2010-09-10 15:03:56 +00:00
|
|
|
/*-
|
|
|
|
* Copyright (c) 2009 Yahoo! Inc.
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
* Copyright (c) 2011, 2012 LSI Corp.
|
2012-01-26 18:17:21 +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.
|
|
|
|
*
|
|
|
|
* LSI MPT-Fusion Host Adapter FreeBSD
|
|
|
|
*
|
|
|
|
* $FreeBSD$
|
|
|
|
*/
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
|
|
|
/* Communications core for LSI MPT2 */
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* TODO Move headers to mpsvar */
|
2010-09-10 15:03:56 +00:00
|
|
|
#include <sys/types.h>
|
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/systm.h>
|
|
|
|
#include <sys/kernel.h>
|
|
|
|
#include <sys/selinfo.h>
|
|
|
|
#include <sys/module.h>
|
|
|
|
#include <sys/bus.h>
|
|
|
|
#include <sys/conf.h>
|
|
|
|
#include <sys/bio.h>
|
|
|
|
#include <sys/malloc.h>
|
|
|
|
#include <sys/uio.h>
|
|
|
|
#include <sys/sysctl.h>
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
#include <sys/endian.h>
|
2012-01-26 18:17:21 +00:00
|
|
|
#include <sys/queue.h>
|
|
|
|
#include <sys/kthread.h>
|
|
|
|
#include <sys/taskqueue.h>
|
|
|
|
#include <sys/sbuf.h>
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
#include <machine/bus.h>
|
|
|
|
#include <machine/resource.h>
|
|
|
|
#include <sys/rman.h>
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
#include <machine/stdarg.h>
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
#include <cam/cam.h>
|
|
|
|
#include <cam/cam_ccb.h>
|
2012-01-26 18:17:21 +00:00
|
|
|
#include <cam/cam_xpt.h>
|
2010-09-10 15:03:56 +00:00
|
|
|
#include <cam/cam_debug.h>
|
|
|
|
#include <cam/cam_sim.h>
|
|
|
|
#include <cam/cam_xpt_sim.h>
|
|
|
|
#include <cam/cam_xpt_periph.h>
|
|
|
|
#include <cam/cam_periph.h>
|
|
|
|
#include <cam/scsi/scsi_all.h>
|
|
|
|
#include <cam/scsi/scsi_message.h>
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
#if __FreeBSD_version >= 900026
|
|
|
|
#include <cam/scsi/smp_all.h>
|
|
|
|
#endif
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
#include <dev/mps/mpi/mpi2_type.h>
|
|
|
|
#include <dev/mps/mpi/mpi2.h>
|
|
|
|
#include <dev/mps/mpi/mpi2_ioc.h>
|
|
|
|
#include <dev/mps/mpi/mpi2_sas.h>
|
|
|
|
#include <dev/mps/mpi/mpi2_cnfg.h>
|
|
|
|
#include <dev/mps/mpi/mpi2_init.h>
|
2012-01-26 18:17:21 +00:00
|
|
|
#include <dev/mps/mpi/mpi2_tool.h>
|
|
|
|
#include <dev/mps/mps_ioctl.h>
|
2010-09-10 15:03:56 +00:00
|
|
|
#include <dev/mps/mpsvar.h>
|
|
|
|
#include <dev/mps/mps_table.h>
|
2012-01-26 18:17:21 +00:00
|
|
|
#include <dev/mps/mps_sas.h>
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
#define MPSSAS_DISCOVERY_TIMEOUT 20
|
|
|
|
#define MPSSAS_MAX_DISCOVERY_TIMEOUTS 10 /* 200 seconds */
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* static array to check SCSI OpCode for EEDP protection bits
|
|
|
|
*/
|
|
|
|
#define PRO_R MPI2_SCSIIO_EEDPFLAGS_CHECK_REMOVE_OP
|
|
|
|
#define PRO_W MPI2_SCSIIO_EEDPFLAGS_INSERT_OP
|
|
|
|
#define PRO_V MPI2_SCSIIO_EEDPFLAGS_INSERT_OP
|
|
|
|
static uint8_t op_code_prot[256] = {
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, PRO_W, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
|
|
|
|
0, 0, 0, PRO_W, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, PRO_R, 0, PRO_W, 0, 0, 0, PRO_W, PRO_V,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
2010-09-10 15:03:56 +00:00
|
|
|
};
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
MALLOC_DEFINE(M_MPSSAS, "MPSSAS", "MPS SAS memory");
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
static void mpssas_discovery_timeout(void *data);
|
|
|
|
static void mpssas_remove_device(struct mps_softc *, struct mps_command *);
|
|
|
|
static void mpssas_remove_complete(struct mps_softc *, struct mps_command *);
|
|
|
|
static void mpssas_action(struct cam_sim *sim, union ccb *ccb);
|
|
|
|
static void mpssas_poll(struct cam_sim *sim);
|
|
|
|
static void mpssas_scsiio_timeout(void *data);
|
|
|
|
static void mpssas_abort_complete(struct mps_softc *sc, struct mps_command *cm);
|
2012-01-26 18:17:21 +00:00
|
|
|
static void mpssas_direct_drive_io(struct mpssas_softc *sassc,
|
|
|
|
struct mps_command *cm, union ccb *ccb);
|
2010-09-10 15:03:56 +00:00
|
|
|
static void mpssas_action_scsiio(struct mpssas_softc *, union ccb *);
|
|
|
|
static void mpssas_scsiio_complete(struct mps_softc *, struct mps_command *);
|
2012-01-26 18:17:21 +00:00
|
|
|
static void mpssas_action_resetdev(struct mpssas_softc *, union ccb *);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
#if __FreeBSD_version >= 900026
|
|
|
|
static void mpssas_smpio_complete(struct mps_softc *sc, struct mps_command *cm);
|
|
|
|
static void mpssas_send_smpcmd(struct mpssas_softc *sassc, union ccb *ccb,
|
|
|
|
uint64_t sasaddr);
|
|
|
|
static void mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb);
|
2012-01-26 18:17:21 +00:00
|
|
|
#endif //FreeBSD_version >= 900026
|
2010-09-10 15:03:56 +00:00
|
|
|
static void mpssas_resetdev_complete(struct mps_softc *, struct mps_command *);
|
2012-01-26 18:17:21 +00:00
|
|
|
static int mpssas_send_abort(struct mps_softc *sc, struct mps_command *tm, struct mps_command *cm);
|
|
|
|
static int mpssas_send_reset(struct mps_softc *sc, struct mps_command *tm, uint8_t type);
|
|
|
|
static void mpssas_async(void *callback_arg, uint32_t code,
|
|
|
|
struct cam_path *path, void *arg);
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#if (__FreeBSD_version < 901503) || \
|
|
|
|
((__FreeBSD_version >= 1000000) && (__FreeBSD_version < 1000006))
|
|
|
|
static void mpssas_check_eedp(struct mps_softc *sc, struct cam_path *path,
|
|
|
|
struct ccb_getdev *cgd);
|
2012-01-26 18:17:21 +00:00
|
|
|
static void mpssas_read_cap_done(struct cam_periph *periph, union ccb *done_ccb);
|
|
|
|
#endif
|
|
|
|
static int mpssas_send_portenable(struct mps_softc *sc);
|
|
|
|
static void mpssas_portenable_complete(struct mps_softc *sc,
|
|
|
|
struct mps_command *cm);
|
2010-12-11 00:36:35 +00:00
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
struct mpssas_target *
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_find_target_by_handle(struct mpssas_softc *sassc, int start, uint16_t handle)
|
2010-12-11 00:36:35 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mpssas_target *target;
|
|
|
|
int i;
|
|
|
|
|
2014-04-07 15:13:24 +00:00
|
|
|
for (i = start; i < sassc->maxtargets; i++) {
|
2012-01-26 18:17:21 +00:00
|
|
|
target = &sassc->targets[i];
|
|
|
|
if (target->handle == handle)
|
|
|
|
return (target);
|
2010-12-11 00:36:35 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
return (NULL);
|
2010-12-11 00:36:35 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* we need to freeze the simq during attach and diag reset, to avoid failing
|
|
|
|
* commands before device handles have been found by discovery. Since
|
|
|
|
* discovery involves reading config pages and possibly sending commands,
|
|
|
|
* discovery actions may continue even after we receive the end of discovery
|
|
|
|
* event, so refcount discovery actions instead of assuming we can unfreeze
|
|
|
|
* the simq when we get the event.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
mpssas_startup_increment(struct mpssas_softc *sassc)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sassc->flags & MPSSAS_IN_STARTUP) != 0) {
|
|
|
|
if (sassc->startup_refcount++ == 0) {
|
|
|
|
/* just starting, freeze the simq */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INIT,
|
2012-01-26 18:17:21 +00:00
|
|
|
"%s freezing simq\n", __func__);
|
2014-02-27 21:48:34 +00:00
|
|
|
#if __FreeBSD_version >= 1000039
|
|
|
|
xpt_hold_boot();
|
|
|
|
#endif
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_freeze_simq(sassc->sim, 1);
|
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INIT, "%s refcount %u\n", __func__,
|
2012-01-26 18:17:21 +00:00
|
|
|
sassc->startup_refcount);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
|
|
|
mpssas_startup_decrement(struct mpssas_softc *sassc)
|
|
|
|
{
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sassc->flags & MPSSAS_IN_STARTUP) != 0) {
|
|
|
|
if (--sassc->startup_refcount == 0) {
|
|
|
|
/* finished all discovery-related actions, release
|
|
|
|
* the simq and rescan for the latest topology.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INIT,
|
2012-01-26 18:17:21 +00:00
|
|
|
"%s releasing simq\n", __func__);
|
|
|
|
sassc->flags &= ~MPSSAS_IN_STARTUP;
|
2014-02-27 21:48:34 +00:00
|
|
|
xpt_release_simq(sassc->sim, 1);
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#if __FreeBSD_version >= 1000039
|
|
|
|
xpt_release_boot();
|
|
|
|
#else
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_rescan_target(sassc->sc, NULL);
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#endif
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INIT, "%s refcount %u\n", __func__,
|
2012-01-26 18:17:21 +00:00
|
|
|
sassc->startup_refcount);
|
|
|
|
}
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* LSI's firmware requires us to stop sending commands when we're doing task
|
|
|
|
* management, so refcount the TMs and keep the simq frozen when any are in
|
|
|
|
* use.
|
|
|
|
*/
|
|
|
|
struct mps_command *
|
|
|
|
mpssas_alloc_tm(struct mps_softc *sc)
|
|
|
|
{
|
|
|
|
struct mps_command *tm;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
tm = mps_alloc_high_priority_command(sc);
|
|
|
|
if (tm != NULL) {
|
|
|
|
if (sc->sassc->tm_count++ == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY,
|
|
|
|
"%s freezing simq\n", __func__);
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_freeze_simq(sc->sassc->sim, 1);
|
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY, "%s tm_count %u\n", __func__,
|
2012-01-26 18:17:21 +00:00
|
|
|
sc->sassc->tm_count);
|
|
|
|
}
|
|
|
|
return tm;
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
|
|
|
mpssas_free_tm(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_TRACE, "%s", __func__);
|
2012-01-26 18:17:21 +00:00
|
|
|
if (tm == NULL)
|
|
|
|
return;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* if there are no TMs in use, we can release the simq. We use our
|
|
|
|
* own refcount so that it's easier for a diag reset to cleanup and
|
|
|
|
* release the simq.
|
|
|
|
*/
|
|
|
|
if (--sc->sassc->tm_count == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY, "%s releasing simq\n", __func__);
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_release_simq(sc->sassc->sim, 1);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY, "%s tm_count %u\n", __func__,
|
2012-01-26 18:17:21 +00:00
|
|
|
sc->sassc->tm_count);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_free_high_priority_command(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
|
|
|
mpssas_rescan_target(struct mps_softc *sc, struct mpssas_target *targ)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mpssas_softc *sassc = sc->sassc;
|
|
|
|
path_id_t pathid;
|
|
|
|
target_id_t targetid;
|
|
|
|
union ccb *ccb;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
pathid = cam_sim_path(sassc->sim);
|
|
|
|
if (targ == NULL)
|
|
|
|
targetid = CAM_TARGET_WILDCARD;
|
|
|
|
else
|
|
|
|
targetid = targ - sassc->targets;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Allocate a CCB and schedule a rescan.
|
|
|
|
*/
|
|
|
|
ccb = xpt_alloc_ccb_nowait();
|
|
|
|
if (ccb == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "unable to alloc CCB for rescan\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2013-04-14 09:55:48 +00:00
|
|
|
if (xpt_create_path(&ccb->ccb_h.path, NULL, pathid,
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
targetid, CAM_LUN_WILDCARD) != CAM_REQ_CMP) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "unable to create path for rescan\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_free_ccb(ccb);
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2012-06-29 17:00:52 +00:00
|
|
|
if (targetid == CAM_TARGET_WILDCARD)
|
|
|
|
ccb->ccb_h.func_code = XPT_SCAN_BUS;
|
|
|
|
else
|
|
|
|
ccb->ccb_h.func_code = XPT_SCAN_TGT;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_dprint(sc, MPS_TRACE, "%s targetid %u\n", __func__, targetid);
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
xpt_rescan(ccb);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(struct mps_command *cm, u_int level, const char *fmt, ...)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
struct sbuf sb;
|
|
|
|
va_list ap;
|
|
|
|
char str[192];
|
|
|
|
char path_str[64];
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (cm == NULL)
|
|
|
|
return;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-08-12 19:16:55 +00:00
|
|
|
/* No need to be in here if debugging isn't enabled */
|
2013-08-12 20:17:37 +00:00
|
|
|
if ((cm->cm_sc->mps_debug & level) == 0)
|
2013-08-12 19:16:55 +00:00
|
|
|
return;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
sbuf_new(&sb, str, sizeof(str), 0);
|
|
|
|
|
|
|
|
va_start(ap, fmt);
|
|
|
|
|
|
|
|
if (cm->cm_ccb != NULL) {
|
|
|
|
xpt_path_string(cm->cm_ccb->csio.ccb_h.path, path_str,
|
|
|
|
sizeof(path_str));
|
|
|
|
sbuf_cat(&sb, path_str);
|
|
|
|
if (cm->cm_ccb->ccb_h.func_code == XPT_SCSI_IO) {
|
|
|
|
scsi_command_string(&cm->cm_ccb->csio, &sb);
|
|
|
|
sbuf_printf(&sb, "length %d ",
|
|
|
|
cm->cm_ccb->csio.dxfer_len);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
else {
|
|
|
|
sbuf_printf(&sb, "(noperiph:%s%d:%u:%u:%u): ",
|
|
|
|
cam_sim_name(cm->cm_sc->sassc->sim),
|
|
|
|
cam_sim_unit(cm->cm_sc->sassc->sim),
|
|
|
|
cam_sim_bus(cm->cm_sc->sassc->sim),
|
|
|
|
cm->cm_targ ? cm->cm_targ->tid : 0xFFFFFFFF,
|
|
|
|
cm->cm_lun);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
sbuf_printf(&sb, "SMID %u ", cm->cm_desc.Default.SMID);
|
|
|
|
sbuf_vprintf(&sb, fmt, ap);
|
|
|
|
sbuf_finish(&sb);
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint_field(cm->cm_sc, level, "%s", sbuf_data(&sb));
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
va_end(ap);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
2012-02-09 00:16:12 +00:00
|
|
|
mpssas_remove_volume(struct mps_softc *sc, struct mps_command *tm)
|
2012-01-26 18:17:21 +00:00
|
|
|
{
|
2012-02-09 00:16:12 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
uint16_t handle;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-02-09 00:16:12 +00:00
|
|
|
|
|
|
|
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
|
|
|
|
targ = tm->cm_targ;
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
if (reply == NULL) {
|
|
|
|
/* XXX retry the remove after the diag reset completes? */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"%s NULL reply reseting device 0x%04x\n", __func__, handle);
|
2012-02-09 00:16:12 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
return;
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
if (reply->IOCStatus != MPI2_IOCSTATUS_SUCCESS) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"IOCStatus = 0x%x while resetting device 0x%x\n",
|
2012-02-09 00:16:12 +00:00
|
|
|
reply->IOCStatus, handle);
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
|
|
|
"Reset aborted %u commands\n", reply->TerminationCount);
|
2012-02-09 00:16:12 +00:00
|
|
|
mps_free_reply(sc, tm->cm_reply_data);
|
2012-09-14 21:28:56 +00:00
|
|
|
tm->cm_reply = NULL; /* Ensures the reply won't get re-freed */
|
2012-02-09 00:16:12 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
|
|
|
"clearing target %u handle 0x%04x\n", targ->tid, handle);
|
2012-02-09 00:16:12 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't clear target if remove fails because things will get confusing.
|
|
|
|
* Leave the devname and sasaddr intact so that we know to avoid reusing
|
|
|
|
* this target id if possible, and so we can assign the same target id
|
|
|
|
* to this device if it comes back in the future.
|
|
|
|
*/
|
|
|
|
if (reply->IOCStatus == MPI2_IOCSTATUS_SUCCESS) {
|
|
|
|
targ = tm->cm_targ;
|
|
|
|
targ->handle = 0x0;
|
|
|
|
targ->encl_handle = 0x0;
|
|
|
|
targ->encl_slot = 0x0;
|
|
|
|
targ->exp_dev_handle = 0x0;
|
|
|
|
targ->phy_num = 0x0;
|
|
|
|
targ->linkrate = 0x0;
|
|
|
|
targ->devinfo = 0x0;
|
|
|
|
targ->flags = 0x0;
|
|
|
|
}
|
|
|
|
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* No Need to call "MPI2_SAS_OP_REMOVE_DEVICE" For Volume removal.
|
|
|
|
* Otherwise Volume Delete is same as Bare Drive Removal.
|
|
|
|
*/
|
|
|
|
void
|
|
|
|
mpssas_prepare_volume_remove(struct mpssas_softc *sassc, uint16_t handle)
|
|
|
|
{
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct mps_command *cm;
|
|
|
|
struct mpssas_target *targ = NULL;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
2012-02-09 00:16:12 +00:00
|
|
|
sc = sassc->sc;
|
|
|
|
|
|
|
|
#ifdef WD_SUPPORT
|
|
|
|
/*
|
|
|
|
* If this is a WD controller, determine if the disk should be exposed
|
|
|
|
* to the OS or not. If disk should be exposed, return from this
|
|
|
|
* function without doing anything.
|
|
|
|
*/
|
|
|
|
if (sc->WD_available && (sc->WD_hide_expose ==
|
|
|
|
MPS_WD_EXPOSE_ALWAYS)) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
#endif //WD_SUPPORT
|
|
|
|
|
|
|
|
targ = mpssas_find_target_by_handle(sassc, 0, handle);
|
|
|
|
if (targ == NULL) {
|
|
|
|
/* FIXME: what is the action? */
|
|
|
|
/* We don't know about this device? */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s %d : invalid handle 0x%x \n", __func__,__LINE__, handle);
|
2012-02-09 00:16:12 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
targ->flags |= MPSSAS_TARGET_INREMOVAL;
|
|
|
|
|
|
|
|
cm = mpssas_alloc_tm(sc);
|
|
|
|
if (cm == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: command alloc failure\n", __func__);
|
2012-02-09 00:16:12 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
mpssas_rescan_target(sc, targ);
|
|
|
|
|
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)cm->cm_req;
|
|
|
|
req->DevHandle = targ->handle;
|
|
|
|
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
|
|
|
|
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
|
|
|
|
|
|
|
|
/* SAS Hard Link Reset / SATA Link Reset */
|
|
|
|
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
|
|
|
|
|
|
|
|
cm->cm_targ = targ;
|
|
|
|
cm->cm_data = NULL;
|
|
|
|
cm->cm_desc.HighPriority.RequestFlags =
|
|
|
|
MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
|
|
|
|
cm->cm_complete = mpssas_remove_volume;
|
|
|
|
cm->cm_complete_data = (void *)(uintptr_t)handle;
|
|
|
|
mps_map_command(sc, cm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2012-01-26 18:17:21 +00:00
|
|
|
* The MPT2 firmware performs debounce on the link to avoid transient link
|
|
|
|
* errors and false removals. When it does decide that link has been lost
|
|
|
|
* and a device need to go away, it expects that the host will perform a
|
|
|
|
* target reset and then an op remove. The reset has the side-effect of
|
|
|
|
* aborting any outstanding requests for the device, which is required for
|
|
|
|
* the op-remove to succeed. It's not clear if the host should check for
|
|
|
|
* the device coming back alive after the reset.
|
2010-09-10 15:03:56 +00:00
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
|
|
|
mpssas_prepare_remove(struct mpssas_softc *sassc, uint16_t handle)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct mps_command *cm;
|
|
|
|
struct mpssas_target *targ = NULL;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
sc = sassc->sc;
|
|
|
|
|
|
|
|
targ = mpssas_find_target_by_handle(sassc, 0, handle);
|
|
|
|
if (targ == NULL) {
|
|
|
|
/* FIXME: what is the action? */
|
2010-09-10 15:03:56 +00:00
|
|
|
/* We don't know about this device? */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s : invalid handle 0x%x \n", __func__, handle);
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->flags |= MPSSAS_TARGET_INREMOVAL;
|
|
|
|
|
|
|
|
cm = mpssas_alloc_tm(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
if (cm == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: command alloc failure\n", __func__);
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
mpssas_rescan_target(sc, targ);
|
2011-02-18 17:01:57 +00:00
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)cm->cm_req;
|
2011-02-18 17:01:57 +00:00
|
|
|
memset(req, 0, sizeof(*req));
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(targ->handle);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
|
|
|
|
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
|
|
|
|
|
|
|
|
/* SAS Hard Link Reset / SATA Link Reset */
|
|
|
|
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_targ = targ;
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_data = NULL;
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_complete = mpssas_remove_device;
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_complete_data = (void *)(uintptr_t)handle;
|
|
|
|
mps_map_command(sc, cm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_remove_device(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
|
|
|
|
MPI2_SAS_IOUNIT_CONTROL_REQUEST *req;
|
|
|
|
struct mpssas_target *targ;
|
2011-02-18 17:01:57 +00:00
|
|
|
struct mps_command *next_cm;
|
2010-09-10 15:03:56 +00:00
|
|
|
uint16_t handle;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
|
|
|
|
targ = tm->cm_targ;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: cm_flags = %#x for remove of handle %#04x! "
|
|
|
|
"This should not happen!\n", __func__, tm->cm_flags,
|
|
|
|
handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (reply == NULL) {
|
|
|
|
/* XXX retry the remove after the diag reset completes? */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"%s NULL reply reseting device 0x%04x\n", __func__, handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if (le16toh(reply->IOCStatus) != MPI2_IOCSTATUS_SUCCESS) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"IOCStatus = 0x%x while resetting device 0x%x\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(reply->IOCStatus), handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "Reset aborted %u commands\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le32toh(reply->TerminationCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_free_reply(sc, tm->cm_reply_data);
|
2012-09-14 21:28:56 +00:00
|
|
|
tm->cm_reply = NULL; /* Ensures the reply won't get re-freed */
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
/* Reuse the existing command */
|
2012-01-26 18:17:21 +00:00
|
|
|
req = (MPI2_SAS_IOUNIT_CONTROL_REQUEST *)tm->cm_req;
|
2011-02-18 17:01:57 +00:00
|
|
|
memset(req, 0, sizeof(*req));
|
2010-09-10 15:03:56 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SAS_IO_UNIT_CONTROL;
|
|
|
|
req->Operation = MPI2_SAS_OP_REMOVE_DEVICE;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_data = NULL;
|
|
|
|
tm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
|
|
|
|
tm->cm_complete = mpssas_remove_complete;
|
|
|
|
tm->cm_complete_data = (void *)(uintptr_t)handle;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_map_command(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "clearing target %u handle 0x%04x\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->tid, handle);
|
|
|
|
TAILQ_FOREACH_SAFE(tm, &targ->commands, cm_link, next_cm) {
|
2011-02-18 17:01:57 +00:00
|
|
|
union ccb *ccb;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "Completing missed command %p\n", tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
ccb = tm->cm_complete_data;
|
2011-02-18 17:01:57 +00:00
|
|
|
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_scsiio_complete(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_remove_complete(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
|
|
|
MPI2_SAS_IOUNIT_CONTROL_REPLY *reply;
|
2012-01-26 18:17:21 +00:00
|
|
|
uint16_t handle;
|
|
|
|
struct mpssas_target *targ;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
struct mpssas_lun *lun;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
reply = (MPI2_SAS_IOUNIT_CONTROL_REPLY *)tm->cm_reply;
|
|
|
|
handle = (uint16_t)(uintptr_t)tm->cm_complete_data;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
|
|
|
*/
|
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
|
|
|
"%s: cm_flags = %#x for remove of handle %#04x! "
|
2012-01-26 18:17:21 +00:00
|
|
|
"This should not happen!\n", __func__, tm->cm_flags,
|
|
|
|
handle);
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (reply == NULL) {
|
|
|
|
/* most likely a chip reset */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"%s NULL reply removing device 0x%04x\n", __func__, handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
|
|
|
"%s on handle 0x%04x, IOCStatus= 0x%x\n", __func__,
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
handle, le16toh(reply->IOCStatus));
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Don't clear target if remove fails because things will get confusing.
|
|
|
|
* Leave the devname and sasaddr intact so that we know to avoid reusing
|
|
|
|
* this target id if possible, and so we can assign the same target id
|
|
|
|
* to this device if it comes back in the future.
|
|
|
|
*/
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if (le16toh(reply->IOCStatus) == MPI2_IOCSTATUS_SUCCESS) {
|
2012-01-26 18:17:21 +00:00
|
|
|
targ = tm->cm_targ;
|
|
|
|
targ->handle = 0x0;
|
|
|
|
targ->encl_handle = 0x0;
|
|
|
|
targ->encl_slot = 0x0;
|
|
|
|
targ->exp_dev_handle = 0x0;
|
|
|
|
targ->phy_num = 0x0;
|
|
|
|
targ->linkrate = 0x0;
|
|
|
|
targ->devinfo = 0x0;
|
2012-02-09 00:16:12 +00:00
|
|
|
targ->flags = 0x0;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
|
|
|
|
while(!SLIST_EMPTY(&targ->luns)) {
|
|
|
|
lun = SLIST_FIRST(&targ->luns);
|
|
|
|
SLIST_REMOVE_HEAD(&targ->luns, lun_link);
|
|
|
|
free(lun, M_MPT2);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
mpssas_register_events(struct mps_softc *sc)
|
|
|
|
{
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
u32 events[MPI2_EVENT_NOTIFY_EVENTMASK_WORDS];
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
bzero(events, 16);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_DEVICE_STATUS_CHANGE);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_DISCOVERY);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_BROADCAST_PRIMITIVE);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_INIT_DEVICE_STATUS_CHANGE);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_INIT_TABLE_OVERFLOW);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_TOPOLOGY_CHANGE_LIST);
|
|
|
|
setbit(events, MPI2_EVENT_SAS_ENCL_DEVICE_STATUS_CHANGE);
|
2012-01-26 18:17:21 +00:00
|
|
|
setbit(events, MPI2_EVENT_IR_CONFIGURATION_CHANGE_LIST);
|
|
|
|
setbit(events, MPI2_EVENT_IR_VOLUME);
|
|
|
|
setbit(events, MPI2_EVENT_IR_PHYSICAL_DISK);
|
|
|
|
setbit(events, MPI2_EVENT_IR_OPERATION_STATUS);
|
|
|
|
setbit(events, MPI2_EVENT_LOG_ENTRY_ADDED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
mps_register_events(sc, events, mpssas_evt_handler, NULL,
|
|
|
|
&sc->sassc->mpssas_eh);
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
mps_attach_sas(struct mps_softc *sc)
|
|
|
|
{
|
|
|
|
struct mpssas_softc *sassc;
|
2012-01-26 18:17:21 +00:00
|
|
|
cam_status status;
|
|
|
|
int unit, error = 0;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
sassc = malloc(sizeof(struct mpssas_softc), M_MPT2, M_WAITOK|M_ZERO);
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if(!sassc) {
|
|
|
|
device_printf(sc->mps_dev, "Cannot allocate memory %s %d\n",
|
|
|
|
__func__, __LINE__);
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
2014-04-07 15:13:24 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX MaxTargets could change during a reinit. Since we don't
|
|
|
|
* resize the targets[] array during such an event, cache the value
|
|
|
|
* of MaxTargets here so that we don't get into trouble later. This
|
|
|
|
* should move into the reinit logic.
|
|
|
|
*/
|
|
|
|
sassc->maxtargets = sc->facts->MaxTargets;
|
2010-09-10 15:03:56 +00:00
|
|
|
sassc->targets = malloc(sizeof(struct mpssas_target) *
|
2014-04-07 15:13:24 +00:00
|
|
|
sassc->maxtargets, M_MPT2, M_WAITOK|M_ZERO);
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if(!sassc->targets) {
|
|
|
|
device_printf(sc->mps_dev, "Cannot allocate memory %s %d\n",
|
|
|
|
__func__, __LINE__);
|
|
|
|
free(sassc, M_MPT2);
|
|
|
|
return (ENOMEM);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
sc->sassc = sassc;
|
|
|
|
sassc->sc = sc;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sassc->devq = cam_simq_alloc(sc->num_reqs)) == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "Cannot allocate SIMQ\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
error = ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
unit = device_get_unit(sc->mps_dev);
|
2010-09-10 15:03:56 +00:00
|
|
|
sassc->sim = cam_sim_alloc(mpssas_action, mpssas_poll, "mps", sassc,
|
2012-01-26 18:17:21 +00:00
|
|
|
unit, &sc->mps_mtx, sc->num_reqs, sc->num_reqs, sassc->devq);
|
2010-09-10 15:03:56 +00:00
|
|
|
if (sassc->sim == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "Cannot allocate SIM\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
error = EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
TAILQ_INIT(&sassc->ev_queue);
|
|
|
|
|
|
|
|
/* Initialize taskqueue for Event Handling */
|
|
|
|
TASK_INIT(&sassc->ev_task, 0, mpssas_firmware_event_work, sc);
|
|
|
|
sassc->ev_tq = taskqueue_create("mps_taskq", M_NOWAIT | M_ZERO,
|
|
|
|
taskqueue_thread_enqueue, &sassc->ev_tq);
|
|
|
|
|
|
|
|
/* Run the task queue with lowest priority */
|
|
|
|
taskqueue_start_threads(&sassc->ev_tq, 1, 255, "%s taskq",
|
|
|
|
device_get_nameunit(sc->mps_dev));
|
|
|
|
|
|
|
|
mps_lock(sc);
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
/*
|
|
|
|
* XXX There should be a bus for every port on the adapter, but since
|
|
|
|
* we're just going to fake the topology for now, we'll pretend that
|
|
|
|
* everything is just a target on a single bus.
|
|
|
|
*/
|
|
|
|
if ((error = xpt_bus_register(sassc->sim, sc->mps_dev, 0)) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "Error %d registering SCSI bus\n",
|
2010-09-10 15:03:56 +00:00
|
|
|
error);
|
|
|
|
mps_unlock(sc);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
* Assume that discovery events will start right away.
|
|
|
|
*
|
|
|
|
* Hold off boot until discovery is complete.
|
2010-09-10 15:03:56 +00:00
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
sassc->flags |= MPSSAS_IN_STARTUP | MPSSAS_IN_DISCOVERY;
|
|
|
|
sc->sassc->startup_refcount = 0;
|
2014-02-27 21:48:34 +00:00
|
|
|
mpssas_startup_increment(sassc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
callout_init(&sassc->discovery_callout, 1 /*mpsafe*/);
|
|
|
|
sassc->discovery_timeouts = 0;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
sassc->tm_count = 0;
|
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
/*
|
|
|
|
* Register for async events so we can determine the EEDP
|
|
|
|
* capabilities of devices.
|
|
|
|
*/
|
|
|
|
status = xpt_create_path(&sassc->path, /*periph*/NULL,
|
|
|
|
cam_sim_path(sc->sassc->sim), CAM_TARGET_WILDCARD,
|
|
|
|
CAM_LUN_WILDCARD);
|
2012-01-26 18:17:21 +00:00
|
|
|
if (status != CAM_REQ_CMP) {
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
mps_printf(sc, "Error %#x creating sim path\n", status);
|
|
|
|
sassc->path = NULL;
|
|
|
|
} else {
|
|
|
|
int event;
|
|
|
|
|
|
|
|
#if (__FreeBSD_version >= 1000006) || \
|
|
|
|
((__FreeBSD_version >= 901503) && (__FreeBSD_version < 1000000))
|
|
|
|
event = AC_ADVINFO_CHANGED;
|
|
|
|
#else
|
|
|
|
event = AC_FOUND_DEVICE;
|
2012-01-26 18:17:21 +00:00
|
|
|
#endif
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
status = xpt_register_async(event, mpssas_async, sc,
|
|
|
|
sassc->path);
|
|
|
|
if (status != CAM_REQ_CMP) {
|
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"Error %#x registering async handler for "
|
|
|
|
"AC_ADVINFO_CHANGED events\n", status);
|
|
|
|
xpt_free_path(sassc->path);
|
|
|
|
sassc->path = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (status != CAM_REQ_CMP) {
|
|
|
|
/*
|
|
|
|
* EEDP use is the exception, not the rule.
|
|
|
|
* Warn the user, but do not fail to attach.
|
|
|
|
*/
|
|
|
|
mps_printf(sc, "EEDP capabilities disabled.\n");
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
mps_unlock(sc);
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
mpssas_register_events(sc);
|
|
|
|
out:
|
|
|
|
if (error)
|
|
|
|
mps_detach_sas(sc);
|
|
|
|
return (error);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
mps_detach_sas(struct mps_softc *sc)
|
|
|
|
{
|
|
|
|
struct mpssas_softc *sassc;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
struct mpssas_lun *lun, *lun_tmp;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
int i;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
if (sc->sassc == NULL)
|
|
|
|
return (0);
|
|
|
|
|
|
|
|
sassc = sc->sassc;
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_deregister_events(sc, sassc->mpssas_eh);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Drain and free the event handling taskqueue with the lock
|
|
|
|
* unheld so that any parallel processing tasks drain properly
|
|
|
|
* without deadlocking.
|
|
|
|
*/
|
|
|
|
if (sassc->ev_tq != NULL)
|
|
|
|
taskqueue_free(sassc->ev_tq);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
/* Make sure CAM doesn't wedge if we had to bail out early. */
|
|
|
|
mps_lock(sc);
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* Deregister our async handler */
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
if (sassc->path != NULL) {
|
|
|
|
xpt_register_async(0, mpssas_async, sc, sassc->path);
|
|
|
|
xpt_free_path(sassc->path);
|
|
|
|
sassc->path = NULL;
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (sassc->flags & MPSSAS_IN_STARTUP)
|
|
|
|
xpt_release_simq(sassc->sim, 1);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
if (sassc->sim != NULL) {
|
|
|
|
xpt_bus_deregister(cam_sim_path(sassc->sim));
|
|
|
|
cam_sim_free(sassc->sim, FALSE);
|
|
|
|
}
|
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
sassc->flags |= MPSSAS_SHUTDOWN;
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_unlock(sc);
|
|
|
|
|
|
|
|
if (sassc->devq != NULL)
|
2010-09-10 15:03:56 +00:00
|
|
|
cam_simq_free(sassc->devq);
|
|
|
|
|
2014-04-07 15:13:24 +00:00
|
|
|
for(i=0; i< sassc->maxtargets ;i++) {
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
targ = &sassc->targets[i];
|
|
|
|
SLIST_FOREACH_SAFE(lun, &targ->luns, lun_link, lun_tmp) {
|
|
|
|
free(lun, M_MPT2);
|
|
|
|
}
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
free(sassc->targets, M_MPT2);
|
|
|
|
free(sassc, M_MPT2);
|
|
|
|
sc->sassc = NULL;
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
2010-09-10 15:03:56 +00:00
|
|
|
mpssas_discovery_end(struct mpssas_softc *sassc)
|
|
|
|
{
|
|
|
|
struct mps_softc *sc = sassc->sc;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
if (sassc->flags & MPSSAS_DISCOVERY_TIMEOUT_PENDING)
|
|
|
|
callout_stop(&sassc->discovery_callout);
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_discovery_timeout(void *data)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mpssas_softc *sassc = data;
|
|
|
|
struct mps_softc *sc;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
sc = sassc->sc;
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_lock(sc);
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"Timeout waiting for discovery, interrupts may not be working!\n");
|
|
|
|
sassc->flags &= ~MPSSAS_DISCOVERY_TIMEOUT_PENDING;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* Poll the hardware for events in case interrupts aren't working */
|
|
|
|
mps_intr_locked(sc);
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"Finished polling after discovery timeout at %d\n", ticks);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
if ((sassc->flags & MPSSAS_IN_DISCOVERY) == 0) {
|
|
|
|
mpssas_discovery_end(sassc);
|
|
|
|
} else {
|
|
|
|
if (sassc->discovery_timeouts < MPSSAS_MAX_DISCOVERY_TIMEOUTS) {
|
|
|
|
sassc->flags |= MPSSAS_DISCOVERY_TIMEOUT_PENDING;
|
|
|
|
callout_reset(&sassc->discovery_callout,
|
|
|
|
MPSSAS_DISCOVERY_TIMEOUT * hz,
|
|
|
|
mpssas_discovery_timeout, sassc);
|
|
|
|
sassc->discovery_timeouts++;
|
|
|
|
} else {
|
|
|
|
mps_dprint(sassc->sc, MPS_FAULT,
|
|
|
|
"Discovery timed out, continuing.\n");
|
|
|
|
sassc->flags &= ~MPSSAS_IN_DISCOVERY;
|
|
|
|
mpssas_discovery_end(sassc);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
mps_unlock(sc);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_action(struct cam_sim *sim, union ccb *ccb)
|
|
|
|
{
|
|
|
|
struct mpssas_softc *sassc;
|
|
|
|
|
|
|
|
sassc = cam_sim_softc(sim);
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
|
|
|
mps_dprint(sassc->sc, MPS_TRACE, "ccb func_code 0x%x\n",
|
2010-09-10 15:03:56 +00:00
|
|
|
ccb->ccb_h.func_code);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
switch (ccb->ccb_h.func_code) {
|
|
|
|
case XPT_PATH_INQ:
|
|
|
|
{
|
|
|
|
struct ccb_pathinq *cpi = &ccb->cpi;
|
|
|
|
|
|
|
|
cpi->version_num = 1;
|
|
|
|
cpi->hba_inquiry = PI_SDTR_ABLE|PI_TAG_ABLE|PI_WIDE_16;
|
|
|
|
cpi->target_sprt = 0;
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#if __FreeBSD_version >= 1000039
|
|
|
|
cpi->hba_misc = PIM_NOBUSRESET | PIM_UNMAPPED | PIM_NOSCAN;
|
|
|
|
#else
|
2013-03-28 11:24:30 +00:00
|
|
|
cpi->hba_misc = PIM_NOBUSRESET | PIM_UNMAPPED;
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#endif
|
2010-09-10 15:03:56 +00:00
|
|
|
cpi->hba_eng_cnt = 0;
|
2014-04-07 15:13:24 +00:00
|
|
|
cpi->max_target = sassc->maxtargets - 1;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
cpi->max_lun = 255;
|
2014-04-07 15:13:24 +00:00
|
|
|
cpi->initiator_id = sassc->maxtargets - 1;
|
2010-09-10 15:03:56 +00:00
|
|
|
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
|
|
|
|
strncpy(cpi->hba_vid, "LSILogic", 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 = 150000;
|
|
|
|
cpi->transport = XPORT_SAS;
|
|
|
|
cpi->transport_version = 0;
|
|
|
|
cpi->protocol = PROTO_SCSI;
|
|
|
|
cpi->protocol_version = SCSI_REV_SPC;
|
2012-01-26 18:17:21 +00:00
|
|
|
#if __FreeBSD_version >= 800001
|
|
|
|
/*
|
|
|
|
* XXX KDM where does this number come from?
|
|
|
|
*/
|
|
|
|
cpi->maxio = 256 * 1024;
|
|
|
|
#endif
|
2010-09-10 15:03:56 +00:00
|
|
|
cpi->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
case XPT_GET_TRAN_SETTINGS:
|
|
|
|
{
|
|
|
|
struct ccb_trans_settings *cts;
|
|
|
|
struct ccb_trans_settings_sas *sas;
|
|
|
|
struct ccb_trans_settings_scsi *scsi;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
|
|
|
|
cts = &ccb->cts;
|
|
|
|
sas = &cts->xport_specific.sas;
|
|
|
|
scsi = &cts->proto_specific.scsi;
|
|
|
|
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(cts->ccb_h.target_id < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in XPT_GET_TRANS_SETTINGS\n",
|
|
|
|
cts->ccb_h.target_id));
|
2010-09-10 15:03:56 +00:00
|
|
|
targ = &sassc->targets[cts->ccb_h.target_id];
|
|
|
|
if (targ->handle == 0x0) {
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
cts->ccb_h.status = CAM_SEL_TIMEOUT;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
cts->protocol_version = SCSI_REV_SPC2;
|
|
|
|
cts->transport = XPORT_SAS;
|
|
|
|
cts->transport_version = 0;
|
|
|
|
|
|
|
|
sas->valid = CTS_SAS_VALID_SPEED;
|
|
|
|
switch (targ->linkrate) {
|
|
|
|
case 0x08:
|
|
|
|
sas->bitrate = 150000;
|
|
|
|
break;
|
|
|
|
case 0x09:
|
|
|
|
sas->bitrate = 300000;
|
|
|
|
break;
|
|
|
|
case 0x0a:
|
|
|
|
sas->bitrate = 600000;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
sas->valid = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
cts->protocol = PROTO_SCSI;
|
|
|
|
scsi->valid = CTS_SCSI_VALID_TQ;
|
|
|
|
scsi->flags = CTS_SCSI_FLAGS_TAG_ENB;
|
|
|
|
|
|
|
|
cts->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
case XPT_CALC_GEOMETRY:
|
|
|
|
cam_calc_geometry(&ccb->ccg, /*extended*/1);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
break;
|
|
|
|
case XPT_RESET_DEV:
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_XINFO, "mpssas_action XPT_RESET_DEV\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
mpssas_action_resetdev(sassc, ccb);
|
|
|
|
return;
|
|
|
|
case XPT_RESET_BUS:
|
|
|
|
case XPT_ABORT:
|
|
|
|
case XPT_TERM_IO:
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_XINFO,
|
|
|
|
"mpssas_action faking success for abort or reset\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
break;
|
|
|
|
case XPT_SCSI_IO:
|
|
|
|
mpssas_action_scsiio(sassc, ccb);
|
|
|
|
return;
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
#if __FreeBSD_version >= 900026
|
|
|
|
case XPT_SMP_IO:
|
|
|
|
mpssas_action_smpio(sassc, ccb);
|
|
|
|
return;
|
2012-01-26 18:17:21 +00:00
|
|
|
#endif
|
2010-09-10 15:03:56 +00:00
|
|
|
default:
|
|
|
|
ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
xpt_done(ccb);
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_announce_reset(struct mps_softc *sc, uint32_t ac_code,
|
|
|
|
target_id_t target_id, lun_id_t lun_id)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
path_id_t path_id = cam_sim_path(sc->sassc->sim);
|
|
|
|
struct cam_path *path;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-10-30 14:04:47 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "%s code %x target %d lun %jx\n", __func__,
|
|
|
|
ac_code, target_id, (uintmax_t)lun_id);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (xpt_create_path(&path, NULL,
|
|
|
|
path_id, target_id, lun_id) != CAM_REQ_CMP) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "unable to create path for reset "
|
2012-01-26 18:17:21 +00:00
|
|
|
"notification\n");
|
|
|
|
return;
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_async(ac_code, path, NULL);
|
|
|
|
xpt_free_path(path);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
|
|
|
mpssas_complete_all_commands(struct mps_softc *sc)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
|
|
|
struct mps_command *cm;
|
2012-01-26 18:17:21 +00:00
|
|
|
int i;
|
|
|
|
int completed;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* complete all commands with a NULL reply */
|
|
|
|
for (i = 1; i < sc->num_reqs; i++) {
|
|
|
|
cm = &sc->commands[i];
|
|
|
|
cm->cm_reply = NULL;
|
|
|
|
completed = 0;
|
|
|
|
|
|
|
|
if (cm->cm_flags & MPS_CM_FLAGS_POLLED)
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_COMPLETE;
|
|
|
|
|
|
|
|
if (cm->cm_complete != NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completing cm %p state %x ccb %p for diag reset\n",
|
|
|
|
cm, cm->cm_state, cm->cm_ccb);
|
|
|
|
|
|
|
|
cm->cm_complete(sc, cm);
|
|
|
|
completed = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (cm->cm_flags & MPS_CM_FLAGS_WAKEUP) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"waking up cm %p state %x ccb %p for diag reset\n",
|
|
|
|
cm, cm->cm_state, cm->cm_ccb);
|
|
|
|
wakeup(cm);
|
|
|
|
completed = 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((completed == 0) && (cm->cm_state != MPS_CM_STATE_FREE)) {
|
|
|
|
/* this should never happen, but if it does, log */
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"cm %p state %x flags 0x%x ccb %p during diag "
|
|
|
|
"reset\n", cm, cm->cm_state, cm->cm_flags,
|
|
|
|
cm->cm_ccb);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
void
|
|
|
|
mpssas_handle_reinit(struct mps_softc *sc)
|
|
|
|
{
|
|
|
|
int i;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* Go back into startup mode and freeze the simq, so that CAM
|
|
|
|
* doesn't send any commands until after we've rediscovered all
|
|
|
|
* targets and found the proper device handles for them.
|
|
|
|
*
|
|
|
|
* After the reset, portenable will trigger discovery, and after all
|
|
|
|
* discovery-related activities have finished, the simq will be
|
|
|
|
* released.
|
2010-09-17 21:53:56 +00:00
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INIT, "%s startup\n", __func__);
|
2012-01-26 18:17:21 +00:00
|
|
|
sc->sassc->flags |= MPSSAS_IN_STARTUP;
|
|
|
|
sc->sassc->flags |= MPSSAS_IN_DISCOVERY;
|
2014-02-27 21:48:34 +00:00
|
|
|
mpssas_startup_increment(sc->sassc);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* notify CAM of a bus reset */
|
|
|
|
mpssas_announce_reset(sc, AC_BUS_RESET, CAM_TARGET_WILDCARD,
|
|
|
|
CAM_LUN_WILDCARD);
|
|
|
|
|
|
|
|
/* complete and cleanup after all outstanding commands */
|
|
|
|
mpssas_complete_all_commands(sc);
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INIT,
|
|
|
|
"%s startup %u tm %u after command completion\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
__func__, sc->sassc->startup_refcount, sc->sassc->tm_count);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* zero all the target handles, since they may change after the
|
|
|
|
* reset, and we have to rediscover all the targets and use the new
|
|
|
|
* handles.
|
|
|
|
*/
|
2014-04-07 15:13:24 +00:00
|
|
|
for (i = 0; i < sc->sassc->maxtargets; i++) {
|
2012-01-26 18:17:21 +00:00
|
|
|
if (sc->sassc->targets[i].outstanding != 0)
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INIT, "target %u outstanding %u\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
i, sc->sassc->targets[i].outstanding);
|
|
|
|
sc->sassc->targets[i].handle = 0x0;
|
|
|
|
sc->sassc->targets[i].exp_dev_handle = 0x0;
|
|
|
|
sc->sassc->targets[i].outstanding = 0;
|
|
|
|
sc->sassc->targets[i].flags = MPSSAS_TARGET_INDIAGRESET;
|
|
|
|
}
|
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
|
|
|
mpssas_tm_timeout(void *data)
|
|
|
|
{
|
|
|
|
struct mps_command *tm = data;
|
|
|
|
struct mps_softc *sc = tm->cm_sc;
|
|
|
|
|
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_INFO|MPS_RECOVERY,
|
|
|
|
"task mgmt %p timed out\n", tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_reinit(sc);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_logical_unit_reset_complete(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
|
2010-09-10 15:03:56 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
2012-01-26 18:17:21 +00:00
|
|
|
unsigned int cm_count = 0;
|
|
|
|
struct mps_command *cm;
|
|
|
|
struct mpssas_target *targ;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
callout_stop(&tm->cm_callout);
|
|
|
|
|
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
|
|
|
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
targ = tm->cm_targ;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
2013-07-19 00:12:41 +00:00
|
|
|
* XXXSL So should it be an assertion?
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "%s: cm_flags = %#x for LUN reset! "
|
2012-01-26 18:17:21 +00:00
|
|
|
"This should not happen!\n", __func__, tm->cm_flags);
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (reply == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
|
|
|
"NULL reset reply for tm %p\n", tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
|
|
|
|
/* this completion was due to a reset, just cleanup */
|
|
|
|
targ->flags &= ~MPSSAS_TARGET_INRESET;
|
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* we should have gotten a reply. */
|
|
|
|
mps_reinit(sc);
|
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"logical unit reset status 0x%x code 0x%x count %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(reply->IOCStatus), le32toh(reply->ResponseCode),
|
|
|
|
le32toh(reply->TerminationCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/* See if there are any outstanding commands for this LUN.
|
|
|
|
* This could be made more efficient by using a per-LU data
|
|
|
|
* structure of some sort.
|
|
|
|
*/
|
|
|
|
TAILQ_FOREACH(cm, &targ->commands, cm_link) {
|
|
|
|
if (cm->cm_lun == tm->cm_lun)
|
|
|
|
cm_count++;
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (cm_count == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY|MPS_INFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"logical unit %u finished recovery after reset\n",
|
|
|
|
tm->cm_lun, tm);
|
|
|
|
|
|
|
|
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
|
|
|
|
tm->cm_lun);
|
|
|
|
|
|
|
|
/* we've finished recovery for this logical unit. check and
|
|
|
|
* see if some other logical unit has a timedout command
|
|
|
|
* that needs to be processed.
|
|
|
|
*/
|
|
|
|
cm = TAILQ_FIRST(&targ->timedout_commands);
|
|
|
|
if (cm) {
|
|
|
|
mpssas_send_abort(sc, tm, cm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* if we still have commands for this LUN, the reset
|
|
|
|
* effectively failed, regardless of the status reported.
|
|
|
|
* Escalate to a target reset.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"logical unit reset complete for tm %p, but still have %u command(s)\n",
|
|
|
|
tm, cm_count);
|
|
|
|
mpssas_send_reset(sc, tm,
|
|
|
|
MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_target_reset_complete(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mpssas_target *targ;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
callout_stop(&tm->cm_callout);
|
|
|
|
|
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
|
|
|
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
targ = tm->cm_targ;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
|
|
|
*/
|
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,"%s: cm_flags = %#x for target reset! "
|
2012-01-26 18:17:21 +00:00
|
|
|
"This should not happen!\n", __func__, tm->cm_flags);
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (reply == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
|
|
|
"NULL reset reply for tm %p\n", tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
|
|
|
|
/* this completion was due to a reset, just cleanup */
|
|
|
|
targ->flags &= ~MPSSAS_TARGET_INRESET;
|
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* we should have gotten a reply. */
|
|
|
|
mps_reinit(sc);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"target reset status 0x%x code 0x%x count %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(reply->IOCStatus), le32toh(reply->ResponseCode),
|
|
|
|
le32toh(reply->TerminationCount));
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->flags &= ~MPSSAS_TARGET_INRESET;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (targ->outstanding == 0) {
|
|
|
|
/* we've finished recovery for this target and all
|
|
|
|
* of its logical units.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY|MPS_INFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"recovery finished after target reset\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
|
|
|
|
CAM_LUN_WILDCARD);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* after a target reset, if this target still has
|
|
|
|
* outstanding commands, the reset effectively failed,
|
|
|
|
* regardless of the status reported. escalate.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"target reset complete for tm %p, but still have %u command(s)\n",
|
|
|
|
tm, targ->outstanding);
|
|
|
|
mps_reinit(sc);
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
#define MPS_RESET_TIMEOUT 30
|
|
|
|
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
static int
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_send_reset(struct mps_softc *sc, struct mps_command *tm, uint8_t type)
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mpssas_target *target;
|
|
|
|
int err;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
target = tm->cm_targ;
|
|
|
|
if (target->handle == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,"%s null devhandle for target_id %d\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
__func__, target->tid);
|
|
|
|
return -1;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(target->handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
|
|
|
|
req->TaskType = type;
|
|
|
|
|
|
|
|
if (type == MPI2_SCSITASKMGMT_TASKTYPE_LOGICAL_UNIT_RESET) {
|
|
|
|
/* XXX Need to handle invalid LUNs */
|
|
|
|
MPS_SET_LUN(req->LUN, tm->cm_lun);
|
|
|
|
tm->cm_targ->logical_unit_resets++;
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY|MPS_INFO,
|
|
|
|
"sending logical unit reset\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_complete = mpssas_logical_unit_reset_complete;
|
|
|
|
}
|
|
|
|
else if (type == MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET) {
|
|
|
|
/* Target reset method = SAS Hard Link Reset / SATA Link Reset */
|
|
|
|
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
|
|
|
|
tm->cm_targ->target_resets++;
|
|
|
|
tm->cm_targ->flags |= MPSSAS_TARGET_INRESET;
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY|MPS_INFO,
|
|
|
|
"sending target reset\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_complete = mpssas_target_reset_complete;
|
|
|
|
}
|
|
|
|
else {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "unexpected reset type 0x%x\n", type);
|
2012-01-26 18:17:21 +00:00
|
|
|
return -1;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_data = NULL;
|
|
|
|
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
|
|
|
|
tm->cm_complete_data = (void *)tm;
|
|
|
|
|
|
|
|
callout_reset(&tm->cm_callout, MPS_RESET_TIMEOUT * hz,
|
|
|
|
mpssas_tm_timeout, tm);
|
|
|
|
|
|
|
|
err = mps_map_command(sc, tm);
|
|
|
|
if (err)
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"error %d sending reset type %u\n",
|
|
|
|
err, type);
|
|
|
|
|
|
|
|
return err;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_abort_complete(struct mps_softc *sc, struct mps_command *tm)
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mps_command *cm;
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *reply;
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mpssas_target *targ;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
callout_stop(&tm->cm_callout);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
|
|
|
reply = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
targ = tm->cm_targ;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2010-09-16 22:54:56 +00:00
|
|
|
/*
|
2012-01-26 18:17:21 +00:00
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
2010-09-16 22:54:56 +00:00
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"cm_flags = %#x for abort %p TaskMID %u!\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
tm->cm_flags, tm, le16toh(req->TaskMID));
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (reply == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"NULL abort reply for tm %p TaskMID %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
tm, le16toh(req->TaskMID));
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
|
|
|
|
/* this completion was due to a reset, just cleanup */
|
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* we should have gotten a reply. */
|
|
|
|
mps_reinit(sc);
|
|
|
|
}
|
|
|
|
return;
|
|
|
|
}
|
2010-09-16 22:54:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"abort TaskMID %u status 0x%x code 0x%x count %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(req->TaskMID),
|
|
|
|
le16toh(reply->IOCStatus), le32toh(reply->ResponseCode),
|
|
|
|
le32toh(reply->TerminationCount));
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
cm = TAILQ_FIRST(&tm->cm_targ->timedout_commands);
|
|
|
|
if (cm == NULL) {
|
|
|
|
/* if there are no more timedout commands, we're done with
|
|
|
|
* error recovery for this target.
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"finished recovery after aborting TaskMID %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(req->TaskMID));
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
targ->tm = NULL;
|
|
|
|
mpssas_free_tm(sc, tm);
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
else if (le16toh(req->TaskMID) != cm->cm_desc.Default.SMID) {
|
2012-01-26 18:17:21 +00:00
|
|
|
/* abort success, but we have more timedout commands to abort */
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"continuing recovery after aborting TaskMID %u\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(req->TaskMID));
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
mpssas_send_abort(sc, tm, cm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* we didn't get a command completion, so the abort
|
|
|
|
* failed as far as we're concerned. escalate.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"abort failed for TaskMID %u tm %p\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(req->TaskMID), tm);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
mpssas_send_reset(sc, tm,
|
|
|
|
MPI2_SCSITASKMGMT_TASKTYPE_LOGICAL_UNIT_RESET);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
#define MPS_ABORT_TIMEOUT 5
|
|
|
|
|
|
|
|
static int
|
|
|
|
mpssas_send_abort(struct mps_softc *sc, struct mps_command *tm, struct mps_command *cm)
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
int err;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ = cm->cm_targ;
|
|
|
|
if (targ->handle == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,"%s null devhandle for target_id %d\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
__func__, cm->cm_ccb->ccb_h.target_id);
|
|
|
|
return -1;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY|MPS_INFO,
|
|
|
|
"Aborting command %p\n", cm);
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(targ->handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
|
|
|
|
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_ABORT_TASK;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* XXX Need to handle invalid LUNs */
|
|
|
|
MPS_SET_LUN(req->LUN, cm->cm_ccb->ccb_h.target_lun);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->TaskMID = htole16(cm->cm_desc.Default.SMID);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_data = NULL;
|
|
|
|
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
|
|
|
|
tm->cm_complete = mpssas_abort_complete;
|
|
|
|
tm->cm_complete_data = (void *)tm;
|
|
|
|
tm->cm_targ = cm->cm_targ;
|
|
|
|
tm->cm_lun = cm->cm_lun;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
callout_reset(&tm->cm_callout, MPS_ABORT_TIMEOUT * hz,
|
|
|
|
mpssas_tm_timeout, tm);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->aborts++;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
err = mps_map_command(sc, tm);
|
|
|
|
if (err)
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(tm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"error %d sending abort for cm %p SMID %u\n",
|
|
|
|
err, cm, req->TaskMID);
|
|
|
|
return err;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
|
|
|
mpssas_scsiio_timeout(void *data)
|
|
|
|
{
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct mps_command *cm;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
|
|
|
|
cm = (struct mps_command *)data;
|
|
|
|
sc = cm->cm_sc;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "Timeout checking cm %p\n", sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Run the interrupt handler to make sure it's not pending. This
|
|
|
|
* isn't perfect because the command could have already completed
|
|
|
|
* and been re-used, though this is unlikely.
|
|
|
|
*/
|
|
|
|
mps_intr_locked(sc);
|
|
|
|
if (cm->cm_state == MPS_CM_STATE_FREE) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_XINFO,
|
|
|
|
"SCSI command %p almost timed out\n", cm);
|
2012-01-26 18:17:21 +00:00
|
|
|
return;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (cm->cm_ccb == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "command timeout with NULL ccb\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
return;
|
|
|
|
}
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_INFO, "command timeout cm %p ccb %p\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
cm, cm->cm_ccb);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ = cm->cm_targ;
|
|
|
|
targ->timeouts++;
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* XXX first, check the firmware state, to see if it's still
|
|
|
|
* operational. if not, do a diag reset.
|
|
|
|
*/
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_ccb->ccb_h.status = CAM_CMD_TIMEOUT;
|
|
|
|
cm->cm_state = MPS_CM_STATE_TIMEDOUT;
|
|
|
|
TAILQ_INSERT_TAIL(&targ->timedout_commands, cm, cm_recovery);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (targ->tm != NULL) {
|
|
|
|
/* target already in recovery, just queue up another
|
|
|
|
* timedout command to be processed later.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY,
|
|
|
|
"queued timedout cm %p for processing by tm %p\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
cm, targ->tm);
|
|
|
|
}
|
|
|
|
else if ((targ->tm = mpssas_alloc_tm(sc)) != NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY, "timedout cm %p allocated tm %p\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
cm, targ->tm);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/* start recovery by aborting the first timedout command */
|
|
|
|
mpssas_send_abort(sc, targ->tm, cm);
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
/* XXX queue this target up for recovery once a TM becomes
|
|
|
|
* available. The firmware only has a limited number of
|
|
|
|
* HighPriority credits for the high priority requests used
|
|
|
|
* for task management, and we ran out.
|
|
|
|
*
|
|
|
|
* Isilon: don't worry about this for now, since we have
|
|
|
|
* more credits than disks in an enclosure, and limit
|
|
|
|
* ourselves to one TM per target for recovery.
|
|
|
|
*/
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_RECOVERY,
|
|
|
|
"timedout cm %p failed to allocate a tm\n", cm);
|
Turn on serialization of task management commands going down to the
controller, but make it optional.
After a problem report from Andrew Boyer, it looks like the LSI
chip may have issues (the watchdog timer fired) if too many aborts
are sent down to the chip at the same time. We know that task
management commands are serialized, and although the manual doesn't
say it, it may be a good idea to just send one at a time.
But, since I'm not certain that this is necessary, add a tunable
and sysctl variable (hw.mps.%d.allow_multiple_tm_cmds) to control
the driver's behavior.
mps.c: Add support for the sysctl and tunable, and add a
comment about the possible return values to
mps_map_command().
mps_sas.c: Run all task management commands through two new
routines, mpssas_issue_tm_request() and
mpssas_complete_tm_request().
This allows us to optionally serialize task
management commands. Also, change things so that
the response to a task management command always
comes back through the callback. (Before it could
come via the callback or the return value.)
mpsvar.h: Add softc variables for the list of active task
management commands, the number of active commands,
and whether we should allow multiple active task
management commands. Add an active command flag.
mps.4: Describe the new sysctl/loader tunable variable.
Sponsored by: Spectra Logic Corporation
2010-10-07 21:56:10 +00:00
|
|
|
}
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_action_scsiio(struct mpssas_softc *sassc, union ccb *ccb)
|
|
|
|
{
|
|
|
|
MPI2_SCSI_IO_REQUEST *req;
|
|
|
|
struct ccb_scsiio *csio;
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct mpssas_target *targ;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mpssas_lun *lun;
|
2010-09-10 15:03:56 +00:00
|
|
|
struct mps_command *cm;
|
2012-01-26 18:17:21 +00:00
|
|
|
uint8_t i, lba_byte, *ref_tag_addr;
|
|
|
|
uint16_t eedp_flags;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
uint32_t mpi_control;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
sc = sassc->sc;
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
csio = &ccb->csio;
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(csio->ccb_h.target_id < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in XPT_SCSI_IO\n",
|
|
|
|
csio->ccb_h.target_id));
|
2010-09-10 15:03:56 +00:00
|
|
|
targ = &sassc->targets[csio->ccb_h.target_id];
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_TRACE, "ccb %p target flag %x\n", ccb, targ->flags);
|
2010-09-10 15:03:56 +00:00
|
|
|
if (targ->handle == 0x0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "%s NULL handle for target %u\n",
|
2012-01-26 18:17:21 +00:00
|
|
|
__func__, csio->ccb_h.target_id);
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
csio->ccb_h.status = CAM_SEL_TIMEOUT;
|
2012-01-26 18:17:21 +00:00
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
2012-02-09 00:16:12 +00:00
|
|
|
if (targ->flags & MPS_TARGET_FLAGS_RAID_COMPONENT) {
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "%s Raid component no SCSI IO "
|
|
|
|
"supported %u\n", __func__, csio->ccb_h.target_id);
|
2012-02-09 00:16:12 +00:00
|
|
|
csio->ccb_h.status = CAM_TID_INVALID;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
/*
|
|
|
|
* Sometimes, it is possible to get a command that is not "In
|
|
|
|
* Progress" and was actually aborted by the upper layer. Check for
|
|
|
|
* this here and complete the command without error.
|
|
|
|
*/
|
|
|
|
if (ccb->ccb_h.status != CAM_REQ_INPROG) {
|
|
|
|
mps_dprint(sc, MPS_TRACE, "%s Command is not in progress for "
|
|
|
|
"target %u\n", __func__, csio->ccb_h.target_id);
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* If devinfo is 0 this will be a volume. In that case don't tell CAM
|
|
|
|
* that the volume has timed out. We want volumes to be enumerated
|
|
|
|
* until they are deleted/removed, not just failed.
|
|
|
|
*/
|
|
|
|
if (targ->flags & MPSSAS_TARGET_INREMOVAL) {
|
|
|
|
if (targ->devinfo == 0)
|
|
|
|
csio->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
else
|
|
|
|
csio->ccb_h.status = CAM_SEL_TIMEOUT;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((sc->mps_flags & MPS_FLAGS_SHUTDOWN) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INFO, "%s shutting down\n", __func__);
|
2012-01-26 18:17:21 +00:00
|
|
|
csio->ccb_h.status = CAM_TID_INVALID;
|
2010-09-10 15:03:56 +00:00
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
cm = mps_alloc_command(sc);
|
|
|
|
if (cm == NULL) {
|
|
|
|
if ((sassc->flags & MPSSAS_QUEUE_FROZEN) == 0) {
|
|
|
|
xpt_freeze_simq(sassc->sim, 1);
|
|
|
|
sassc->flags |= MPSSAS_QUEUE_FROZEN;
|
|
|
|
}
|
|
|
|
ccb->ccb_h.status &= ~CAM_SIM_QUEUED;
|
|
|
|
ccb->ccb_h.status |= CAM_REQUEUE_REQ;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
req = (MPI2_SCSI_IO_REQUEST *)cm->cm_req;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
bzero(req, sizeof(*req));
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(targ->handle);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SCSI_IO_REQUEST;
|
|
|
|
req->MsgFlags = 0;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->SenseBufferLowAddress = htole32(cm->cm_sense_busaddr);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->SenseBufferLength = MPS_SENSE_LEN;
|
|
|
|
req->SGLFlags = 0;
|
|
|
|
req->ChainOffset = 0;
|
|
|
|
req->SGLOffset0 = 24; /* 32bit word offset to the SGL */
|
|
|
|
req->SGLOffset1= 0;
|
|
|
|
req->SGLOffset2= 0;
|
|
|
|
req->SGLOffset3= 0;
|
|
|
|
req->SkipCount = 0;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DataLength = htole32(csio->dxfer_len);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->BidirectionalDataLength = 0;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->IoFlags = htole16(csio->cdb_len);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->EEDPFlags = 0;
|
|
|
|
|
|
|
|
/* Note: BiDirectional transfers are not supported */
|
|
|
|
switch (csio->ccb_h.flags & CAM_DIR_MASK) {
|
|
|
|
case CAM_DIR_IN:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control = MPI2_SCSIIO_CONTROL_READ;
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_DATAIN;
|
|
|
|
break;
|
|
|
|
case CAM_DIR_OUT:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control = MPI2_SCSIIO_CONTROL_WRITE;
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_DATAOUT;
|
|
|
|
break;
|
|
|
|
case CAM_DIR_NONE:
|
|
|
|
default:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control = MPI2_SCSIIO_CONTROL_NODATATRANSFER;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
if (csio->cdb_len == 32)
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
|
2010-09-10 15:03:56 +00:00
|
|
|
/*
|
|
|
|
* It looks like the hardware doesn't require an explicit tag
|
|
|
|
* number for each transaction. SAM Task Management not supported
|
|
|
|
* at the moment.
|
|
|
|
*/
|
|
|
|
switch (csio->tag_action) {
|
|
|
|
case MSG_HEAD_OF_Q_TAG:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= MPI2_SCSIIO_CONTROL_HEADOFQ;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case MSG_ORDERED_Q_TAG:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= MPI2_SCSIIO_CONTROL_ORDEREDQ;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case MSG_ACA_TASK:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= MPI2_SCSIIO_CONTROL_ACAQ;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case CAM_TAG_ACTION_NONE:
|
|
|
|
case MSG_SIMPLE_Q_TAG:
|
|
|
|
default:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= MPI2_SCSIIO_CONTROL_SIMPLEQ;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
mpi_control |= sc->mapping_table[csio->ccb_h.target_id].TLR_bits;
|
|
|
|
req->Control = htole32(mpi_control);
|
2010-12-11 00:36:35 +00:00
|
|
|
if (MPS_SET_LUN(req->LUN, csio->ccb_h.target_lun) != 0) {
|
2010-09-10 15:03:56 +00:00
|
|
|
mps_free_command(sc, cm);
|
|
|
|
ccb->ccb_h.status = CAM_LUN_INVALID;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (csio->ccb_h.flags & CAM_CDB_POINTER)
|
|
|
|
bcopy(csio->cdb_io.cdb_ptr, &req->CDB.CDB32[0], csio->cdb_len);
|
|
|
|
else
|
|
|
|
bcopy(csio->cdb_io.cdb_bytes, &req->CDB.CDB32[0],csio->cdb_len);
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->IoFlags = htole16(csio->cdb_len);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
/*
|
2012-01-26 18:17:21 +00:00
|
|
|
* Check if EEDP is supported and enabled. If it is then check if the
|
|
|
|
* SCSI opcode could be using EEDP. If so, make sure the LUN exists and
|
|
|
|
* is formatted for EEDP support. If all of this is true, set CDB up
|
|
|
|
* for EEDP transfer.
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
*/
|
2012-01-26 18:17:21 +00:00
|
|
|
eedp_flags = op_code_prot[req->CDB.CDB32[0]];
|
|
|
|
if (sc->eedp_enabled && eedp_flags) {
|
|
|
|
SLIST_FOREACH(lun, &targ->luns, lun_link) {
|
|
|
|
if (lun->lun_id == csio->ccb_h.target_lun) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((lun != NULL) && (lun->eedp_formatted)) {
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->EEDPBlockSize = htole16(lun->eedp_block_size);
|
2012-01-26 18:17:21 +00:00
|
|
|
eedp_flags |= (MPI2_SCSIIO_EEDPFLAGS_INC_PRI_REFTAG |
|
|
|
|
MPI2_SCSIIO_EEDPFLAGS_CHECK_REFTAG |
|
|
|
|
MPI2_SCSIIO_EEDPFLAGS_CHECK_GUARD);
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->EEDPFlags = htole16(eedp_flags);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If CDB less than 32, fill in Primary Ref Tag with
|
|
|
|
* low 4 bytes of LBA. If CDB is 32, tag stuff is
|
|
|
|
* already there. Also, set protection bit. FreeBSD
|
|
|
|
* currently does not support CDBs bigger than 16, but
|
|
|
|
* the code doesn't hurt, and will be here for the
|
|
|
|
* future.
|
|
|
|
*/
|
|
|
|
if (csio->cdb_len != 32) {
|
|
|
|
lba_byte = (csio->cdb_len == 16) ? 6 : 2;
|
|
|
|
ref_tag_addr = (uint8_t *)&req->CDB.EEDP32.
|
|
|
|
PrimaryReferenceTag;
|
|
|
|
for (i = 0; i < 4; i++) {
|
|
|
|
*ref_tag_addr =
|
|
|
|
req->CDB.CDB32[lba_byte + i];
|
|
|
|
ref_tag_addr++;
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->CDB.EEDP32.PrimaryReferenceTag =
|
|
|
|
htole32(req->CDB.EEDP32.PrimaryReferenceTag);
|
2012-01-26 18:17:21 +00:00
|
|
|
req->CDB.EEDP32.PrimaryApplicationTagMask =
|
|
|
|
0xFFFF;
|
|
|
|
req->CDB.CDB32[1] = (req->CDB.CDB32[1] & 0x1F) |
|
|
|
|
0x20;
|
|
|
|
} else {
|
|
|
|
eedp_flags |=
|
|
|
|
MPI2_SCSIIO_EEDPFLAGS_INC_PRI_APPTAG;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->EEDPFlags = htole16(eedp_flags);
|
2012-01-26 18:17:21 +00:00
|
|
|
req->CDB.CDB32[10] = (req->CDB.CDB32[10] &
|
|
|
|
0x1F) | 0x20;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_length = csio->dxfer_len;
|
2013-02-12 16:57:20 +00:00
|
|
|
if (cm->cm_length != 0) {
|
|
|
|
cm->cm_data = ccb;
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_USE_CCB;
|
|
|
|
} else {
|
|
|
|
cm->cm_data = NULL;
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_sge = &req->SGL;
|
|
|
|
cm->cm_sglsize = (32 - 24) * 4;
|
|
|
|
cm->cm_desc.SCSIIO.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_SCSI_IO;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
cm->cm_desc.SCSIIO.DevHandle = htole16(targ->handle);
|
2010-09-10 15:03:56 +00:00
|
|
|
cm->cm_complete = mpssas_scsiio_complete;
|
|
|
|
cm->cm_complete_data = ccb;
|
|
|
|
cm->cm_targ = targ;
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_lun = csio->ccb_h.target_lun;
|
|
|
|
cm->cm_ccb = ccb;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* If HBA is a WD and the command is not for a retry, try to build a
|
|
|
|
* direct I/O message. If failed, or the command is for a retry, send
|
|
|
|
* the I/O to the IR volume itself.
|
|
|
|
*/
|
|
|
|
if (sc->WD_valid_config) {
|
|
|
|
if (ccb->ccb_h.status != MPS_WD_RETRY) {
|
|
|
|
mpssas_direct_drive_io(sassc, cm, ccb);
|
|
|
|
} else {
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INPROG;
|
|
|
|
}
|
|
|
|
}
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
callout_reset(&cm->cm_callout, (ccb->ccb_h.timeout * hz) / 1000,
|
|
|
|
mpssas_scsiio_timeout, cm);
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ->issued++;
|
|
|
|
targ->outstanding++;
|
|
|
|
TAILQ_INSERT_TAIL(&targ->commands, cm, cm_link);
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
ccb->ccb_h.status |= CAM_SIM_QUEUED;
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_XINFO, "%s cm %p ccb %p outstanding %u\n",
|
|
|
|
__func__, cm, ccb, targ->outstanding);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
mps_map_command(sc, cm);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-02-09 00:16:12 +00:00
|
|
|
static void
|
|
|
|
mps_response_code(struct mps_softc *sc, u8 response_code)
|
|
|
|
{
|
|
|
|
char *desc;
|
|
|
|
|
|
|
|
switch (response_code) {
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_TM_COMPLETE:
|
|
|
|
desc = "task management request completed";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_INVALID_FRAME:
|
|
|
|
desc = "invalid frame";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_TM_NOT_SUPPORTED:
|
|
|
|
desc = "task management request not supported";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_TM_FAILED:
|
|
|
|
desc = "task management request failed";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_TM_SUCCEEDED:
|
|
|
|
desc = "task management request succeeded";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_TM_INVALID_LUN:
|
|
|
|
desc = "invalid lun";
|
|
|
|
break;
|
|
|
|
case 0xA:
|
|
|
|
desc = "overlapped tag attempted";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSITASKMGMT_RSP_IO_QUEUED_ON_IOC:
|
|
|
|
desc = "task queued, however not sent to target";
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
desc = "unknown";
|
|
|
|
break;
|
|
|
|
}
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "response_code(0x%01x): %s\n",
|
2012-02-09 00:16:12 +00:00
|
|
|
response_code, desc);
|
|
|
|
}
|
|
|
|
/**
|
|
|
|
* mps_sc_failed_io_info - translated non-succesfull SCSI_IO request
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
mps_sc_failed_io_info(struct mps_softc *sc, struct ccb_scsiio *csio,
|
|
|
|
Mpi2SCSIIOReply_t *mpi_reply)
|
|
|
|
{
|
|
|
|
u32 response_info;
|
|
|
|
u8 *response_bytes;
|
|
|
|
u16 ioc_status = le16toh(mpi_reply->IOCStatus) &
|
|
|
|
MPI2_IOCSTATUS_MASK;
|
|
|
|
u8 scsi_state = mpi_reply->SCSIState;
|
|
|
|
u8 scsi_status = mpi_reply->SCSIStatus;
|
|
|
|
char *desc_ioc_state = NULL;
|
|
|
|
char *desc_scsi_status = NULL;
|
|
|
|
char *desc_scsi_state = sc->tmp_string;
|
|
|
|
u32 log_info = le32toh(mpi_reply->IOCLogInfo);
|
|
|
|
|
|
|
|
if (log_info == 0x31170000)
|
|
|
|
return;
|
|
|
|
|
|
|
|
switch (ioc_status) {
|
|
|
|
case MPI2_IOCSTATUS_SUCCESS:
|
|
|
|
desc_ioc_state = "success";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_INVALID_FUNCTION:
|
|
|
|
desc_ioc_state = "invalid function";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR:
|
|
|
|
desc_ioc_state = "scsi recovered error";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_INVALID_DEVHANDLE:
|
|
|
|
desc_ioc_state = "scsi invalid dev handle";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_DEVICE_NOT_THERE:
|
|
|
|
desc_ioc_state = "scsi device not there";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_DATA_OVERRUN:
|
|
|
|
desc_ioc_state = "scsi data overrun";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_DATA_UNDERRUN:
|
|
|
|
desc_ioc_state = "scsi data underrun";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_IO_DATA_ERROR:
|
|
|
|
desc_ioc_state = "scsi io data error";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_PROTOCOL_ERROR:
|
|
|
|
desc_ioc_state = "scsi protocol error";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_TASK_TERMINATED:
|
|
|
|
desc_ioc_state = "scsi task terminated";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_RESIDUAL_MISMATCH:
|
|
|
|
desc_ioc_state = "scsi residual mismatch";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_TASK_MGMT_FAILED:
|
|
|
|
desc_ioc_state = "scsi task mgmt failed";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_IOC_TERMINATED:
|
|
|
|
desc_ioc_state = "scsi ioc terminated";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_EXT_TERMINATED:
|
|
|
|
desc_ioc_state = "scsi ext terminated";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_EEDP_GUARD_ERROR:
|
|
|
|
desc_ioc_state = "eedp guard error";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_EEDP_REF_TAG_ERROR:
|
|
|
|
desc_ioc_state = "eedp ref tag error";
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_EEDP_APP_TAG_ERROR:
|
|
|
|
desc_ioc_state = "eedp app tag error";
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
desc_ioc_state = "unknown";
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (scsi_status) {
|
|
|
|
case MPI2_SCSI_STATUS_GOOD:
|
|
|
|
desc_scsi_status = "good";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_CHECK_CONDITION:
|
|
|
|
desc_scsi_status = "check condition";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_CONDITION_MET:
|
|
|
|
desc_scsi_status = "condition met";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_BUSY:
|
|
|
|
desc_scsi_status = "busy";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_INTERMEDIATE:
|
|
|
|
desc_scsi_status = "intermediate";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_INTERMEDIATE_CONDMET:
|
|
|
|
desc_scsi_status = "intermediate condmet";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_RESERVATION_CONFLICT:
|
|
|
|
desc_scsi_status = "reservation conflict";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_COMMAND_TERMINATED:
|
|
|
|
desc_scsi_status = "command terminated";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_TASK_SET_FULL:
|
|
|
|
desc_scsi_status = "task set full";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_ACA_ACTIVE:
|
|
|
|
desc_scsi_status = "aca active";
|
|
|
|
break;
|
|
|
|
case MPI2_SCSI_STATUS_TASK_ABORTED:
|
|
|
|
desc_scsi_status = "task aborted";
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
desc_scsi_status = "unknown";
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
desc_scsi_state[0] = '\0';
|
|
|
|
if (!scsi_state)
|
|
|
|
desc_scsi_state = " ";
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_RESPONSE_INFO_VALID)
|
|
|
|
strcat(desc_scsi_state, "response info ");
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_TERMINATED)
|
|
|
|
strcat(desc_scsi_state, "state terminated ");
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_NO_SCSI_STATUS)
|
|
|
|
strcat(desc_scsi_state, "no status ");
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_AUTOSENSE_FAILED)
|
|
|
|
strcat(desc_scsi_state, "autosense failed ");
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_AUTOSENSE_VALID)
|
|
|
|
strcat(desc_scsi_state, "autosense valid ");
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "\thandle(0x%04x), ioc_status(%s)(0x%04x)\n",
|
|
|
|
le16toh(mpi_reply->DevHandle), desc_ioc_state, ioc_status);
|
2012-02-09 00:16:12 +00:00
|
|
|
/* We can add more detail about underflow data here
|
|
|
|
* TO-DO
|
|
|
|
* */
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "\tscsi_status(%s)(0x%02x), "
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
"scsi_state(%s)(0x%02x)\n", desc_scsi_status, scsi_status,
|
|
|
|
desc_scsi_state, scsi_state);
|
2012-02-09 00:16:12 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
if (sc->mps_debug & MPS_XINFO &&
|
2012-02-09 00:16:12 +00:00
|
|
|
scsi_state & MPI2_SCSI_STATE_AUTOSENSE_VALID) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "-> Sense Buffer Data : Start :\n");
|
2012-02-09 00:16:12 +00:00
|
|
|
scsi_sense_print(csio);
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "-> Sense Buffer Data : End :\n");
|
2012-02-09 00:16:12 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (scsi_state & MPI2_SCSI_STATE_RESPONSE_INFO_VALID) {
|
|
|
|
response_info = le32toh(mpi_reply->ResponseInfo);
|
|
|
|
response_bytes = (u8 *)&response_info;
|
|
|
|
mps_response_code(sc,response_bytes[0]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
static void
|
|
|
|
mpssas_scsiio_complete(struct mps_softc *sc, struct mps_command *cm)
|
|
|
|
{
|
|
|
|
MPI2_SCSI_IO_REPLY *rep;
|
|
|
|
union ccb *ccb;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct ccb_scsiio *csio;
|
2010-09-10 15:03:56 +00:00
|
|
|
struct mpssas_softc *sassc;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct scsi_vpd_supported_page_list *vpd_list = NULL;
|
|
|
|
u8 *TLR_bits, TLR_on;
|
|
|
|
int dir = 0, i;
|
|
|
|
u16 alloc_len;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_dprint(sc, MPS_TRACE,
|
2013-07-19 00:12:41 +00:00
|
|
|
"cm %p SMID %u ccb %p reply %p outstanding %u\n", cm,
|
|
|
|
cm->cm_desc.Default.SMID, cm->cm_ccb, cm->cm_reply,
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_targ->outstanding);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
callout_stop(&cm->cm_callout);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
|
|
|
sassc = sc->sassc;
|
|
|
|
ccb = cm->cm_complete_data;
|
2012-01-26 18:17:21 +00:00
|
|
|
csio = &ccb->csio;
|
2010-09-10 15:03:56 +00:00
|
|
|
rep = (MPI2_SCSI_IO_REPLY *)cm->cm_reply;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
/*
|
|
|
|
* XXX KDM if the chain allocation fails, does it matter if we do
|
|
|
|
* the sync and unload here? It is simpler to do it in every case,
|
|
|
|
* assuming it doesn't cause problems.
|
|
|
|
*/
|
2010-09-10 15:03:56 +00:00
|
|
|
if (cm->cm_data != NULL) {
|
|
|
|
if (cm->cm_flags & MPS_CM_FLAGS_DATAIN)
|
|
|
|
dir = BUS_DMASYNC_POSTREAD;
|
|
|
|
else if (cm->cm_flags & MPS_CM_FLAGS_DATAOUT)
|
2012-10-22 03:00:37 +00:00
|
|
|
dir = BUS_DMASYNC_POSTWRITE;
|
2010-09-10 15:03:56 +00:00
|
|
|
bus_dmamap_sync(sc->buffer_dmat, cm->cm_dmamap, dir);
|
|
|
|
bus_dmamap_unload(sc->buffer_dmat, cm->cm_dmamap);
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_targ->completed++;
|
|
|
|
cm->cm_targ->outstanding--;
|
|
|
|
TAILQ_REMOVE(&cm->cm_targ->commands, cm, cm_link);
|
Fix mps(4) driver breakage that came in in change 253550 that
manifested itself in out of chain frame conditions.
When the driver ran out of chain frames, the request in question
would get completed early, and go through mpssas_scsiio_complete().
In mpssas_scsiio_complete(), the negation of the CAM status values
(CAM_STATUS_MASK | CAM_SIM_QUEUED) was ORed in instead of being
ANDed in. This resulted in a bogus CAM CCB status value. This
didn't show up in the non-error case, because the status was reset
to something valid (e.g. CAM_REQ_CMP) later on in the function.
But in the error case, such as when the driver ran out of chain
frames, the CAM_REQUEUE_REQ status was ORed in to the bogus status
value. This led to the CAM transport layer repeatedly releasing
the SIM queue, because it though that the CAM_RELEASE_SIMQ flag had
been set. The symptom was messages like this on the console when
INVARIANTS were enabled:
xpt_release_simq: requested 1 > present 0
xpt_release_simq: requested 1 > present 0
xpt_release_simq: requested 1 > present 0
mps_sas.c: In mpssas_scsiio_complete(), use &= to take status
bits out. |= adds them in.
In the error case in mpssas_scsiio_complete(), set
the status to CAM_REQUEUE_REQ, don't OR it in.
MFC after: 3 days
Sponsored by: Spectra Logic
2013-08-21 21:30:56 +00:00
|
|
|
ccb->ccb_h.status &= ~(CAM_STATUS_MASK | CAM_SIM_QUEUED);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
if (cm->cm_state == MPS_CM_STATE_TIMEDOUT) {
|
|
|
|
TAILQ_REMOVE(&cm->cm_targ->timedout_commands, cm, cm_recovery);
|
|
|
|
if (cm->cm_reply != NULL)
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completed timedout cm %p ccb %p during recovery "
|
|
|
|
"ioc %x scsi %x state %x xfer %u\n",
|
|
|
|
cm, cm->cm_ccb,
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(rep->IOCStatus), rep->SCSIStatus, rep->SCSIState,
|
|
|
|
le32toh(rep->TransferCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
else
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completed timedout cm %p ccb %p during recovery\n",
|
|
|
|
cm, cm->cm_ccb);
|
|
|
|
} else if (cm->cm_targ->tm != NULL) {
|
|
|
|
if (cm->cm_reply != NULL)
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completed cm %p ccb %p during recovery "
|
|
|
|
"ioc %x scsi %x state %x xfer %u\n",
|
|
|
|
cm, cm->cm_ccb,
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(rep->IOCStatus), rep->SCSIStatus, rep->SCSIState,
|
|
|
|
le32toh(rep->TransferCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
else
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completed cm %p ccb %p during recovery\n",
|
|
|
|
cm, cm->cm_ccb);
|
|
|
|
} else if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_RECOVERY,
|
2012-01-26 18:17:21 +00:00
|
|
|
"reset completed cm %p ccb %p\n",
|
|
|
|
cm, cm->cm_ccb);
|
|
|
|
}
|
|
|
|
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
|
|
|
/*
|
|
|
|
* We ran into an error after we tried to map the command,
|
|
|
|
* so we're getting a callback without queueing the command
|
|
|
|
* to the hardware. So we set the status here, and it will
|
|
|
|
* be retained below. We'll go through the "fast path",
|
|
|
|
* because there can be no reply when we haven't actually
|
|
|
|
* gone out to the hardware.
|
|
|
|
*/
|
Fix mps(4) driver breakage that came in in change 253550 that
manifested itself in out of chain frame conditions.
When the driver ran out of chain frames, the request in question
would get completed early, and go through mpssas_scsiio_complete().
In mpssas_scsiio_complete(), the negation of the CAM status values
(CAM_STATUS_MASK | CAM_SIM_QUEUED) was ORed in instead of being
ANDed in. This resulted in a bogus CAM CCB status value. This
didn't show up in the non-error case, because the status was reset
to something valid (e.g. CAM_REQ_CMP) later on in the function.
But in the error case, such as when the driver ran out of chain
frames, the CAM_REQUEUE_REQ status was ORed in to the bogus status
value. This led to the CAM transport layer repeatedly releasing
the SIM queue, because it though that the CAM_RELEASE_SIMQ flag had
been set. The symptom was messages like this on the console when
INVARIANTS were enabled:
xpt_release_simq: requested 1 > present 0
xpt_release_simq: requested 1 > present 0
xpt_release_simq: requested 1 > present 0
mps_sas.c: In mpssas_scsiio_complete(), use &= to take status
bits out. |= adds them in.
In the error case in mpssas_scsiio_complete(), set
the status to CAM_REQUEUE_REQ, don't OR it in.
MFC after: 3 days
Sponsored by: Spectra Logic
2013-08-21 21:30:56 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQUEUE_REQ;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Currently the only error included in the mask is
|
|
|
|
* MPS_CM_FLAGS_CHAIN_FAILED, which means we're out of
|
|
|
|
* chain frames. We need to freeze the queue until we get
|
|
|
|
* a command that completed without this error, which will
|
|
|
|
* hopefully have some chain frames attached that we can
|
|
|
|
* use. If we wanted to get smarter about it, we would
|
|
|
|
* only unfreeze the queue in this condition when we're
|
|
|
|
* sure that we're getting some chain frames back. That's
|
|
|
|
* probably unnecessary.
|
|
|
|
*/
|
|
|
|
if ((sassc->flags & MPSSAS_QUEUE_FROZEN) == 0) {
|
|
|
|
xpt_freeze_simq(sassc->sim, 1);
|
|
|
|
sassc->flags |= MPSSAS_QUEUE_FROZEN;
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "Error sending command, "
|
2011-02-25 17:30:25 +00:00
|
|
|
"freezing SIM queue\n");
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Take the fast path to completion */
|
|
|
|
if (cm->cm_reply == NULL) {
|
2011-02-18 17:01:57 +00:00
|
|
|
if ((ccb->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_INPROG) {
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((sc->mps_flags & MPS_FLAGS_DIAGRESET) != 0)
|
|
|
|
ccb->ccb_h.status = CAM_SCSI_BUS_RESET;
|
|
|
|
else {
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
ccb->csio.scsi_status = SCSI_STATUS_OK;
|
|
|
|
}
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
if (sassc->flags & MPSSAS_QUEUE_FROZEN) {
|
|
|
|
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
|
|
|
|
sassc->flags &= ~MPSSAS_QUEUE_FROZEN;
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
"Unfreezing SIM queue\n");
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* There are two scenarios where the status won't be
|
|
|
|
* CAM_REQ_CMP. The first is if MPS_CM_FLAGS_ERROR_MASK is
|
|
|
|
* set, the second is in the MPS_FLAGS_DIAGRESET above.
|
|
|
|
*/
|
|
|
|
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
|
|
|
|
/*
|
|
|
|
* Freeze the dev queue so that commands are
|
|
|
|
* executed in the correct order with after error
|
|
|
|
* recovery.
|
|
|
|
*/
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
ccb->ccb_h.status |= CAM_DEV_QFRZN;
|
|
|
|
xpt_freeze_devq(ccb->ccb_h.path, /*count*/ 1);
|
2011-02-18 17:01:57 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
mps_free_command(sc, cm);
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_XINFO,
|
|
|
|
"ioc %x scsi %x state %x xfer %u\n",
|
|
|
|
le16toh(rep->IOCStatus), rep->SCSIStatus, rep->SCSIState,
|
|
|
|
le32toh(rep->TransferCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is a Direct Drive I/O, reissue the I/O to the original IR
|
|
|
|
* Volume if an error occurred (normal I/O retry). Use the original
|
|
|
|
* CCB, but set a flag that this will be a retry so that it's sent to
|
|
|
|
* the original volume. Free the command but reuse the CCB.
|
|
|
|
*/
|
|
|
|
if (cm->cm_flags & MPS_CM_FLAGS_DD_IO) {
|
|
|
|
mps_free_command(sc, cm);
|
|
|
|
ccb->ccb_h.status = MPS_WD_RETRY;
|
|
|
|
mpssas_action_scsiio(sassc, ccb);
|
|
|
|
return;
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
switch (le16toh(rep->IOCStatus) & MPI2_IOCSTATUS_MASK) {
|
2010-09-10 15:03:56 +00:00
|
|
|
case MPI2_IOCSTATUS_SCSI_DATA_UNDERRUN:
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
csio->resid = cm->cm_length - le32toh(rep->TransferCount);
|
2010-09-10 15:03:56 +00:00
|
|
|
/* FALLTHROUGH */
|
|
|
|
case MPI2_IOCSTATUS_SUCCESS:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR:
|
2012-01-26 18:17:21 +00:00
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if ((le16toh(rep->IOCStatus) & MPI2_IOCSTATUS_MASK) ==
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_IOCSTATUS_SCSI_RECOVERED_ERROR)
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_XINFO, "recovered error\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/* Completion failed at the transport level. */
|
|
|
|
if (rep->SCSIState & (MPI2_SCSI_STATE_NO_SCSI_STATUS |
|
|
|
|
MPI2_SCSI_STATE_TERMINATED)) {
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* In a modern packetized environment, an autosense failure
|
|
|
|
* implies that there's not much else that can be done to
|
|
|
|
* recover the command.
|
|
|
|
*/
|
|
|
|
if (rep->SCSIState & MPI2_SCSI_STATE_AUTOSENSE_FAILED) {
|
|
|
|
ccb->ccb_h.status = CAM_AUTOSENSE_FAIL;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* CAM doesn't care about SAS Response Info data, but if this is
|
|
|
|
* the state check if TLR should be done. If not, clear the
|
|
|
|
* TLR_bits for the target.
|
|
|
|
*/
|
|
|
|
if ((rep->SCSIState & MPI2_SCSI_STATE_RESPONSE_INFO_VALID) &&
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
((le32toh(rep->ResponseInfo) & MPI2_SCSI_RI_MASK_REASONCODE) ==
|
2012-01-26 18:17:21 +00:00
|
|
|
MPS_SCSI_RI_INVALID_FRAME)) {
|
|
|
|
sc->mapping_table[csio->ccb_h.target_id].TLR_bits =
|
|
|
|
(u8)MPI2_SCSIIO_CONTROL_NO_TLR;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Intentionally override the normal SCSI status reporting
|
|
|
|
* for these two cases. These are likely to happen in a
|
|
|
|
* multi-initiator environment, and we want to make sure that
|
|
|
|
* CAM retries these commands rather than fail them.
|
|
|
|
*/
|
|
|
|
if ((rep->SCSIStatus == MPI2_SCSI_STATUS_COMMAND_TERMINATED) ||
|
|
|
|
(rep->SCSIStatus == MPI2_SCSI_STATUS_TASK_ABORTED)) {
|
|
|
|
ccb->ccb_h.status = CAM_REQ_ABORTED;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Handle normal status and sense */
|
|
|
|
csio->scsi_status = rep->SCSIStatus;
|
|
|
|
if (rep->SCSIStatus == MPI2_SCSI_STATUS_GOOD)
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
else
|
|
|
|
ccb->ccb_h.status = CAM_SCSI_STATUS_ERROR;
|
|
|
|
|
|
|
|
if (rep->SCSIState & MPI2_SCSI_STATE_AUTOSENSE_VALID) {
|
|
|
|
int sense_len, returned_sense_len;
|
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
returned_sense_len = min(le32toh(rep->SenseCount),
|
2012-01-26 18:17:21 +00:00
|
|
|
sizeof(struct scsi_sense_data));
|
|
|
|
if (returned_sense_len < ccb->csio.sense_len)
|
|
|
|
ccb->csio.sense_resid = ccb->csio.sense_len -
|
|
|
|
returned_sense_len;
|
|
|
|
else
|
|
|
|
ccb->csio.sense_resid = 0;
|
|
|
|
|
|
|
|
sense_len = min(returned_sense_len,
|
|
|
|
ccb->csio.sense_len - ccb->csio.sense_resid);
|
|
|
|
bzero(&ccb->csio.sense_data,
|
2012-06-25 05:41:16 +00:00
|
|
|
sizeof(ccb->csio.sense_data));
|
2012-01-26 18:17:21 +00:00
|
|
|
bcopy(cm->cm_sense, &ccb->csio.sense_data, sense_len);
|
|
|
|
ccb->ccb_h.status |= CAM_AUTOSNS_VALID;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if this is an INQUIRY command. If it's a VPD inquiry,
|
|
|
|
* and it's page code 0 (Supported Page List), and there is
|
|
|
|
* inquiry data, and this is for a sequential access device, and
|
|
|
|
* the device is an SSP target, and TLR is supported by the
|
|
|
|
* controller, turn the TLR_bits value ON if page 0x90 is
|
|
|
|
* supported.
|
|
|
|
*/
|
|
|
|
if ((csio->cdb_io.cdb_bytes[0] == INQUIRY) &&
|
|
|
|
(csio->cdb_io.cdb_bytes[1] & SI_EVPD) &&
|
|
|
|
(csio->cdb_io.cdb_bytes[2] == SVPD_SUPPORTED_PAGE_LIST) &&
|
2013-03-28 11:24:30 +00:00
|
|
|
((csio->ccb_h.flags & CAM_DATA_MASK) == CAM_DATA_VADDR) &&
|
2014-05-08 20:28:22 +00:00
|
|
|
(csio->data_ptr != NULL) &&
|
|
|
|
((csio->data_ptr[0] & 0x1f) == T_SEQUENTIAL) &&
|
|
|
|
(sc->control_TLR) &&
|
2012-01-26 18:17:21 +00:00
|
|
|
(sc->mapping_table[csio->ccb_h.target_id].device_info &
|
|
|
|
MPI2_SAS_DEVICE_INFO_SSP_TARGET)) {
|
|
|
|
vpd_list = (struct scsi_vpd_supported_page_list *)
|
|
|
|
csio->data_ptr;
|
|
|
|
TLR_bits = &sc->mapping_table[csio->ccb_h.target_id].
|
|
|
|
TLR_bits;
|
|
|
|
*TLR_bits = (u8)MPI2_SCSIIO_CONTROL_NO_TLR;
|
|
|
|
TLR_on = (u8)MPI2_SCSIIO_CONTROL_TLR_ON;
|
|
|
|
alloc_len = ((u16)csio->cdb_io.cdb_bytes[3] << 8) +
|
|
|
|
csio->cdb_io.cdb_bytes[4];
|
2014-05-08 20:28:22 +00:00
|
|
|
alloc_len -= csio->resid;
|
2012-01-26 18:17:21 +00:00
|
|
|
for (i = 0; i < MIN(vpd_list->length, alloc_len); i++) {
|
|
|
|
if (vpd_list->list[i] == 0x90) {
|
|
|
|
*TLR_bits = TLR_on;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_INVALID_DEVHANDLE:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_DEVICE_NOT_THERE:
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* If devinfo is 0 this will be a volume. In that case don't
|
|
|
|
* tell CAM that the volume is not there. We want volumes to
|
|
|
|
* be enumerated until they are deleted/removed, not just
|
|
|
|
* failed.
|
|
|
|
*/
|
|
|
|
if (cm->cm_targ->devinfo == 0)
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
else
|
|
|
|
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
|
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_INVALID_SGL:
|
|
|
|
mps_print_scsiio_cmd(sc, cm);
|
|
|
|
ccb->ccb_h.status = CAM_UNREC_HBA_ERROR;
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_SCSI_TASK_TERMINATED:
|
|
|
|
/*
|
|
|
|
* This is one of the responses that comes back when an I/O
|
|
|
|
* has been aborted. If it is because of a timeout that we
|
|
|
|
* initiated, just set the status to CAM_CMD_TIMEOUT.
|
|
|
|
* Otherwise set it to CAM_REQ_ABORTED. The effect on the
|
|
|
|
* command is the same (it gets retried, subject to the
|
|
|
|
* retry counter), the only difference is what gets printed
|
|
|
|
* on the console.
|
|
|
|
*/
|
|
|
|
if (cm->cm_state == MPS_CM_STATE_TIMEDOUT)
|
|
|
|
ccb->ccb_h.status = CAM_CMD_TIMEOUT;
|
|
|
|
else
|
|
|
|
ccb->ccb_h.status = CAM_REQ_ABORTED;
|
|
|
|
break;
|
2012-01-26 18:17:21 +00:00
|
|
|
case MPI2_IOCSTATUS_SCSI_DATA_OVERRUN:
|
|
|
|
/* resid is ignored for this condition */
|
|
|
|
csio->resid = 0;
|
|
|
|
ccb->ccb_h.status = CAM_DATA_RUN_ERR;
|
|
|
|
break;
|
2010-09-10 15:03:56 +00:00
|
|
|
case MPI2_IOCSTATUS_SCSI_IOC_TERMINATED:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_EXT_TERMINATED:
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Since these are generally external (i.e. hopefully
|
|
|
|
* transient transport-related) errors, retry these without
|
|
|
|
* decrementing the retry count.
|
|
|
|
*/
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQUEUE_REQ;
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_INFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"terminated ioc %x scsi %x state %x xfer %u\n",
|
2013-07-19 00:12:41 +00:00
|
|
|
le16toh(rep->IOCStatus), rep->SCSIStatus, rep->SCSIState,
|
|
|
|
le32toh(rep->TransferCount));
|
2010-09-10 15:03:56 +00:00
|
|
|
break;
|
|
|
|
case MPI2_IOCSTATUS_INVALID_FUNCTION:
|
|
|
|
case MPI2_IOCSTATUS_INTERNAL_ERROR:
|
|
|
|
case MPI2_IOCSTATUS_INVALID_VPID:
|
|
|
|
case MPI2_IOCSTATUS_INVALID_FIELD:
|
|
|
|
case MPI2_IOCSTATUS_INVALID_STATE:
|
|
|
|
case MPI2_IOCSTATUS_OP_STATE_NOT_SUPPORTED:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_IO_DATA_ERROR:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_PROTOCOL_ERROR:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_RESIDUAL_MISMATCH:
|
|
|
|
case MPI2_IOCSTATUS_SCSI_TASK_MGMT_FAILED:
|
|
|
|
default:
|
2013-07-19 00:12:41 +00:00
|
|
|
mpssas_log_command(cm, MPS_XINFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"completed ioc %x scsi %x state %x xfer %u\n",
|
2013-07-19 00:12:41 +00:00
|
|
|
le16toh(rep->IOCStatus), rep->SCSIStatus, rep->SCSIState,
|
|
|
|
le32toh(rep->TransferCount));
|
2012-01-26 18:17:21 +00:00
|
|
|
csio->resid = cm->cm_length;
|
2010-09-10 15:03:56 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
2012-01-26 18:17:21 +00:00
|
|
|
break;
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
2012-02-09 00:16:12 +00:00
|
|
|
|
|
|
|
mps_sc_failed_io_info(sc,csio,rep);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if (sassc->flags & MPSSAS_QUEUE_FROZEN) {
|
|
|
|
ccb->ccb_h.status |= CAM_RELEASE_SIMQ;
|
|
|
|
sassc->flags &= ~MPSSAS_QUEUE_FROZEN;
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "Command completed, "
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
"unfreezing SIM queue\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
if ((ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
|
|
|
|
ccb->ccb_h.status |= CAM_DEV_QFRZN;
|
|
|
|
xpt_freeze_devq(ccb->ccb_h.path, /*count*/ 1);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_free_command(sc, cm);
|
|
|
|
xpt_done(ccb);
|
|
|
|
}
|
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
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
/* All Request reached here are Endian safe */
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
|
|
|
mpssas_direct_drive_io(struct mpssas_softc *sassc, struct mps_command *cm,
|
|
|
|
union ccb *ccb) {
|
|
|
|
pMpi2SCSIIORequest_t pIO_req;
|
|
|
|
struct mps_softc *sc = sassc->sc;
|
|
|
|
uint64_t virtLBA;
|
|
|
|
uint32_t physLBA, stripe_offset, stripe_unit;
|
|
|
|
uint32_t io_size, column;
|
|
|
|
uint8_t *ptrLBA, lba_idx, physLBA_byte, *CDB;
|
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
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* If this is a valid SCSI command (Read6, Read10, Read16, Write6,
|
|
|
|
* Write10, or Write16), build a direct I/O message. Otherwise, the I/O
|
|
|
|
* will be sent to the IR volume itself. Since Read6 and Write6 are a
|
|
|
|
* bit different than the 10/16 CDBs, handle them separately.
|
|
|
|
*/
|
|
|
|
pIO_req = (pMpi2SCSIIORequest_t)cm->cm_req;
|
|
|
|
CDB = pIO_req->CDB.CDB32;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Handle 6 byte CDBs.
|
|
|
|
*/
|
|
|
|
if ((pIO_req->DevHandle == sc->DD_dev_handle) && ((CDB[0] == READ_6) ||
|
|
|
|
(CDB[0] == WRITE_6))) {
|
|
|
|
/*
|
|
|
|
* Get the transfer size in blocks.
|
|
|
|
*/
|
|
|
|
io_size = (cm->cm_length >> sc->DD_block_exponent);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Get virtual LBA given in the CDB.
|
|
|
|
*/
|
|
|
|
virtLBA = ((uint64_t)(CDB[1] & 0x1F) << 16) |
|
|
|
|
((uint64_t)CDB[2] << 8) | (uint64_t)CDB[3];
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Check that LBA range for I/O does not exceed volume's
|
|
|
|
* MaxLBA.
|
|
|
|
*/
|
|
|
|
if ((virtLBA + (uint64_t)io_size - 1) <=
|
|
|
|
sc->DD_max_lba) {
|
|
|
|
/*
|
|
|
|
* Check if the I/O crosses a stripe boundary. If not,
|
|
|
|
* translate the virtual LBA to a physical LBA and set
|
|
|
|
* the DevHandle for the PhysDisk to be used. If it
|
|
|
|
* does cross a boundry, do normal I/O. To get the
|
|
|
|
* right DevHandle to use, get the map number for the
|
|
|
|
* column, then use that map number to look up the
|
|
|
|
* DevHandle of the PhysDisk.
|
|
|
|
*/
|
|
|
|
stripe_offset = (uint32_t)virtLBA &
|
|
|
|
(sc->DD_stripe_size - 1);
|
|
|
|
if ((stripe_offset + io_size) <= sc->DD_stripe_size) {
|
|
|
|
physLBA = (uint32_t)virtLBA >>
|
|
|
|
sc->DD_stripe_exponent;
|
|
|
|
stripe_unit = physLBA / sc->DD_num_phys_disks;
|
|
|
|
column = physLBA % sc->DD_num_phys_disks;
|
|
|
|
pIO_req->DevHandle =
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
htole16(sc->DD_column_map[column].dev_handle);
|
|
|
|
/* ???? Is this endian safe*/
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_desc.SCSIIO.DevHandle =
|
|
|
|
pIO_req->DevHandle;
|
|
|
|
|
|
|
|
physLBA = (stripe_unit <<
|
|
|
|
sc->DD_stripe_exponent) + stripe_offset;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[1];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 16);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[2];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 8);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[3];
|
|
|
|
physLBA_byte = (uint8_t)physLBA;
|
|
|
|
*ptrLBA = physLBA_byte;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Set flag that Direct Drive I/O is
|
|
|
|
* being done.
|
|
|
|
*/
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
/*
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
* Handle 10, 12 or 16 byte CDBs.
|
2012-01-26 18:17:21 +00:00
|
|
|
*/
|
|
|
|
if ((pIO_req->DevHandle == sc->DD_dev_handle) && ((CDB[0] == READ_10) ||
|
|
|
|
(CDB[0] == WRITE_10) || (CDB[0] == READ_16) ||
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
(CDB[0] == WRITE_16) || (CDB[0] == READ_12) ||
|
|
|
|
(CDB[0] == WRITE_12))) {
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* For 16-byte CDB's, verify that the upper 4 bytes of the CDB
|
|
|
|
* are 0. If not, this is accessing beyond 2TB so handle it in
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
* the else section. 10-byte and 12-byte CDB's are OK.
|
|
|
|
* FreeBSD sends very rare 12 byte READ/WRITE, but driver is
|
|
|
|
* ready to accept 12byte CDB for Direct IOs.
|
2012-01-26 18:17:21 +00:00
|
|
|
*/
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if ((CDB[0] == READ_10 || CDB[0] == WRITE_10) ||
|
|
|
|
(CDB[0] == READ_12 || CDB[0] == WRITE_12) ||
|
2012-01-26 18:17:21 +00:00
|
|
|
!(CDB[2] | CDB[3] | CDB[4] | CDB[5])) {
|
|
|
|
/*
|
|
|
|
* Get the transfer size in blocks.
|
|
|
|
*/
|
|
|
|
io_size = (cm->cm_length >> sc->DD_block_exponent);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get virtual LBA. Point to correct lower 4 bytes of
|
|
|
|
* LBA in the CDB depending on command.
|
|
|
|
*/
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
lba_idx = ((CDB[0] == READ_12) ||
|
|
|
|
(CDB[0] == WRITE_12) ||
|
|
|
|
(CDB[0] == READ_10) ||
|
|
|
|
(CDB[0] == WRITE_10))? 2 : 6;
|
2012-01-26 18:17:21 +00:00
|
|
|
virtLBA = ((uint64_t)CDB[lba_idx] << 24) |
|
|
|
|
((uint64_t)CDB[lba_idx + 1] << 16) |
|
|
|
|
((uint64_t)CDB[lba_idx + 2] << 8) |
|
|
|
|
(uint64_t)CDB[lba_idx + 3];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that LBA range for I/O does not exceed volume's
|
|
|
|
* MaxLBA.
|
|
|
|
*/
|
|
|
|
if ((virtLBA + (uint64_t)io_size - 1) <=
|
|
|
|
sc->DD_max_lba) {
|
|
|
|
/*
|
|
|
|
* Check if the I/O crosses a stripe boundary.
|
|
|
|
* If not, translate the virtual LBA to a
|
|
|
|
* physical LBA and set the DevHandle for the
|
|
|
|
* PhysDisk to be used. If it does cross a
|
|
|
|
* boundry, do normal I/O. To get the right
|
|
|
|
* DevHandle to use, get the map number for the
|
|
|
|
* column, then use that map number to look up
|
|
|
|
* the DevHandle of the PhysDisk.
|
|
|
|
*/
|
|
|
|
stripe_offset = (uint32_t)virtLBA &
|
|
|
|
(sc->DD_stripe_size - 1);
|
|
|
|
if ((stripe_offset + io_size) <=
|
|
|
|
sc->DD_stripe_size) {
|
|
|
|
physLBA = (uint32_t)virtLBA >>
|
|
|
|
sc->DD_stripe_exponent;
|
|
|
|
stripe_unit = physLBA /
|
|
|
|
sc->DD_num_phys_disks;
|
|
|
|
column = physLBA %
|
|
|
|
sc->DD_num_phys_disks;
|
|
|
|
pIO_req->DevHandle =
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
htole16(sc->DD_column_map[column].
|
|
|
|
dev_handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_desc.SCSIIO.DevHandle =
|
|
|
|
pIO_req->DevHandle;
|
|
|
|
|
|
|
|
physLBA = (stripe_unit <<
|
|
|
|
sc->DD_stripe_exponent) +
|
|
|
|
stripe_offset;
|
|
|
|
ptrLBA =
|
|
|
|
&pIO_req->CDB.CDB32[lba_idx];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 24);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA =
|
|
|
|
&pIO_req->CDB.CDB32[lba_idx + 1];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 16);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA =
|
|
|
|
&pIO_req->CDB.CDB32[lba_idx + 2];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 8);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA =
|
|
|
|
&pIO_req->CDB.CDB32[lba_idx + 3];
|
|
|
|
physLBA_byte = (uint8_t)physLBA;
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set flag that Direct Drive I/O is
|
|
|
|
* being done.
|
|
|
|
*/
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* 16-byte CDB and the upper 4 bytes of the CDB are not
|
|
|
|
* 0. Get the transfer size in blocks.
|
|
|
|
*/
|
|
|
|
io_size = (cm->cm_length >> sc->DD_block_exponent);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get virtual LBA.
|
|
|
|
*/
|
|
|
|
virtLBA = ((uint64_t)CDB[2] << 54) |
|
|
|
|
((uint64_t)CDB[3] << 48) |
|
|
|
|
((uint64_t)CDB[4] << 40) |
|
|
|
|
((uint64_t)CDB[5] << 32) |
|
|
|
|
((uint64_t)CDB[6] << 24) |
|
|
|
|
((uint64_t)CDB[7] << 16) |
|
|
|
|
((uint64_t)CDB[8] << 8) |
|
|
|
|
(uint64_t)CDB[9];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that LBA range for I/O does not exceed volume's
|
|
|
|
* MaxLBA.
|
|
|
|
*/
|
|
|
|
if ((virtLBA + (uint64_t)io_size - 1) <=
|
|
|
|
sc->DD_max_lba) {
|
|
|
|
/*
|
|
|
|
* Check if the I/O crosses a stripe boundary.
|
|
|
|
* If not, translate the virtual LBA to a
|
|
|
|
* physical LBA and set the DevHandle for the
|
|
|
|
* PhysDisk to be used. If it does cross a
|
|
|
|
* boundry, do normal I/O. To get the right
|
|
|
|
* DevHandle to use, get the map number for the
|
|
|
|
* column, then use that map number to look up
|
|
|
|
* the DevHandle of the PhysDisk.
|
|
|
|
*/
|
|
|
|
stripe_offset = (uint32_t)virtLBA &
|
|
|
|
(sc->DD_stripe_size - 1);
|
|
|
|
if ((stripe_offset + io_size) <=
|
|
|
|
sc->DD_stripe_size) {
|
|
|
|
physLBA = (uint32_t)(virtLBA >>
|
|
|
|
sc->DD_stripe_exponent);
|
|
|
|
stripe_unit = physLBA /
|
|
|
|
sc->DD_num_phys_disks;
|
|
|
|
column = physLBA %
|
|
|
|
sc->DD_num_phys_disks;
|
|
|
|
pIO_req->DevHandle =
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
htole16(sc->DD_column_map[column].
|
|
|
|
dev_handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
cm->cm_desc.SCSIIO.DevHandle =
|
|
|
|
pIO_req->DevHandle;
|
|
|
|
|
|
|
|
physLBA = (stripe_unit <<
|
|
|
|
sc->DD_stripe_exponent) +
|
|
|
|
stripe_offset;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set upper 4 bytes of LBA to 0. We
|
|
|
|
* assume that the phys disks are less
|
|
|
|
* than 2 TB's in size. Then, set the
|
|
|
|
* lower 4 bytes.
|
|
|
|
*/
|
|
|
|
pIO_req->CDB.CDB32[2] = 0;
|
|
|
|
pIO_req->CDB.CDB32[3] = 0;
|
|
|
|
pIO_req->CDB.CDB32[4] = 0;
|
|
|
|
pIO_req->CDB.CDB32[5] = 0;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[6];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 24);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[7];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 16);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[8];
|
|
|
|
physLBA_byte = (uint8_t)(physLBA >> 8);
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
ptrLBA = &pIO_req->CDB.CDB32[9];
|
|
|
|
physLBA_byte = (uint8_t)physLBA;
|
|
|
|
*ptrLBA = physLBA_byte;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set flag that Direct Drive I/O is
|
|
|
|
* being done.
|
|
|
|
*/
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_DD_IO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
#if __FreeBSD_version >= 900026
|
|
|
|
static void
|
|
|
|
mpssas_smpio_complete(struct mps_softc *sc, struct mps_command *cm)
|
|
|
|
{
|
|
|
|
MPI2_SMP_PASSTHROUGH_REPLY *rpl;
|
|
|
|
MPI2_SMP_PASSTHROUGH_REQUEST *req;
|
|
|
|
uint64_t sasaddr;
|
|
|
|
union ccb *ccb;
|
|
|
|
|
|
|
|
ccb = cm->cm_complete_data;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and SMP
|
|
|
|
* commands require two S/G elements only. That should be handled
|
|
|
|
* in the standard request size.
|
|
|
|
*/
|
|
|
|
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,"%s: cm_flags = %#x on SMP request!\n",
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
__func__, cm->cm_flags);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
goto bailout;
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
rpl = (MPI2_SMP_PASSTHROUGH_REPLY *)cm->cm_reply;
|
|
|
|
if (rpl == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "%s: NULL cm_reply!\n", __func__);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
|
|
|
|
req = (MPI2_SMP_PASSTHROUGH_REQUEST *)cm->cm_req;
|
|
|
|
sasaddr = le32toh(req->SASAddress.Low);
|
|
|
|
sasaddr |= ((uint64_t)(le32toh(req->SASAddress.High))) << 32;
|
|
|
|
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
if ((le16toh(rpl->IOCStatus) & MPI2_IOCSTATUS_MASK) !=
|
|
|
|
MPI2_IOCSTATUS_SUCCESS ||
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
rpl->SASStatus != MPI2_SASSTATUS_SUCCESS) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "%s: IOCStatus %04x SASStatus %02x\n",
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
__func__, le16toh(rpl->IOCStatus), rpl->SASStatus);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "%s: SMP request to SAS address "
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
"%#jx completed successfully\n", __func__,
|
|
|
|
(uintmax_t)sasaddr);
|
|
|
|
|
|
|
|
if (ccb->smpio.smp_response[2] == SMP_FR_ACCEPTED)
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
|
|
|
else
|
|
|
|
ccb->ccb_h.status = CAM_SMP_STATUS_ERROR;
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
/*
|
|
|
|
* We sync in both directions because we had DMAs in the S/G list
|
|
|
|
* in both directions.
|
|
|
|
*/
|
|
|
|
bus_dmamap_sync(sc->buffer_dmat, cm->cm_dmamap,
|
|
|
|
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
|
|
|
|
bus_dmamap_unload(sc->buffer_dmat, cm->cm_dmamap);
|
|
|
|
mps_free_command(sc, cm);
|
|
|
|
xpt_done(ccb);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_send_smpcmd(struct mpssas_softc *sassc, union ccb *ccb, uint64_t sasaddr)
|
|
|
|
{
|
|
|
|
struct mps_command *cm;
|
|
|
|
uint8_t *request, *response;
|
|
|
|
MPI2_SMP_PASSTHROUGH_REQUEST *req;
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct sglist *sg;
|
|
|
|
int error;
|
|
|
|
|
|
|
|
sc = sassc->sc;
|
|
|
|
sg = NULL;
|
|
|
|
error = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX We don't yet support physical addresses here.
|
|
|
|
*/
|
2013-02-12 16:57:20 +00:00
|
|
|
switch ((ccb->ccb_h.flags & CAM_DATA_MASK)) {
|
|
|
|
case CAM_DATA_PADDR:
|
|
|
|
case CAM_DATA_SG_PADDR:
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: physical addresses not supported\n", __func__);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
2013-02-12 16:57:20 +00:00
|
|
|
case CAM_DATA_SG:
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
/*
|
|
|
|
* The chip does not support more than one buffer for the
|
|
|
|
* request or response.
|
|
|
|
*/
|
|
|
|
if ((ccb->smpio.smp_request_sglist_cnt > 1)
|
|
|
|
|| (ccb->smpio.smp_response_sglist_cnt > 1)) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: multiple request or response "
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
"buffer segments not supported for SMP\n",
|
|
|
|
__func__);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The CAM_SCATTER_VALID flag was originally implemented
|
|
|
|
* for the XPT_SCSI_IO CCB, which only has one data pointer.
|
|
|
|
* We have two. So, just take that flag to mean that we
|
|
|
|
* might have S/G lists, and look at the S/G segment count
|
|
|
|
* to figure out whether that is the case for each individual
|
|
|
|
* buffer.
|
|
|
|
*/
|
|
|
|
if (ccb->smpio.smp_request_sglist_cnt != 0) {
|
|
|
|
bus_dma_segment_t *req_sg;
|
|
|
|
|
|
|
|
req_sg = (bus_dma_segment_t *)ccb->smpio.smp_request;
|
2012-10-02 23:04:12 +00:00
|
|
|
request = (uint8_t *)(uintptr_t)req_sg[0].ds_addr;
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
} else
|
|
|
|
request = ccb->smpio.smp_request;
|
|
|
|
|
|
|
|
if (ccb->smpio.smp_response_sglist_cnt != 0) {
|
|
|
|
bus_dma_segment_t *rsp_sg;
|
|
|
|
|
|
|
|
rsp_sg = (bus_dma_segment_t *)ccb->smpio.smp_response;
|
2012-10-02 23:04:12 +00:00
|
|
|
response = (uint8_t *)(uintptr_t)rsp_sg[0].ds_addr;
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
} else
|
|
|
|
response = ccb->smpio.smp_response;
|
2013-02-12 16:57:20 +00:00
|
|
|
break;
|
|
|
|
case CAM_DATA_VADDR:
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
request = ccb->smpio.smp_request;
|
|
|
|
response = ccb->smpio.smp_response;
|
2013-02-12 16:57:20 +00:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
cm = mps_alloc_command(sc);
|
|
|
|
if (cm == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: cannot allocate command\n", __func__);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
req = (MPI2_SMP_PASSTHROUGH_REQUEST *)cm->cm_req;
|
|
|
|
bzero(req, sizeof(*req));
|
|
|
|
req->Function = MPI2_FUNCTION_SMP_PASSTHROUGH;
|
|
|
|
|
|
|
|
/* Allow the chip to use any route to this SAS address. */
|
|
|
|
req->PhysicalPort = 0xff;
|
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->RequestDataLength = htole16(ccb->smpio.smp_request_len);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
req->SGLFlags =
|
|
|
|
MPI2_SGLFLAGS_SYSTEM_ADDRESS_SPACE | MPI2_SGLFLAGS_SGL_TYPE_MPI;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "%s: sending SMP request to SAS "
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
"address %#jx\n", __func__, (uintmax_t)sasaddr);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
|
|
|
|
mpi_init_sge(cm, req, &req->SGL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Set up a uio to pass into mps_map_command(). This allows us to
|
|
|
|
* do one map command, and one busdma call in there.
|
|
|
|
*/
|
|
|
|
cm->cm_uio.uio_iov = cm->cm_iovec;
|
|
|
|
cm->cm_uio.uio_iovcnt = 2;
|
|
|
|
cm->cm_uio.uio_segflg = UIO_SYSSPACE;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The read/write flag isn't used by busdma, but set it just in
|
|
|
|
* case. This isn't exactly accurate, either, since we're going in
|
|
|
|
* both directions.
|
|
|
|
*/
|
|
|
|
cm->cm_uio.uio_rw = UIO_WRITE;
|
|
|
|
|
|
|
|
cm->cm_iovec[0].iov_base = request;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
cm->cm_iovec[0].iov_len = le16toh(req->RequestDataLength);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
cm->cm_iovec[1].iov_base = response;
|
|
|
|
cm->cm_iovec[1].iov_len = ccb->smpio.smp_response_len;
|
|
|
|
|
|
|
|
cm->cm_uio.uio_resid = cm->cm_iovec[0].iov_len +
|
|
|
|
cm->cm_iovec[1].iov_len;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Trigger a warning message in mps_data_cb() for the user if we
|
|
|
|
* wind up exceeding two S/G segments. The chip expects one
|
|
|
|
* segment for the request and another for the response.
|
|
|
|
*/
|
|
|
|
cm->cm_max_segs = 2;
|
|
|
|
|
|
|
|
cm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
|
|
|
|
cm->cm_complete = mpssas_smpio_complete;
|
|
|
|
cm->cm_complete_data = ccb;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Tell the mapping code that we're using a uio, and that this is
|
|
|
|
* an SMP passthrough request. There is a little special-case
|
|
|
|
* logic there (in mps_data_cb()) to handle the bidirectional
|
|
|
|
* transfer.
|
|
|
|
*/
|
|
|
|
cm->cm_flags |= MPS_CM_FLAGS_USE_UIO | MPS_CM_FLAGS_SMP_PASS |
|
|
|
|
MPS_CM_FLAGS_DATAIN | MPS_CM_FLAGS_DATAOUT;
|
|
|
|
|
|
|
|
/* The chip data format is little endian. */
|
|
|
|
req->SASAddress.High = htole32(sasaddr >> 32);
|
|
|
|
req->SASAddress.Low = htole32(sasaddr);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* XXX Note that we don't have a timeout/abort mechanism here.
|
|
|
|
* From the manual, it looks like task management requests only
|
|
|
|
* work for SCSI IO and SATA passthrough requests. We may need to
|
|
|
|
* have a mechanism to retry requests in the event of a chip reset
|
|
|
|
* at least. Hopefully the chip will insure that any errors short
|
|
|
|
* of that are relayed back to the driver.
|
|
|
|
*/
|
|
|
|
error = mps_map_command(sc, cm);
|
|
|
|
if ((error != 0) && (error != EINPROGRESS)) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: error %d returned from mps_map_command()\n",
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
__func__, error);
|
|
|
|
goto bailout_error;
|
|
|
|
}
|
|
|
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
bailout_error:
|
|
|
|
mps_free_command(sc, cm);
|
|
|
|
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
|
|
|
|
{
|
|
|
|
struct mps_softc *sc;
|
|
|
|
struct mpssas_target *targ;
|
|
|
|
uint64_t sasaddr = 0;
|
|
|
|
|
|
|
|
sc = sassc->sc;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure the target exists.
|
|
|
|
*/
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(ccb->ccb_h.target_id < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in XPT_SMP_IO\n", ccb->ccb_h.target_id));
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
targ = &sassc->targets[ccb->ccb_h.target_id];
|
|
|
|
if (targ->handle == 0x0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: target %d does not exist!\n", __func__,
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
ccb->ccb_h.target_id);
|
|
|
|
ccb->ccb_h.status = CAM_SEL_TIMEOUT;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If this device has an embedded SMP target, we'll talk to it
|
|
|
|
* directly.
|
|
|
|
* figure out what the expander's address is.
|
|
|
|
*/
|
|
|
|
if ((targ->devinfo & MPI2_SAS_DEVICE_INFO_SMP_TARGET) != 0)
|
|
|
|
sasaddr = targ->sasaddr;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we don't have a SAS address for the expander yet, try
|
|
|
|
* grabbing it from the page 0x83 information cached in the
|
|
|
|
* transport layer for this target. LSI expanders report the
|
|
|
|
* expander SAS address as the port-associated SAS address in
|
|
|
|
* Inquiry VPD page 0x83. Maxim expanders don't report it in page
|
|
|
|
* 0x83.
|
|
|
|
*
|
|
|
|
* XXX KDM disable this for now, but leave it commented out so that
|
|
|
|
* it is obvious that this is another possible way to get the SAS
|
|
|
|
* address.
|
|
|
|
*
|
|
|
|
* The parent handle method below is a little more reliable, and
|
|
|
|
* the other benefit is that it works for devices other than SES
|
|
|
|
* devices. So you can send a SMP request to a da(4) device and it
|
|
|
|
* will get routed to the expander that device is attached to.
|
|
|
|
* (Assuming the da(4) device doesn't contain an SMP target...)
|
|
|
|
*/
|
|
|
|
#if 0
|
|
|
|
if (sasaddr == 0)
|
|
|
|
sasaddr = xpt_path_sas_addr(ccb->ccb_h.path);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we still don't have a SAS address for the expander, look for
|
|
|
|
* the parent device of this device, which is probably the expander.
|
|
|
|
*/
|
|
|
|
if (sasaddr == 0) {
|
2012-01-26 18:17:21 +00:00
|
|
|
#ifdef OLD_MPS_PROBE
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
struct mpssas_target *parent_target;
|
2012-01-26 18:17:21 +00:00
|
|
|
#endif
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
|
|
|
|
if (targ->parent_handle == 0x0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: handle %d does not have a valid "
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
"parent handle!\n", __func__, targ->handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
#ifdef OLD_MPS_PROBE
|
|
|
|
parent_target = mpssas_find_target_by_handle(sassc, 0,
|
|
|
|
targ->parent_handle);
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
|
|
|
|
if (parent_target == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: handle %d does not have a valid "
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
"parent target!\n", __func__, targ->handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((parent_target->devinfo &
|
|
|
|
MPI2_SAS_DEVICE_INFO_SMP_TARGET) == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: handle %d parent %d does not "
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
"have an SMP target!\n", __func__,
|
|
|
|
targ->handle, parent_target->handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
sasaddr = parent_target->sasaddr;
|
2012-01-26 18:17:21 +00:00
|
|
|
#else /* OLD_MPS_PROBE */
|
|
|
|
if ((targ->parent_devinfo &
|
|
|
|
MPI2_SAS_DEVICE_INFO_SMP_TARGET) == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: handle %d parent %d does not "
|
2012-01-26 18:17:21 +00:00
|
|
|
"have an SMP target!\n", __func__,
|
|
|
|
targ->handle, targ->parent_handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
|
|
|
|
}
|
|
|
|
if (targ->parent_sasaddr == 0x0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: handle %d parent handle %d does "
|
2012-01-26 18:17:21 +00:00
|
|
|
"not have a valid SAS address!\n",
|
|
|
|
__func__, targ->handle, targ->parent_handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
|
|
|
|
sasaddr = targ->parent_sasaddr;
|
|
|
|
#endif /* OLD_MPS_PROBE */
|
|
|
|
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (sasaddr == 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_INFO,
|
|
|
|
"%s: unable to find SAS address for handle %d\n",
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
__func__, targ->handle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_INVALID;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
mpssas_send_smpcmd(sassc, ccb, sasaddr);
|
|
|
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
bailout:
|
|
|
|
xpt_done(ccb);
|
|
|
|
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
#endif //__FreeBSD_version >= 900026
|
Add Serial Management Protocol (SMP) passthrough support to CAM.
This includes support in the kernel, camcontrol(8), libcam and the mps(4)
driver for SMP passthrough.
The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00
to determine supported pages, and will now fetch page 0x83 in addition to
page 0x80 if supported.
Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is
intended for SMP requests and responses. The ADVINFO is currently used to
fetch cached VPD page 0x83 data from the transport layer, but is intended
to be extensible to fetch other types of device-specific data.
SMP-only devices are not currently represented in the CAM topology, and so
the current semantics are that the SIM will route SMP CCBs to either the
addressed device, if it contains an SMP target, or its parent, if it
contains an SMP target. (This is noted in cam_ccb.h, since it will change
later once we have the ability to have SMP-only devices in CAM's topology.)
smp_all.c,
smp_all.h: New helper routines for SMP. This includes
SMP request building routines, response parsing
routines, error decoding routines, and structure
definitions for a number of SMP commands.
libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality
is available to userland applications.
camcontrol.8,
camcontrol.c: Add smp passthrough support to camcontrol. Several
new subcommands are now available:
'smpcmd' functions much like 'cmd', except that it
allows the user to send generic SMP commands.
'smprg' sends the SMP report general command, and
displays the decoded output. It will automatically
fetch extended output if it is available.
'smppc' sends the SMP phy control command, with any
number of potential options. Among other things,
this allows the user to reset a phy on a SAS
expander, or disable a phy on an expander.
'smpmaninfo' sends the SMP report manufacturer
information and displays the decoded output.
'smpphylist' displays a list of phys on an
expander, and the CAM devices attached to those
phys, if any.
cam.h,
cam.c: Add a status value for SMP errors
(CAM_SMP_STATUS_ERROR).
Add a missing description for CAM_SCSI_IT_NEXUS_LOST.
Add support for SMP commands to cam_error_string().
cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP
commands are by nature bi-directional, and we may
need to support bi-directional SCSI commands later.
Add the XPT_SMP_IO CCB. Since SMP commands are
bi-directional, there are pointers for both the
request and response.
Add a fill routine for SMP CCBs.
Add the XPT_GDEV_ADVINFO CCB. This is currently
used to fetch cached page 0x83 data from the
transport later, but is extensible to fetch many
other types of data.
cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO
and XPT_GDEV_ADVINFO CCBs.
cam_xpt.c: Add support for executing XPT_SMP_IO CCBs.
cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct
cam_ed.
scsi_all.c: Add scsi_get_sas_addr(), a function that parses
VPD page 0x83 data and pulls out a SAS address.
scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a
prototype for scsi_get_sas_addr().
scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and
XPT_GDEV_ADVINFO CCBs.
scsi_xpt.c: In the SCSI probe code, first ask the device for
VPD page 0x00. If any VPD pages are supported,
that page is required to be implemented. Based on
the response, we may probe for the serial number
(page 0x80) or device id (page 0x83).
Add support for the XPT_GDEV_ADVINFO CCB.
sys/conf/files: Add smp_all.c.
mps.c: Add support for passing in a uio in mps_map_command(),
so we can map a S/G list at once.
Add support for SMP passthrough commands in
mps_data_cb(). SMP is a special case, because the
first buffer in the S/G list is outbound and the
second buffer is inbound.
Add support for warning the user if the busdma code
comes back with more buffers than will work for the
command. This will, for example, help the user
determine why an SMP command failed if busdma comes
back with three buffers.
mps_pci.c: Add sys/uio.h.
mps_sas.c: Add the SAS address and the parent handle to the
list of fields we pull from device page 0 and cache
in struct mpssas_target. These are needed for SMP
passthrough.
Add support for the XPT_SMP_IO CCB. For now, this
CCB is routed to the addressed device if it supports
SMP, or to its parent if it does not and the parent
does. This is necessary because CAM does not
currently support SMP-only nodes in the topology.
Make SMP passthrough support conditional on
__FreeBSD_version >= 900026. This will make it
easier to MFC this change to the driver without
MFCing the CAM changes as well.
mps_user.c: Un-staticize mpi_init_sge() so we can use it for
the SMP passthrough code.
mpsvar.h: Add a uio and iovecs into struct mps_command for
SMP passthrough commands.
Add a cm_max_segs field to struct mps_command so
that we can warn the user if busdma comes back with
too many segments.
Clear the cm_reply when a command gets freed. If
it is not cleared, reply frames will eventually get
freed into the pool multiple times and corrupt the
pool. (This fix is from scottl.)
Add a prototype for mpi_init_sge().
sys/param.h: Bump __FreeBSD_version to 900026 for the for the
inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO
CAM CCBs.
2010-11-30 22:39:46 +00:00
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
static void
|
|
|
|
mpssas_action_resetdev(struct mpssas_softc *sassc, union ccb *ccb)
|
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
2010-09-10 15:03:56 +00:00
|
|
|
struct mps_softc *sc;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mps_command *tm;
|
2010-09-10 15:03:56 +00:00
|
|
|
struct mpssas_target *targ;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sassc->sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sassc->sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(ccb->ccb_h.target_id < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in XPT_RESET_DEV\n",
|
|
|
|
ccb->ccb_h.target_id));
|
2012-01-26 18:17:21 +00:00
|
|
|
sc = sassc->sc;
|
|
|
|
tm = mps_alloc_command(sc);
|
|
|
|
if (tm == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
2013-07-30 17:35:06 +00:00
|
|
|
"command alloc failure in mpssas_action_resetdev\n");
|
2010-09-10 15:03:56 +00:00
|
|
|
ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
|
|
|
|
xpt_done(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
targ = &sassc->targets[ccb->ccb_h.target_id];
|
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
req->DevHandle = htole16(targ->handle);
|
2010-09-10 15:03:56 +00:00
|
|
|
req->Function = MPI2_FUNCTION_SCSI_TASK_MGMT;
|
|
|
|
req->TaskType = MPI2_SCSITASKMGMT_TASKTYPE_TARGET_RESET;
|
|
|
|
|
|
|
|
/* SAS Hard Link Reset / SATA Link Reset */
|
|
|
|
req->MsgFlags = MPI2_SCSITASKMGMT_MSGFLAGS_LINK_RESET;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
tm->cm_data = NULL;
|
|
|
|
tm->cm_desc.HighPriority.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY;
|
|
|
|
tm->cm_complete = mpssas_resetdev_complete;
|
|
|
|
tm->cm_complete_data = ccb;
|
2012-08-01 12:24:13 +00:00
|
|
|
tm->cm_targ = targ;
|
2012-01-26 18:17:21 +00:00
|
|
|
mps_map_command(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_resetdev_complete(struct mps_softc *sc, struct mps_command *tm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
|
|
|
MPI2_SCSI_TASK_MANAGE_REPLY *resp;
|
|
|
|
union ccb *ccb;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
mtx_assert(&sc->mps_mtx, MA_OWNED);
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
resp = (MPI2_SCSI_TASK_MANAGE_REPLY *)tm->cm_reply;
|
|
|
|
ccb = tm->cm_complete_data;
|
2010-09-10 15:03:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* task management commands don't have S/G lists.
|
|
|
|
*/
|
|
|
|
if ((tm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
MPI2_SCSI_TASK_MANAGE_REQUEST *req;
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
req = (MPI2_SCSI_TASK_MANAGE_REQUEST *)tm->cm_req;
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"%s: cm_flags = %#x for reset of handle %#04x! "
|
2012-01-26 18:17:21 +00:00
|
|
|
"This should not happen!\n", __func__, tm->cm_flags,
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
req->DevHandle);
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
goto bailout;
|
|
|
|
}
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
|
|
|
"%s: IOCStatus = 0x%x ResponseCode = 0x%x\n", __func__,
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
le16toh(resp->IOCStatus), le32toh(resp->ResponseCode));
|
2010-09-10 15:03:56 +00:00
|
|
|
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
if (le32toh(resp->ResponseCode) == MPI2_SCSITASKMGMT_RSP_TM_COMPLETE) {
|
2010-09-10 15:03:56 +00:00
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP;
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_announce_reset(sc, AC_SENT_BDR, tm->cm_targ->tid,
|
|
|
|
CAM_LUN_WILDCARD);
|
|
|
|
}
|
2010-09-10 15:03:56 +00:00
|
|
|
else
|
|
|
|
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
|
|
|
|
|
Fix several issues with the mps(4) driver.
When the driver ran out of DMA chaining buffers, it kept the timeout for
the I/O, and I/O would stall.
The driver was not freezing the device queue on errors.
mps.c: Pull command completion logic into a separate
function, and call the callback/wakeup for commands
that are never sent due to lack of chain buffers.
Add a number of extra diagnostic sysctl variables.
Handle pre-hardware errors for configuration I/O.
This doesn't panic the system, but it will fail the
configuration I/O and there is no retry mechanism.
So the device probe will not succeed. This should
be a very uncommon situation, however.
mps_sas.c: Freeze the SIM queue when we run out of chain
buffers, and unfreeze it when more commands
complete.
Freeze the device queue when errors occur, so that
CAM can insure proper command ordering.
Report pre-hardware errors for task management
commands. In general, that shouldn't be possible
because task management commands don't have S/G
lists, and that is currently the only error path
before we get to the hardware.
Handle pre-hardware errors (like out of chain
elements) for SMP requests. That shouldn't happen
either, since we should have enough space for two
S/G elements in the standard request.
For commands that end with
MPI2_IOCSTATUS_SCSI_IOC_TERMINATED and
MPI2_IOCSTATUS_SCSI_EXT_TERMINATED, return them
with CAM_REQUEUE_REQ to retry them unconditionally.
These seem to be related to back end, transport
related problems that are hopefully transient. We
don't want to go through the retry count for
something that is not a permanent error.
Keep track of the number of outstanding I/Os.
mpsvar.h: Track the number of free chain elements.
Add variables for the number of outstanding I/Os,
and I/O high water mark.
Add variables to track the number of free chain
buffers and the chain low water mark, as well as
the number of chain allocation failures.
Add I/O state flags and an attach done flag.
MFC after: 3 days
2011-02-18 17:06:06 +00:00
|
|
|
bailout:
|
2010-09-16 22:54:56 +00:00
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_free_tm(sc, tm);
|
2010-09-10 15:03:56 +00:00
|
|
|
xpt_done(ccb);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_poll(struct cam_sim *sim)
|
|
|
|
{
|
|
|
|
struct mpssas_softc *sassc;
|
|
|
|
|
|
|
|
sassc = cam_sim_softc(sim);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
if (sassc->sc->mps_debug & MPS_TRACE) {
|
|
|
|
/* frequent debug messages during a panic just slow
|
|
|
|
* everything down too much.
|
|
|
|
*/
|
|
|
|
mps_printf(sassc->sc, "%s clearing MPS_TRACE\n", __func__);
|
|
|
|
sassc->sc->mps_debug &= ~MPS_TRACE;
|
|
|
|
}
|
|
|
|
|
2010-09-10 15:03:56 +00:00
|
|
|
mps_intr_locked(sassc->sc);
|
|
|
|
}
|
|
|
|
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
|
|
|
mpssas_async(void *callback_arg, uint32_t code, struct cam_path *path,
|
|
|
|
void *arg)
|
|
|
|
{
|
|
|
|
struct mps_softc *sc;
|
|
|
|
|
|
|
|
sc = (struct mps_softc *)callback_arg;
|
|
|
|
|
|
|
|
switch (code) {
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#if (__FreeBSD_version >= 1000006) || \
|
|
|
|
((__FreeBSD_version >= 901503) && (__FreeBSD_version < 1000000))
|
2012-01-26 18:17:21 +00:00
|
|
|
case AC_ADVINFO_CHANGED: {
|
|
|
|
struct mpssas_target *target;
|
|
|
|
struct mpssas_softc *sassc;
|
|
|
|
struct scsi_read_capacity_data_long rcap_buf;
|
|
|
|
struct ccb_dev_advinfo cdai;
|
|
|
|
struct mpssas_lun *lun;
|
|
|
|
lun_id_t lunid;
|
|
|
|
int found_lun;
|
|
|
|
uintptr_t buftype;
|
|
|
|
|
|
|
|
buftype = (uintptr_t)arg;
|
|
|
|
|
|
|
|
found_lun = 0;
|
|
|
|
sassc = sc->sassc;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We're only interested in read capacity data changes.
|
|
|
|
*/
|
|
|
|
if (buftype != CDAI_TYPE_RCAPLONG)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We should have a handle for this, but check to make sure.
|
|
|
|
*/
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(xpt_path_target_id(path) < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in mpssas_async\n",
|
|
|
|
xpt_path_target_id(path)));
|
2012-01-26 18:17:21 +00:00
|
|
|
target = &sassc->targets[xpt_path_target_id(path)];
|
|
|
|
if (target->handle == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
lunid = xpt_path_lun_id(path);
|
|
|
|
|
|
|
|
SLIST_FOREACH(lun, &target->luns, lun_link) {
|
|
|
|
if (lun->lun_id == lunid) {
|
|
|
|
found_lun = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (found_lun == 0) {
|
|
|
|
lun = malloc(sizeof(struct mpssas_lun), M_MPT2,
|
|
|
|
M_NOWAIT | M_ZERO);
|
|
|
|
if (lun == NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "Unable to alloc "
|
2012-01-26 18:17:21 +00:00
|
|
|
"LUN for EEDP support.\n");
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
lun->lun_id = lunid;
|
|
|
|
SLIST_INSERT_HEAD(&target->luns, lun, lun_link);
|
|
|
|
}
|
|
|
|
|
|
|
|
bzero(&rcap_buf, sizeof(rcap_buf));
|
|
|
|
xpt_setup_ccb(&cdai.ccb_h, path, CAM_PRIORITY_NORMAL);
|
|
|
|
cdai.ccb_h.func_code = XPT_DEV_ADVINFO;
|
|
|
|
cdai.ccb_h.flags = CAM_DIR_IN;
|
|
|
|
cdai.buftype = CDAI_TYPE_RCAPLONG;
|
|
|
|
cdai.flags = 0;
|
|
|
|
cdai.bufsiz = sizeof(rcap_buf);
|
|
|
|
cdai.buf = (uint8_t *)&rcap_buf;
|
|
|
|
xpt_action((union ccb *)&cdai);
|
|
|
|
if ((cdai.ccb_h.status & CAM_DEV_QFRZN) != 0)
|
|
|
|
cam_release_devq(cdai.ccb_h.path,
|
|
|
|
0, 0, 0, FALSE);
|
|
|
|
|
|
|
|
if (((cdai.ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP)
|
|
|
|
&& (rcap_buf.prot & SRC16_PROT_EN)) {
|
|
|
|
lun->eedp_formatted = TRUE;
|
|
|
|
lun->eedp_block_size = scsi_4btoul(rcap_buf.length);
|
|
|
|
} else {
|
|
|
|
lun->eedp_formatted = FALSE;
|
|
|
|
lun->eedp_block_size = 0;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#else
|
|
|
|
case AC_FOUND_DEVICE: {
|
|
|
|
struct ccb_getdev *cgd;
|
|
|
|
|
|
|
|
cgd = arg;
|
|
|
|
mpssas_check_eedp(sc, path, cgd);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
#endif
|
2012-01-26 18:17:21 +00:00
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#if (__FreeBSD_version < 901503) || \
|
|
|
|
((__FreeBSD_version >= 1000000) && (__FreeBSD_version < 1000006))
|
2012-01-26 18:17:21 +00:00
|
|
|
static void
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
mpssas_check_eedp(struct mps_softc *sc, struct cam_path *path,
|
|
|
|
struct ccb_getdev *cgd)
|
2012-01-26 18:17:21 +00:00
|
|
|
{
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
struct mpssas_softc *sassc = sc->sassc;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct ccb_scsiio *csio;
|
|
|
|
struct scsi_read_capacity_16 *scsi_cmd;
|
|
|
|
struct scsi_read_capacity_eedp *rcap_buf;
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
path_id_t pathid;
|
2012-01-26 18:17:21 +00:00
|
|
|
target_id_t targetid;
|
|
|
|
lun_id_t lunid;
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
union ccb *ccb;
|
|
|
|
struct cam_path *local_path;
|
2012-01-26 18:17:21 +00:00
|
|
|
struct mpssas_target *target;
|
|
|
|
struct mpssas_lun *lun;
|
|
|
|
uint8_t found_lun;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
char path_str[64];
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
sassc = sc->sassc;
|
|
|
|
pathid = cam_sim_path(sassc->sim);
|
|
|
|
targetid = xpt_path_target_id(path);
|
|
|
|
lunid = xpt_path_lun_id(path);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(targetid < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in mpssas_check_eedp\n",
|
|
|
|
targetid));
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
target = &sassc->targets[targetid];
|
|
|
|
if (target->handle == 0x0)
|
|
|
|
return;
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
/*
|
|
|
|
* Determine if the device is EEDP capable.
|
|
|
|
*
|
|
|
|
* If this flag is set in the inquiry data,
|
|
|
|
* the device supports protection information,
|
|
|
|
* and must support the 16 byte read
|
|
|
|
* capacity command, otherwise continue without
|
|
|
|
* sending read cap 16
|
|
|
|
*/
|
|
|
|
if ((cgd->inq_data.spc3_flags & SPC3_SID_PROTECT) == 0)
|
|
|
|
return;
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
/*
|
|
|
|
* Issue a READ CAPACITY 16 command. This info
|
|
|
|
* is used to determine if the LUN is formatted
|
|
|
|
* for EEDP support.
|
|
|
|
*/
|
|
|
|
ccb = xpt_alloc_ccb_nowait();
|
|
|
|
if (ccb == NULL) {
|
|
|
|
mps_dprint(sc, MPS_ERROR, "Unable to alloc CCB "
|
|
|
|
"for EEDP support.\n");
|
|
|
|
return;
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
if (xpt_create_path(&local_path, xpt_periph,
|
|
|
|
pathid, targetid, lunid) != CAM_REQ_CMP) {
|
|
|
|
mps_dprint(sc, MPS_ERROR, "Unable to create "
|
|
|
|
"path for EEDP support\n");
|
|
|
|
xpt_free_ccb(ccb);
|
|
|
|
return;
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
/*
|
|
|
|
* If LUN is already in list, don't create a new
|
|
|
|
* one.
|
|
|
|
*/
|
|
|
|
found_lun = FALSE;
|
|
|
|
SLIST_FOREACH(lun, &target->luns, lun_link) {
|
|
|
|
if (lun->lun_id == lunid) {
|
|
|
|
found_lun = TRUE;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!found_lun) {
|
|
|
|
lun = malloc(sizeof(struct mpssas_lun), M_MPT2,
|
|
|
|
M_NOWAIT | M_ZERO);
|
|
|
|
if (lun == NULL) {
|
|
|
|
mps_dprint(sc, MPS_ERROR,
|
|
|
|
"Unable to alloc LUN for EEDP support.\n");
|
|
|
|
xpt_free_path(local_path);
|
|
|
|
xpt_free_ccb(ccb);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
lun->lun_id = lunid;
|
|
|
|
SLIST_INSERT_HEAD(&target->luns, lun,
|
|
|
|
lun_link);
|
|
|
|
}
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
xpt_path_string(local_path, path_str, sizeof(path_str));
|
|
|
|
mps_dprint(sc, MPS_INFO, "Sending read cap: path %s handle %d\n",
|
|
|
|
path_str, target->handle);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
/*
|
|
|
|
* Issue a READ CAPACITY 16 command for the LUN.
|
|
|
|
* The mpssas_read_cap_done function will load
|
|
|
|
* the read cap info into the LUN struct.
|
|
|
|
*/
|
|
|
|
rcap_buf = malloc(sizeof(struct scsi_read_capacity_eedp),
|
|
|
|
M_MPT2, M_NOWAIT | M_ZERO);
|
|
|
|
if (rcap_buf == NULL) {
|
|
|
|
mps_dprint(sc, MPS_FAULT,
|
|
|
|
"Unable to alloc read capacity buffer for EEDP support.\n");
|
|
|
|
xpt_free_path(ccb->ccb_h.path);
|
|
|
|
xpt_free_ccb(ccb);
|
|
|
|
return;
|
2012-01-26 18:17:21 +00:00
|
|
|
}
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
xpt_setup_ccb(&ccb->ccb_h, local_path, CAM_PRIORITY_XPT);
|
|
|
|
csio = &ccb->csio;
|
|
|
|
csio->ccb_h.func_code = XPT_SCSI_IO;
|
|
|
|
csio->ccb_h.flags = CAM_DIR_IN;
|
|
|
|
csio->ccb_h.retry_count = 4;
|
|
|
|
csio->ccb_h.cbfcnp = mpssas_read_cap_done;
|
|
|
|
csio->ccb_h.timeout = 60000;
|
|
|
|
csio->data_ptr = (uint8_t *)rcap_buf;
|
|
|
|
csio->dxfer_len = sizeof(struct scsi_read_capacity_eedp);
|
|
|
|
csio->sense_len = MPS_SENSE_LEN;
|
|
|
|
csio->cdb_len = sizeof(*scsi_cmd);
|
|
|
|
csio->tag_action = MSG_SIMPLE_Q_TAG;
|
|
|
|
|
|
|
|
scsi_cmd = (struct scsi_read_capacity_16 *)&csio->cdb_io.cdb_bytes;
|
|
|
|
bzero(scsi_cmd, sizeof(*scsi_cmd));
|
|
|
|
scsi_cmd->opcode = 0x9E;
|
|
|
|
scsi_cmd->service_action = SRC16_SERVICE_ACTION;
|
|
|
|
((uint8_t *)scsi_cmd)[13] = sizeof(struct scsi_read_capacity_eedp);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
ccb->ccb_h.ppriv_ptr1 = sassc;
|
|
|
|
xpt_action(ccb);
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
static void
|
|
|
|
mpssas_read_cap_done(struct cam_periph *periph, union ccb *done_ccb)
|
|
|
|
{
|
|
|
|
struct mpssas_softc *sassc;
|
|
|
|
struct mpssas_target *target;
|
|
|
|
struct mpssas_lun *lun;
|
|
|
|
struct scsi_read_capacity_eedp *rcap_buf;
|
|
|
|
|
|
|
|
if (done_ccb == NULL)
|
|
|
|
return;
|
2012-02-09 00:16:12 +00:00
|
|
|
|
|
|
|
/* Driver need to release devq, it Scsi command is
|
|
|
|
* generated by driver internally.
|
|
|
|
* Currently there is a single place where driver
|
|
|
|
* calls scsi command internally. In future if driver
|
|
|
|
* calls more scsi command internally, it needs to release
|
|
|
|
* devq internally, since those command will not go back to
|
|
|
|
* cam_periph.
|
|
|
|
*/
|
|
|
|
if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) ) {
|
|
|
|
done_ccb->ccb_h.status &= ~CAM_DEV_QFRZN;
|
|
|
|
xpt_release_devq(done_ccb->ccb_h.path,
|
|
|
|
/*count*/ 1, /*run_queue*/TRUE);
|
|
|
|
}
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
rcap_buf = (struct scsi_read_capacity_eedp *)done_ccb->csio.data_ptr;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get the LUN ID for the path and look it up in the LUN list for the
|
|
|
|
* target.
|
|
|
|
*/
|
|
|
|
sassc = (struct mpssas_softc *)done_ccb->ccb_h.ppriv_ptr1;
|
2014-04-07 15:13:24 +00:00
|
|
|
KASSERT(done_ccb->ccb_h.target_id < sassc->maxtargets,
|
|
|
|
("Target %d out of bounds in mpssas_read_cap_done\n",
|
|
|
|
done_ccb->ccb_h.target_id));
|
2012-01-26 18:17:21 +00:00
|
|
|
target = &sassc->targets[done_ccb->ccb_h.target_id];
|
|
|
|
SLIST_FOREACH(lun, &target->luns, lun_link) {
|
|
|
|
if (lun->lun_id != done_ccb->ccb_h.target_lun)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Got the LUN in the target's LUN list. Fill it in
|
|
|
|
* with EEDP info. If the READ CAP 16 command had some
|
|
|
|
* SCSI error (common if command is not supported), mark
|
|
|
|
* the lun as not supporting EEDP and set the block size
|
|
|
|
* to 0.
|
|
|
|
*/
|
|
|
|
if (((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP)
|
|
|
|
|| (done_ccb->csio.scsi_status != SCSI_STATUS_OK)) {
|
|
|
|
lun->eedp_formatted = FALSE;
|
|
|
|
lun->eedp_block_size = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (rcap_buf->protect & 0x01) {
|
Merge in phase 14+ -> 16 mps driver fixes from LSI:
---------------------------------------------------------------
System panics during a Port reset with ouststanding I/O
---------------------------------------------------------------
It is possible to call mps_mapping_free_memory after this
memory is already freed, causing a panic. Removed this extra
call to mps_mappiing_free_memory and call mps_mapping_exit
in place of the mps_mapping_free_memory call so that any
outstanding mapping items can be flushed before memory is
freed.
---------------------------------------------------------------
Correct memory leak during a Port reset with ouststanding I/O
---------------------------------------------------------------
In mps_reinit function, the mapping memory was not being
freed before being re-allocated. Added line to call the
memory free function for mapping memory.
---------------------------------------------------------------
Use CAM_SIM_QUEUED flag in Driver IO path.
---------------------------------------------------------------
This flag informs the XPT that successful abort of a CCB
requires an abort ccb to be issued to the SIM. While
processing SCSI IO's, set the CAM_SIM_QUEUED flag in the
status for the IO. When the command completes, clear this
flag.
---------------------------------------------------------------
Check for CAM_REQ_INPROG in I/O path.
---------------------------------------------------------------
Added a check in mpssas_action_scsiio for the In Progress
status for the IO. If this flag is set, the IO has already
been aborted by the upper layer (before CAM_SIM_QUEUED was
set) and there is no need to send the IO. The request will
be completed without error.
---------------------------------------------------------------
Improve "doorbell handshake method" for mps_get_iocfacts
---------------------------------------------------------------
Removed call to get Port Facts since this information is
not used currently.
Added mps_iocfacts_allocate function to allocate memory
that is based on IOC Facts data. Added mps_iocfacts_free
function to free memory that is based on IOC Facts data.
Both of the functions are used when a Diag Reset is performed
or when the driver is attached/detached. This is needed in
case IOC Facts changes after a Diag Reset, which could
happen if FW is upgraded.
Moved call of mps_bases_static_config_pages from the attach
routine to after the IOC is ready to process accesses based
on the new memory allocations (instead of polling through
the Doorbell).
---------------------------------------------------------------
Set TimeStamp in INIT message in millisecond format Set the IOC
---------------------------------------------------------------
---------------------------------------------------------------
Prefer mps_wait_command to mps_request_polled
---------------------------------------------------------------
Instead of using mps_request_polled, call mps_wait_command
whenever possible. Change the mps_wait_command function to
check the current context and either use interrupt context
or poll if required by using the pause or DELAY function.
Added a check after waiting 50mSecs to see if the command
has timed out. This is only done if polliing, the msleep
command will automatically timeout if the command has taken
too long to complete.
---------------------------------------------------------------
Integrated RAID: Volume Activation Failed error message is
displayed though the volume has been activated.
---------------------------------------------------------------
Instead of failing an IOCTL request that does not have a
large enough buffer to hold the complete reply, copy as
much data from the reply as possible into the user's buffer
and log a message saying that the user's buffer was smaller
than the returned data.
---------------------------------------------------------------
mapping_add_new_device failure due to persistent table FULL
---------------------------------------------------------------
When a new device is added, if it is determined that the
device persistent table is being used and is full, instead
of displaying a message for this condition every time, only
log a message if the MPS_INFO bit is set in the debug_flags.
Submitted by: LSI
MFC after: 1 week
2013-07-22 18:41:53 +00:00
|
|
|
mps_dprint(sassc->sc, MPS_INFO, "LUN %d for "
|
|
|
|
"target ID %d is formatted for EEDP "
|
|
|
|
"support.\n", done_ccb->ccb_h.target_lun,
|
|
|
|
done_ccb->ccb_h.target_id);
|
2012-01-26 18:17:21 +00:00
|
|
|
lun->eedp_formatted = TRUE;
|
|
|
|
lun->eedp_block_size = scsi_4btoul(rcap_buf->length);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Finished with this CCB and path.
|
|
|
|
free(rcap_buf, M_MPT2);
|
|
|
|
xpt_free_path(done_ccb->ccb_h.path);
|
|
|
|
xpt_free_ccb(done_ccb);
|
|
|
|
}
|
CAM and mps(4) driver scanning changes.
Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM
not to perform a rescan on a bus when it is registered.
We now use this flag in the mps(4) driver. Since it knows what
devices it has attached, it is more efficient for it to just issue
a target rescan on the targets that are attached.
Also, remove the private rescan thread from the mps(4) driver in
favor of the rescan thread already built into CAM. Without this
change, but with the change above, the MPS scanner could run before
or during CAM's initial setup, which would cause duplicate device
reprobes and announcements.
sys/param.h:
Bump __FreeBSD_version to 1000039 for the inclusion of the
PIM_RESCAN CAM path inquiry flag.
sys/cam/cam_ccb.h:
sys/cam/cam_xpt.c:
Added a PIM_NOSCAN flag. If a SIM sets this in the path
inquiry ccb, then CAM won't rescan the bus in
xpt_bus_regsister.
sys/dev/mps/mps_sas.c
For versions of FreeBSD that have the PIM_NOSCAN path
inquiry flag, don't freeze the sim queue during scanning,
because CAM won't be scanning this bus. Instead, hold
up the boot. Don't call mpssas_rescan_target in
mpssas_startup_decrement; it's redundant and I don't
know why it was in there.
Set PIM_NOSCAN in path inquiry CCBs.
Remove methods related to the internal rescan daemon.
Always use async events to trigger a probe for EEDP support.
In older versions of FreeBSD where AC_ADVINFO_CHANGED is
not available, use AC_FOUND_DEVICE and issue the
necessary READ CAPACITY manually.
Provide a path to xpt_register_async() so that we only
receive events for our own SCSI domain.
Improve error reporting in cases where setup for EEDP
detection fails.
sys/dev/mps/mps_sas.h:
Remove softc flags and data related to the scanner thread.
sys/dev/mps/mps_sas_lsi.c:
Unconditionally rescan the target whenever a device is added.
Sponsored by: Spectra Logic
MFC after: 1 week
2013-07-22 18:37:07 +00:00
|
|
|
#endif /* (__FreeBSD_version < 901503) || \
|
|
|
|
((__FreeBSD_version >= 1000000) && (__FreeBSD_version < 1000006)) */
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
int
|
|
|
|
mpssas_startup(struct mps_softc *sc)
|
|
|
|
{
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Send the port enable message and set the wait_for_port_enable flag.
|
|
|
|
* This flag helps to keep the simq frozen until all discovery events
|
|
|
|
* are processed.
|
|
|
|
*/
|
|
|
|
sc->wait_for_port_enable = 1;
|
|
|
|
mpssas_send_portenable(sc);
|
|
|
|
return (0);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
mpssas_send_portenable(struct mps_softc *sc)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_PORT_ENABLE_REQUEST *request;
|
|
|
|
struct mps_command *cm;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
|
|
|
|
if ((cm = mps_alloc_command(sc)) == NULL)
|
|
|
|
return (EBUSY);
|
|
|
|
request = (MPI2_PORT_ENABLE_REQUEST *)cm->cm_req;
|
|
|
|
request->Function = MPI2_FUNCTION_PORT_ENABLE;
|
|
|
|
request->MsgFlags = 0;
|
|
|
|
request->VP_ID = 0;
|
|
|
|
cm->cm_desc.Default.RequestFlags = MPI2_REQ_DESCRIPT_FLAGS_DEFAULT_TYPE;
|
|
|
|
cm->cm_complete = mpssas_portenable_complete;
|
|
|
|
cm->cm_data = NULL;
|
|
|
|
cm->cm_sge = NULL;
|
|
|
|
|
|
|
|
mps_map_command(sc, cm);
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO,
|
2012-01-26 18:17:21 +00:00
|
|
|
"mps_send_portenable finished cm %p req %p complete %p\n",
|
|
|
|
cm, cm->cm_req, cm->cm_complete);
|
|
|
|
return (0);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_portenable_complete(struct mps_softc *sc, struct mps_command *cm)
|
2010-09-10 15:03:56 +00:00
|
|
|
{
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_PORT_ENABLE_REPLY *reply;
|
|
|
|
struct mpssas_softc *sassc;
|
|
|
|
|
2013-07-19 00:12:41 +00:00
|
|
|
MPS_FUNCTRACE(sc);
|
2012-01-26 18:17:21 +00:00
|
|
|
sassc = sc->sassc;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Currently there should be no way we can hit this case. It only
|
|
|
|
* happens when we have a failure to allocate chain frames, and
|
|
|
|
* port enable commands don't have S/G lists.
|
|
|
|
*/
|
|
|
|
if ((cm->cm_flags & MPS_CM_FLAGS_ERROR_MASK) != 0) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_ERROR, "%s: cm_flags = %#x for port enable! "
|
2012-01-26 18:17:21 +00:00
|
|
|
"This should not happen!\n", __func__, cm->cm_flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
reply = (MPI2_PORT_ENABLE_REPLY *)cm->cm_reply;
|
|
|
|
if (reply == NULL)
|
|
|
|
mps_dprint(sc, MPS_FAULT, "Portenable NULL reply\n");
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
else if (le16toh(reply->IOCStatus & MPI2_IOCSTATUS_MASK) !=
|
2012-01-26 18:17:21 +00:00
|
|
|
MPI2_IOCSTATUS_SUCCESS)
|
|
|
|
mps_dprint(sc, MPS_FAULT, "Portenable failed\n");
|
|
|
|
|
|
|
|
mps_free_command(sc, cm);
|
|
|
|
if (sc->mps_ich.ich_arg != NULL) {
|
2013-07-19 00:12:41 +00:00
|
|
|
mps_dprint(sc, MPS_XINFO, "disestablish config intrhook\n");
|
2012-01-26 18:17:21 +00:00
|
|
|
config_intrhook_disestablish(&sc->mps_ich);
|
|
|
|
sc->mps_ich.ich_arg = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get WarpDrive info after discovery is complete but before the scan
|
|
|
|
* starts. At this point, all devices are ready to be exposed to the
|
|
|
|
* OS. If devices should be hidden instead, take them out of the
|
|
|
|
* 'targets' array before the scan. The devinfo for a disk will have
|
|
|
|
* some info and a volume's will be 0. Use that to remove disks.
|
|
|
|
*/
|
|
|
|
mps_wd_config_pages(sc);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Done waiting for port enable to complete. Decrement the refcount.
|
|
|
|
* If refcount is 0, discovery is complete and a rescan of the bus can
|
|
|
|
* take place. Since the simq was explicitly frozen before port
|
|
|
|
* enable, it must be explicitly released here to keep the
|
|
|
|
* freeze/release count in sync.
|
|
|
|
*/
|
|
|
|
sc->wait_for_port_enable = 0;
|
|
|
|
sc->port_enable_complete = 1;
|
Bring in LSI's latest mps(4) 6Gb SAS and WarpDrive driver, version
14.00.00.01-fbsd.
Their description of the changes is as follows:
1. Copyright contents has been changed in all respective .c
and .h files
2. Support for WRITE12 and READ12 for direct-io (warpdrive only)
has been added.
3. Driver has added checks to see if Drive has READ_CAP_16
support before sending it down to the device.
If SPC3_SID_PROTECT flag is set in the inquiry data, the
device supports protection information, and must support
the 16 byte read capacity command, otherwise continue without
sending read cap 16. This will optimize driver performance,
since it will not send READ_CAP_16 to the drive which does
not have support of READ_CAP_16.
4. With new approach, "MPTIOCTL_RESET_ADAPTER" IOCTL will not
use DELAY() which is busy loop implementation.
It will use <msleep> (Better way to sleep without busy
loop). Also from the HBA reset code path and some other
places, DELAY() is replaced with msleep() or "pause()",
which is based on sleep/wakeup style calls. Driver use
msleep()/pause() instead of DELAY based on CAN_SLEEP/NO_SLEEP
flags to avoid busy loop which is not required all the
time.e.a
a. While driver is getting loaded, driver calls most of the
commands with NO_SLEEP.
b. When Driver is functional and it needs Reinit of HBA,
CAN_SLEEP flag is used.
5. <mpslsi> driver is not Endian safe. It will not work on Big
Endian machines like Sparc and PowerPC platforms because it
assumes it is running on a Little Endian machine.
Driver code is modified such way that it does not assume CPU
arch is Little Endian.
a. All places where Driver interacts from HBA to Host, it
converts Little Endian format to CPU format.
b. All places where Driver interacts from Host to HBA, it
converts CPU format to Little Endian.
6. Findout memory leaks in FreeBSD Driver and resolve those,
such as memory leak in targ's luns creation/deletion.
Also added additional checks to see memory allocation
success/fail.
7. Add loginfo prints as debug message, i.e. When FW sends any
loginfo, Driver should print those as debug message.
This will help for debugging purpose.
8. There is possibility to get config request timeout. Current
driver is able to detect config request timetout, but it does
not do anything on config_request timeout. Driver should
call mps_reinit() if any request_poll (which is called as
part of config_request) is time out.
9. cdb length check is required for 32 byte CDB. Add correct mpi
control value for 32 bit CDB as below while submitting SCSI IO
Request to controller.
mpi_control |= 4 << MPI2_SCSIIO_CONTROL_ADDCDBLEN_SHIFT;
10. Check the actual status of Message unit reset
(mps_message_unit_reset).Previously FreeBSD Driver just writes
MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET and never check the ack
(it just wait for 50 millisecond). So, Driver now check the
status of "MPI2_FUNCTION_IOC_MESSAGE_UNIT_RESET" after writing
it to the FW.
Now it also checking for whether doorbell ack uses msleep with
proper sleep flags, instead of <DELAY>.
11. Previously CAM does not detect Multi-Lun Devices. In order to
detect Multi-Lun Devices by CAM the driver needs following change
set:
a. There is "max_lun" field which Driver need to set based on
hw/fw support. Currently LSI released driver does not set
this field.
b. Default of "max_lun" should not be 0 in OS, but it is
currently set to 0 in CAM layer.
c. Export max_lun capacity to 255
12. Driver will not reset target info after port enable complete and
also do Device removal when Device remove from FW. The detail
description is as follows
a. When Driver receive WD PD add events, it will add all
information in driver local data structure.
b. Only for WD, we have below checks after port enable
completes, where driver clear off all information retrieved
at #1.
if ((sc->WD_available &&
(sc->WD_hide_expose == MPS_WD_HIDE_ALWAYS)) ||
(sc->WD_valid_config && (sc->WD_hide_expose ==
MPS_WD_HIDE_IF_VOLUME)) {
// clear off target data structure.
}
It is mainly not to attach PDs to OS.
FreeBSD does bus rescan as older Parallel scsi style. So Driver
needs to handle which Drive is visible to OS. That is a reason
we have to clear off targ information for PDs.
Again, above logic was implemented long time ago. Similar concept
we have for non-wd also. For that, LSI have introduced different
logic to hide PDs.
Eventually, because of above gap, when Phy goes offline, we
observe below failure. That is what Driver is not doing complete
removal of device with FW. (which was pointed by Scott)
Apr 5 02:39:24 Freebsd7 kernel: mpslsi0: mpssas_prepare_remove
Apr 5 02:39:24 Freebsd7 kernel: mpssas_prepare_remove 497 : invalid handle 0xe
Now Driver will not reset target info after port enable complete
and also will do Device removal when Device remove from FW.
13. Returning "CAM_SEL_TIMEOUT" instead of "CAM_TID_INVALID"
error code on request to the Target IDs that have no devices
conected at that moment. As if "CAM_TID_INVALID" error code
is returned to the CAM Layaer then it results in a huge chain
of errors in verbose kernel messages on boot and every
hot-plug event.
Submitted by: Sreekanth Reddy <Sreekanth.Reddy@lsi.com>
MFC after: 3 days
2012-06-28 03:48:54 +00:00
|
|
|
wakeup(&sc->port_enable_complete);
|
2012-01-26 18:17:21 +00:00
|
|
|
mpssas_startup_decrement(sassc);
|
2010-09-10 15:03:56 +00:00
|
|
|
}
|
|
|
|
|
2013-08-09 01:09:02 +00:00
|
|
|
int
|
|
|
|
mpssas_check_id(struct mpssas_softc *sassc, int id)
|
|
|
|
{
|
|
|
|
struct mps_softc *sc = sassc->sc;
|
|
|
|
char *ids;
|
|
|
|
char *name;
|
|
|
|
|
|
|
|
ids = &sc->exclude_ids[0];
|
|
|
|
while((name = strsep(&ids, ",")) != NULL) {
|
|
|
|
if (name[0] == '\0')
|
|
|
|
continue;
|
|
|
|
if (strtol(name, NULL, 0) == (long)id)
|
|
|
|
return (1);
|
|
|
|
}
|
|
|
|
|
|
|
|
return (0);
|
|
|
|
}
|