Re: [PATCH RFC 3/4] xfs: reuse xfs_file_iomap_begin_delay() for cow fork delalloc

[Christoph Hellwig <hch@infradead.org>]

> +	/*
> +	 * Search for a preexisting extent. COW fork allocation may still be
> +	 * required for reflink inodes if the data extent is shared.
> +	 */
>  	xfs_bmap_search_extents(ip, offset_fsb, XFS_DATA_FORK, &eof, &idx,
>  			&got, &prev);
>  	imap = got;

Maybe we should look up directly into imap and now duplicate that
information for imap and got?

>  		if (xfs_is_reflink_inode(ip)) {
> +			bool		shared, trimmed;
>  
> +			/*
> +			 * Assume the data extent is shared if an extent exists
> +			 * in the cow fork.
> +			 */
> +			xfs_trim_extent(&imap, offset_fsb,
> +					end_fsb - offset_fsb);
> +			xfs_bmap_search_extents(ip, imap.br_startoff,
> +					XFS_COW_FORK, &eof, &idx, &got, &prev);
> +			if (!eof && got.br_startoff <= imap.br_startoff) {
> +				trace_xfs_reflink_cow_found(ip, &got);
> +				xfs_trim_extent(&imap, got.br_startoff,
> +						got.br_blockcount);
> +				goto done;
> +			}
> +
> +			/*
> +			 * No existing cow fork extent. Now we have to actually
> +			 * check if the data extent is shared and trim the
> +			 * mapping to the next (un)shared boundary.
> +			 */
> +			error = xfs_reflink_trim_around_shared(ip, &imap,
> +						       &shared, &trimmed);
>  			if (error)
>  				goto out_unlock;
> +			if (!shared)
> +				goto done;
> +
> +			end_fsb = imap.br_startoff + imap.br_blockcount;

I think the code for looking at the COW fork should go into a little
helper, even if it means passing a few arguments to it.

> +			fork = XFS_COW_FORK;
> +		} else {
> +			trace_xfs_iomap_found(ip, offset, count, 0, &imap);
> +			goto done;
>  		}

And while we're at it - try to always handle the a that ends with a
goto first, that helps to reduce the indentation level.