Re: [PATCH 3/5] fs: make BLKFLSBUF's wait for a superblock killable

[Jan Kara]

On Wed 21-10-26 16:46:59, Josef Bacik wrote:
> fs_bdev_sync() syncs the filesystems on a block device for BLKFLSBUF.  It
> can already be killed while it waits for one which is being unmounted.  For
> one which is still being mounted it waits for SB_BORN, and then for
> s_umount, in super_lock(), which can't be killed.  If the mount is stuck on
> that very device, so is the ioctl.
> 
> Use super_lock_killable() and stop the walk when the task is killed, as for
> the unmount.  All BLKFLSBUF does afterwards is invalidate_bdev(), which
> leaves dirty pages and pages under writeback alone.
> 
> Signed-off-by: Josef Bacik <josef@xxxxxxxxxxxxxx>

Christian has just pulled some changes in this area so please rebase on top
of them. Otherwise the changes look ok to me.

								Honza

> ---
>  fs/super.c | 20 ++++++++++++--------
>  1 file changed, 12 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/super.c b/fs/super.c
> index 40f92ffa39c1..ecba64ed0c2a 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -1676,17 +1676,21 @@ static void fs_bdev_sync(struct block_device *bdev)
>  
>  	for (sb_dev = super_dev_first(dev); sb_dev; sb_dev = super_dev_next(sb_dev)) {
>  		struct super_block *sb = sb_dev->sd_sb;
> +		int ret;
>  
> -		if (!super_lock_shared(sb)) {
> -			if (super_wait_shutdown_killable(sb)) {
> -				super_dev_put(sb_dev);
> -				return;
> -			}
> +		ret = super_lock_killable(sb, false);
> +		if (!ret) {
> +			if (sb->s_root && (sb->s_flags & SB_ACTIVE))
> +				sync_filesystem(sb);
> +			super_unlock_shared(sb);
>  			continue;
>  		}
> -		if (sb->s_root && (sb->s_flags & SB_ACTIVE))
> -			sync_filesystem(sb);
> -		super_unlock_shared(sb);
> +		if (ret == -ENOENT)
> +			ret = super_wait_shutdown_killable(sb);
> +		if (ret) {
> +			super_dev_put(sb_dev);
> +			return;
> +		}
>  	}
>  }
>  
> 
> -- 
> 2.55.0
> 
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR