[PATCH] fs: push i_mutex and filemap_write_and_wait down into ->fsync() handlers

[Christoph Hellwig]

On Sun, Jul 24, 2011 at 12:02:12AM +0200, Markus Trippelsdorf wrote:
> The following patch fixes the problem for me:

Good catch, can you send it with a proper signoff and description?

> 
> diff --git a/fs/xfs/linux-2.6/xfs_file.c b/fs/xfs/linux-2.6/xfs_file.c
> index cca00f4..2e23e3d 100644
> --- a/fs/xfs/linux-2.6/xfs_file.c
> +++ b/fs/xfs/linux-2.6/xfs_file.c
> @@ -881,11 +881,14 @@ xfs_file_aio_write(
>  	/* Handle various SYNC-type writes */
>  	if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) {
>  		loff_t end = pos + ret - 1;
> +		int error;
>  
>  		xfs_rw_iunlock(ip, iolock);
> -		ret = -xfs_file_fsync(file, pos, end,
> +		error = -xfs_file_fsync(file, pos, end,
>  				      (file->f_flags & __O_SYNC) ? 0 : 1);

In addition xfs_file_fsync actually returns a normal negative error,
but that bug already existed before the fsync prototype change.  If you
resend the patch, can you fix it as well?

>  		xfs_rw_ilock(ip, iolock);
> +		if (error)
> +			ret = error;
>  	}