[dm-devel] [PATCH 1/7] dm: Fix a (theoretical?) race condition in rq_completed()

[Mike Snitzer]

On Tue, Nov 15 2016 at  6:32pm -0500,
Bart Van Assche <bart.vanassche at sandisk.com> wrote:

> It is required to hold the queue lock when calling blk_run_queue_async()
> to avoid that a race between blk_run_queue_async() and
> blk_cleanup_queue() is triggered.
> 
> Signed-off-by: Bart Van Assche <bart.vanassche at sandisk.com>

I picked this patch up earlier today, see:
https://git.kernel.org/cgit/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=dm-4.10&id=d15bb3a6467e102e60d954aadda5fb19ce6fd8ec

But you your "(theoretical?)", I'd really expected you to have realized
an actual the need for this change...

Mike

> ---
>  drivers/md/dm-rq.c | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c
> index f9f37ad..7df7948 100644
> --- a/drivers/md/dm-rq.c
> +++ b/drivers/md/dm-rq.c
> @@ -210,6 +210,9 @@ static void rq_end_stats(struct mapped_device *md, struct request *orig)
>   */
>  static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
>  {
> +	struct request_queue *q = md->queue;
> +	unsigned long flags;
> +
>  	atomic_dec(&md->pending[rw]);
>  
>  	/* nudge anyone waiting on suspend queue */
> @@ -222,8 +225,11 @@ static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
>  	 * back into ->request_fn() could deadlock attempting to grab the
>  	 * queue lock again.
>  	 */
> -	if (!md->queue->mq_ops && run_queue)
> -		blk_run_queue_async(md->queue);
> +	if (!q->mq_ops && run_queue) {
> +		spin_lock_irqsave(q->queue_lock, flags);
> +		blk_run_queue_async(q);
> +		spin_unlock_irqrestore(q->queue_lock, flags);
> +	}
>  
>  	/*
>  	 * dm_put() must be at the end of this function. See the comment above
> -- 
> 2.10.1
> 
> --
> dm-devel mailing list
> dm-devel at redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel