Re: [PATCH 3/7] xfs: handle zeroing in xfs_file_iomap_begin_delay

[Christoph Hellwig <hch@lst.de>]

On Mon, Oct 01, 2018 at 10:20:05AM -0400, Brian Foster wrote:
> >  	eof = !xfs_iext_lookup_extent(ip, ifp, offset_fsb, &icur, &got);
> > -	if (!eof && got.br_startoff <= offset_fsb) {
> > -		if (xfs_is_reflink_inode(ip)) {
> > +	if (eof)
> > +		got.br_startoff = maxbytes_fsb;
> 
> What's the purpose of this? Can't we just continue to use eof in the
> logic below and report holes up through the requested range (offset +
> length) just like the other branch does (via xfs_bmapi_read())?

Setting the startblock makes the code a lot simpler.  But yes,
we could limit the hole to the reporting range.

> 
> > +	if (got.br_startoff <= offset_fsb) {
> > +		if (xfs_is_reflink_inode(ip) &&
> > +		    ((flags & IOMAP_WRITE) ||
> > +		     got.br_state != XFS_EXT_UNWRITTEN)) {
> 
> I think a small comment is useful here due to the implicit logic. For
> example:
> 
> /* reservation is required for writes and zeroing over normal extents */

Ok.