Re: [PATCH 2/2] xfs: don't unconditionally clear the reflink flag on zero-block files

[Christoph Hellwig <hch@infradead.org>]

On Mon, Sep 04, 2017 at 08:49:44AM -0700, Darrick J. Wong wrote:
> If we have speculative cow preallocations hanging around in the cow
> fork, don't let a truncate operation clear the reflink flag because if
> we do then there's a chance we'll forget to free those extents when we
> destroy the incore inode.
> 
> Reported-by: Amir Goldstein <amir73il@gmail.com>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
>  fs/xfs/xfs_inode.c |    8 +++++---
>  1 file changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> index 5599dda..4ec5b7f 100644
> --- a/fs/xfs/xfs_inode.c
> +++ b/fs/xfs/xfs_inode.c
> @@ -1624,10 +1624,12 @@ xfs_itruncate_extents(
>  		goto out;
>  
>  	/*
> -	 * Clear the reflink flag if we truncated everything.
> +	 * Clear the reflink flag if there are no data fork blocks and
> +	 * there are no extents staged in the cow fork.
>  	 */
> -	if (ip->i_d.di_nblocks == 0 && xfs_is_reflink_inode(ip)) {
> -		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
> +	if (xfs_is_reflink_inode(ip) && ip->i_cnextents == 0) {
> +		if (ip->i_d.di_nblocks == 0)
> +			ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
>  		xfs_inode_clear_cowblocks_tag(ip);

This part looks ok:

Reviewed-by: Christoph Hellwig <hch@lst.de>

But while looking at that I decided I'd want to take a look at
other places where we potentially clear XFS_DIFLAG2_REFLINK.

In xfs_ioctl_setattr_xflags it seems like it's not currently safe
for setting the RT flag vs having COW extents?  I also think we'd
need the iolock for every change to the rt flag, so this probably
needs a bigger rework..

swapext also doesn't seem to flush out existing cow extents.