xfs: apply rt extent alignment constraints to CoW extsize hint

The copy-on-write extent size hint is subject to the same alignment
constraints as the regular extent size hint.  Since we're in the process
of adding reflink (and therefore CoW) to the realtime device, we must
apply the same scattered rextsize alignment validation strategies to
both hints to deal with the possibility of rextsize changing.

Therefore, fix the inode validator to perform rextsize alignment checks
on regular realtime files, and to remove misaligned directory hints.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>

3 files changed, 49 insertions, 7 deletions

diff --git a/fs/xfs/libxfs/xfs_inode_buf.c b/fs/xfs/libxfs/xfs_inode_buf.c
index c33552b974f5..6dc396cfc369 100644
--- a/fs/xfs/libxfs/xfs_inode_buf.c
+++ b/fs/xfs/libxfs/xfs_inode_buf.c
@@ -760,11 +760,29 @@ xfs_inode_validate_cowextsize(
 	bool				rt_flag;
 	bool				hint_flag;
 	uint32_t			cowextsize_bytes;
+	uint32_t			blocksize_bytes;
 
 	rt_flag = (flags & XFS_DIFLAG_REALTIME);
 	hint_flag = (flags2 & XFS_DIFLAG2_COWEXTSIZE);
 	cowextsize_bytes = XFS_FSB_TO_B(mp, cowextsize);
 
+	/*
+	 * Similar to extent size hints, a directory can be configured to
+	 * propagate realtime status and a CoW extent size hint to newly
+	 * created files even if there is no realtime device, and the hints on
+	 * disk can become misaligned if the sysadmin changes the rt extent
+	 * size while adding the realtime device.
+	 *
+	 * Therefore, we can only enforce the rextsize alignment check against
+	 * regular realtime files, and rely on callers to decide when alignment
+	 * checks are appropriate, and fix things up as needed.
+	 */
+
+	if (rt_flag)
+		blocksize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
+	else
+		blocksize_bytes = mp->m_sb.sb_blocksize;
+
 	if (hint_flag && !xfs_has_reflink(mp))
 		return __this_address;
 
@@ -778,16 +796,13 @@ xfs_inode_validate_cowextsize(
 	if (mode && !hint_flag && cowextsize != 0)
 		return __this_address;
 
-	if (hint_flag && rt_flag)
-		return __this_address;
-
-	if (cowextsize_bytes % mp->m_sb.sb_blocksize)
+	if (cowextsize_bytes % blocksize_bytes)
 		return __this_address;
 
 	if (cowextsize > MAXEXTLEN)
 		return __this_address;
 
-	if (cowextsize > mp->m_sb.sb_agblocks / 2)
+	if (!rt_flag && cowextsize > mp->m_sb.sb_agblocks / 2)
 		return __this_address;
 
 	return NULL;
diff --git a/fs/xfs/libxfs/xfs_trans_inode.c b/fs/xfs/libxfs/xfs_trans_inode.c
index 6a3a869635bf..d8bb3aa49179 100644
--- a/fs/xfs/libxfs/xfs_trans_inode.c
+++ b/fs/xfs/libxfs/xfs_trans_inode.c
@@ -160,6 +160,20 @@ xfs_trans_log_inode(
 	}
 
 	/*
+	 * Inode verifiers do not check that the CoW extent size hint is an
+	 * integer multiple of the rt extent size on a directory with both
+	 * rtinherit and cowextsize flags set.  If we're logging a directory
+	 * that is misconfigured in this way, clear the hint.
+	 */
+	if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
+	    (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE) &&
+	    (ip->i_cowextsize % ip->i_mount->m_sb.sb_rextsize) > 0) {
+		ip->i_diflags2 &= ~XFS_DIFLAG2_COWEXTSIZE;
+		ip->i_cowextsize = 0;
+		flags |= XFS_ILOG_CORE;
+	}
+
+	/*
 	 * Record the specific change for fdatasync optimisation. This allows
 	 * fdatasync to skip log forces for inodes that are only timestamp
 	 * dirty.
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 0fe260d4d785..4ccdcde80467 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -1092,8 +1092,21 @@ xfs_fill_fsxattr(
 		}
 	}
 
-	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
-		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
+	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE) {
+		/*
+		 * Don't let a misaligned CoW extent size hint on a directory
+		 * escape to userspace if it won't pass the setattr checks
+		 * later.
+		 */
+		if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
+		    ip->i_cowextsize % mp->m_sb.sb_rextsize > 0) {
+			fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
+			fa->fsx_cowextsize = 0;
+		} else {
+			fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
+		}
+	}
+
 	fa->fsx_projid = ip->i_projid;
 	if (ifp && !xfs_need_iread_extents(ifp))
 		fa->fsx_nextents = xfs_iext_count(ifp);