Ocfs2 devel - Sep 2020 - [PATCH] ocfs2: fix potential soft lockup during fstrim

When we discard unused blocks on a mounted ocfs2 filesystem, fstrim
handles each block goup with locking/unlocking global bitmap meta-file
repeatedly. we should let fstrim thread take a break(if need) between
unlock and lock, this will avoid the potential soft lockup problem,
and also gives the upper applications more IO opportunities, these
applications are not blocked for too long at writing files.

Signed-off-by: Gang He <ghe at suse.com>
---
 fs/ocfs2/alloc.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
index 4c1b90442d6f..2cf9321919b5 100644
--- a/fs/ocfs2/alloc.c
+++ b/fs/ocfs2/alloc.c
@@ -7654,8 +7654,10 @@ int ocfs2_trim_mainbm(struct super_block *sb, struct
fstrim_range *range)
 	 * main_bm related locks for avoiding the current IO starve, then go to
 	 * trim the next group
 	 */
-	if (ret >= 0 && group <= last_group)
+	if (ret >= 0 && group <= last_group) {
+		cond_resched();
 		goto next_group;
+	}
 out:
 	range->len = trimmed * sb->s_blocksize;
 	return ret;
-- 
2.21.0

On 2020/9/27 09:58, Gang He wrote:
> When we discard unused blocks on a mounted ocfs2 filesystem, fstrim
> handles each block goup with locking/unlocking global bitmap meta-file
> repeatedly. we should let fstrim thread take a break(if need) between
> unlock and lock, this will avoid the potential soft lockup problem,
> and also gives the upper applications more IO opportunities, these
> applications are not blocked for too long at writing files.
> 
> Signed-off-by: Gang He <ghe at suse.com>
It makes sense.
Reviewed-by: Joseph Qi <joseph.qi at linux.alibaba.com>
> ---
>  fs/ocfs2/alloc.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
> index 4c1b90442d6f..2cf9321919b5 100644
> --- a/fs/ocfs2/alloc.c
> +++ b/fs/ocfs2/alloc.c
> @@ -7654,8 +7654,10 @@ int ocfs2_trim_mainbm(struct super_block *sb, struct
fstrim_range *range)
>  	 * main_bm related locks for avoiding the current IO starve, then go to
>  	 * trim the next group
>  	 */
> -	if (ret >= 0 && group <= last_group)
> +	if (ret >= 0 && group <= last_group) {
> +		cond_resched();
>  		goto next_group;
> +	}
>  out:
>  	range->len = trimmed * sb->s_blocksize;
>  	return ret;
>