[PATCH V2] block: mark GFP_NOIO around sysfs ->store()

[PATCH V2] block: mark GFP_NOIO around sysfs ->store()

[Ming Lei]

sysfs ->store is called with queue freezed, meantime we have several
->store() callbacks(update_nr_requests, wbt, scheduler) to allocate
memory with GFP_KERNEL which may run into direct reclaim code path,
then potential deadlock can be caused.

Fix the issue by marking NOIO around sysfs ->store()

Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reported-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Closes: https://lore.kernel.org/linux-block/ead7c5ce5138912c1f3179d62370b84a64014a38.camel@linux.intel.com/
Fixes: bd166ef183c2 ("blk-mq-sched: add framework for MQ capable IO schedulers")
Cc: stable@vger.kernel.org
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
V2:
	- add Closes & Reviewed-by & Fixes tag

 block/blk-sysfs.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index e828be777206..e09b455874bf 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -681,6 +681,7 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr,
 	struct queue_sysfs_entry *entry = to_queue(attr);
 	struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj);
 	struct request_queue *q = disk->queue;
+	unsigned int noio_flag;
 	ssize_t res;
 
 	if (!entry->store_limit && !entry->store)
@@ -711,7 +712,9 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr,
 
 	mutex_lock(&q->sysfs_lock);
 	blk_mq_freeze_queue(q);
+	noio_flag = memalloc_noio_save();
 	res = entry->store(disk, page, length);
+	memalloc_noio_restore(noio_flag);
 	blk_mq_unfreeze_queue(q);
 	mutex_unlock(&q->sysfs_lock);
 	return res;
-- 
2.44.0

Re: [PATCH V2] block: mark GFP_NOIO around sysfs ->store()

[Thomas Hellström]

Hi!

On Mon, 2025-01-13 at 16:41 +0800, Ming Lei wrote:
> sysfs ->store is called with queue freezed, meantime we have several
> ->store() callbacks(update_nr_requests, wbt, scheduler) to allocate
> memory with GFP_KERNEL which may run into direct reclaim code path,
> then potential deadlock can be caused.
> 
> Fix the issue by marking NOIO around sysfs ->store()
> 
> Reviewed-by: Christoph Hellwig <hch@lst.de>
> Reviewed-by: John Garry <john.g.garry@oracle.com>
> Reported-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Closes:
> https://lore.kernel.org/linux-block/ead7c5ce5138912c1f3179d62370b84a64014a38.camel@linux.intel.com/
> Fixes: bd166ef183c2 ("blk-mq-sched: add framework for MQ capable IO
> schedulers")

Does this fix also the #2 lockdep splat in that email?
Thanks,

Thomas

Re: [PATCH V2] block: mark GFP_NOIO around sysfs ->store()

[Ming Lei]

On Mon, Jan 13, 2025 at 09:50:37AM +0100, Thomas Hellström wrote:
> Hi!
> 
> On Mon, 2025-01-13 at 16:41 +0800, Ming Lei wrote:
> > sysfs ->store is called with queue freezed, meantime we have several
> > ->store() callbacks(update_nr_requests, wbt, scheduler) to allocate
> > memory with GFP_KERNEL which may run into direct reclaim code path,
> > then potential deadlock can be caused.
> > 
> > Fix the issue by marking NOIO around sysfs ->store()
> > 
> > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > Reviewed-by: John Garry <john.g.garry@oracle.com>
> > Reported-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> > Closes:
> > https://lore.kernel.org/linux-block/ead7c5ce5138912c1f3179d62370b84a64014a38.camel@linux.intel.com/
> > Fixes: bd166ef183c2 ("blk-mq-sched: add framework for MQ capable IO
> > schedulers")
> 
> Does this fix also the #2 lockdep splat in that email?

No.

The #2 splat fix has been merged to for-6.14/block, and this patch only
covers the one reported in the Closes link.

https://lore.kernel.org/linux-block/20250110054726.1499538-1-hch@lst.de/


Thanks, 
Ming

Re: [PATCH V2] block: mark GFP_NOIO around sysfs ->store()

[Thomas Hellström]

On Mon, 2025-01-13 at 17:01 +0800, Ming Lei wrote:
> On Mon, Jan 13, 2025 at 09:50:37AM +0100, Thomas Hellström wrote:
> > Hi!
> > 
> > On Mon, 2025-01-13 at 16:41 +0800, Ming Lei wrote:
> > > sysfs ->store is called with queue freezed, meantime we have
> > > several
> > > ->store() callbacks(update_nr_requests, wbt, scheduler) to
> > > allocate
> > > memory with GFP_KERNEL which may run into direct reclaim code
> > > path,
> > > then potential deadlock can be caused.
> > > 
> > > Fix the issue by marking NOIO around sysfs ->store()
> > > 
> > > Reviewed-by: Christoph Hellwig <hch@lst.de>
> > > Reviewed-by: John Garry <john.g.garry@oracle.com>
> > > Reported-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> > > Closes:
> > > https://lore.kernel.org/linux-block/ead7c5ce5138912c1f3179d62370b84a64014a38.camel@linux.intel.com/
> > > Fixes: bd166ef183c2 ("blk-mq-sched: add framework for MQ capable
> > > IO
> > > schedulers")
> > 
> > Does this fix also the #2 lockdep splat in that email?
> 
> No.
> 
> The #2 splat fix has been merged to for-6.14/block, and this patch
> only
> covers the one reported in the Closes link.

I actually reported two new splats in the Closes link.

(The second was found when using the suggested lockdep priming, but
would ofc emerge sooner or later without it). I'm pretty sure
Christoph's series was applied when that patch emerged, but I can retry
if you want.

Thanks,
Thomas



> 
> https://lore.kernel.org/linux-block/20250110054726.1499538-1-hch@lst.de/
> 
> 
> Thanks, 
> Ming
>