xfs: resample the data fork mapping after cycling ILOCK

commit-author Darrick J. Wong <djwong@kernel.org>
commit 2f4acd0
upstream-diff |
	Between this kernel and the recent change xfs_reflink_allocate_cow()
	was refactored into two separate functions in d621133
	xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork
	In addition after fixing the bug the offset_fsb and count_fsb need to
	be refreshed as a part of the above commit it creates
	xfs_reflink_convert_unwritten() which refreshes prior to returns.

xfs_reflink_fill_{cow_hole,delalloc} are both presented with an inode,
a data fork mapping, and a cow fork mapping.  Unfortunately, these two
helpers cycle the ILOCK to grab a transaction, which means that the
mappings are stale as soon as we reacquire the ILOCK.  Currently we
refresh the cow fork mapping by re-calling xfs_find_trim_cow_extent, but
we don't refresh the data fork mapping beforehand, which means that the
xfs_bmap_trim_cow in that function queries the refcount btree about the
wrong physical blocks and returns an inaccurate value in *shared.

If *shared is now false, the directio write proceeds with a stale data
fork mapping.  Fix this by querying the data fork mapping if the
sequence counter changes across the ILOCK cycle.

	Cc: hch@lst.de
	Cc: stable@vger.kernel.org # v4.11
Fixes: 3c68d44 ("xfs: allocate direct I/O COW blocks in iomap_begin")
	Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
	Reviewed-by: Christoph Hellwig <hch@lst.de>
	Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
	Signed-off-by: Carlos Maiolino <cem@kernel.org>
Signed-off-by: Jonathan Maple <jmaple@ciq.com>

Author: PlaidCat

fs/xfs/xfs_reflink.c

@@ -357,6 +357,7 @@ xfs_reflink_allocate_cow(
 	int			nimaps, error = 0;
 	bool			found;
 	xfs_filblks_t		resaligned;
+	unsigned int		seq_before = READ_ONCE(ip->i_df.if_seq);
 	xfs_extlen_t		resblks = 0;
 
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
@@ -385,6 +386,24 @@ xfs_reflink_allocate_cow(
 
 	*lockmode = XFS_ILOCK_EXCL;
 
+	/*
+	 * The data fork mapping may have changed while we dropped the ILOCK
+	 * (a racing O_DIRECT writer under IOLOCK_SHARED can complete a full
+	 * CoW cycle including xfs_reflink_end_cow(), which remaps this offset
+	 * and drops the refcount of the old shared block).  Re-read it so the
+	 * shared-status recheck below and the caller's in-place iomap both
+	 * operate on the current mapping rather than a stale physical block.
+	 */
+	if (seq_before != READ_ONCE(ip->i_df.if_seq)) {
+		nimaps = 1;
+		error = xfs_bmapi_read(ip, imap->br_startoff,
+				imap->br_blockcount, imap, &nimaps, 0);
+		if (error)
+			goto out_trans_cancel;
+		offset_fsb = imap->br_startoff;
+		count_fsb = imap->br_blockcount;
+	}
+
 	/*
 	 * Check for an overlapping extent again now that we dropped the ilock.
 	 */