Re: [PATCH 4/4] change thaw_super() to re-acquire s_writers.lock_map

[Jan Kara <jack@suse.cz>]

On Mon 20-07-15 19:01:09, Oleg Nesterov wrote:
> Change thaw_super() to re-acquire the "write" locks we are going to
> release. This way s_op->unfreeze_fs(sb) is called with these locks
> held as it seen by lockdep, and this matches the reality. This adds
> another trivial helper, sb_freeze_acquire().
> 
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>

Ah, I was confused for a moment why this doesn't trigger a lockdep warning
because of lock inversion between s_umount and freeze protection but you
use trylock in sb_freeze_acquire(). So things are fine.

You can add:

Reviewed-by: Jan Kara <jack@suse.com>

								Honza

> ---
>  fs/super.c |   12 ++++++++++++
>  1 files changed, 12 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/super.c b/fs/super.c
> index e7ea9f1..b4db3ee 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -1263,6 +1263,14 @@ static void sb_freeze_release(struct super_block *sb)
>  		rwsem_release(sb->s_writers.lock_map + level, 1, _THIS_IP_);
>  }
>  
> +static void sb_freeze_acquire(struct super_block *sb)
> +{
> +	int level;
> +
> +	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
> +		rwsem_acquire(sb->s_writers.lock_map + level, 0, 1, _THIS_IP_);
> +}
> +
>  /**
>   * freeze_super - lock the filesystem and force it into a consistent state
>   * @sb: the super to lock
> @@ -1386,16 +1394,20 @@ int thaw_super(struct super_block *sb)
>  	if (sb->s_flags & MS_RDONLY)
>  		goto out;
>  
> +	sb_freeze_acquire(sb);
> +
>  	if (sb->s_op->unfreeze_fs) {
>  		error = sb->s_op->unfreeze_fs(sb);
>  		if (error) {
>  			printk(KERN_ERR
>  				"VFS:Filesystem thaw failed\n");
> +			sb_freeze_release(sb);
>  			up_write(&sb->s_umount);
>  			return error;
>  		}
>  	}
>  
> +	sb_freeze_release(sb);
>  out:
>  	sb->s_writers.frozen = SB_UNFROZEN;
>  	smp_wmb();
> -- 
> 1.5.5.1
> 
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR