[PATCH 3/4] [PATCH 2/5] xfs: cleanup ->sync_fs

[Alex Elder]

Christoph Hellwig wrote:
> Sort out ->sync_fs to not perform a superblock writeback for the wait = 0 case
> as that is just an optional first pass and the superblock will be written back
> properly in the next call with wait = 1.  Instead perform an opportunistic
> quota writeback to have less work later.  Also remove the freeze special case
> as we do a proper wait = 1 call in the freeze code anyway.
> 
> Also rename the function to xfs_fs_sync_fs to match the normal naming
> convention, update comments and avoid calling into the laptop_mode logic on
> an error.
> 
> 
> Signed-off-by: Christoph Hellwig <hch at lst.de>

Looks good.

Reviewed-by: Alex Elder <aelder at sgi.com>
 
> Index: linux-2.6/fs/xfs/linux-2.6/xfs_super.c
> ===================================================================
> --- linux-2.6.orig/fs/xfs/linux-2.6/xfs_super.c	2009-08-26 20:13:54.609362865 -0300
> +++ linux-2.6/fs/xfs/linux-2.6/xfs_super.c	2009-08-26 20:18:36.065357266 -0300
> @@ -1144,7 +1144,7 @@ xfs_fs_put_super(
>  }
> 
>  STATIC int
> -xfs_fs_sync_super(
> +xfs_fs_sync_fs(
>  	struct super_block	*sb,
>  	int			wait)
>  {
> @@ -1152,23 +1152,23 @@ xfs_fs_sync_super(
>  	int			error;
> 
>  	/*
> -	 * Treat a sync operation like a freeze.  This is to work
> -	 * around a race in sync_inodes() which works in two phases
> -	 * - an asynchronous flush, which can write out an inode
> -	 * without waiting for file size updates to complete, and a
> -	 * synchronous flush, which wont do anything because the
> -	 * async flush removed the inode's dirty flag.  Also
> -	 * sync_inodes() will not see any files that just have
> -	 * outstanding transactions to be flushed because we don't
> -	 * dirty the Linux inode until after the transaction I/O
> -	 * completes.
> +	 * Not much we can do for the first async pass.  Writing out the
> +	 * superblock would be counter-productive as we are going to redirty
> +	 * when writing out other data and metadata (and writing out a single
> +	 * block is quite fast anyway).
> +	 *
> +	 * Try to asynchronously kick off quota syncing at least.
>  	 */
> -	if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE))
> -		error = xfs_quiesce_data(mp);
> -	else
> -		error = xfs_sync_fsdata(mp, 0);
> +	if (!wait) {
> +		xfs_qm_sync(mp, SYNC_TRYLOCK);
> +		return 0;
> +	}
> +
> +	error = xfs_quiesce_data(mp);
> +	if (error)
> +		return -error;
> 
> -	if (unlikely(laptop_mode)) {
> +	if (laptop_mode) {
>  		int	prev_sync_seq = mp->m_sync_seq;
> 
>  		/*
> @@ -1187,7 +1187,7 @@ xfs_fs_sync_super(
>  				mp->m_sync_seq != prev_sync_seq);
>  	}
> 
> -	return -error;
> +	return 0;
>  }
> 
>  STATIC int
> @@ -1561,7 +1561,7 @@ static struct super_operations xfs_super
>  	.write_inode		= xfs_fs_write_inode,
>  	.clear_inode		= xfs_fs_clear_inode,
>  	.put_super		= xfs_fs_put_super,
> -	.sync_fs		= xfs_fs_sync_super,
> +	.sync_fs		= xfs_fs_sync_fs,
>  	.freeze_fs		= xfs_fs_freeze,
>  	.statfs			= xfs_fs_statfs,
>  	.remount_fs		= xfs_fs_remount,
> 
> _______________________________________________
> xfs mailing list
> xfs at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs