[PATCH] dm-bufio: break out of both loops when freeing data

[PATCH] dm-bufio: break out of both loops when freeing data

[Mikulas Patocka]

When we free the required number of entries (nr_to_scan becomes zero), we
must break out of both loops. The code exited only the inner loop and
continued in the outer loop.

Also, move dm_bufio_cond_resched to the inner loop, so that scanning the
list doesn't result in scheduling latency.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>

---
 drivers/md/dm-bufio.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

Index: linux-3.17-rc7/drivers/md/dm-bufio.c
===================================================================
--- linux-3.17-rc7.orig/drivers/md/dm-bufio.c	2014-10-01 19:20:36.000000000 +0200
+++ linux-3.17-rc7/drivers/md/dm-bufio.c	2014-10-01 19:22:50.000000000 +0200
@@ -1473,10 +1473,12 @@ static long __scan(struct dm_bufio_clien
 		list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) {
 			freed += __cleanup_old_buffer(b, gfp_mask, 0);
 			if (!--nr_to_scan)
-				break;
+				goto brk;
+			dm_bufio_cond_resched();
 		}
-		dm_bufio_cond_resched();
 	}
+
+brk:
 	return freed;
 }
 

Re: [PATCH] dm-bufio: break out of both loops when freeing data

[Mikulas Patocka]

On Wed, 1 Oct 2014, Mikulas Patocka wrote:

> When we free the required number of entries (nr_to_scan becomes zero), we
> must break out of both loops. The code exited only the inner loop and
> continued in the outer loop.
> 
> Also, move dm_bufio_cond_resched to the inner loop, so that scanning the
> list doesn't result in scheduling latency.
> 
> Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>

I forgot to add:

Cc: stable@vger.kernel.org	# 3.2+

> ---
>  drivers/md/dm-bufio.c |    6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> Index: linux-3.17-rc7/drivers/md/dm-bufio.c
> ===================================================================
> --- linux-3.17-rc7.orig/drivers/md/dm-bufio.c	2014-10-01 19:20:36.000000000 +0200
> +++ linux-3.17-rc7/drivers/md/dm-bufio.c	2014-10-01 19:22:50.000000000 +0200
> @@ -1473,10 +1473,12 @@ static long __scan(struct dm_bufio_clien
>  		list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) {
>  			freed += __cleanup_old_buffer(b, gfp_mask, 0);
>  			if (!--nr_to_scan)
> -				break;
> +				goto brk;
> +			dm_bufio_cond_resched();
>  		}
> -		dm_bufio_cond_resched();
>  	}
> +
> +brk:
>  	return freed;
>  }
>  
> 

Re: dm-bufio: break out of both loops when freeing data

[Mike Snitzer]

On Wed, Oct 01 2014 at  1:35pm -0400,
Mikulas Patocka <mpatocka@redhat.com> wrote:

> 
> 
> On Wed, 1 Oct 2014, Mikulas Patocka wrote:
> 
> > When we free the required number of entries (nr_to_scan becomes zero), we
> > must break out of both loops. The code exited only the inner loop and
> > continued in the outer loop.
> > 
> > Also, move dm_bufio_cond_resched to the inner loop, so that scanning the
> > list doesn't result in scheduling latency.
> > 
> > Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
> 
> I forgot to add:
> 
> Cc: stable@vger.kernel.org	# 3.2+

I tweaked the header, used a return rather than goto, and staged in
linux-next for 3.18.  See:
https://git.kernel.org/cgit/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=for-next&id=04c308f43a90a9b3b84c344b324d6af29288da05