The ALUA handler maps a 0 value (no implicit transition timeout provided
by the target) to the ALUA_FAILOVER_TIMEOUT constant, currently 60
seconds. This means the kernel already does not accept an infinite
transition time.
However, 60 seconds is insufficient for some arrays that may take longer
to complete ALUA transitions. Since the highest value allowed by the
SCSI specification for the implicit transition timeout is a single byte
(255 seconds), change the default to 255. This way, when a target does
not provide an explicit transition timeout, we default to the maximum
value the spec allows rather than an arbitrary 60 second limit.
Co-developed-by: Krishna Kant <krishna.kant@purestorage.com>
Signed-off-by: Krishna Kant <krishna.kant@purestorage.com>
Co-developed-by: Riya Savla <rsavla@purestorage.com>
Signed-off-by: Riya Savla <rsavla@purestorage.com>
Signed-off-by: Brian Bunker <brian@purestorage.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://patch.msgid.link/20260416165512.26497-2-brian@purestorage.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
This was done entirely with mindless brute force, using
git grep -l '\<k[vmz]*alloc_objs*(.*, GFP_KERNEL)' |
xargs sed -i 's/\(alloc_objs*(.*\), GFP_KERNEL)/\1)/'
to convert the new alloc_obj() users that had a simple GFP_KERNEL
argument to just drop that argument.
Note that due to the extreme simplicity of the scripting, any slightly
more complex cases spread over multiple lines would not be triggered:
they definitely exist, but this covers the vast bulk of the cases, and
the resulting diff is also then easier to check automatically.
For the same reason the 'flex' versions will be done as a separate
conversion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.
alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.
This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.
This change adds a new WQ_PERCPU flag to explicitly request
alloc_workqueue() to be per-cpu when WQ_UNBOUND has not been specified.
With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.
Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Link: https://patch.msgid.link/20251031095643.74246-4-marco.crivellari@suse.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
asm/unaligned.h is always an include of asm-generic/unaligned.h;
might as well move that thing to linux/unaligned.h and include
that - there's nothing arch-specific in that header.
auto-generated by the following:
for i in `git grep -l -w asm/unaligned.h`; do
sed -i -e "s/asm\/unaligned.h/linux\/unaligned.h/" $i
done
for i in `git grep -l -w asm-generic/unaligned.h`; do
sed -i -e "s/asm-generic\/unaligned.h/linux\/unaligned.h/" $i
done
git mv include/asm-generic/unaligned.h include/linux/unaligned.h
git mv tools/include/asm-generic/unaligned.h tools/include/linux/unaligned.h
sed -i -e "/unaligned.h/d" include/asm-generic/Kbuild
sed -i -e "s/__ASM_GENERIC/__LINUX/" include/linux/unaligned.h tools/include/linux/unaligned.h
The workqueue maintainer wants to remove the create*_workqueue() macros
because these macros always set the WQ_MEM_RECLAIM flag and because these
only support literal workqueue names. Hence this patch that replaces the
create*_workqueue() invocations with the definition of this macro. The
WQ_MEM_RECLAIM flag has been retained because I think that flag is necessary
for workqueues created by storage drivers. This patch has been generated by
running spatch and git clang-format. spatch has been invoked as follows:
spatch --in-place --sp-file expand-create-workqueue.spatch $(git grep -lEw 'create_(freezable_|singlethread_|)workqueue' */scsi */ufs)
The contents of the expand-create-workqueue.spatch file is as follows:
@@
expression name;
@@
-create_workqueue(name)
+alloc_workqueue("%s", WQ_MEM_RECLAIM, 1, name)
@@
expression name;
@@
-create_freezable_workqueue(name)
+alloc_workqueue("%s", WQ_FREEZABLE | WQ_UNBOUND | WQ_MEM_RECLAIM, 1, name)
@@
expression name;
@@
-create_singlethread_workqueue(name)
+alloc_ordered_workqueue("%s", WQ_MEM_RECLAIM, name)
Reviewed-by: Peter Wang <peter.wang@mediatek.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240822195944.654691-2-bvanassche@acm.org
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
When a host is configured with a few LUNs and I/O is running, injecting FC
faults repeatedly leads to path recovery problems. The LUNs have 4 paths
each and 3 of them come back active after say an FC fault which makes 2 of
the paths go down, instead of all 4. This happens after several iterations
of continuous FC faults.
Reason here is that we're returning an I/O error whenever we're
encountering sense code 06/04/0a (LOGICAL UNIT NOT ACCESSIBLE, ASYMMETRIC
ACCESS STATE TRANSITION) instead of retrying.
[mwilck: The original patch was developed by Rajashekhar M A and Hannes
Reinecke. I moved the code to alua_check_sense() as suggested by Mike
Christie [1]. Evan Milne had raised the question whether pg->state should
be set to transitioning in the UA case [2]. I believe that doing this is
correct. SCSI_ACCESS_STATE_TRANSITIONING by itself doesn't cause I/O
errors. Our handler schedules an RTPG, which will only result in an I/O
error condition if the transitioning timeout expires.]
[1] https://lore.kernel.org/all/0bc96e82-fdda-4187-148d-5b34f81d4942@oracle.com/
[2] https://lore.kernel.org/all/CAGtn9r=kicnTDE2o7Gt5Y=yoidHYD7tG8XdMHEBJTBraVEoOCw@mail.gmail.com/
Co-developed-by: Rajashekhar M A <rajs@netapp.com>
Co-developed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Martin Wilck <martin.wilck@suse.com>
Link: https://lore.kernel.org/r/20240514140344.19538-1-mwilck@suse.com
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
If send_mode_select retries scsi_execute_cmd it will leave err set to
SCSI_DH_RETRY/SCSI_DH_IMM_RETRY. If on the retry, the command is
successful, then SCSI_DH_RETRY/SCSI_DH_IMM_RETRY will be returned to the
scsi_dh activation caller. On the retry, we will then detect the previous
MODE SELECT had worked, and so we will return success.
This patch has us return the correct return value, so we can avoid the
extra scsi_dh activation call and to avoid failures if the caller had hit
its activation retry limit and does not end up retrying.
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Link: https://lore.kernel.org/r/20231004210013.5601-5-michael.christie@oracle.com
Reviewed-by: Martin Wilck <mwilck@suse.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Pull SCSI updates from James Bottomley:
"Updates to the usual drivers (ufs, lpfc, qla2xxx, libsas).
The major core change is a rework to remove the two helpers around
scsi_execute_cmd and use it as the only submission interface along
with other minor fixes and updates"
* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (142 commits)
scsi: ufs: core: Fix an error handling path in ufshcd_read_desc_param()
scsi: ufs: core: Fix device management cmd timeout flow
scsi: aic94xx: Add missing check for dma_map_single()
scsi: smartpqi: Replace one-element array with flexible-array member
scsi: mpt3sas: Fix a memory leak
scsi: qla2xxx: Remove the unused variable wwn
scsi: ufs: core: Fix kernel-doc syntax
scsi: ufs: core: Add hibernation callbacks
scsi: snic: Fix memory leak with using debugfs_lookup()
scsi: ufs: core: Limit DMA alignment check
scsi: Documentation: Correct spelling
scsi: Documentation: Correct spelling
scsi: target: Documentation: Correct spelling
scsi: aacraid: Allocate cmd_priv with scsicmd
scsi: ufs: qcom: dt-bindings: Add SM8550 compatible string
scsi: ufs: ufs-qcom: Clear qunipro_g4_sel for HW version major 5
scsi: ufs: qcom: fix platform_msi_domain_free_irqs() reference
scsi: ufs: core: Enable DMA clustering
scsi: ufs: exynos: Fix the maximum segment size
scsi: ufs: exynos: Fix DMA alignment for PAGE_SIZE != 4096
...
The might_sleep() annotation in alua_rtpg_queue() is not correct since the
command completion code may call this function from atomic context.
Calling alua_rtpg_queue() from atomic context in the command completion
path is fine since request submitters must hold an sdev reference until
command execution has completed. This patch fixes the following kernel
complaint:
BUG: sleeping function called from invalid context at drivers/scsi/device_handler/scsi_dh_alua.c:992
Call Trace:
dump_stack_lvl+0xac/0x100
__might_resched+0x284/0x2c8
alua_rtpg_queue+0x3c/0x98 [scsi_dh_alua]
alua_check+0x122/0x250 [scsi_dh_alua]
alua_check_sense+0x172/0x228 [scsi_dh_alua]
scsi_check_sense+0x8a/0x2e0
scsi_decide_disposition+0x286/0x298
scsi_complete+0x6a/0x108
blk_complete_reqs+0x6e/0x88
__do_softirq+0x13e/0x6b8
__irq_exit_rcu+0x14a/0x170
irq_exit_rcu+0x22/0x50
do_ext_irq+0x10a/0x1d0
Link: https://lore.kernel.org/r/20230118180557.1212577-1-bvanassche@acm.org
Reported-by: Steffen Maier <maier@linux.ibm.com>
Reviewed-by: Martin Wilck <mwilck@suse.com>
Tested-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Since commit f93ed747e2 ("scsi: core: Release SCSI devices
synchronously"), scsi_device_put() might sleep. Avoid calling it from
alua_rtpg_queue() with the pg_lock held. The lock only pretects h->pg,
anyway. To avoid the pg being freed under us, because of a race with
another thread, take a temporary reference. In alua_rtpg_queue(), verify
that the pg still belongs to the sdev being passed before actually queueing
the RTPG.
This patch fixes the following smatch warning:
drivers/scsi/device_handler/scsi_dh_alua.c:1013 alua_rtpg_queue() warn: sleeping in atomic context
alua_check_vpd() <- disables preempt
-> alua_rtpg_queue()
-> scsi_device_put()
Cc: Martin Wilck <mwilck@suse.com>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Sachin Sant <sachinp@linux.ibm.com>
Cc: Benjamin Block <bblock@linux.ibm.com>
Suggested-by: Martin Wilck <mwilck@suse.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20221117183626.2656196-3-bvanassche@acm.org
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The handling of the ALUA transitioning state is currently broken. When a
target goes into this state, it is expected that the target is allowed to
stay in this state for the implicit transition timeout without a path
failure. The handler has this logic, but it gets skipped currently.
When the target transitions, there is in-flight I/O from the initiator. The
first of these responses from the target will be a unit attention letting
the initiator know that the ALUA state has changed. The remaining
in-flight I/Os, before the initiator finds out that the portal state has
changed, will return not ready, ALUA state is transitioning. The portal
state will change to SCSI_ACCESS_STATE_TRANSITIONING. This will lead to all
new I/O immediately failing the path unexpectedly. The path failure happens
in less than a second instead of the expected successes until the
transition timer is exceeded.
Allow I/Os to continue while the path is in the ALUA transitioning
state. The handler already takes care of a target that stays in the
transitioning state for too long by changing the state to ALUA state
standby once the transition timeout is exceeded at which point the path
will fail.
Link: https://lore.kernel.org/r/CAHZQxy+4sTPz9+pY3=7VJH+CLUJsDct81KtnR2be8ycN5mhqTg@mail.gmail.com
Reviewed-by: Hannes Reinecke <hare@suse.de>
Acked-by: Krishna Kant <krishna.kant@purestorage.com>
Acked-by: Seamus Connor <sconnor@purestorage.com>
Signed-off-by: Brian Bunker <brian@purestorage.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
