xfs: Add support FALLOC_FL_INSERT_RANGE for fallocate

This patch implements fallocate's FALLOC_FL_INSERT_RANGE for XFS.

1) Make sure that both offset and len are block size aligned.
2) Update the i_size of inode by len bytes.
3) Compute the file's logical block number against offset. If the computed
   block number is not the starting block of the extent, split the extent
   such that the block number is the starting block of the extent.
4) Shift all the extents which are lying bewteen [offset, last allocated extent]
   towards right by len bytes. This step will make a hole of len bytes
   at offset.
Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Ashish Sangwan <a.sangwan@samsung.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>

fs/xfs/libxfs/xfs_bmap.h

@@ -166,6 +166,11 @@ static inline void xfs_bmap_init(xfs_bmap_free_t *flp, xfs_fsblock_t *fbp)
  */
 #define XFS_BMAP_MAX_SHIFT_EXTENTS	1
 
+enum shift_direction {
+	SHIFT_LEFT = 0,
+	SHIFT_RIGHT,
+};
+
 #ifdef DEBUG
 void	xfs_bmap_trace_exlist(struct xfs_inode *ip, xfs_extnum_t cnt,
 		int whichfork, unsigned long caller_ip);
@@ -211,8 +216,10 @@ int	xfs_check_nostate_extents(struct xfs_ifork *ifp, xfs_extnum_t idx,
 		xfs_extnum_t num);
 uint	xfs_default_attroffset(struct xfs_inode *ip);
 int	xfs_bmap_shift_extents(struct xfs_trans *tp, struct xfs_inode *ip,
-		xfs_fileoff_t start_fsb, xfs_fileoff_t offset_shift_fsb,
-		int *done, xfs_fileoff_t *next_fsb, xfs_fsblock_t *firstblock,
-		struct xfs_bmap_free *flist, int num_exts);
+		xfs_fileoff_t *next_fsb, xfs_fileoff_t offset_shift_fsb,
+		int *done, xfs_fileoff_t stop_fsb, xfs_fsblock_t *firstblock,
+		struct xfs_bmap_free *flist, enum shift_direction direction,
+		int num_exts);
+int	xfs_bmap_split_extent(struct xfs_inode *ip, xfs_fileoff_t split_offset);
 
 #endif	/* __XFS_BMAP_H__ */

fs/xfs/xfs_bmap_util.c

@@ -1376,22 +1376,19 @@ xfs_zero_file_space(
 }
 
 /*
- * xfs_collapse_file_space()
- *	This routine frees disk space and shift extent for the given file.
- *	The first thing we do is to free data blocks in the specified range
- *	by calling xfs_free_file_space(). It would also sync dirty data
- *	and invalidate page cache over the region on which collapse range
- *	is working. And Shift extent records to the left to cover a hole.
- * RETURNS:
- *	0 on success
- *	errno on error
- *
+ * @next_fsb will keep track of the extent currently undergoing shift.
+ * @stop_fsb will keep track of the extent at which we have to stop.
+ * If we are shifting left, we will start with block (offset + len) and
+ * shift each extent till last extent.
+ * If we are shifting right, we will start with last extent inside file space
+ * and continue until we reach the block corresponding to offset.
  */
 int
-xfs_collapse_file_space(
-	struct xfs_inode	*ip,
-	xfs_off_t		offset,
-	xfs_off_t		len)
+xfs_shift_file_space(
+	struct xfs_inode        *ip,
+	xfs_off_t               offset,
+	xfs_off_t               len,
+	enum shift_direction	direction)
 {
 	int			done = 0;
 	struct xfs_mount	*mp = ip->i_mount;
@@ -1400,21 +1397,26 @@ xfs_collapse_file_space(
 	struct xfs_bmap_free	free_list;
 	xfs_fsblock_t		first_block;
 	int			committed;
-	xfs_fileoff_t		start_fsb;
+	xfs_fileoff_t		stop_fsb;
 	xfs_fileoff_t		next_fsb;
 	xfs_fileoff_t		shift_fsb;
 
-	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
+	ASSERT(direction == SHIFT_LEFT || direction == SHIFT_RIGHT);
 
-	trace_xfs_collapse_file_space(ip);
+	if (direction == SHIFT_LEFT) {
+		next_fsb = XFS_B_TO_FSB(mp, offset + len);
+		stop_fsb = XFS_B_TO_FSB(mp, VFS_I(ip)->i_size);
+	} else {
+		/*
+		 * If right shift, delegate the work of initialization of
+		 * next_fsb to xfs_bmap_shift_extent as it has ilock held.
+		 */
+		next_fsb = NULLFSBLOCK;
+		stop_fsb = XFS_B_TO_FSB(mp, offset);
+	}
 
-	next_fsb = XFS_B_TO_FSB(mp, offset + len);
 	shift_fsb = XFS_B_TO_FSB(mp, len);
 
-	error = xfs_free_file_space(ip, offset, len);
-	if (error)
-		return error;
-
 	/*
 	 * Trim eofblocks to avoid shifting uninitialized post-eof preallocation
 	 * into the accessible region of the file.
@@ -1427,20 +1429,28 @@ xfs_collapse_file_space(
 
 	/*
 	 * Writeback and invalidate cache for the remainder of the file as we're
-	 * about to shift down every extent from the collapse range to EOF. The
-	 * free of the collapse range above might have already done some of
-	 * this, but we shouldn't rely on it to do anything outside of the range
-	 * that was freed.
+	 * about to shift down every extent from offset to EOF.
 	 */
 	error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
-					     offset + len, -1);
+					     offset, -1);
 	if (error)
 		return error;
 	error = invalidate_inode_pages2_range(VFS_I(ip)->i_mapping,
-					(offset + len) >> PAGE_CACHE_SHIFT, -1);
+					offset >> PAGE_CACHE_SHIFT, -1);
 	if (error)
 		return error;
 
+	/*
+	 * The extent shiting code works on extent granularity. So, if
+	 * stop_fsb is not the starting block of extent, we need to split
+	 * the extent at stop_fsb.
+	 */
+	if (direction == SHIFT_RIGHT) {
+		error = xfs_bmap_split_extent(ip, stop_fsb);
+		if (error)
+			return error;
+	}
+
 	while (!error && !done) {
 		tp = xfs_trans_alloc(mp, XFS_TRANS_DIOSTRAT);
 		/*
@@ -1464,7 +1474,7 @@ xfs_collapse_file_space(
 		if (error)
 			goto out;
 
-		xfs_trans_ijoin(tp, ip, 0);
+		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
 
 		xfs_bmap_init(&free_list, &first_block);
 
@@ -1472,10 +1482,9 @@ xfs_collapse_file_space(
 		 * We are using the write transaction in which max 2 bmbt
 		 * updates are allowed
 		 */
-		start_fsb = next_fsb;
-		error = xfs_bmap_shift_extents(tp, ip, start_fsb, shift_fsb,
-				&done, &next_fsb, &first_block, &free_list,
-				XFS_BMAP_MAX_SHIFT_EXTENTS);
+		error = xfs_bmap_shift_extents(tp, ip, &next_fsb, shift_fsb,
+				&done, stop_fsb, &first_block, &free_list,
+				direction, XFS_BMAP_MAX_SHIFT_EXTENTS);
 		if (error)
 			goto out;
 
@@ -1484,17 +1493,69 @@ xfs_collapse_file_space(
 			goto out;
 
 		error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
-		xfs_iunlock(ip, XFS_ILOCK_EXCL);
 	}
 
 	return error;
 
 out:
 	xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
-	xfs_iunlock(ip, XFS_ILOCK_EXCL);
 	return error;
 }
 
+/*
+ * xfs_collapse_file_space()
+ *	This routine frees disk space and shift extent for the given file.
+ *	The first thing we do is to free data blocks in the specified range
+ *	by calling xfs_free_file_space(). It would also sync dirty data
+ *	and invalidate page cache over the region on which collapse range
+ *	is working. And Shift extent records to the left to cover a hole.
+ * RETURNS:
+ *	0 on success
+ *	errno on error
+ *
+ */
+int
+xfs_collapse_file_space(
+	struct xfs_inode	*ip,
+	xfs_off_t		offset,
+	xfs_off_t		len)
+{
+	int error;
+
+	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
+	trace_xfs_collapse_file_space(ip);
+
+	error = xfs_free_file_space(ip, offset, len);
+	if (error)
+		return error;
+
+	return xfs_shift_file_space(ip, offset, len, SHIFT_LEFT);
+}
+
+/*
+ * xfs_insert_file_space()
+ *	This routine create hole space by shifting extents for the given file.
+ *	The first thing we do is to sync dirty data and invalidate page cache
+ *	over the region on which insert range is working. And split an extent
+ *	to two extents at given offset by calling xfs_bmap_split_extent.
+ *	And shift all extent records which are laying between [offset,
+ *	last allocated extent] to the right to reserve hole range.
+ * RETURNS:
+ *	0 on success
+ *	errno on error
+ */
+int
+xfs_insert_file_space(
+	struct xfs_inode	*ip,
+	loff_t			offset,
+	loff_t			len)
+{
+	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
+	trace_xfs_insert_file_space(ip);
+
+	return xfs_shift_file_space(ip, offset, len, SHIFT_RIGHT);
+}
+
 /*
  * We need to check that the format of the data fork in the temporary inode is
  * valid for the target inode before doing the swap. This is not a problem with

fs/xfs/xfs_bmap_util.h

@@ -63,6 +63,8 @@ int	xfs_zero_file_space(struct xfs_inode *ip, xfs_off_t offset,
 			    xfs_off_t len);
 int	xfs_collapse_file_space(struct xfs_inode *, xfs_off_t offset,
 				xfs_off_t len);
+int	xfs_insert_file_space(struct xfs_inode *, xfs_off_t offset,
+				xfs_off_t len);
 
 /* EOF block manipulation functions */
 bool	xfs_can_free_eofblocks(struct xfs_inode *ip, bool force);

fs/xfs/xfs_file.c

@@ -816,6 +816,11 @@ xfs_file_write_iter(
 	return ret;
 }
 
+#define	XFS_FALLOC_FL_SUPPORTED						\
+		(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |		\
+		 FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |	\
+		 FALLOC_FL_INSERT_RANGE)
+
 STATIC long
 xfs_file_fallocate(
 	struct file		*file,
@@ -829,11 +834,11 @@ xfs_file_fallocate(
 	enum xfs_prealloc_flags	flags = 0;
 	uint			iolock = XFS_IOLOCK_EXCL;
 	loff_t			new_size = 0;
+	bool			do_file_insert = 0;
 
 	if (!S_ISREG(inode->i_mode))
 		return -EINVAL;
-	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
-		     FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE))
+	if (mode & ~XFS_FALLOC_FL_SUPPORTED)
 		return -EOPNOTSUPP;
 
 	xfs_ilock(ip, iolock);
@@ -867,6 +872,27 @@ xfs_file_fallocate(
 		error = xfs_collapse_file_space(ip, offset, len);
 		if (error)
 			goto out_unlock;
+	} else if (mode & FALLOC_FL_INSERT_RANGE) {
+		unsigned blksize_mask = (1 << inode->i_blkbits) - 1;
+
+		new_size = i_size_read(inode) + len;
+		if (offset & blksize_mask || len & blksize_mask) {
+			error = -EINVAL;
+			goto out_unlock;
+		}
+
+		/* check the new inode size does not wrap through zero */
+		if (new_size > inode->i_sb->s_maxbytes) {
+			error = -EFBIG;
+			goto out_unlock;
+		}
+
+		/* Offset should be less than i_size */
+		if (offset >= i_size_read(inode)) {
+			error = -EINVAL;
+			goto out_unlock;
+		}
+		do_file_insert = 1;
 	} else {
 		flags |= XFS_PREALLOC_SET;
 
@@ -901,8 +927,19 @@ xfs_file_fallocate(
 		iattr.ia_valid = ATTR_SIZE;
 		iattr.ia_size = new_size;
 		error = xfs_setattr_size(ip, &iattr);
+		if (error)
+			goto out_unlock;
 	}
 
+	/*
+	 * Perform hole insertion now that the file size has been
+	 * updated so that if we crash during the operation we don't
+	 * leave shifted extents past EOF and hence losing access to
+	 * the data that is contained within them.
+	 */
+	if (do_file_insert)
+		error = xfs_insert_file_space(ip, offset, len);
+
 out_unlock:
 	xfs_iunlock(ip, iolock);
 	return error;

fs/xfs/xfs_trace.h

@@ -664,6 +664,7 @@ DEFINE_INODE_EVENT(xfs_alloc_file_space);
 DEFINE_INODE_EVENT(xfs_free_file_space);
 DEFINE_INODE_EVENT(xfs_zero_file_space);
 DEFINE_INODE_EVENT(xfs_collapse_file_space);
+DEFINE_INODE_EVENT(xfs_insert_file_space);
 DEFINE_INODE_EVENT(xfs_readdir);
 #ifdef CONFIG_XFS_POSIX_ACL
 DEFINE_INODE_EVENT(xfs_get_acl);