Merge two commits that had dependencies on other 5.11 trees (the block
and the irq trees respectively).
- We reverted a megaraid_sas change in 5.10 due to missing block
layer plumbing. Now that this is in place, reinstate the change.
- The hisi_sas driver had a dependency on a driver core irq change
that went in through Thomas' tree.
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
As a performance enhancement, make the completion queue interrupts managed.
In addition, in commit bf0beec060 ("blk-mq: drain I/O when all CPUs in a
hctx are offline"), CPU hotplug for MQ devices using managed interrupts is
made safe. So expose HW queues to blk-mq to take advantage of this.
Flag Scsi_host.host_tagset is also set to ensure that the HBA is not sent
more commands than it can handle. However the driver still does not use
request tag for IPTT as there are many HW bugs means that special rules
apply for IPTT allocation.
Link: https://lore.kernel.org/r/1606905417-183214-6-git-send-email-john.garry@huawei.com
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull SCSI updates from James Bottomley:
"This consists of the usual driver updates (ufs, qla2xxx, smartpqi,
target, zfcp, fnic, mpt3sas, ibmvfc) plus a load of cleanups, a major
power management rework and a load of assorted minor updates.
There are a few core updates (formatting fixes being the big one) but
nothing major this cycle"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (279 commits)
scsi: mpt3sas: Update driver version to 36.100.00.00
scsi: mpt3sas: Handle trigger page after firmware update
scsi: mpt3sas: Add persistent MPI trigger page
scsi: mpt3sas: Add persistent SCSI sense trigger page
scsi: mpt3sas: Add persistent Event trigger page
scsi: mpt3sas: Add persistent Master trigger page
scsi: mpt3sas: Add persistent trigger pages support
scsi: mpt3sas: Sync time periodically between driver and firmware
scsi: qla2xxx: Update version to 10.02.00.104-k
scsi: qla2xxx: Fix device loss on 4G and older HBAs
scsi: qla2xxx: If fcport is undergoing deletion complete I/O with retry
scsi: qla2xxx: Fix the call trace for flush workqueue
scsi: qla2xxx: Fix flash update in 28XX adapters on big endian machines
scsi: qla2xxx: Handle aborts correctly for port undergoing deletion
scsi: qla2xxx: Fix N2N and NVMe connect retry failure
scsi: qla2xxx: Fix FW initialization error on big endian machines
scsi: qla2xxx: Fix crash during driver load on big endian machines
scsi: qla2xxx: Fix compilation issue in PPC systems
scsi: qla2xxx: Don't check for fw_started while posting NVMe command
scsi: qla2xxx: Tear down session if FW say it is down
...
hisi_sas_task_exec() uses preemptible() to see if it's safe to block. This
does not work for CONFIG_PREEMPT_COUNT=n kernels in which preemptible()
always returns 0.
The problem is masked when enabling some of the common Kconfig.debug
options (like CONFIG_DEBUG_ATOMIC_SLEEP), as they implicitly enable the
preemption counter.
In general, driver leaf functions should not make logic decisions based on
the context they're called from. The caller should be the entity
responsible for explicitly indicating context.
Since hisi_sas_task_exec() already has a gfp_t flags parameter, use it as
the explicit context marker.
Link: https://lore.kernel.org/r/20201126132952.2287996-3-bigeasy@linutronix.de
Fixes: 214e702d4b ("scsi: hisi_sas: Adjust task reject period during host reset")
Fixes: 550c0d89d5 ("scsi: hisi_sas: Replace in_softirq() check in hisi_sas_task_exec()")
Cc: Xiaofei Tan <tanxiaofei@huawei.com>
Cc: Xiang Chen <chenxiang66@hisilicon.com>
Cc: John Garry <john.garry@huawei.com>
Acked-by: John Garry <john.garry@huawei.com>
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In commit 8d98416a55 ("scsi: hisi_sas: Switch v3 hw to MQ"), the dispatch
function was changed to choose the delivery queue based on the request tag
HW queue index.
This heavily degrades performance for v2 hw, since the HW queues are not
exposed there, and, as such, HW queue #0 is used for every command.
Revert to previous behaviour for when nr_hw_queues is not set, that being
to choose the HW queue based on target device index.
Link: https://lore.kernel.org/r/1602750425-240341-1-git-send-email-john.garry@huawei.com
Fixes: 8d98416a55 ("scsi: hisi_sas: Switch v3 hw to MQ")
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull SCSI updates from James Bottomley:
"The usual driver updates (ufs, qla2xxx, tcmu, ibmvfc, lpfc, smartpqi,
hisi_sas, qedi, qedf, mpt3sas) and minor bug fixes.
There are only three core changes: adding sense codes, cleaning up
noretry and adding an option for limitless retries"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (226 commits)
scsi: hisi_sas: Recover PHY state according to the status before reset
scsi: hisi_sas: Filter out new PHY up events during suspend
scsi: hisi_sas: Add device link between SCSI devices and hisi_hba
scsi: hisi_sas: Add check for methods _PS0 and _PR0
scsi: hisi_sas: Add controller runtime PM support for v3 hw
scsi: hisi_sas: Switch to new framework to support suspend and resume
scsi: hisi_sas: Use hisi_hba->cq_nvecs for calling calling synchronize_irq()
scsi: qedf: Remove redundant assignment to variable 'rc'
scsi: lpfc: Remove unneeded variable 'status' in lpfc_fcp_cpu_map_store()
scsi: snic: Convert to use DEFINE_SEQ_ATTRIBUTE macro
scsi: qla4xxx: Delete unneeded variable 'status' in qla4xxx_process_ddb_changed
scsi: sun_esp: Use module_platform_driver to simplify the code
scsi: sun3x_esp: Use module_platform_driver to simplify the code
scsi: sni_53c710: Use module_platform_driver to simplify the code
scsi: qlogicpti: Use module_platform_driver to simplify the code
scsi: mac_esp: Use module_platform_driver to simplify the code
scsi: jazz_esp: Use module_platform_driver to simplify the code
scsi: mvumi: Fix error return in mvumi_io_attach()
scsi: lpfc: Drop nodelist reference on error in lpfc_gen_req()
scsi: be2iscsi: Fix a theoretical leak in beiscsi_create_eqs()
...
Currently sas_resume_ha() is called while resuming the controller to wait
for all suspended PHYs to come up and all the libsas events to be
completed.
There is a scenario which will cause task hung: For direct attach with two
disks connected with two PHYs, disable phy0 before suspending the disk on
phy1 and the controller, then enable phy0 and resume the controller, and
task hung occurs as follows:
[ 591.901463] hisi_sas_v3_hw 0000:b4:02.0: resuming from operating state [D0]
[ 593.113525] hisi_sas_v3_hw 0000:b4:02.0: neither _PS0 nor _PR0 is defined
[ 593.120301] hisi_sas_v3_hw 0000:b4:02.0: waiting up to 25 seconds for 1 phy to resume
[ 593.120836] hisi_sas_v3_hw 0000:b4:02.0: phyup: phy0 link_rate=10(sata)
[ 593.134680] hisi_sas_v3_hw 0000:b4:02.0: phyup: phy1 link_rate=10(sata)
[ 593.134733] sas: phy-2:0 added to port-2:0, phy_mask:0x1 (5000000000000200)
[ 593.148350] sas: DOING DISCOVERY on port 0, pid:948
[ 593.153227] hisi_sas_v3_hw 0000:b4:02.0: dev[3:5] found
[ 593.159840] sas: Enter sas_scsi_recover_host busy: 0 failed: 0
[ 593.165663] sas: ata7: end_device-2:0: dev error handler
[ 593.165730] sas: ata2: end_device-2:1: dev error handler
[ 593.172532] hisi_sas_v3_hw 0000:b4:02.0: phydown: phy0 phy_state=0x2
[ 593.182570] hisi_sas_v3_hw 0000:b4:02.0: ignore flutter phy0 down
[ 593.331277] hisi_sas_v3_hw 0000:b4:02.0: phyup: phy0 link_rate=10(sata)
[ 593.498956] ata7.00: ATA-11: SAMSUNG MZ7LH960HAJR-00005, HXT7404Q, max UDMA/133
[ 593.506235] ata7.00: 1875385008 sectors, multi 16: LBA48 NCQ (depth 32)
[ 593.514295] ata7.00: configured for UDMA/133
[ 593.518557] sas: --- Exit sas_scsi_recover_host: busy: 0 failed: 0 tries: 1
[ 593.528613] sas: ata7: end_device-2:0: model:SAMSUNG MZ7LH960HAJR-00005
serial:S45NNA0M712225
[ 593.537520] device_link_add 316: dev=2:0:2:0 supplier:2 consumer:0
[ 593.543674] device_link_add 324
[ 593.546801] device_link_add 352
[ 593.549930] device_link_add 406
[ 593.553058] device_link_add 440: dev=2:0:2:0 supplier:2 consumer:0
[ 593.559208] device_link_add 444
[ 593.562335] device_link_add 455
[ 593.565517] scsi 2:0:2:0: Direct-Access ATA SAMSUNG MZ7LH960 404Q PQ: 0
ANSI: 5
[ 620.057464] phy-2:1: resume timeout
[ 738.841445] INFO: task kworker/u256:0:8 blocked for more than 120 seconds.
[ 738.848295] Not tainted 5.8.0-rc1-76154-g0d52b59-dirty #744
[ 738.854361] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 738.862155] kworker/u256:0 D 0 8 2 0x00000028
[ 738.867626] Workqueue: 0000:b4:02.0_event_q sas_port_event_worker
[ 738.873693] Call trace:
[ 738.876133] __switch_to+0xf4/0x148
[ 738.879613] __schedule+0x270/0x5d8
[ 738.883091] schedule+0x78/0x110
[ 738.886307] schedule_timeout+0x1ac/0x280
[ 738.890299] wait_for_completion+0x94/0x138
[ 738.894472] flush_workqueue+0x114/0x438
[ 738.898377] sas_porte_bytes_dmaed+0x400/0x500
[ 738.902801] sas_port_event_worker+0x28/0x40
[ 738.907053] process_one_work+0x1e8/0x360
[ 738.911046] worker_thread+0x44/0x478
[ 738.914698] kthread+0x150/0x158
[ 738.917915] ret_from_fork+0x10/0x1c
[ 738.921534] INFO: task kworker/u256:1:948 blocked for more than 120 seconds.
[ 738.928550] Not tainted 5.8.0-rc1-76154-g0d52b59-dirty #744
[ 738.934614] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 738.942408] kworker/u256:1 D 0 948 2 0x00000028
[ 738.947873] Workqueue: 0000:b4:02.0_disco_q sas_discover_domain
[ 738.953766] Call trace:
[ 738.956203] __switch_to+0xf4/0x148
[ 738.959678] __schedule+0x270/0x5d8
[ 738.963152] schedule+0x78/0x110
[ 738.966368] rpm_resume+0xcc/0x550
[ 738.969757] __pm_runtime_resume+0x3c/0x88
[ 738.973836] rpm_get_suppliers+0x50/0x148
[ 738.977829] __pm_runtime_set_status+0x124/0x2f0
[ 738.982427] scsi_sysfs_add_sdev+0x1a0/0x2a8
[ 738.986679] scsi_probe_and_add_lun+0x888/0xab0
[ 738.991190] __scsi_scan_target+0xec/0x520
[ 738.995268] scsi_scan_target+0x11c/0x128
[ 738.999261] sas_rphy_add+0x15c/0x1e8
[ 739.002907] sas_probe_devices+0xe4/0x150
[ 739.006899] sas_discover_domain+0x33c/0x588
[ 739.011150] process_one_work+0x1e8/0x360
[ 739.015143] worker_thread+0x44/0x478
[ 739.018789] kthread+0x150/0x158
[ 739.022003] ret_from_fork+0x10/0x1c
...
If an extra phy0 up happens during resume of the SAS controller, it will
emit a new libsas event (event PORTE_BYTES_DMAED and event
DISCE_DISCOVER_DOMAIN). We will call function scsi_sysfs_add_sdev() in
event DISCE_DISCOVER_DOMAIN, which will call __pm_runtime_set_status() to
resume supplier (host controller). For runtime PM core, if device is in the
resuming state, the later resume request of the device will wait for
previous resume request to complete synchronously. At that point in time
the state of the controller is still resuming as it waits for all libsas
events to be completed, while libsas event DISCE_DISCOVER_DOMAIN is blocked
as the state of the controller is resuming which causes a deadlock.
To avoid the issue, filter out new PHY up events while the controller is
suspended.
Link: https://lore.kernel.org/r/1601649038-25534-7-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Add BIST support for phy FFE (Feed forward equalizer) setting. The user can
configure FFE through the new debugfs interface.
FFE is a parameter used for link layer control. It will affect the link
quality between the SAS controller and the backplane. In the BIST test, the
FFE interface is provided to assist board testers in optimizing link
parameters.
The modification of the FFE parameter will affect the test after BIST or
the normal running of the board. The user should save the initial FFE
values and restore them after BIST test is complete.
Link: https://lore.kernel.org/r/1598958790-232272-6-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
We found out that after phy up, the hardware reports another oob interrupt
but did not follow a phy up interrupt:
oob ready -> phy up -> DEV found -> oob read -> wait phy up -> timeout
We run link reset when wait phy up timeout, and it send a normal disk into
reset processing. So we made some circumvention action in the code, so that
this abnormal oob interrupt will not start the timer to wait for phy up.
Link: https://lore.kernel.org/r/1589552025-165012-2-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently IRQ_EFFECTIVE_AFF_MASK is enabled for ARM_GIC and ARM_GIC3, so it
only allows a single target CPU in the affinity mask to process interrupts
and also interrupt thread, and the performance of using threaded irq is
almost the same as tasklet. But if the config is not enabled, the interrupt
thread will be allowed all the CPUs in the affinity mask. At that situation
it improves the performance (about 20%).
Note: IRQ_EFFECTIVE_AFF_MASK is configured differently for different
architecture chip, and it seems to be better to make it be configured
easily.
Link: https://lore.kernel.org/r/1579522957-4393-2-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The number of phy down reflects the quality of the link between SAS
controller and disk. In order to allow the user to confirm the link quality
of the system, we record the number of phy down for each phy.
The user can check the current phy down count by reading the debugfs file
corresponding to the specific phy, or clear the phy down count by writing 0
to the debugfs file.
Link: https://lore.kernel.org/r/1571926105-74636-19-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Although if the debugfs initialization fails, we will delete the debugfs
folder of hisi_sas, but we did not consider the scenario where debugfs was
successfully initialized, but the probe failed for other reasons. We found
out that hisi_sas folder is still remain after the probe failed.
When probe fail, we should delete debugfs folder to avoid the above issue.
Link: https://lore.kernel.org/r/1571926105-74636-18-git-send-email-john.garry@huawei.com
Signed-off-by: Luo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
For IOs from upper layer, preemption may be disabled as it may be called by
function __blk_mq_delay_run_hw_queue which will call get_cpu() (it disables
preemption). So if flags HISI_SAS_REJECT_CMD_BIT is set in function
hisi_sas_task_exec(), it may disable preempt twice after down() and up()
which will cause following call trace:
BUG: scheduling while atomic: fio/60373/0x00000002
Call trace:
dump_backtrace+0x0/0x150
show_stack+0x24/0x30
dump_stack+0xa0/0xc4
__schedule_bug+0x68/0x88
__schedule+0x4b8/0x548
schedule+0x40/0xd0
schedule_timeout+0x200/0x378
__down+0x78/0xc8
down+0x54/0x70
hisi_sas_task_exec.isra.10+0x598/0x8d8 [hisi_sas_main]
hisi_sas_queue_command+0x28/0x38 [hisi_sas_main]
sas_queuecommand+0x168/0x1b0 [libsas]
scsi_queue_rq+0x2ac/0x980
blk_mq_dispatch_rq_list+0xb0/0x550
blk_mq_do_dispatch_sched+0x6c/0x110
blk_mq_sched_dispatch_requests+0x114/0x1d8
__blk_mq_run_hw_queue+0xb8/0x130
__blk_mq_delay_run_hw_queue+0x1c0/0x220
blk_mq_run_hw_queue+0xb0/0x128
blk_mq_sched_insert_requests+0xdc/0x208
blk_mq_flush_plug_list+0x1b4/0x3a0
blk_flush_plug_list+0xdc/0x110
blk_finish_plug+0x3c/0x50
blkdev_direct_IO+0x404/0x550
generic_file_read_iter+0x9c/0x848
blkdev_read_iter+0x50/0x78
aio_read+0xc8/0x170
io_submit_one+0x1fc/0x8d8
__arm64_sys_io_submit+0xdc/0x280
el0_svc_common.constprop.0+0xe0/0x1e0
el0_svc_handler+0x34/0x90
el0_svc+0x10/0x14
...
To solve the issue, check preemptible() to avoid disabling preempt multiple
when flag HISI_SAS_REJECT_CMD_BIT is set.
Link: https://lore.kernel.org/r/1571926105-74636-5-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When injecting 2bit ecc errors, it will cause confusion inside SAS
controller which needs host reset to recover it. If a device is gone at the
same times inject 2bit ecc errors, we may not receive the ITCT interrupt so
it will wait for completion in clear_itct_v3_hw() all the time. And host
reset will also not occur because it can't require hisi_hba->sem, so the
system will be suspended.
To solve the issue, use wait_for_completion_timeout() instead of
wait_for_completion(), and also don't mark the gone device as
SAS_PHY_UNUSED when device gone.
Link: https://lore.kernel.org/r/1571926105-74636-4-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Fix sparse warnings:
drivers/scsi/hisi_sas/hisi_sas_main.c:3686:6:
warning: symbol 'hisi_sas_debugfs_release' was not declared. Should it be static?
drivers/scsi/hisi_sas/hisi_sas_main.c:3708:5:
warning: symbol 'hisi_sas_debugfs_alloc' was not declared. Should it be static?
drivers/scsi/hisi_sas/hisi_sas_main.c:3799:6:
warning: symbol 'hisi_sas_debugfs_bist_init' was not declared. Should it be static?
Link: https://lore.kernel.org/r/20190923054035.19036-1-yuehaibing@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Acked-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When device gone, it will check whether it is during reset, if not, it will
send internal task abort. Before internal task abort returned, reset
begins, and it will check whether SAS_PHY_UNUSED is set, if not, it will
call hisi_sas_init_device(), but at that time domain_device may already be
freed or part of it is freed, so it may referenece null pointer in
hisi_sas_init_device(). It may occur as follows:
thread0 thread1
hisi_sas_dev_gone()
check whether in RESET(no)
internal task abort
reset prep
soft_reset
... (part of reset_done)
internal task abort failed
release resource anyway
clear_itct
device->lldd_dev=NULL
hisi_sas_reset_init_all_device
check sas_dev->dev_type is SAS_PHY_UNUSED and
!device
set dev_type SAS_PHY_UNUSED
sas_free_device
hisi_sas_init_device
...
Semaphore hisi_hba.sema is used to sync the processes of device gone and
host reset.
To solve the issue, expand the scope that semaphore protects and let them
never occur together.
And also some places will check whether domain_device is NULL to judge
whether the device is gone. So when device gone, need to clear
sas_dev->sas_device.
Link: https://lore.kernel.org/r/1567774537-20003-14-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Currently the NCQ tag is only assigned for FPDMA READ and FPDMA WRITE
commands, and for other NCQ commands (such as FPDMA SEND), their NCQ tags
are set in the delivery command to 0.
So for all the NCQ commands, we also need to assign normal NCQ tag for
them, so drop the command type check in hisi_sas_get_ncq_tag() [drop
hisi_sas_get_ncq_tag() altogether actually], and always use the ATA command
NCQ tag when appropriate.
Link: https://lore.kernel.org/r/1567774537-20003-8-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When init device for SAS disks, it will send TMF IO to clear disks. At that
time TMF IO is broken by some operations such as injecting controller reset
from HW RAs event, the TMF IO will be timeout, and at last device will be
gone. Print is as followed:
hisi_sas_v3_hw 0000:74:02.0: dev[240:1] found
...
hisi_sas_v3_hw 0000:74:02.0: controller resetting...
hisi_sas_v3_hw 0000:74:02.0: phyup: phy7 link_rate=10(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy0 link_rate=9(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy1 link_rate=9(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy2 link_rate=9(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy3 link_rate=9(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy6 link_rate=10(sata)
hisi_sas_v3_hw 0000:74:02.0: phyup: phy5 link_rate=11
hisi_sas_v3_hw 0000:74:02.0: phyup: phy4 link_rate=11
hisi_sas_v3_hw 0000:74:02.0: controller reset complete
hisi_sas_v3_hw 0000:74:02.0: abort tmf: TMF task timeout and not done
hisi_sas_v3_hw 0000:74:02.0: dev[240:1] is gone
sas: driver on host 0000:74:02.0 cannot handle device 5000c500a75a860d,
error:5
To improve the reliability, retry TMF IO max of 3 times for SAS disks which
is the same as softreset does.
Link: https://lore.kernel.org/r/1567774537-20003-6-git-send-email-john.garry@huawei.com
Signed-off-by: Xiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
