[6.0,479/862] sbitmap: fix possible io hung due to lost wakeup

Message ID 20221019083311.114449669@linuxfoundation.org
State New
Headers
Series None |

Commit Message

Greg KH Oct. 19, 2022, 8:29 a.m. UTC
  From: Yu Kuai <yukuai3@huawei.com>

[ Upstream commit 040b83fcecfb86f3225d3a5de7fd9b3fbccf83b4 ]

There are two problems can lead to lost wakeup:

1) invalid wakeup on the wrong waitqueue:

For example, 2 * wake_batch tags are put, while only wake_batch threads
are woken:

__sbq_wake_up
 atomic_cmpxchg -> reset wait_cnt
			__sbq_wake_up -> decrease wait_cnt
			...
			__sbq_wake_up -> wait_cnt is decreased to 0 again
			 atomic_cmpxchg
			 sbq_index_atomic_inc -> increase wake_index
			 wake_up_nr -> wake up and waitqueue might be empty
 sbq_index_atomic_inc -> increase again, one waitqueue is skipped
 wake_up_nr -> invalid wake up because old wakequeue might be empty

To fix the problem, increasing 'wake_index' before resetting 'wait_cnt'.

2) 'wait_cnt' can be decreased while waitqueue is empty

As pointed out by Jan Kara, following race is possible:

CPU1				CPU2
__sbq_wake_up			 __sbq_wake_up
 sbq_wake_ptr()			 sbq_wake_ptr() -> the same
 wait_cnt = atomic_dec_return()
 /* decreased to 0 */
 sbq_index_atomic_inc()
 /* move to next waitqueue */
 atomic_set()
 /* reset wait_cnt */
 wake_up_nr()
 /* wake up on the old waitqueue */
				 wait_cnt = atomic_dec_return()
				 /*
				  * decrease wait_cnt in the old
				  * waitqueue, while it can be
				  * empty.
				  */

Fix the problem by waking up before updating 'wake_index' and
'wait_cnt'.

With this patch, noted that 'wait_cnt' is still decreased in the old
empty waitqueue, however, the wakeup is redirected to a active waitqueue,
and the extra decrement on the old empty waitqueue is not handled.

Fixes: 88459642cba4 ("blk-mq: abstract tag allocation out into sbitmap library")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220803121504.212071-1-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 lib/sbitmap.c | 55 ++++++++++++++++++++++++++++++---------------------
 1 file changed, 33 insertions(+), 22 deletions(-)
  

Comments

Hugh Dickins Oct. 19, 2022, 3:06 p.m. UTC | #1
On Wed, 19 Oct 2022, Greg Kroah-Hartman wrote:

> From: Yu Kuai <yukuai3@huawei.com>
> 
> [ Upstream commit 040b83fcecfb86f3225d3a5de7fd9b3fbccf83b4 ]
> 
> There are two problems can lead to lost wakeup:
> 
> 1) invalid wakeup on the wrong waitqueue:
> 
> For example, 2 * wake_batch tags are put, while only wake_batch threads
> are woken:
> 
> __sbq_wake_up
>  atomic_cmpxchg -> reset wait_cnt
> 			__sbq_wake_up -> decrease wait_cnt
> 			...
> 			__sbq_wake_up -> wait_cnt is decreased to 0 again
> 			 atomic_cmpxchg
> 			 sbq_index_atomic_inc -> increase wake_index
> 			 wake_up_nr -> wake up and waitqueue might be empty
>  sbq_index_atomic_inc -> increase again, one waitqueue is skipped
>  wake_up_nr -> invalid wake up because old wakequeue might be empty
> 
> To fix the problem, increasing 'wake_index' before resetting 'wait_cnt'.
> 
> 2) 'wait_cnt' can be decreased while waitqueue is empty
> 
> As pointed out by Jan Kara, following race is possible:
> 
> CPU1				CPU2
> __sbq_wake_up			 __sbq_wake_up
>  sbq_wake_ptr()			 sbq_wake_ptr() -> the same
>  wait_cnt = atomic_dec_return()
>  /* decreased to 0 */
>  sbq_index_atomic_inc()
>  /* move to next waitqueue */
>  atomic_set()
>  /* reset wait_cnt */
>  wake_up_nr()
>  /* wake up on the old waitqueue */
> 				 wait_cnt = atomic_dec_return()
> 				 /*
> 				  * decrease wait_cnt in the old
> 				  * waitqueue, while it can be
> 				  * empty.
> 				  */
> 
> Fix the problem by waking up before updating 'wake_index' and
> 'wait_cnt'.
> 
> With this patch, noted that 'wait_cnt' is still decreased in the old
> empty waitqueue, however, the wakeup is redirected to a active waitqueue,
> and the extra decrement on the old empty waitqueue is not handled.
> 
> Fixes: 88459642cba4 ("blk-mq: abstract tag allocation out into sbitmap library")
> Signed-off-by: Yu Kuai <yukuai3@huawei.com>
> Reviewed-by: Jan Kara <jack@suse.cz>
> Link: https://lore.kernel.org/r/20220803121504.212071-1-yukuai1@huaweicloud.com
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> Signed-off-by: Sasha Levin <sashal@kernel.org>

I have no authority on linux-block, but I'll say NAK to this one
(and 517/862), and let Jens and Jan overrule me if they disagree.

This was the first of several 6.1-rc1 commits which had given me lost
wakeups never suffered before; was not tagged Cc stable; and (unless I've
missed it on lore) never had AUTOSEL posted to linux-block or linux-kernel.

Hugh

> ---
>  lib/sbitmap.c | 55 ++++++++++++++++++++++++++++++---------------------
>  1 file changed, 33 insertions(+), 22 deletions(-)
> 
> diff --git a/lib/sbitmap.c b/lib/sbitmap.c
> index 29eb0484215a..1f31147872e6 100644
> --- a/lib/sbitmap.c
> +++ b/lib/sbitmap.c
> @@ -611,32 +611,43 @@ static bool __sbq_wake_up(struct sbitmap_queue *sbq)
>  		return false;
>  
>  	wait_cnt = atomic_dec_return(&ws->wait_cnt);
> -	if (wait_cnt <= 0) {
> -		int ret;
> +	/*
> +	 * For concurrent callers of this, callers should call this function
> +	 * again to wakeup a new batch on a different 'ws'.
> +	 */
> +	if (wait_cnt < 0 || !waitqueue_active(&ws->wait))
> +		return true;
>  
> -		wake_batch = READ_ONCE(sbq->wake_batch);
> +	if (wait_cnt > 0)
> +		return false;
>  
> -		/*
> -		 * Pairs with the memory barrier in sbitmap_queue_resize() to
> -		 * ensure that we see the batch size update before the wait
> -		 * count is reset.
> -		 */
> -		smp_mb__before_atomic();
> +	wake_batch = READ_ONCE(sbq->wake_batch);
>  
> -		/*
> -		 * For concurrent callers of this, the one that failed the
> -		 * atomic_cmpxhcg() race should call this function again
> -		 * to wakeup a new batch on a different 'ws'.
> -		 */
> -		ret = atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wake_batch);
> -		if (ret == wait_cnt) {
> -			sbq_index_atomic_inc(&sbq->wake_index);
> -			wake_up_nr(&ws->wait, wake_batch);
> -			return false;
> -		}
> +	/*
> +	 * Wake up first in case that concurrent callers decrease wait_cnt
> +	 * while waitqueue is empty.
> +	 */
> +	wake_up_nr(&ws->wait, wake_batch);
>  
> -		return true;
> -	}
> +	/*
> +	 * Pairs with the memory barrier in sbitmap_queue_resize() to
> +	 * ensure that we see the batch size update before the wait
> +	 * count is reset.
> +	 *
> +	 * Also pairs with the implicit barrier between decrementing wait_cnt
> +	 * and checking for waitqueue_active() to make sure waitqueue_active()
> +	 * sees result of the wakeup if atomic_dec_return() has seen the result
> +	 * of atomic_set().
> +	 */
> +	smp_mb__before_atomic();
> +
> +	/*
> +	 * Increase wake_index before updating wait_cnt, otherwise concurrent
> +	 * callers can see valid wait_cnt in old waitqueue, which can cause
> +	 * invalid wakeup on the old waitqueue.
> +	 */
> +	sbq_index_atomic_inc(&sbq->wake_index);
> +	atomic_set(&ws->wait_cnt, wake_batch);
>  
>  	return false;
>  }
> -- 
> 2.35.1
> 
> 
> 
>
  
Greg KH Oct. 19, 2022, 5:25 p.m. UTC | #2
On Wed, Oct 19, 2022 at 08:06:26AM -0700, Hugh Dickins wrote:
> On Wed, 19 Oct 2022, Greg Kroah-Hartman wrote:
> 
> > From: Yu Kuai <yukuai3@huawei.com>
> > 
> > [ Upstream commit 040b83fcecfb86f3225d3a5de7fd9b3fbccf83b4 ]
> > 
> > There are two problems can lead to lost wakeup:
> > 
> > 1) invalid wakeup on the wrong waitqueue:
> > 
> > For example, 2 * wake_batch tags are put, while only wake_batch threads
> > are woken:
> > 
> > __sbq_wake_up
> >  atomic_cmpxchg -> reset wait_cnt
> > 			__sbq_wake_up -> decrease wait_cnt
> > 			...
> > 			__sbq_wake_up -> wait_cnt is decreased to 0 again
> > 			 atomic_cmpxchg
> > 			 sbq_index_atomic_inc -> increase wake_index
> > 			 wake_up_nr -> wake up and waitqueue might be empty
> >  sbq_index_atomic_inc -> increase again, one waitqueue is skipped
> >  wake_up_nr -> invalid wake up because old wakequeue might be empty
> > 
> > To fix the problem, increasing 'wake_index' before resetting 'wait_cnt'.
> > 
> > 2) 'wait_cnt' can be decreased while waitqueue is empty
> > 
> > As pointed out by Jan Kara, following race is possible:
> > 
> > CPU1				CPU2
> > __sbq_wake_up			 __sbq_wake_up
> >  sbq_wake_ptr()			 sbq_wake_ptr() -> the same
> >  wait_cnt = atomic_dec_return()
> >  /* decreased to 0 */
> >  sbq_index_atomic_inc()
> >  /* move to next waitqueue */
> >  atomic_set()
> >  /* reset wait_cnt */
> >  wake_up_nr()
> >  /* wake up on the old waitqueue */
> > 				 wait_cnt = atomic_dec_return()
> > 				 /*
> > 				  * decrease wait_cnt in the old
> > 				  * waitqueue, while it can be
> > 				  * empty.
> > 				  */
> > 
> > Fix the problem by waking up before updating 'wake_index' and
> > 'wait_cnt'.
> > 
> > With this patch, noted that 'wait_cnt' is still decreased in the old
> > empty waitqueue, however, the wakeup is redirected to a active waitqueue,
> > and the extra decrement on the old empty waitqueue is not handled.
> > 
> > Fixes: 88459642cba4 ("blk-mq: abstract tag allocation out into sbitmap library")
> > Signed-off-by: Yu Kuai <yukuai3@huawei.com>
> > Reviewed-by: Jan Kara <jack@suse.cz>
> > Link: https://lore.kernel.org/r/20220803121504.212071-1-yukuai1@huaweicloud.com
> > Signed-off-by: Jens Axboe <axboe@kernel.dk>
> > Signed-off-by: Sasha Levin <sashal@kernel.org>
> 
> I have no authority on linux-block, but I'll say NAK to this one
> (and 517/862), and let Jens and Jan overrule me if they disagree.
> 
> This was the first of several 6.1-rc1 commits which had given me lost
> wakeups never suffered before; was not tagged Cc stable; and (unless I've
> missed it on lore) never had AUTOSEL posted to linux-block or linux-kernel.

Ok, thanks for the review.  I'll drop both of the sbitmap.c changes and
if people report issues and want them back, I'll be glad to revisit them
then.

greg k-h
  
Jens Axboe Oct. 19, 2022, 5:37 p.m. UTC | #3
On 10/19/22 10:25 AM, Greg Kroah-Hartman wrote:
> On Wed, Oct 19, 2022 at 08:06:26AM -0700, Hugh Dickins wrote:
>> On Wed, 19 Oct 2022, Greg Kroah-Hartman wrote:
>>
>>> From: Yu Kuai <yukuai3@huawei.com>
>>>
>>> [ Upstream commit 040b83fcecfb86f3225d3a5de7fd9b3fbccf83b4 ]
>>>
>>> There are two problems can lead to lost wakeup:
>>>
>>> 1) invalid wakeup on the wrong waitqueue:
>>>
>>> For example, 2 * wake_batch tags are put, while only wake_batch threads
>>> are woken:
>>>
>>> __sbq_wake_up
>>>  atomic_cmpxchg -> reset wait_cnt
>>> 			__sbq_wake_up -> decrease wait_cnt
>>> 			...
>>> 			__sbq_wake_up -> wait_cnt is decreased to 0 again
>>> 			 atomic_cmpxchg
>>> 			 sbq_index_atomic_inc -> increase wake_index
>>> 			 wake_up_nr -> wake up and waitqueue might be empty
>>>  sbq_index_atomic_inc -> increase again, one waitqueue is skipped
>>>  wake_up_nr -> invalid wake up because old wakequeue might be empty
>>>
>>> To fix the problem, increasing 'wake_index' before resetting 'wait_cnt'.
>>>
>>> 2) 'wait_cnt' can be decreased while waitqueue is empty
>>>
>>> As pointed out by Jan Kara, following race is possible:
>>>
>>> CPU1				CPU2
>>> __sbq_wake_up			 __sbq_wake_up
>>>  sbq_wake_ptr()			 sbq_wake_ptr() -> the same
>>>  wait_cnt = atomic_dec_return()
>>>  /* decreased to 0 */
>>>  sbq_index_atomic_inc()
>>>  /* move to next waitqueue */
>>>  atomic_set()
>>>  /* reset wait_cnt */
>>>  wake_up_nr()
>>>  /* wake up on the old waitqueue */
>>> 				 wait_cnt = atomic_dec_return()
>>> 				 /*
>>> 				  * decrease wait_cnt in the old
>>> 				  * waitqueue, while it can be
>>> 				  * empty.
>>> 				  */
>>>
>>> Fix the problem by waking up before updating 'wake_index' and
>>> 'wait_cnt'.
>>>
>>> With this patch, noted that 'wait_cnt' is still decreased in the old
>>> empty waitqueue, however, the wakeup is redirected to a active waitqueue,
>>> and the extra decrement on the old empty waitqueue is not handled.
>>>
>>> Fixes: 88459642cba4 ("blk-mq: abstract tag allocation out into sbitmap library")
>>> Signed-off-by: Yu Kuai <yukuai3@huawei.com>
>>> Reviewed-by: Jan Kara <jack@suse.cz>
>>> Link: https://lore.kernel.org/r/20220803121504.212071-1-yukuai1@huaweicloud.com
>>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>> Signed-off-by: Sasha Levin <sashal@kernel.org>
>>
>> I have no authority on linux-block, but I'll say NAK to this one
>> (and 517/862), and let Jens and Jan overrule me if they disagree.
>>
>> This was the first of several 6.1-rc1 commits which had given me lost
>> wakeups never suffered before; was not tagged Cc stable; and (unless I've
>> missed it on lore) never had AUTOSEL posted to linux-block or linux-kernel.
> 
> Ok, thanks for the review.  I'll drop both of the sbitmap.c changes and
> if people report issues and want them back, I'll be glad to revisit them
> then.

Sorry for being late, did see Hugh respond to the original auto-select
as well, and was surprised to see it moving forward after that. Let's
please drop them for now.
  

Patch

diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index 29eb0484215a..1f31147872e6 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -611,32 +611,43 @@  static bool __sbq_wake_up(struct sbitmap_queue *sbq)
 		return false;
 
 	wait_cnt = atomic_dec_return(&ws->wait_cnt);
-	if (wait_cnt <= 0) {
-		int ret;
+	/*
+	 * For concurrent callers of this, callers should call this function
+	 * again to wakeup a new batch on a different 'ws'.
+	 */
+	if (wait_cnt < 0 || !waitqueue_active(&ws->wait))
+		return true;
 
-		wake_batch = READ_ONCE(sbq->wake_batch);
+	if (wait_cnt > 0)
+		return false;
 
-		/*
-		 * Pairs with the memory barrier in sbitmap_queue_resize() to
-		 * ensure that we see the batch size update before the wait
-		 * count is reset.
-		 */
-		smp_mb__before_atomic();
+	wake_batch = READ_ONCE(sbq->wake_batch);
 
-		/*
-		 * For concurrent callers of this, the one that failed the
-		 * atomic_cmpxhcg() race should call this function again
-		 * to wakeup a new batch on a different 'ws'.
-		 */
-		ret = atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wake_batch);
-		if (ret == wait_cnt) {
-			sbq_index_atomic_inc(&sbq->wake_index);
-			wake_up_nr(&ws->wait, wake_batch);
-			return false;
-		}
+	/*
+	 * Wake up first in case that concurrent callers decrease wait_cnt
+	 * while waitqueue is empty.
+	 */
+	wake_up_nr(&ws->wait, wake_batch);
 
-		return true;
-	}
+	/*
+	 * Pairs with the memory barrier in sbitmap_queue_resize() to
+	 * ensure that we see the batch size update before the wait
+	 * count is reset.
+	 *
+	 * Also pairs with the implicit barrier between decrementing wait_cnt
+	 * and checking for waitqueue_active() to make sure waitqueue_active()
+	 * sees result of the wakeup if atomic_dec_return() has seen the result
+	 * of atomic_set().
+	 */
+	smp_mb__before_atomic();
+
+	/*
+	 * Increase wake_index before updating wait_cnt, otherwise concurrent
+	 * callers can see valid wait_cnt in old waitqueue, which can cause
+	 * invalid wakeup on the old waitqueue.
+	 */
+	sbq_index_atomic_inc(&sbq->wake_index);
+	atomic_set(&ws->wait_cnt, wake_batch);
 
 	return false;
 }