[PATCH] bcache: Fix a sleep-in-atomic bug

[PATCH] bcache: Fix a sleep-in-atomic bug

[Jia-Ju Bai]

The driver may sleep under a spin lock, and the function call path is:
journal_wait_for_write (acquire the lock by spin_lock)
  closure_sync
    schedule --> may sleep

To fix it, the lock is released before "closure_sync", and the lock is 
acquired again after this function.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
---
 drivers/md/bcache/journal.c |    1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c
index 1198e53..ad47c36 100644
--- a/drivers/md/bcache/journal.c
+++ b/drivers/md/bcache/journal.c
@@ -724,6 +724,7 @@ static struct journal_write *journal_wait_for_write(struct cache_set *c,
 			btree_flush_write(c);
 		}
 
+		spin_unlock(&c->journal.lock);
 		closure_sync(&cl);
 		spin_lock(&c->journal.lock);
 		wait = true;
-- 
1.7.9.5

Re: [PATCH] bcache: Fix a sleep-in-atomic bug

[Kent Overstreet]

On Wed, May 31, 2017 at 03:23:40PM +0800, Jia-Ju Bai wrote:
> The driver may sleep under a spin lock, and the function call path is:
> journal_wait_for_write (acquire the lock by spin_lock)
>   closure_sync
>     schedule --> may sleep

This patch is incorrect, you've introduced a double unlock.

Did you actually observe a sleep in atomic?

> 
> To fix it, the lock is released before "closure_sync", and the lock is 
> acquired again after this function.
> 
> Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
> ---
>  drivers/md/bcache/journal.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c
> index 1198e53..ad47c36 100644
> --- a/drivers/md/bcache/journal.c
> +++ b/drivers/md/bcache/journal.c
> @@ -724,6 +724,7 @@ static struct journal_write *journal_wait_for_write(struct cache_set *c,
>  			btree_flush_write(c);
>  		}
>  
> +		spin_unlock(&c->journal.lock);
>  		closure_sync(&cl);
>  		spin_lock(&c->journal.lock);
>  		wait = true;
> -- 
> 1.7.9.5
> 
> 

Re: [PATCH] bcache: Fix a sleep-in-atomic bug

[Jia-Ju Bai]

On 05/31/2017 03:23 PM, Jia-Ju Bai wrote:
> The driver may sleep under a spin lock, and the function call path is:
> journal_wait_for_write (acquire the lock by spin_lock)
>    closure_sync
>      schedule -->  may sleep
>
> To fix it, the lock is released before "closure_sync", and the lock is
> acquired again after this function.
>
> Signed-off-by: Jia-Ju Bai<baijiaju1990@163.com>
> ---
>   drivers/md/bcache/journal.c |    1 +
>   1 file changed, 1 insertion(+)
>
> diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c
> index 1198e53..ad47c36 100644
> --- a/drivers/md/bcache/journal.c
> +++ b/drivers/md/bcache/journal.c
> @@ -724,6 +724,7 @@ static struct journal_write *journal_wait_for_write(struct cache_set *c,
>   			btree_flush_write(c);
>   		}
>
> +		spin_unlock(&c->journal.lock);
>   		closure_sync(&cl);
>   		spin_lock(&c->journal.lock);
>   		wait = true;

Sorry, my patch is not correct, and it will cause double unlock.
Please ignore my patch.


Jia-Ju Bai