xfs: make COW fork unwritten extent conversions more robust
authorChristoph Hellwig <hch@lst.de>
Mon, 18 Feb 2019 17:38:48 +0000 (09:38 -0800)
committerDarrick J. Wong <darrick.wong@oracle.com>
Thu, 21 Feb 2019 15:55:07 +0000 (07:55 -0800)
commit26b91c728b2d15952432371dc2b6ba1dda1fb61f
treea666a88de0bbe438c2e7371910f5f63d1d54cece
parentdb46e604adf8c923214a63b46e87ca2411d3d580
xfs: make COW fork unwritten extent conversions more robust

If we have racing buffered and direct I/O COW fork extents under
writeback can have been moved to the data fork by the time we call
xfs_reflink_convert_cow from xfs_submit_ioend.  This would be mostly
harmless as the block numbers don't change by this move, except for
the fact that xfs_bmapi_write will crash or trigger asserts when
not finding existing extents, even despite trying to paper over this
with the XFS_BMAPI_CONVERT_ONLY flag.

Instead of special casing non-transaction conversions in the already
way too complicated xfs_bmapi_write just add a new helper for the much
simpler non-transactional COW fork case, which simplify ignores not
found extents.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/libxfs/xfs_bmap.c
fs/xfs/libxfs/xfs_bmap.h
fs/xfs/xfs_reflink.c