Merge branch 'xfs-shift-extents-rework' into for-next

fs/xfs/libxfs/xfs_bmap.h

@@ -178,9 +178,8 @@ 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,
-		int *done, xfs_fileoff_t start_fsb,
-		xfs_fileoff_t offset_shift_fsb, xfs_extnum_t *current_ext,
-		xfs_fsblock_t *firstblock, struct xfs_bmap_free	*flist,
-		int num_exts);
+		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);
 
 #endif	/* __XFS_BMAP_H__ */

fs/xfs/xfs_aops.c

@@ -434,10 +434,22 @@ xfs_start_page_writeback(
 {
 	ASSERT(PageLocked(page));
 	ASSERT(!PageWriteback(page));
-	if (clear_dirty)
+
+	/*
+	 * if the page was not fully cleaned, we need to ensure that the higher
+	 * layers come back to it correctly. That means we need to keep the page
+	 * dirty, and for WB_SYNC_ALL writeback we need to ensure the
+	 * PAGECACHE_TAG_TOWRITE index mark is not removed so another attempt to
+	 * write this page in this writeback sweep will be made.
+	 */
+	if (clear_dirty) {
 		clear_page_dirty_for_io(page);
 		set_page_writeback(page);
+	} else
+		set_page_writeback_keepwrite(page);
+
 	unlock_page(page);
+
 	/* If no buffers on the page are to be written, finish it here */
 	if (!buffers)
 		end_page_writeback(page);

fs/xfs/xfs_bmap_util.c

@@ -1205,6 +1205,7 @@ xfs_free_file_space(
 	xfs_bmap_free_t		free_list;
 	xfs_bmbt_irec_t		imap;
 	xfs_off_t		ioffset;
+	xfs_off_t		iendoffset;
 	xfs_extlen_t		mod=0;
 	xfs_mount_t		*mp;
 	int			nimap;
@@ -1233,12 +1234,13 @@ xfs_free_file_space(
 	inode_dio_wait(VFS_I(ip));
 
 	rounding = max_t(xfs_off_t, 1 << mp->m_sb.sb_blocklog, PAGE_CACHE_SIZE);
-	ioffset = offset & ~(rounding - 1);
-	error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
-					      ioffset, -1);
+	ioffset = round_down(offset, rounding);
+	iendoffset = round_up(offset + len, rounding) - 1;
+	error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, ioffset,
+					     iendoffset);
 	if (error)
 		goto out;
-	truncate_pagecache_range(VFS_I(ip), ioffset, -1);
+	truncate_pagecache_range(VFS_I(ip), ioffset, iendoffset);
 
 	/*
 	 * Need to zero the stuff we're not freeing, on disk.
@@ -1456,41 +1458,47 @@ xfs_collapse_file_space(
 	struct xfs_mount	*mp = ip->i_mount;
 	struct xfs_trans	*tp;
 	int			error;
-	xfs_extnum_t		current_ext = 0;
 	struct xfs_bmap_free	free_list;
 	xfs_fsblock_t		first_block;
 	int			committed;
 	xfs_fileoff_t		start_fsb;
+	xfs_fileoff_t		next_fsb;
 	xfs_fileoff_t		shift_fsb;
 
 	ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL));
 
 	trace_xfs_collapse_file_space(ip);
 
-	start_fsb = XFS_B_TO_FSB(mp, offset + len);
+	next_fsb = XFS_B_TO_FSB(mp, offset + len);
 	shift_fsb = XFS_B_TO_FSB(mp, len);
 
-	/*
-	 * Writeback the entire file and force remove any post-eof blocks. The
-	 * writeback prevents changes to the extent list via concurrent
-	 * writeback and the eofblocks trim prevents the extent shift algorithm
-	 * from running into a post-eof delalloc extent.
-	 *
-	 * XXX: This is a temporary fix until the extent shift loop below is
-	 * converted to use offsets and lookups within the ILOCK rather than
-	 * carrying around the index into the extent list for the next
-	 * iteration.
-	 */
-	error = filemap_write_and_wait(VFS_I(ip)->i_mapping);
+	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.
+	 */
 	if (xfs_can_free_eofblocks(ip, true)) {
 		error = xfs_free_eofblocks(mp, ip, false);
 		if (error)
 			return error;
 	}
 
-	error = xfs_free_file_space(ip, offset, len);
+	/*
+	 * 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.
+	 */
+	error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
+					     offset + len, -1);
+	if (error)
+		return error;
+	error = invalidate_inode_pages2_range(VFS_I(ip)->i_mapping,
+					(offset + len) >> PAGE_CACHE_SHIFT, -1);
 	if (error)
 		return error;
 
@@ -1525,9 +1533,9 @@ xfs_collapse_file_space(
 		 * We are using the write transaction in which max 2 bmbt
 		 * updates are allowed
 		 */
-		error = xfs_bmap_shift_extents(tp, ip, &done, start_fsb,
-					       shift_fsb, &current_ext,
-					       &first_block, &free_list,
+		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);
 		if (error)
 			goto out;