commit 176ddd89171ddcf661862d90c5d257877f7326d6 upstream.
When the cache_type for the SCSI device is changed, the SCSI layer issues a
MODE_SELECT command. The caching mode details are communicated via a
request buffer associated with the SCSI command with data direction set as
DMA_TO_DEVICE (scsi_mode_select()). When this command reaches the libata
layer, as a part of generic initial setup, libata layer sets up the
scatterlist for the command using the SCSI command (ata_scsi_qc_new()).
This command is then translated by the libata layer into
ATA_CMD_SET_FEATURES (ata_scsi_mode_select_xlat()). The libata layer treats
this as a non-data command (ata_mselect_caching()), since it only needs an
ATA taskfile to pass the caching on/off information to the device. It does
not need the scatterlist that has been setup, so it does not perform
dma_map_sg() on the scatterlist (ata_qc_issue()). Unfortunately, when this
command reaches the libsas layer (sas_ata_qc_issue()), libsas layer sees it
as a non-data command with a scatterlist. It cannot extract the correct DMA
length since the scatterlist has not been mapped with dma_map_sg() for a
DMA operation. When this partially constructed SAS task reaches pm80xx
LLDD, it results in the following warning:
"pm80xx_chip_sata_req 6058: The sg list address
start_addr=0x0000000000000000 data_len=0x0end_addr_high=0xffffffff
end_addr_low=0xffffffff has crossed 4G boundary"
Update libsas to handle ATA non-data commands separately so num_scatter and
total_xfer_len remain 0.
Link: https://lore.kernel.org/r/20210318225632.2481291-1-jollys@google.com
Fixes: 53de092f47 ("scsi: libsas: Set data_dir as DMA_NONE if libata marks qc as NODATA")
Tested-by: Luo Jiaxing <luojiaxing@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Signed-off-by: Jolly Shah <jollys@google.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 5cd0f6f57639c5afbb36100c69281fee82c95ee7 ]
rport_dev_loss_timedout() sets the rport state to SRP_PORT_LOST and the
SCSI target state to SDEV_TRANSPORT_OFFLINE. If this races with
srp_reconnect_work(), a warning is printed:
Mar 27 18:48:07 ictm1604s01h4 kernel: dev_loss_tmo expired for SRP port-18:1 / host18.
Mar 27 18:48:07 ictm1604s01h4 kernel: ------------[ cut here ]------------
Mar 27 18:48:07 ictm1604s01h4 kernel: scsi_internal_device_block(18:0:0:100) failed: ret = -22
Mar 27 18:48:07 ictm1604s01h4 kernel: Call Trace:
Mar 27 18:48:07 ictm1604s01h4 kernel: ? scsi_target_unblock+0x50/0x50 [scsi_mod]
Mar 27 18:48:07 ictm1604s01h4 kernel: starget_for_each_device+0x80/0xb0 [scsi_mod]
Mar 27 18:48:07 ictm1604s01h4 kernel: target_block+0x24/0x30 [scsi_mod]
Mar 27 18:48:07 ictm1604s01h4 kernel: device_for_each_child+0x57/0x90
Mar 27 18:48:07 ictm1604s01h4 kernel: srp_reconnect_rport+0xe4/0x230 [scsi_transport_srp]
Mar 27 18:48:07 ictm1604s01h4 kernel: srp_reconnect_work+0x40/0xc0 [scsi_transport_srp]
Avoid this by not trying to block targets for rports in SRP_PORT_LOST
state.
Link: https://lore.kernel.org/r/20210401091105.8046-1-mwilck@suse.com
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin Wilck <mwilck@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1235fc569e0bf541ddda0a1224d4c6fa6d914890 ]
ufshcd_tmc_handler() calls blk_mq_tagset_busy_iter(fn = ufshcd_compl_tm()),
but since blk_mq_tagset_busy_iter() only iterates over all reserved tags
and requests which are not in IDLE state, ufshcd_compl_tm() never gets a
chance to run. Thus, TMR always ends up with completion timeout. Fix it by
calling blk_mq_start_request() in __ufshcd_issue_tm_cmd().
Link: https://lore.kernel.org/r/1617262750-4864-2-git-send-email-cang@codeaurora.org
Fixes: 69a6c269c0 ("scsi: ufs: Use blk_{get,put}_request() to allocate and free TMFs")
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 39c0c8553bfb5a3d108aa47f1256076d507605e3 ]
Calling vha->hw->tgt.tgt_ops->free_cmd() from qlt_xmit_response() is wrong
since the command for which a response is sent must remain valid until the
SCSI target core calls .release_cmd(). It has been observed that the
following scenario triggers a kernel crash:
- qlt_xmit_response() calls qlt_check_reserve_free_req()
- qlt_check_reserve_free_req() returns -EAGAIN
- qlt_xmit_response() calls vha->hw->tgt.tgt_ops->free_cmd(cmd)
- transport_handle_queue_full() tries to retransmit the response
Fix this crash by reverting the patch that introduced it.
Link: https://lore.kernel.org/r/20210320232359.941-2-bvanassche@acm.org
Fixes: 0dcec41acb ("scsi: qla2xxx: Make sure that aborted commands are freed")
Cc: Quinn Tran <qutran@marvell.com>
Cc: Mike Christie <michael.christie@oracle.com>
Reviewed-by: Daniel Wagner <dwagner@suse.de>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit feb18e900f0048001ff375dca639eaa327ab3c1b ]
mvsas calls the non _gfp version of the libsas event notifiers API, leading
to the buggy call chains below:
mvsas/mv_sas.c: mvs_work_queue() [process context]
spin_lock_irqsave(mvs_info::lock, )
-> libsas/sas_event.c: sas_notify_phy_event()
-> sas_alloc_event()
-> in_interrupt() = false
-> invalid GFP_KERNEL allocation
-> libsas/sas_event.c: sas_notify_port_event()
-> sas_alloc_event()
-> in_interrupt() = false
-> invalid GFP_KERNEL allocation
Use the new event notifiers API instead, which requires callers to
explicitly pass the gfp_t memory allocation flags.
Below are context analysis for the modified functions:
=> mvs_bytes_dmaed():
Since it is invoked from both process and atomic contexts, let its callers
pass the gfp_t flags. Call chains:
scsi_scan.c: do_scsi_scan_host() [has msleep()]
-> shost->hostt->scan_start()
-> [mvsas/mv_init.c: Scsi_Host::scsi_host_template .scan_start = mvs_scan_start()]
-> mvsas/mv_sas.c: mvs_scan_start()
-> mvs_bytes_dmaed(..., GFP_KERNEL)
mvsas/mv_sas.c: mvs_work_queue()
spin_lock_irqsave(mvs_info::lock,)
-> mvs_bytes_dmaed(..., GFP_ATOMIC)
mvsas/mv_64xx.c: mvs_64xx_isr() || mvsas/mv_94xx.c: mvs_94xx_isr()
-> mvsas/mv_chips.h: mvs_int_full()
-> mvsas/mv_sas.c: mvs_int_port()
-> mvs_bytes_dmaed(..., GFP_ATOMIC);
=> mvs_work_queue():
Invoked from process context, but it calls all the libsas event notifier
APIs under a spin_lock_irqsave(). Pass GFP_ATOMIC.
Link: https://lore.kernel.org/r/20210118100955.1761652-5-a.darwish@linutronix.de
Fixes: 1c393b970e ("scsi: libsas: Use dynamic alloced work to avoid sas event lost")
Cc: Jason Yan <yanaijie@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c2d0f1a65ab9fbabebb463bf36f50ea8f4633386 ]
sas_alloc_event() uses in_interrupt() to decide which allocation should be
used.
The usage of in_interrupt() in drivers is phased out and Linus clearly
requested that code which changes behaviour depending on context should
either be separated or the context be conveyed in an argument passed by the
caller, which usually knows the context.
The in_interrupt() check is also only partially correct, because it fails
to choose the correct code path when just preemption or interrupts are
disabled. For example, as in the following call chain:
mvsas/mv_sas.c: mvs_work_queue() [process context]
spin_lock_irqsave(mvs_info::lock, )
-> libsas/sas_event.c: sas_notify_phy_event()
-> sas_alloc_event()
-> in_interrupt() = false
-> invalid GFP_KERNEL allocation
-> libsas/sas_event.c: sas_notify_port_event()
-> sas_alloc_event()
-> in_interrupt() = false
-> invalid GFP_KERNEL allocation
Introduce sas_alloc_event_gfp(), sas_notify_port_event_gfp(), and
sas_notify_phy_event_gfp(), which all behave like the non _gfp() variants
but use a caller-passed GFP mask for allocations.
For bisectability, all callers will be modified first to pass GFP context,
then the non _gfp() libsas API variants will be modified to take a gfp_t by
default.
Link: https://lore.kernel.org/r/20210118100955.1761652-4-a.darwish@linutronix.de
Fixes: 1c393b970e ("scsi: libsas: Use dynamic alloced work to avoid sas event lost")
Cc: Jason Yan <yanaijie@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 121181f3f839c29d8dd9fdc3cc9babbdc74227f8 ]
LLDDs report events to libsas with .notify_port_event and .notify_phy_event
callbacks.
These callbacks are fixed and so there is no reason why the functions
cannot be called directly, so do that.
This neatens the code slightly, makes it more obvious, and reduces function
pointer usage, which is generally a good thing. Downside is that there are
2x more symbol exports.
[a.darwish@linutronix.de: Remove the now unused "sas_ha" local variables]
Link: https://lore.kernel.org/r/20210118100955.1761652-3-a.darwish@linutronix.de
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: John Garry <john.garry@huawei.com>
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1b5d2793283dcb97b401b3b2c02b8a94eee29af1 ]
Every PM8001_<FOO>_DBG macro uses an internal call to pm8001_printk.
Convert all uses of:
PM8001_<FOO>_DBG(hba, pm8001_printk(fmt, ...))
to
pm8001_dbg(hba, <FOO>, fmt, ...)
so the visual complexity of each macro is reduced.
The repetitive macro definitions are converted to a single pm8001_dbg and
the level is concatenated using PM8001_##level##_LOGGING for the specific
level test.
Done with coccinelle, checkpatch and a little typing of the new macro
definition.
Miscellanea:
- Coalesce formats
- Realign arguments
- Add missing terminating newlines to formats
- Remove trailing spaces from formats
- Change defective loop with printk(KERN_INFO... to emit a 16 byte hex
block to %p16h
Link: https://lore.kernel.org/r/49f36a93af7752b613d03c89a87078243567fd9a.1605914030.git.joe@perches.com
Reported-by: kernel test robot <lkp@intel.com>
Acked-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1f889b58716a5f5e3e4fe0e6742c1a4472f29ac1 ]
A use-after-free or null-pointer error occurs when the 251-byte response
data is copied from IOMB buffer to response message buffer in function
pm8001_mpi_get_nvmd_resp().
After sending the command get_nvmd_data(), the caller begins to sleep by
calling wait_for_complete() and waits for the wake-up from calling
complete() in pm8001_mpi_get_nvmd_resp(). Due to unexpected events (e.g.,
interrupt), if response buffer gets freed before memcpy(), a use-after-free
error will occur. To fix this, the complete() should be called after
memcpy().
Link: https://lore.kernel.org/r/20201102165528.26510-5-Viswas.G@microchip.com.com
Acked-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: yuuzheng <yuuzheng@google.com>
Signed-off-by: Viswas G <Viswas.G@microchip.com>
Signed-off-by: Ruksar Devadi <Ruksar.devadi@microchip.com>
Signed-off-by: Radha Ramachandran <radha@google.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 4a2efd4b89fcaa6e9a7b4ce49a441afaacba00ea ]
Incorrect value of the running_req was causing the driver unload to be
stuck during the SAS lldd_dev_gone notification handling. During SATA I/O
completion, for some error status values, the driver schedules the event
handler and running_req is decremented from that. However, there are some
other error status values (like IO_DS_IN_RECOVERY,
IO_XFER_ERR_LAST_PIO_DATAIN_CRC_ERR) where the I/O has already been
completed by fw/driver so running_req is not decremented.
Also during NCQ error handling, driver itself will initiate READ_LOG_EXT
and ABORT_ALL. When libsas/libata initiate READ_LOG_EXT (0x2F), driver
increments running_req. This will be completed by the driver in
pm80xx_chip_sata_req(), but running_req was not decremented.
Link: https://lore.kernel.org/r/20201102165528.26510-3-Viswas.G@microchip.com.com
Acked-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: Viswas G <Viswas.G@microchip.com>
Signed-off-by: Ruksar Devadi <Ruksar.devadi@microchip.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 7640e1eb8c5de33dafa6c68fd4389214ff9ec1f9 ]
Driver submits all internal requests (like abort_task, event acknowledgment
etc.) through inbound queue 0. While submitting those, driver does not
acquire any lock and this may lead to a race when there is an I/O request
coming in on CPU0 and submitted through inbound queue 0. To avoid this,
lock acquisition has been moved to pm8001_mpi_build_cmd(). All command
submission will go through this path.
Link: https://lore.kernel.org/r/20201102165528.26510-2-Viswas.G@microchip.com.com
Acked-by: Jack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: peter chang <dpf@google.com>
Signed-off-by: Viswas G <Viswas.G@microchip.com>
Signed-off-by: Ruksar Devadi <Ruksar.devadi@microchip.com>
Signed-off-by: Radha Ramachandran <radha@google.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d28d48c699779973ab9a3bd0e5acfa112bd4fdef ]
If iscsi_prep_scsi_cmd_pdu() fails we try to add it back to the cmdqueue,
but we leave it partially setup. We don't have functions that can undo the
pdu and init task setup. We only have cleanup_task which can clean up both
parts. So this has us just fail the cmd and go through the standard cleanup
routine and then have the SCSI midlayer retry it like is done when it fails
in the queuecommand path.
Link: https://lore.kernel.org/r/20210207044608.27585-2-michael.christie@oracle.com
Reviewed-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit a2fca52ee640a04112ed9d9a137c940ea6ad288e ]
Kernel stack violation when getting unit_descriptor/wb_buf_alloc_units from
rpmb LUN. The reason is that the unit descriptor length is different per
LU.
The length of Normal LU is 45 while the one of rpmb LU is 35.
int ufshcd_read_desc_param(struct ufs_hba *hba, ...)
{
param_offset=41;
param_size=4;
buff_len=45;
...
buff_len=35 by rpmb LU;
if (is_kmalloc) {
/* Make sure we don't copy more data than available */
if (param_offset + param_size > buff_len)
param_size = buff_len - param_offset;
--> param_size = 250;
memcpy(param_read_buf, &desc_buf[param_offset], param_size);
--> memcpy(param_read_buf, desc_buf+41, 250);
[ 141.868974][ T9174] Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: wb_buf_alloc_units_show+0x11c/0x11c
}
}
Link: https://lore.kernel.org/r/20210111095927.1830311-1-jaegeuk@kernel.org
Reviewed-by: Avri Altman <avri.altman@wdc.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 2b2bfc8aa519f696087475ed8e8c61850c673272 ]
Some SoCs require a single scatterlist entry for smaller than page size,
i.e. 4KB. When dispatching commands with more than one scatterlist entry
under 4KB in size the following behavior is observed:
A command to read a block range is dispatched with two scatterlist entries
that are named AAA and BBB. After dispatching, the host builds two PRDT
entries and during transmission, device sends just one DATA IN because
device doesn't care about host DMA. The host then transfers the combined
amount of data from start address of the area named AAA. As a consequence,
the area that follows AAA in memory would be corrupted.
|<------------->|
+-------+------------ +-------+
+ AAA + (corrupted) ... + BBB +
+-------+------------ +-------+
To avoid this we need to enforce page size alignment for sg entries.
Link: https://lore.kernel.org/r/56dddef94f60bd9466fd77e69f64bbbd657ed2a1.1611026909.git.kwmad.kim@samsung.com
Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b1d0d2eb89d4e3a25b212a9d836587503537067e ]
The UniPro specification states that attribute IDs of the following
parameters are vendor-specific so some SoCs could have no regions at the
defined addresses:
- DME_LocalFC0ProtectionTimeOutVal
- DME_LocalTC0ReplayTimeOutVal
- DME_LocalAFC0ReqTimeOutVal
In addition, the following parameters should be set considering the
compatibility between host and device.
- PA_PWRMODEUSERDATA0
- PA_PWRMODEUSERDATA1
- PA_PWRMODEUSERDATA2
- PA_PWRMODEUSERDATA3
- PA_PWRMODEUSERDATA4
- PA_PWRMODEUSERDATA5
Introduce a quirk to allow vendor drivers to override the UniPro defaults.
Link: https://lore.kernel.org/r/1fedd3dea0ccc980913a5995a10510d86a5b01b9.1608513782.git.kwmad.kim@samsung.com
Acked-by: Avri Altman <Avri.Altman@wdc.com>
Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit f9dbdf97a5bd92b1a49cee3d591b55b11fd7a6d5 upstream.
Open-iSCSI sends passthrough PDUs over netlink, but the kernel should be
verifying that the provided PDU header and data lengths fall within the
netlink message to prevent accessing beyond that in memory.
Cc: stable@vger.kernel.org
Reported-by: Adam Nichols <adam@grimm-co.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Reviewed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Chris Leech <cleech@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ec98ea7070e94cc25a422ec97d1421e28d97b7ee upstream.
As the iSCSI parameters are exported back through sysfs, it should be
enforcing that they never are more than PAGE_SIZE (which should be more
than enough) before accepting updates through netlink.
Change all iSCSI sysfs attributes to use sysfs_emit().
Cc: stable@vger.kernel.org
Reported-by: Adam Nichols <adam@grimm-co.com>
Reviewed-by: Lee Duncan <lduncan@suse.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Chris Leech <cleech@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 688e8128b7a92df982709a4137ea4588d16f24aa upstream.
Protect the iSCSI transport handle, available in sysfs, by requiring
CAP_SYS_ADMIN to read it. Also protect the netlink socket by restricting
reception of messages to ones sent with CAP_SYS_ADMIN. This disables
normal users from being able to end arbitrary iSCSI sessions.
Cc: stable@vger.kernel.org
Reported-by: Adam Nichols <adam@grimm-co.com>
Reviewed-by: Chris Leech <cleech@redhat.com>
Reviewed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Lee Duncan <lduncan@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 8c65830ae1629b03e5d65e9aafae7e2cf5f8b743 ]
In testing, in a configuration with Redfish and native NVMe multipath when
an EEH is injected, a kernel oops is being encountered:
(unreliable)
lpfc_nvme_ls_req+0x328/0x720 [lpfc]
__nvme_fc_send_ls_req.constprop.13+0x1d8/0x3d0 [nvme_fc]
nvme_fc_create_association+0x224/0xd10 [nvme_fc]
nvme_fc_reset_ctrl_work+0x110/0x154 [nvme_fc]
process_one_work+0x304/0x5d
the NBMe transport is issuing a Disconnect LS request, which the driver
receives and tries to post but the work queue used by the driver is already
being torn down by the eeh.
Fix by validating the validity of the work queue before proceeding with the
LS transmit.
Link: https://lore.kernel.org/r/20210127221601.84878-1-jsmart2021@gmail.com
Reviewed-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 764907293edc1af7ac857389af9dc858944f53dc ]
While testing live partition mobility, we have observed occasional crashes
of the Linux partition. What we've seen is that during the live migration,
for specific configurations with large amounts of memory, slow network
links, and workloads that are changing memory a lot, the partition can end
up being suspended for 30 seconds or longer. This resulted in the following
scenario:
CPU 0 CPU 1
------------------------------- ----------------------------------
scsi_queue_rq migration_store
-> blk_mq_start_request -> rtas_ibm_suspend_me
-> blk_add_timer -> on_each_cpu(rtas_percpu_suspend_me
_______________________________________V
|
V
-> IPI from CPU 1
-> rtas_percpu_suspend_me
-> __rtas_suspend_last_cpu
-- Linux partition suspended for > 30 seconds --
-> for_each_online_cpu(cpu)
plpar_hcall_norets(H_PROD
-> scsi_dispatch_cmd
-> scsi_times_out
-> scsi_abort_command
-> queue_delayed_work
-> ibmvfc_queuecommand_lck
-> ibmvfc_send_event
-> ibmvfc_send_crq
- returns H_CLOSED
<- returns SCSI_MLQUEUE_HOST_BUSY
-> __blk_mq_requeue_request
-> scmd_eh_abort_handler
-> scsi_try_to_abort_cmd
- returns SUCCESS
-> scsi_queue_insert
Normally, the SCMD_STATE_COMPLETE bit would protect against the command
completion and the timeout, but that doesn't work here, since we don't
check that at all in the SCSI_MLQUEUE_HOST_BUSY path.
In this case we end up calling scsi_queue_insert on a request that has
already been queued, or possibly even freed, and we crash.
The patch below simply increases the default I/O timeout to avoid this race
condition. This is also the timeout value that nearly all IBM SAN storage
recommends setting as the default value.
Link: https://lore.kernel.org/r/1610463998-19791-1-git-send-email-brking@linux.vnet.ibm.com
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b2b0f16fa65e910a3ec8771206bb49ee87a54ac5 ]
A race condition exists between the response handler getting called because
of exchange_mgr_reset() (which clears out all the active XIDs) and the
response we get via an interrupt.
Sequence of events:
rport ba0200: Port timeout, state PLOGI
rport ba0200: Port entered PLOGI state from PLOGI state
xid 1052: Exchange timer armed : 20000 msecs xid timer armed here
rport ba0200: Received LOGO request while in state PLOGI
rport ba0200: Delete port
rport ba0200: work event 3
rport ba0200: lld callback ev 3
bnx2fc: rport_event_hdlr: event = 3, port_id = 0xba0200
bnx2fc: ba0200 - rport not created Yet!!
/* Here we reset any outstanding exchanges before
freeing rport using the exch_mgr_reset() */
xid 1052: Exchange timer canceled
/* Here we got two responses for one xid */
xid 1052: invoking resp(), esb 20000000 state 3
xid 1052: invoking resp(), esb 20000000 state 3
xid 1052: fc_rport_plogi_resp() : ep->resp_active 2
xid 1052: fc_rport_plogi_resp() : ep->resp_active 2
Skip the response if the exchange is already completed.
Link: https://lore.kernel.org/r/20201215194731.2326-1-jhasan@marvell.com
Signed-off-by: Javed Hasan <jhasan@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit eeb1b55b6e25c5f7265ff45cd050f3bc2cc423a4 ]
When non-fatal error like line-reset happens, ufshcd_err_handler() starts
to abort tasks by ufshcd_try_to_abort_task(). When it tries to issue a task
management request, we hit two warnings:
WARNING: CPU: 7 PID: 7 at block/blk-core.c:630 blk_get_request+0x68/0x70
WARNING: CPU: 4 PID: 157 at block/blk-mq-tag.c:82 blk_mq_get_tag+0x438/0x46c
After fixing the above warnings we hit another tm_cmd timeout which may be
caused by unstable controller state:
__ufshcd_issue_tm_cmd: task management cmd 0x80 timed-out
Then, ufshcd_err_handler() enters full reset, and kernel gets stuck. It
turned out ufshcd_print_trs() printed too many messages on console which
requires CPU locks. Likewise hba->silence_err_logs, we need to avoid too
verbose messages. This is actually not an error case.
Link: https://lore.kernel.org/r/20210107185316.788815-3-jaegeuk@kernel.org
Fixes: 69a6c269c0 ("scsi: ufs: Use blk_{get,put}_request() to allocate and free TMFs")
Reviewed-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
