void
ciss_nop_complete(struct ciss_request *cr)
{
struct ciss_softc *sc;
Don't repeat error logging about NOP message sending if ciss_report_request() return an error (which is most likely data underrun). Noticed by: Mark Atkinson MFC after: 1 week
2008-01-28 16:21:34 +00:00
static int first_time = 1;
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
sc = cr->cr_sc;
if (ciss_report_request(cr, NULL, NULL) != 0) {
Don't repeat error logging about NOP message sending if ciss_report_request() return an error (which is most likely data underrun). Noticed by: Mark Atkinson MFC after: 1 week
2008-01-28 16:21:34 +00:00
if (first_time == 1) {
first_time = 0;
ciss_printf(sc, "SENDING NOP MESSAGE FAILED (not logging anymore)\n");
}
Fix NOP message sending in ciss_periodic() which causes panic with option INVARIANTS. Reviewed by: simokawa Tested by: noriyosi_kawano MFC after: 1 week
2007-11-05 13:54:23 +00:00
}
ciss_release_request(cr);
}
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
/************************************************************************
* Disable the adapter.
*
* The all requests in completed queue is failed with hardware error.
* This will cause failover in a multipath configuration.
*/
static void
ciss_disable_adapter(struct ciss_softc *sc)
{
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
cr_qhead_t qh;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_error_info *ce;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
int i;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
CISS_TL_SIMPLE_DISABLE_INTERRUPTS(sc);
pci_disable_busmaster(sc->ciss_dev);
sc->ciss_flags &= ~CISS_FLAG_RUNNING;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
for (i = 1; i < sc->ciss_max_requests; i++) {
cr = &sc->ciss_request[i];
if ((cr->cr_flags & CISS_REQ_BUSY) == 0)
continue;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
ce = (struct ciss_error_info *)&(cc->sg[0]);
ce->command_status = CISS_CMD_STATUS_HARDWARE_ERROR;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
ciss_enqueue_complete(cr, &qh);
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
}
for (;;) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if ((cr = ciss_dequeue_complete(sc, &qh)) == NULL)
Add NOP-message polling to ciss_periodic(). Disable adapter by detecting adapter is dead. Tested by: Masaki YATSU(on RELENG_6) Reviewed by: scottl MFC after: 1 week
2007-10-13 05:45:45 +00:00
break;
/*
* If the request has a callback, invoke it.
*/
if (cr->cr_complete != NULL) {
cr->cr_complete(cr);
continue;
}
/*
* If someone is sleeping on this request, wake them up.
*/
if (cr->cr_flags & CISS_REQ_SLEEP) {
cr->cr_flags &= ~CISS_REQ_SLEEP;
wakeup(cr);
continue;
}
}
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Request a notification response from the adapter.
*
* If (cr) is NULL, this is the first request of the adapter, so
* reset the adapter's message pointer and start with the oldest
* message available.
*/
static void
ciss_notify_event(struct ciss_softc *sc)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_notify_cdb *cnc;
int error;
debug_called(1);
cr = sc->ciss_periodic_notify;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* get a request if we don't already have one */
if (cr == NULL) {
if ((error = ciss_get_request(sc, &cr)) != 0) {
debug(0, "can't get notify event request");
goto out;
}
sc->ciss_periodic_notify = cr;
cr->cr_complete = ciss_notify_complete;
debug(1, "acquired request %d", cr->cr_tag);
}
White space cleanup
2003-12-13 07:54:07 +00:00
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
/*
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* Get a databuffer if we don't already have one, note that the
* adapter command wants a larger buffer than the actual
* structure.
*/
if (cr->cr_data == NULL) {
if ((cr->cr_data = malloc(CISS_NOTIFY_DATA_SIZE, CISS_MALLOC_CLASS, M_NOWAIT)) == NULL) {
debug(0, "can't get notify event request buffer");
error = ENOMEM;
goto out;
}
cr->cr_length = CISS_NOTIFY_DATA_SIZE;
}
/* re-setup the request's command (since we never release it) XXX overkill*/
ciss_preen_command(cr);
/* (re)build the notify event command */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.address.physical.mode = CISS_HDR_ADDRESS_MODE_PERIPHERAL;
cc->header.address.physical.bus = 0;
cc->header.address.physical.target = 0;
cc->cdb.cdb_length = sizeof(*cnc);
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_READ;
cc->cdb.timeout = 0; /* no timeout, we hope */
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cnc = (struct ciss_notify_cdb *)&(cc->cdb.cdb[0]);
bzero(cr->cr_data, CISS_NOTIFY_DATA_SIZE);
cnc->opcode = CISS_OPCODE_READ;
cnc->command = CISS_COMMAND_NOTIFY_ON_EVENT;
cnc->timeout = 0; /* no timeout, we hope */
cnc->synchronous = 0;
cnc->ordered = 0;
cnc->seek_to_oldest = 0;
Report only new new events when initially attaching to the controller.
2004-04-19 00:57:29 +00:00
if ((sc->ciss_flags & CISS_FLAG_RUNNING) == 0)
cnc->new_only = 1;
else
cnc->new_only = 0;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cnc->length = htonl(CISS_NOTIFY_DATA_SIZE);
/* submit the request */
error = ciss_start(cr);
out:
if (error) {
if (cr != NULL) {
if (cr->cr_data != NULL)
free(cr->cr_data, CISS_MALLOC_CLASS);
ciss_release_request(cr);
}
sc->ciss_periodic_notify = NULL;
debug(0, "can't submit notify event request");
sc->ciss_flags &= ~CISS_FLAG_NOTIFY_OK;
} else {
debug(1, "notify event submitted");
sc->ciss_flags |= CISS_FLAG_NOTIFY_OK;
}
}
static void
ciss_notify_complete(struct ciss_request *cr)
{
struct ciss_command *cc;
struct ciss_notify *cn;
struct ciss_softc *sc;
int scsi_status;
int command_status;
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cn = (struct ciss_notify *)cr->cr_data;
sc = cr->cr_sc;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Report request results, decode status.
*/
ciss_report_request(cr, &command_status, &scsi_status);
/*
* Abort the chain on a fatal error.
*
* XXX which of these are actually errors?
*/
if ((command_status != CISS_CMD_STATUS_SUCCESS) &&
(command_status != CISS_CMD_STATUS_TARGET_STATUS) &&
(command_status != CISS_CMD_STATUS_TIMEOUT)) { /* XXX timeout? */
ciss_printf(sc, "fatal error in Notify Event request (%s)\n",
ciss_name_command_status(command_status));
ciss_release_request(cr);
sc->ciss_flags &= ~CISS_FLAG_NOTIFY_OK;
return;
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
/*
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
* If the adapter gave us a text message, print it.
*/
if (cn->message[0] != 0)
ciss_printf(sc, "*** %.80s\n", cn->message);
debug(0, "notify event class %d subclass %d detail %d",
cn->class, cn->subclass, cn->detail);
/*
* If the response indicates that the notifier has been aborted,
* release the notifier command.
*/
if ((cn->class == CISS_NOTIFY_NOTIFIER) &&
(cn->subclass == CISS_NOTIFY_NOTIFIER_STATUS) &&
(cn->detail == 1)) {
debug(0, "notifier exiting");
sc->ciss_flags &= ~CISS_FLAG_NOTIFY_OK;
ciss_release_request(cr);
sc->ciss_periodic_notify = NULL;
wakeup(&sc->ciss_periodic_notify);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
} else {
/* Handle notify events in a kernel thread */
ciss_enqueue_notify(cr);
sc->ciss_periodic_notify = NULL;
wakeup(&sc->ciss_periodic_notify);
wakeup(&sc->ciss_notify);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/*
* Send a new notify event command, if we're not aborting.
*/
if (!(sc->ciss_flags & CISS_FLAG_ABORTING)) {
ciss_notify_event(sc);
}
}
/************************************************************************
* Abort the Notify Event chain.
*
* Note that we can't just abort the command in progress; we have to
* explicitly issue an Abort Notify Event command in order for the
* adapter to clean up correctly.
*
* If we are called with CISS_FLAG_ABORTING set in the adapter softc,
* the chain will not restart itself.
*/
static int
ciss_notify_abort(struct ciss_softc *sc)
{
struct ciss_request *cr;
struct ciss_command *cc;
struct ciss_notify_cdb *cnc;
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
int error, command_status, scsi_status;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(1);
cr = NULL;
error = 0;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* verify that there's an outstanding command */
if (!(sc->ciss_flags & CISS_FLAG_NOTIFY_OK))
goto out;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* get a command to issue the abort with */
if ((error = ciss_get_request(sc, &cr)))
goto out;
/* get a buffer for the result */
if ((cr->cr_data = malloc(CISS_NOTIFY_DATA_SIZE, CISS_MALLOC_CLASS, M_NOWAIT)) == NULL) {
debug(0, "can't get notify event request buffer");
error = ENOMEM;
goto out;
}
cr->cr_length = CISS_NOTIFY_DATA_SIZE;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* build the CDB */
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cc->header.address.physical.mode = CISS_HDR_ADDRESS_MODE_PERIPHERAL;
cc->header.address.physical.bus = 0;
cc->header.address.physical.target = 0;
cc->cdb.cdb_length = sizeof(*cnc);
cc->cdb.type = CISS_CDB_TYPE_COMMAND;
cc->cdb.attribute = CISS_CDB_ATTRIBUTE_SIMPLE;
cc->cdb.direction = CISS_CDB_DIRECTION_READ;
cc->cdb.timeout = 0; /* no timeout, we hope */
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cnc = (struct ciss_notify_cdb *)&(cc->cdb.cdb[0]);
bzero(cnc, sizeof(*cnc));
cnc->opcode = CISS_OPCODE_WRITE;
cnc->command = CISS_COMMAND_ABORT_NOTIFY;
cnc->length = htonl(CISS_NOTIFY_DATA_SIZE);
ciss_print_request(cr);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "Abort Notify Event command failed (%d)\n", error);
goto out;
}
/*
* Check response.
*/
ciss_report_request(cr, &command_status, &scsi_status);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS:
break;
case CISS_CMD_STATUS_INVALID_COMMAND:
/*
* Some older adapters don't support the CISS version of this
* command. Fall back to using the BMIC version.
*/
error = ciss_notify_abort_bmic(sc);
if (error != 0)
goto out;
break;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CISS_CMD_STATUS_TARGET_STATUS:
/*
* This can happen if the adapter thinks there wasn't an outstanding
* Notify Event command but we did. We clean up here.
*/
if (scsi_status == CISS_SCSI_STATUS_CHECK_CONDITION) {
if (sc->ciss_periodic_notify != NULL)
ciss_release_request(sc->ciss_periodic_notify);
error = 0;
goto out;
}
/* FALLTHROUGH */
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
default:
ciss_printf(sc, "Abort Notify Event command failed (%s)\n",
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Sleep waiting for the notifier command to complete. Note
* that if it doesn't, we may end up in a bad situation, since
* the adapter may deliver it later. Also note that the adapter
* requires the Notify Event command to be cancelled in order to
* maintain internal bookkeeping.
*/
while (sc->ciss_periodic_notify != NULL) {
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
error = msleep(&sc->ciss_periodic_notify, &sc->ciss_mtx, PRIBIO, "cissNEA", hz * 5);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
if (error == EWOULDBLOCK) {
ciss_printf(sc, "Notify Event command failed to abort, adapter may wedge.\n");
break;
}
}
out:
/* release the cancel request */
if (cr != NULL) {
if (cr->cr_data != NULL)
free(cr->cr_data, CISS_MALLOC_CLASS);
ciss_release_request(cr);
}
if (error == 0)
sc->ciss_flags &= ~CISS_FLAG_NOTIFY_OK;
return(error);
}
/************************************************************************
* Abort the Notify Event chain using a BMIC command.
*/
static int
ciss_notify_abort_bmic(struct ciss_softc *sc)
{
struct ciss_request *cr;
int error, command_status;
debug_called(1);
cr = NULL;
error = 0;
/* verify that there's an outstanding command */
if (!(sc->ciss_flags & CISS_FLAG_NOTIFY_OK))
goto out;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Build a BMIC command to cancel the Notify on Event command.
*
* Note that we are sending a CISS opcode here. Odd.
*/
if ((error = ciss_get_bmic_request(sc, &cr, CISS_COMMAND_ABORT_NOTIFY,
NULL, 0)) != 0)
goto out;
/*
* Submit the request and wait for it to complete.
*/
if ((error = ciss_synch_request(cr, 60 * 1000)) != 0) {
ciss_printf(sc, "error sending BMIC Cancel Notify on Event command (%d)\n", error);
goto out;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/*
* Check response.
*/
ciss_report_request(cr, &command_status, NULL);
switch(command_status) {
case CISS_CMD_STATUS_SUCCESS:
break;
default:
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
ciss_printf(sc, "error cancelling Notify on Event (%s)\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_name_command_status(command_status));
error = EIO;
goto out;
}
out:
if (cr != NULL)
ciss_release_request(cr);
return(error);
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
/************************************************************************
* Handle rescanning all the logical volumes when a notify event
* causes the drives to come online or offline.
*/
static void
ciss_notify_rescan_logical(struct ciss_softc *sc)
{
struct ciss_lun_report *cll;
struct ciss_ldrive *ld;
int i, j, ndrives;
/*
* We must rescan all logical volumes to get the right logical
* drive address.
*/
cll = ciss_report_luns(sc, CISS_OPCODE_REPORT_LOGICAL_LUNS,
CISS_MAX_LOGICAL);
if (cll == NULL)
return;
ndrives = (ntohl(cll->list_size) / sizeof(union ciss_device_address));
/*
* Delete any of the drives which were destroyed by the
* firmware.
*/
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
for (i = 0; i < sc->ciss_max_logical_bus; i++) {
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
for (j = 0; j < CISS_MAX_LOGICAL; j++) {
ld = &sc->ciss_logical[i][j];
if (ld->cl_update == 0)
continue;
if (ld->cl_status != CISS_LD_ONLINE) {
ciss_cam_rescan_target(sc, i, j);
ld->cl_update = 0;
if (ld->cl_ldrive)
free(ld->cl_ldrive, CISS_MALLOC_CLASS);
if (ld->cl_lstatus)
free(ld->cl_lstatus, CISS_MALLOC_CLASS);
ld->cl_ldrive = NULL;
ld->cl_lstatus = NULL;
}
}
}
/*
* Scan for new drives.
*/
for (i = 0; i < ndrives; i++) {
int bus, target;
bus = CISS_LUN_TO_BUS(cll->lun[i].logical.lun);
target = CISS_LUN_TO_TARGET(cll->lun[i].logical.lun);
ld = &sc->ciss_logical[bus][target];
if (ld->cl_update == 0)
continue;
Reset the update flag when scanning for new devices.
2004-06-22 23:05:06 +00:00
ld->cl_update = 0;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ld->cl_address = cll->lun[i];
ld->cl_controller = &sc->ciss_controllers[bus];
if (ciss_identify_logical(sc, ld) == 0) {
ciss_cam_rescan_target(sc, bus, target);
}
}
free(cll, CISS_MALLOC_CLASS);
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Handle a notify event relating to the status of a logical drive.
*
* XXX need to be able to defer some of these to properly handle
* calling the "ID Physical drive" command, unless the 'extended'
* drive IDs are always in BIG_MAP format.
*/
static void
ciss_notify_logical(struct ciss_softc *sc, struct ciss_notify *cn)
{
struct ciss_ldrive *ld;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
int ostatus, bus, target;
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
debug_called(2);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
bus = cn->device.physical.bus;
target = cn->data.logical_status.logical_drive;
ld = &sc->ciss_logical[bus][target];
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
switch (cn->subclass) {
case CISS_NOTIFY_LOGICAL_STATUS:
switch (cn->detail) {
case 0:
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_name_device(sc, bus, target);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "logical drive %d (%s) changed status %s->%s, spare status 0x%b\n",
cn->data.logical_status.logical_drive, ld->cl_name,
ciss_name_ldrive_status(cn->data.logical_status.previous_state),
ciss_name_ldrive_status(cn->data.logical_status.new_state),
cn->data.logical_status.spare_state,
"\20\1configured\2rebuilding\3failed\4in use\5available\n");
/*
* Update our idea of the drive's status.
*/
ostatus = ciss_decode_ldrive_status(cn->data.logical_status.previous_state);
ld->cl_status = ciss_decode_ldrive_status(cn->data.logical_status.new_state);
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
if (ld->cl_lstatus != NULL)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ld->cl_lstatus->status = cn->data.logical_status.new_state;
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
/*
* Have CAM rescan the drive if its status has changed.
*/
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
if (ostatus != ld->cl_status) {
ld->cl_update = 1;
ciss_notify_rescan_logical(sc);
}
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
case 1: /* logical drive has recognised new media, needs Accept Media Exchange */
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ciss_name_device(sc, bus, target);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "logical drive %d (%s) media exchanged, ready to go online\n",
cn->data.logical_status.logical_drive, ld->cl_name);
There is no need to call ciss_media_accept async anymore since all notify events are handled in a kthread.
2004-06-21 21:56:57 +00:00
ciss_accept_media(sc, ld);
on media exchange, update/rescan the drives. This allows a volume in a failed stated to come back on-line without a reboot.
2004-06-21 22:57:10 +00:00
ld->cl_update = 1;
ld->cl_status = ciss_decode_ldrive_status(cn->data.logical_status.new_state);
ciss_notify_rescan_logical(sc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
case 2:
case 3:
ciss_printf(sc, "rebuild of logical drive %d (%s) failed due to %s error\n",
cn->data.rebuild_aborted.logical_drive,
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ld->cl_name,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
(cn->detail == 2) ? "read" : "write");
break;
}
break;
case CISS_NOTIFY_LOGICAL_ERROR:
if (cn->detail == 0) {
ciss_printf(sc, "FATAL I/O ERROR on logical drive %d (%s), SCSI port %d ID %d\n",
cn->data.io_error.logical_drive,
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ld->cl_name,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
cn->data.io_error.failure_bus,
cn->data.io_error.failure_drive);
/* XXX should we take the drive down at this point, or will we be told? */
}
break;
case CISS_NOTIFY_LOGICAL_SURFACE:
if (cn->detail == 0)
ciss_printf(sc, "logical drive %d (%s) completed consistency initialisation\n",
cn->data.consistency_completed.logical_drive,
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
ld->cl_name);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
}
}
/************************************************************************
* Handle a notify event relating to the status of a physical drive.
*/
static void
ciss_notify_physical(struct ciss_softc *sc, struct ciss_notify *cn)
{
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
/************************************************************************
* Handle a notify event relating to the status of a physical drive.
*/
static void
ciss_notify_hotplug(struct ciss_softc *sc, struct ciss_notify *cn)
{
Free another memory leak when dealing with disk notification.
2006-04-20 03:05:02 +00:00
struct ciss_lun_report *cll = NULL;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
int bus, target;
switch (cn->subclass) {
case CISS_NOTIFY_HOTPLUG_PHYSICAL:
case CISS_NOTIFY_HOTPLUG_NONDISK:
bus = CISS_BIG_MAP_BUS(sc, cn->data.drive.big_physical_drive_number);
target =
CISS_BIG_MAP_TARGET(sc, cn->data.drive.big_physical_drive_number);
if (cn->detail == 0) {
/*
* Mark the device offline so that it'll start producing selection
* timeouts to the upper layer.
*/
- Remove dead code. - Protect against negative values as array indexes. Submitted by: Coverity Prevent analysis tool
2005-02-21 05:17:29 +00:00
if ((bus >= 0) && (target >= 0))
sc->ciss_physical[bus][target].cp_online = 0;
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
} else {
/*
* Rescan the physical lun list for new items
*/
cll = ciss_report_luns(sc, CISS_OPCODE_REPORT_PHYSICAL_LUNS,
CISS_MAX_PHYSICAL);
if (cll == NULL) {
ciss_printf(sc, "Warning, cannot get physical lun list\n");
break;
}
ciss_filter_physical(sc, cll);
}
break;
default:
ciss_printf(sc, "Unknown hotplug event %d\n", cn->subclass);
return;
}
Free another memory leak when dealing with disk notification.
2006-04-20 03:05:02 +00:00
if (cll != NULL)
free(cll, CISS_MALLOC_CLASS);
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
}
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
/************************************************************************
* Handle deferred processing of notify events. Notify events may need
* sleep which is unsafe during an interrupt.
*/
static void
ciss_notify_thread(void *arg)
{
struct ciss_softc *sc;
struct ciss_request *cr;
struct ciss_notify *cn;
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
sc = (struct ciss_softc *)arg;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#if __FreeBSD_version >= 500000
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#endif
for (;;) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
if (STAILQ_EMPTY(&sc->ciss_notify) != 0 &&
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
(sc->ciss_flags & CISS_FLAG_THREAD_SHUT) == 0) {
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
msleep(&sc->ciss_notify, &sc->ciss_mtx, PUSER, "idle", 0);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
}
if (sc->ciss_flags & CISS_FLAG_THREAD_SHUT)
break;
cr = ciss_dequeue_notify(sc);
if (cr == NULL)
panic("cr null");
cn = (struct ciss_notify *)cr->cr_data;
switch (cn->class) {
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
case CISS_NOTIFY_HOTPLUG:
ciss_notify_hotplug(sc, cn);
break;
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
case CISS_NOTIFY_LOGICAL:
ciss_notify_logical(sc, cn);
break;
case CISS_NOTIFY_PHYSICAL:
ciss_notify_physical(sc, cn);
break;
}
ciss_release_request(cr);
}
sc->ciss_notify_thread = NULL;
wakeup(&sc->ciss_notify_thread);
#if __FreeBSD_version >= 500000
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
#endif
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
kproc_exit(0);
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
}
/************************************************************************
* Start the notification kernel thread.
*/
static void
ciss_spawn_notify_thread(struct ciss_softc *sc)
{
move the cleanup of the control device into ciss_free and add some ifdefs for the diffrent kthread_create API between -current and -stable
2004-04-18 02:39:01 +00:00
#if __FreeBSD_version > 500005
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
if (kproc_create((void(*)(void *))ciss_notify_thread, sc,
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
&sc->ciss_notify_thread, 0, 0, "ciss_notify%d",
device_get_unit(sc->ciss_dev)))
move the cleanup of the control device into ciss_free and add some ifdefs for the diffrent kthread_create API between -current and -stable
2004-04-18 02:39:01 +00:00
#else
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
if (kproc_create((void(*)(void *))ciss_notify_thread, sc,
move the cleanup of the control device into ciss_free and add some ifdefs for the diffrent kthread_create API between -current and -stable
2004-04-18 02:39:01 +00:00
&sc->ciss_notify_thread, "ciss_notify%d",
device_get_unit(sc->ciss_dev)))
#endif
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
panic("Could not create notify thread\n");
}
/************************************************************************
* Kill the notification kernel thread.
*/
static void
ciss_kill_notify_thread(struct ciss_softc *sc)
{
if (sc->ciss_notify_thread == NULL)
return;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
Add support for the HP Modular Smart Array 20 & 500 storage arrays. Logical volumes on these devices show up as LUNs behind another controller (also known as proxy controller). In order to issue firmware commands for a volume on a proxy controller, they must be targeted at the address of the proxy controller it is attached to, not the Host/PCI controller. A proxy controller is defined as a device listed in the INQUIRY PHYSICAL LUNS command who's L2 and L3 SCSI addresses are zero. The corresponding address returned defines which "bus" the controller lives on and we use this to create a virtual CAM bus. A logical volume's addresses first byte defines the logical drive number. The second byte defines the bus that it is attached to which corresponds to the BUS of the proxy controller's found or the Host/PCI controller. Change event notification to be handled in its own kernel thread. This is needed since some events may require the driver to sleep on some operations and this cannot be done during interrupt context. With this change, it is now possible to create and destroy logical volumes from FreeBSD, but it requires a native application to construct the proper firmware commands which is not publicly available. Special thanks to John Cagle @ HP for providing remote access to all the hardware and beating on the storage engineers at HP to answer my questions.
2004-04-16 23:00:01 +00:00
sc->ciss_flags |= CISS_FLAG_THREAD_SHUT;
wakeup(&sc->ciss_notify);
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
msleep(&sc->ciss_notify_thread, &sc->ciss_mtx, PUSER, "thtrm", 0);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
/************************************************************************
* Print a request.
*/
static void
ciss_print_request(struct ciss_request *cr)
{
struct ciss_softc *sc;
struct ciss_command *cc;
int i;
sc = cr->cr_sc;
Sync driver with Yahoo: - Implement MSI support (MSIX support was already there) - Use a table to drive MSI/MSIX exceptions - Pre-calculate the command address instead of wasting cycles doing the calculation on every i/o.
2009-09-16 23:27:14 +00:00
cc = cr->cr_cc;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ciss_printf(sc, "REQUEST @ %p\n", cr);
ciss_printf(sc, " data %p/%d tag %d flags %b\n",
cr->cr_data, cr->cr_length, cr->cr_tag, cr->cr_flags,
"\20\1mapped\2sleep\3poll\4dataout\5datain\n");
ciss_printf(sc, " sg list/total %d/%d host tag 0x%x\n",
cc->header.sg_in_list, cc->header.sg_total, cc->header.host_tag);
switch(cc->header.address.mode.mode) {
case CISS_HDR_ADDRESS_MODE_PERIPHERAL:
case CISS_HDR_ADDRESS_MODE_MASK_PERIPHERAL:
ciss_printf(sc, " physical bus %d target %d\n",
cc->header.address.physical.bus, cc->header.address.physical.target);
break;
case CISS_HDR_ADDRESS_MODE_LOGICAL:
ciss_printf(sc, " logical unit %d\n", cc->header.address.logical.lun);
break;
}
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
ciss_printf(sc, " %s cdb length %d type %s attribute %s\n",
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
(cc->cdb.direction == CISS_CDB_DIRECTION_NONE) ? "no-I/O" :
(cc->cdb.direction == CISS_CDB_DIRECTION_READ) ? "READ" :
(cc->cdb.direction == CISS_CDB_DIRECTION_WRITE) ? "WRITE" : "??",
cc->cdb.cdb_length,
(cc->cdb.type == CISS_CDB_TYPE_COMMAND) ? "command" :
(cc->cdb.type == CISS_CDB_TYPE_MESSAGE) ? "message" : "??",
(cc->cdb.attribute == CISS_CDB_ATTRIBUTE_UNTAGGED) ? "untagged" :
(cc->cdb.attribute == CISS_CDB_ATTRIBUTE_SIMPLE) ? "simple" :
(cc->cdb.attribute == CISS_CDB_ATTRIBUTE_HEAD_OF_QUEUE) ? "head-of-queue" :
(cc->cdb.attribute == CISS_CDB_ATTRIBUTE_ORDERED) ? "ordered" :
(cc->cdb.attribute == CISS_CDB_ATTRIBUTE_AUTO_CONTINGENT) ? "auto-contingent" : "??");
ciss_printf(sc, " %*D\n", cc->cdb.cdb_length, &cc->cdb.cdb[0], " ");
if (cc->header.host_tag & CISS_HDR_HOST_TAG_ERROR) {
/* XXX print error info */
} else {
/* since we don't use chained s/g, don't support it here */
for (i = 0; i < cc->header.sg_in_list; i++) {
if ((i % 4) == 0)
ciss_printf(sc, " ");
printf("0x%08x/%d ", (u_int32_t)cc->sg[i].address, cc->sg[i].length);
if ((((i + 1) % 4) == 0) || (i == (cc->header.sg_in_list - 1)))
printf("\n");
}
}
}
/************************************************************************
* Print information about the status of a logical drive.
*/
static void
ciss_print_ldrive(struct ciss_softc *sc, struct ciss_ldrive *ld)
{
int bus, target, i;
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
if (ld->cl_lstatus == NULL) {
printf("does not exist\n");
return;
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/* print drive status */
switch(ld->cl_lstatus->status) {
case CISS_LSTATUS_OK:
printf("online\n");
break;
case CISS_LSTATUS_INTERIM_RECOVERY:
printf("in interim recovery mode\n");
break;
case CISS_LSTATUS_READY_RECOVERY:
printf("ready to begin recovery\n");
break;
case CISS_LSTATUS_RECOVERING:
bus = CISS_BIG_MAP_BUS(sc, ld->cl_lstatus->drive_rebuilding);
target = CISS_BIG_MAP_BUS(sc, ld->cl_lstatus->drive_rebuilding);
printf("being recovered, working on physical drive %d.%d, %u blocks remaining\n",
bus, target, ld->cl_lstatus->blocks_to_recover);
break;
case CISS_LSTATUS_EXPANDING:
printf("being expanded, %u blocks remaining\n",
ld->cl_lstatus->blocks_to_recover);
break;
case CISS_LSTATUS_QUEUED_FOR_EXPANSION:
printf("queued for expansion\n");
break;
case CISS_LSTATUS_FAILED:
printf("queued for expansion\n");
break;
case CISS_LSTATUS_WRONG_PDRIVE:
printf("wrong physical drive inserted\n");
break;
case CISS_LSTATUS_MISSING_PDRIVE:
printf("missing a needed physical drive\n");
break;
case CISS_LSTATUS_BECOMING_READY:
printf("becoming ready\n");
break;
}
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
/* print failed physical drives */
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
for (i = 0; i < CISS_BIG_MAP_ENTRIES / 8; i++) {
bus = CISS_BIG_MAP_BUS(sc, ld->cl_lstatus->drive_failure_map[i]);
target = CISS_BIG_MAP_TARGET(sc, ld->cl_lstatus->drive_failure_map[i]);
if (bus == -1)
continue;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
ciss_printf(sc, "physical drive %d:%d (%x) failed\n", bus, target,
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
ld->cl_lstatus->drive_failure_map[i]);
}
}
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
#ifdef CISS_DEBUG
/************************************************************************
* Print information about the controller/driver.
*/
static void
ciss_print_adapter(struct ciss_softc *sc)
{
make this compile/work with CISS_DEBUG defined.
2004-04-26 19:28:08 +00:00
int i, j;
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
ciss_printf(sc, "ADAPTER:\n");
for (i = 0; i < CISSQ_COUNT; i++) {
ciss_printf(sc, "%s %d/%d\n",
i == 0 ? "free" :
i == 1 ? "busy" : "complete",
sc->ciss_qstat[i].q_length,
sc->ciss_qstat[i].q_max);
}
ciss_printf(sc, "max_requests %d\n", sc->ciss_max_requests);
ciss_printf(sc, "flags %b\n", sc->ciss_flags,
"\20\1notify_ok\2control_open\3aborting\4running\21fake_synch\22bmic_abort\n");
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
for (i = 0; i < sc->ciss_max_logical_bus; i++) {
make this compile/work with CISS_DEBUG defined.
2004-04-26 19:28:08 +00:00
for (j = 0; j < CISS_MAX_LOGICAL; j++) {
ciss_printf(sc, "LOGICAL DRIVE %d: ", i);
ciss_print_ldrive(sc, &sc->ciss_logical[i][j]);
}
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
}
Add SCSI passthrough support to CISS. This allows devices like tape drives that are on a CISS bus to be exported up to CAM and made available as normal devices. This will typically add one or two buses to CAM, which will be numbered starting at 32 to allow room for CISS proxy buses. Also, the CISS firmware usually hides disk devices, but these can also be exposed as 'pass' devices if you set the hw.ciss.expose_hidden_physical tunable. Sponsored by: Tape Laboratories, Inc. MFC After: 3 days
2004-06-21 20:18:40 +00:00
/* XXX Should physical drives be printed out here? */
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
for (i = 1; i < sc->ciss_max_requests; i++)
ciss_print_request(sc->ciss_request + i);
}
/* DDB hook */
Fix a warning when CISS_DEBUG is defined.
2002-10-27 12:05:11 +00:00
static void
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
ciss_print0(void)
{
struct ciss_softc *sc;
Whitespace cleanup.
2004-04-16 21:03:38 +00:00
- Set the maximum number of device transactions to what the driver is limiting it to, not what the device says it can handle. - cl_status is an integer. cl_lstatus is a pointer. - Add some debugging code to dump some things the driver knows about the adapter. - Tell CAM that the adapter can handle more commands when a command completes. This fixes the problem were the SIM would freeze once the driver hit the maximum number of transactions for the device. - Change the vendor string to COMPAQ. - Turn of Synchronize Cache for now. It locks the controller up. Approved by: msmith Obtained from: Yahoo!
2002-04-29 21:46:09 +00:00
sc = devclass_get_softc(devclass_find("ciss"), 0);
if (sc == NULL) {
printf("no ciss controllers\n");
} else {
ciss_print_adapter(sc);
}
}
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
/************************************************************************
* Return a name for a logical drive status value.
*/
static const char *
ciss_name_ldrive_status(int status)
{
switch (status) {
case CISS_LSTATUS_OK:
return("OK");
case CISS_LSTATUS_FAILED:
return("failed");
case CISS_LSTATUS_NOT_CONFIGURED:
return("not configured");
case CISS_LSTATUS_INTERIM_RECOVERY:
return("interim recovery");
case CISS_LSTATUS_READY_RECOVERY:
return("ready for recovery");
case CISS_LSTATUS_RECOVERING:
return("recovering");
case CISS_LSTATUS_WRONG_PDRIVE:
return("wrong physical drive inserted");
case CISS_LSTATUS_MISSING_PDRIVE:
return("missing physical drive");
case CISS_LSTATUS_EXPANDING:
return("expanding");
case CISS_LSTATUS_BECOMING_READY:
return("becoming ready");
case CISS_LSTATUS_QUEUED_FOR_EXPANSION:
return("queued for expansion");
}
return("unknown status");
}
/************************************************************************
* Return an online/offline/nonexistent value for a logical drive
* status value.
*/
static int
ciss_decode_ldrive_status(int status)
{
switch(status) {
case CISS_LSTATUS_NOT_CONFIGURED:
return(CISS_LD_NONEXISTENT);
case CISS_LSTATUS_OK:
case CISS_LSTATUS_INTERIM_RECOVERY:
case CISS_LSTATUS_READY_RECOVERY:
case CISS_LSTATUS_RECOVERING:
case CISS_LSTATUS_EXPANDING:
case CISS_LSTATUS_QUEUED_FOR_EXPANSION:
return(CISS_LD_ONLINE);
case CISS_LSTATUS_FAILED:
case CISS_LSTATUS_WRONG_PDRIVE:
case CISS_LSTATUS_MISSING_PDRIVE:
case CISS_LSTATUS_BECOMING_READY:
default:
return(CISS_LD_OFFLINE);
}
}
/************************************************************************
* Return a name for a logical drive's organisation.
*/
static const char *
ciss_name_ldrive_org(int org)
{
switch(org) {
case CISS_LDRIVE_RAID0:
return("RAID 0");
case CISS_LDRIVE_RAID1:
Cosmetic: make it less confusing when displaying RAID 1 level, that might be 1+0 as well. PR: kern/150936 MFC after: 2 weeks
2010-10-02 08:18:19 +00:00
return("RAID 1(1+0)");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CISS_LDRIVE_RAID4:
return("RAID 4");
case CISS_LDRIVE_RAID5:
return("RAID 5");
Define RAID 5+1 and RAID ADG fault tolerances
2003-12-08 16:07:35 +00:00
case CISS_LDRIVE_RAID51:
return("RAID 5+1");
case CISS_LDRIVE_RAIDADG:
return("RAID ADG");
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
}
return("unkown");
}
/************************************************************************
* Return a name for a command status value.
*/
static const char *
ciss_name_command_status(int status)
{
switch(status) {
case CISS_CMD_STATUS_SUCCESS:
return("success");
case CISS_CMD_STATUS_TARGET_STATUS:
return("target status");
case CISS_CMD_STATUS_DATA_UNDERRUN:
return("data underrun");
case CISS_CMD_STATUS_DATA_OVERRUN:
return("data overrun");
case CISS_CMD_STATUS_INVALID_COMMAND:
return("invalid command");
case CISS_CMD_STATUS_PROTOCOL_ERROR:
return("protocol error");
case CISS_CMD_STATUS_HARDWARE_ERROR:
return("hardware error");
case CISS_CMD_STATUS_CONNECTION_LOST:
return("connection lost");
case CISS_CMD_STATUS_ABORTED:
return("aborted");
case CISS_CMD_STATUS_ABORT_FAILED:
return("abort failed");
case CISS_CMD_STATUS_UNSOLICITED_ABORT:
return("unsolicited abort");
case CISS_CMD_STATUS_TIMEOUT:
return("timeout");
case CISS_CMD_STATUS_UNABORTABLE:
return("unabortable");
}
return("unknown status");
}
/************************************************************************
* Handle an open on the control device.
*/
static int
We no longer need to use d_thread_t, migrate to struct thread *.
2009-05-20 17:29:21 +00:00
ciss_open(struct cdev *dev, int flags, int fmt, struct thread *p)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_softc *sc;
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc = (struct ciss_softc *)dev->si_drv1;
/* we might want to veto if someone already has us open */
White space cleanup
2003-12-13 07:54:07 +00:00
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc->ciss_flags |= CISS_FLAG_CONTROL_OPEN;
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(0);
}
/************************************************************************
* Handle the last close on the control device.
*/
static int
We no longer need to use d_thread_t, migrate to struct thread *.
2009-05-20 17:29:21 +00:00
ciss_close(struct cdev *dev, int flags, int fmt, struct thread *p)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_softc *sc;
debug_called(1);
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc = (struct ciss_softc *)dev->si_drv1;
White space cleanup
2003-12-13 07:54:07 +00:00
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
sc->ciss_flags &= ~CISS_FLAG_CONTROL_OPEN;
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return (0);
}
/********************************************************************************
* Handle adapter-specific control operations.
*
* Note that the API here is compatible with the Linux driver, in order to
* simplify the porting of Compaq's userland tools.
*/
static int
We no longer need to use d_thread_t, migrate to struct thread *.
2009-05-20 17:29:21 +00:00
ciss_ioctl(struct cdev *dev, u_long cmd, caddr_t addr, int32_t flag, struct thread *p)
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
{
struct ciss_softc *sc;
Support passthru ioctl commands from 32bit binaries.
2005-05-18 05:31:34 +00:00
IOCTL_Command_struct *ioc = (IOCTL_Command_struct *)addr;
#ifdef __amd64__
IOCTL_Command_struct32 *ioc32 = (IOCTL_Command_struct32 *)addr;
IOCTL_Command_struct ioc_swab;
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
int error;
debug_called(1);
sc = (struct ciss_softc *)dev->si_drv1;
error = 0;
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_lock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
switch(cmd) {
A number of significant enhancements to the ciss driver: 1. The FreeBSD driver was setting an interrupt coalesce delay of 1000us for reasons that I can only speculate on. This was hurting everything from lame sequential I/O "benchmarks" to legitimate filesystem metadata operations that relied on serialized barrier writes. One of my filesystem tests went from 35s to complete down to 6s. 2. Implemented the Performant transport method. Without the fix in (1), I saw almost no difference. With it, my filesystem tests showed another 5-10% improvement in speed. It was hard to measure CPU utilization in any meaningful way, so it's not clear if there was a benefit there, though there should have been since the interrupt handler was reduced from 2 or more PCI reads down to 1. 3. Implemented MSI-X. Without any docs on this, I was just taking a guess, and it appears to only work with the Performant method. This could be a programming or understanding mistake on my part. While this by itself made almost no difference to performance since the Performant method already eliminated most of the synchronous reads over the PCI bus, it did allow the CISS hardware to stop sharing its interrupt with the USB hardware, which in turn allowed the driver to become decoupled from the Giant-locked USB driver stack. This increased performance by almost 20%. The MSI-X setup was done with 4 vectors allocated, but only 1 vector used since the performant method was told to only use 1 of 4 queues. Fiddling with this might make it work with the simpleq method, not sure. I did not implement MSI since I have no MSI-specific hardware in my test lab. 4. Improved the locking in the driver, trimmed some data structures. This didn't improve test times in any measurable way, but it does look like it gave a minor improvement to CPU usage when many processes/threads were doing I/O in parallel. Again, this was hard to accurately test.
2008-07-11 21:20:51 +00:00
case CCISS_GETQSTATS:
{
union ciss_statrequest *cr = (union ciss_statrequest *)addr;
switch (cr->cs_item) {
case CISSQ_FREE:
case CISSQ_NOTIFY:
bcopy(&sc->ciss_qstat[cr->cs_item], &cr->cs_qstat,
sizeof(struct ciss_qstat));
break;
default:
error = ENOIOCTL;
break;
}
break;
}
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CCISS_GETPCIINFO:
{
cciss_pci_info_struct *pis = (cciss_pci_info_struct *)addr;
pis->bus = pci_get_bus(sc->ciss_dev);
pis->dev_fn = pci_get_slot(sc->ciss_dev);
Update PCI IDs for ciss controllers that are supported by this driver. Submitted by: scott.benesh@hp.com Obtained from: Yahoo! Inc and HP MFC after: 1 week
2012-01-26 17:04:17 +00:00
pis->board_id = (pci_get_subvendor(sc->ciss_dev) << 16) |
pci_get_subdevice(sc->ciss_dev);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
}
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CCISS_GETINTINFO:
{
cciss_coalint_struct *cis = (cciss_coalint_struct *)addr;
cis->delay = sc->ciss_cfg->interrupt_coalesce_delay;
cis->count = sc->ciss_cfg->interrupt_coalesce_count;
break;
}
case CCISS_SETINTINFO:
{
cciss_coalint_struct *cis = (cciss_coalint_struct *)addr;
if ((cis->delay == 0) && (cis->count == 0)) {
error = EINVAL;
break;
}
/*
* XXX apparently this is only safe if the controller is idle,
* we should suspend it before doing this.
*/
sc->ciss_cfg->interrupt_coalesce_delay = cis->delay;
sc->ciss_cfg->interrupt_coalesce_count = cis->count;
if (ciss_update_config(sc))
error = EIO;
/* XXX resume the controller here */
break;
}
case CCISS_GETNODENAME:
bcopy(sc->ciss_cfg->server_name, (NodeName_type *)addr,
sizeof(NodeName_type));
break;
case CCISS_SETNODENAME:
bcopy((NodeName_type *)addr, sc->ciss_cfg->server_name,
sizeof(NodeName_type));
if (ciss_update_config(sc))
error = EIO;
break;
White space cleanup
2003-12-13 07:54:07 +00:00
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CCISS_GETHEARTBEAT:
*(Heartbeat_type *)addr = sc->ciss_cfg->heartbeat;
break;
case CCISS_GETBUSTYPES:
*(BusTypes_type *)addr = sc->ciss_cfg->bus_types;
break;
case CCISS_GETFIRMVER:
bcopy(sc->ciss_id->running_firmware_revision, (FirmwareVer_type *)addr,
sizeof(FirmwareVer_type));
break;
case CCISS_GETDRIVERVER:
*(DriverVer_type *)addr = CISS_DRIVER_VERSION;
break;
case CCISS_REVALIDVOLS:
/*
* This is a bit ugly; to do it "right" we really need
* to find any disks that have changed, kick CAM off them,
* then rescan only these disks. It'd be nice if they
* a) told us which disk(s) they were going to play with,
* and b) which ones had arrived. 8(
*/
break;
Support passthru ioctl commands from 32bit binaries.
2005-05-18 05:31:34 +00:00
#ifdef __amd64__
case CCISS_PASSTHRU32:
ioc_swab.LUN_info = ioc32->LUN_info;
ioc_swab.Request = ioc32->Request;
ioc_swab.error_info = ioc32->error_info;
ioc_swab.buf_size = ioc32->buf_size;
ioc_swab.buf = (u_int8_t *)(uintptr_t)ioc32->buf;
ioc = &ioc_swab;
/* FALLTHROUGH */
#endif
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
case CCISS_PASSTHRU:
Support passthru ioctl commands from 32bit binaries.
2005-05-18 05:31:34 +00:00
error = ciss_user_command(sc, ioc);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
break;
default:
debug(0, "unknown ioctl 0x%lx", cmd);
debug(1, "CCISS_GETPCIINFO: 0x%lx", CCISS_GETPCIINFO);
debug(1, "CCISS_GETINTINFO: 0x%lx", CCISS_GETINTINFO);
debug(1, "CCISS_SETINTINFO: 0x%lx", CCISS_SETINTINFO);
debug(1, "CCISS_GETNODENAME: 0x%lx", CCISS_GETNODENAME);
debug(1, "CCISS_SETNODENAME: 0x%lx", CCISS_SETNODENAME);
debug(1, "CCISS_GETHEARTBEAT: 0x%lx", CCISS_GETHEARTBEAT);
debug(1, "CCISS_GETBUSTYPES: 0x%lx", CCISS_GETBUSTYPES);
debug(1, "CCISS_GETFIRMVER: 0x%lx", CCISS_GETFIRMVER);
debug(1, "CCISS_GETDRIVERVER: 0x%lx", CCISS_GETDRIVERVER);
debug(1, "CCISS_REVALIDVOLS: 0x%lx", CCISS_REVALIDVOLS);
debug(1, "CCISS_PASSTHRU: 0x%lx", CCISS_PASSTHRU);
error = ENOIOCTL;
break;
}
MPSAFE ciss driver
2007-05-01 05:13:15 +00:00
mtx_unlock(&sc->ciss_mtx);
Add the 'ciss' driver, which supports the Compaq SmartRAID 5* family of RAID controllers (5300, 532, 5i, etc.) Thanks to Compaq and Yahoo! for support during the development of this driver. MFC after: 1 week
2001-11-27 23:08:37 +00:00
return(error);
}
Reference in New Issue Copy Permalink