From: Ming Lei <[email protected]>
commit d36a9ea5e7766961e753ee38d4c331bbe6ef659b upstream.
For blk-mq, queue release handler is usually called after
blk_mq_freeze_queue_wait() returns. However, the
q_usage_counter->release() handler may not be run yet at that time, so
this can cause a use-after-free.
Fix the issue by moving percpu_ref_exit() into blk_free_queue_rcu().
Since ->release() is called with rcu read lock held, it is agreed that
the race should be covered in caller per discussion from the two links.
Backport-notes: Not a clean cherry-pick since a lot has changed,
however essentially the same fix.
Reported-by: Zhang Wensheng <[email protected]>
Reported-by: Zhong Jinghua <[email protected]>
Link: https://lore.kernel.org/linux-block/Y5prfOjyyjQKUrtH@T590/T/#u
Link: https://lore.kernel.org/lkml/Y4%2FmzMd4evRg9yDi@fedora/
Cc: Hillf Danton <[email protected]>
Cc: Yu Kuai <[email protected]>
Cc: Dennis Zhou <[email protected]>
Fixes: 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path")
Signed-off-by: Ming Lei <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
Signed-off-by: Saranya Muruganandam <[email protected]>
---
block/blk-core.c | 2 --
block/blk-sysfs.c | 2 ++
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/block/blk-core.c b/block/blk-core.c
index d0d0dd8151f7..e5eeec801f56 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -414,8 +414,6 @@ void blk_cleanup_queue(struct request_queue *q)
blk_mq_sched_free_requests(q);
mutex_unlock(&q->sysfs_lock);
- percpu_ref_exit(&q->q_usage_counter);
-
/* @q is and will stay empty, shutdown and put */
blk_put_queue(q);
}
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 8c5816364dd1..9174137a913c 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -726,6 +726,8 @@ static void blk_free_queue_rcu(struct rcu_head *rcu_head)
{
struct request_queue *q = container_of(rcu_head, struct request_queue,
rcu_head);
+
+ percpu_ref_exit(&q->q_usage_counter);
kmem_cache_free(blk_requestq_cachep, q);
}
--
2.42.0.515.g380fc7ccd1-goog
On Thu, Sep 21, 2023 at 11:20:12AM -0700, Saranya Muruganandam wrote:
> From: Ming Lei <[email protected]>
>
> commit d36a9ea5e7766961e753ee38d4c331bbe6ef659b upstream.
>
> For blk-mq, queue release handler is usually called after
> blk_mq_freeze_queue_wait() returns. However, the
> q_usage_counter->release() handler may not be run yet at that time, so
> this can cause a use-after-free.
>
> Fix the issue by moving percpu_ref_exit() into blk_free_queue_rcu().
> Since ->release() is called with rcu read lock held, it is agreed that
> the race should be covered in caller per discussion from the two links.
>
> Backport-notes: Not a clean cherry-pick since a lot has changed,
> however essentially the same fix.
>
> Reported-by: Zhang Wensheng <[email protected]>
> Reported-by: Zhong Jinghua <[email protected]>
> Link: https://lore.kernel.org/linux-block/Y5prfOjyyjQKUrtH@T590/T/#u
> Link: https://lore.kernel.org/lkml/Y4%2FmzMd4evRg9yDi@fedora/
> Cc: Hillf Danton <[email protected]>
> Cc: Yu Kuai <[email protected]>
> Cc: Dennis Zhou <[email protected]>
> Fixes: 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path")
> Signed-off-by: Ming Lei <[email protected]>
> Link: https://lore.kernel.org/r/[email protected]
> Signed-off-by: Jens Axboe <[email protected]>
> Signed-off-by: Saranya Muruganandam <[email protected]>
> ---
> block/blk-core.c | 2 --
> block/blk-sysfs.c | 2 ++
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/block/blk-core.c b/block/blk-core.c
> index d0d0dd8151f7..e5eeec801f56 100644
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -414,8 +414,6 @@ void blk_cleanup_queue(struct request_queue *q)
> blk_mq_sched_free_requests(q);
> mutex_unlock(&q->sysfs_lock);
>
> - percpu_ref_exit(&q->q_usage_counter);
> -
> /* @q is and will stay empty, shutdown and put */
> blk_put_queue(q);
> }
> diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
> index 8c5816364dd1..9174137a913c 100644
> --- a/block/blk-sysfs.c
> +++ b/block/blk-sysfs.c
> @@ -726,6 +726,8 @@ static void blk_free_queue_rcu(struct rcu_head *rcu_head)
> {
> struct request_queue *q = container_of(rcu_head, struct request_queue,
> rcu_head);
> +
> + percpu_ref_exit(&q->q_usage_counter);
> kmem_cache_free(blk_requestq_cachep, q);
> }
Looks fine.
BTW, you should have provided target stable tree release info, otherwise how
you expect people to review?
Thanks,
Ming
On Thu, Sep 21, 2023 at 11:20:12AM -0700, Saranya Muruganandam wrote:
> From: Ming Lei <[email protected]>
>
> commit d36a9ea5e7766961e753ee38d4c331bbe6ef659b upstream.
>
> For blk-mq, queue release handler is usually called after
> blk_mq_freeze_queue_wait() returns. However, the
> q_usage_counter->release() handler may not be run yet at that time, so
> this can cause a use-after-free.
>
> Fix the issue by moving percpu_ref_exit() into blk_free_queue_rcu().
> Since ->release() is called with rcu read lock held, it is agreed that
> the race should be covered in caller per discussion from the two links.
>
> Backport-notes: Not a clean cherry-pick since a lot has changed,
> however essentially the same fix.
>
> Reported-by: Zhang Wensheng <[email protected]>
> Reported-by: Zhong Jinghua <[email protected]>
> Link: https://lore.kernel.org/linux-block/Y5prfOjyyjQKUrtH@T590/T/#u
> Link: https://lore.kernel.org/lkml/Y4%2FmzMd4evRg9yDi@fedora/
> Cc: Hillf Danton <[email protected]>
> Cc: Yu Kuai <[email protected]>
> Cc: Dennis Zhou <[email protected]>
> Fixes: 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path")
> Signed-off-by: Ming Lei <[email protected]>
> Link: https://lore.kernel.org/r/[email protected]
> Signed-off-by: Jens Axboe <[email protected]>
> Signed-off-by: Saranya Muruganandam <[email protected]>
> ---
> block/blk-core.c | 2 --
> block/blk-sysfs.c | 2 ++
> 2 files changed, 2 insertions(+), 2 deletions(-)
What stable kernel(s) are you expecting this backport to be applied to?
thanks,
greg "not a mind reader" k-h
Apologies for leaving out the stable release info.
This is for both 5.10 and patch applies cleanly for 5.15.
I just sent out a (different) modified patch for 6.1 LTS.
On Fri, Sep 22, 2023 at 2:26 AM Greg KH <[email protected]> wrote:
>
> On Thu, Sep 21, 2023 at 11:20:12AM -0700, Saranya Muruganandam wrote:
> > From: Ming Lei <[email protected]>
> >
> > commit d36a9ea5e7766961e753ee38d4c331bbe6ef659b upstream.
> >
> > For blk-mq, queue release handler is usually called after
> > blk_mq_freeze_queue_wait() returns. However, the
> > q_usage_counter->release() handler may not be run yet at that time, so
> > this can cause a use-after-free.
> >
> > Fix the issue by moving percpu_ref_exit() into blk_free_queue_rcu().
> > Since ->release() is called with rcu read lock held, it is agreed that
> > the race should be covered in caller per discussion from the two links.
> >
> > Backport-notes: Not a clean cherry-pick since a lot has changed,
> > however essentially the same fix.
> >
> > Reported-by: Zhang Wensheng <[email protected]>
> > Reported-by: Zhong Jinghua <[email protected]>
> > Link: https://lore.kernel.org/linux-block/Y5prfOjyyjQKUrtH@T590/T/#u
> > Link: https://lore.kernel.org/lkml/Y4%2FmzMd4evRg9yDi@fedora/
> > Cc: Hillf Danton <[email protected]>
> > Cc: Yu Kuai <[email protected]>
> > Cc: Dennis Zhou <[email protected]>
> > Fixes: 2b0d3d3e4fcf ("percpu_ref: reduce memory footprint of percpu_ref in fast path")
> > Signed-off-by: Ming Lei <[email protected]>
> > Link: https://lore.kernel.org/r/[email protected]
> > Signed-off-by: Jens Axboe <[email protected]>
> > Signed-off-by: Saranya Muruganandam <[email protected]>
> > ---
> > block/blk-core.c | 2 --
> > block/blk-sysfs.c | 2 ++
> > 2 files changed, 2 insertions(+), 2 deletions(-)
>
> What stable kernel(s) are you expecting this backport to be applied to?
>
> thanks,
>
> greg "not a mind reader" k-h
On Fri, Sep 22, 2023 at 11:21:10AM -0700, Saranya Muruganandam wrote:
> Apologies for leaving out the stable release info.
> This is for both 5.10 and patch applies cleanly for 5.15.
>
> I just sent out a (different) modified patch for 6.1 LTS.
All now queued up, thanks.
greg k-h