xfs: update i_size after unwritten conversion in dio completion

Since commit d531d91 ("xfs: always use unwritten extents for
direct I/O writes"), we start allocating unwritten extents for all
direct writes to allow appending aio in XFS.

But for dio writes that could extend file size we update the in-core
inode size first, then convert the unwritten extents to real
allocations at dio completion time in xfs_dio_write_end_io(). Thus a
racing direct read could see the new i_size and find the unwritten
extents first and read zeros instead of actual data, if the direct
writer also takes a shared iolock.

Fix it by updating the in-core inode size after the unwritten extent
conversion. To do this, introduce a new boolean argument to
xfs_iomap_write_unwritten() to tell if we want to update in-core
i_size or not.

Suggested-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

fs/xfs/xfs_aops.c

@@ -343,7 +343,8 @@ xfs_end_io(
 		error = xfs_reflink_end_cow(ip, offset, size);
 		break;
 	case XFS_IO_UNWRITTEN:
-		error = xfs_iomap_write_unwritten(ip, offset, size);
+		/* writeback should never update isize */
+		error = xfs_iomap_write_unwritten(ip, offset, size, false);
 		break;
 	default:
 		ASSERT(!xfs_ioend_is_append(ioend) || ioend->io_append_trans);

fs/xfs/xfs_file.c

@@ -434,7 +434,6 @@ xfs_dio_write_end_io(
 	struct inode		*inode = file_inode(iocb->ki_filp);
 	struct xfs_inode	*ip = XFS_I(inode);
 	loff_t			offset = iocb->ki_pos;
-	bool			update_size = false;
 	int			error = 0;
 
 	trace_xfs_end_io_direct_write(ip, offset, size);
@@ -445,6 +444,21 @@ xfs_dio_write_end_io(
 	if (size <= 0)
 		return size;
 
+	if (flags & IOMAP_DIO_COW) {
+		error = xfs_reflink_end_cow(ip, offset, size);
+		if (error)
+			return error;
+	}
+
+	/*
+	 * Unwritten conversion updates the in-core isize after extent
+	 * conversion but before updating the on-disk size. Updating isize any
+	 * earlier allows a racing dio read to find unwritten extents before
+	 * they are converted.
+	 */
+	if (flags & IOMAP_DIO_UNWRITTEN)
+		return xfs_iomap_write_unwritten(ip, offset, size, true);
+
 	/*
 	 * We need to update the in-core inode size here so that we don't end up
 	 * with the on-disk inode size being outside the in-core inode size. We
@@ -459,20 +473,11 @@ xfs_dio_write_end_io(
 	spin_lock(&ip->i_flags_lock);
 	if (offset + size > i_size_read(inode)) {
 		i_size_write(inode, offset + size);
-		update_size = true;
-	}
-	spin_unlock(&ip->i_flags_lock);
-
-	if (flags & IOMAP_DIO_COW) {
-		error = xfs_reflink_end_cow(ip, offset, size);
-		if (error)
-			return error;
-	}
-
-	if (flags & IOMAP_DIO_UNWRITTEN)
-		error = xfs_iomap_write_unwritten(ip, offset, size);
-	else if (update_size)
+		spin_unlock(&ip->i_flags_lock);
 		error = xfs_setfilesize(ip, offset, size);
+	} else {
+		spin_unlock(&ip->i_flags_lock);
+	}
 
 	return error;
 }

fs/xfs/xfs_iomap.c

@@ -829,7 +829,8 @@ int
 xfs_iomap_write_unwritten(
 	xfs_inode_t	*ip,
 	xfs_off_t	offset,
-	xfs_off_t	count)
+	xfs_off_t	count,
+	bool		update_isize)
 {
 	xfs_mount_t	*mp = ip->i_mount;
 	xfs_fileoff_t	offset_fsb;
@@ -840,6 +841,7 @@ xfs_iomap_write_unwritten(
 	xfs_trans_t	*tp;
 	xfs_bmbt_irec_t imap;
 	struct xfs_defer_ops dfops;
+	struct inode	*inode = VFS_I(ip);
 	xfs_fsize_t	i_size;
 	uint		resblks;
 	int		error;
@@ -899,7 +901,8 @@ xfs_iomap_write_unwritten(
 		i_size = XFS_FSB_TO_B(mp, offset_fsb + count_fsb);
 		if (i_size > offset + count)
 			i_size = offset + count;
-
+		if (update_isize && i_size > i_size_read(inode))
+			i_size_write(inode, i_size);
 		i_size = xfs_new_eof(ip, i_size);
 		if (i_size) {
 			ip->i_d.di_size = i_size;

fs/xfs/xfs_iomap.h

@@ -27,7 +27,7 @@ int xfs_iomap_write_direct(struct xfs_inode *, xfs_off_t, size_t,
 			struct xfs_bmbt_irec *, int);
 int xfs_iomap_write_allocate(struct xfs_inode *, int, xfs_off_t,
 			struct xfs_bmbt_irec *);
-int xfs_iomap_write_unwritten(struct xfs_inode *, xfs_off_t, xfs_off_t);
+int xfs_iomap_write_unwritten(struct xfs_inode *, xfs_off_t, xfs_off_t, bool);
 
 void xfs_bmbt_to_iomap(struct xfs_inode *, struct iomap *,
 		struct xfs_bmbt_irec *);

fs/xfs/xfs_pnfs.c

@@ -274,7 +274,7 @@ xfs_fs_commit_blocks(
 					(end - 1) >> PAGE_SHIFT);
 		WARN_ON_ONCE(error);
 
-		error = xfs_iomap_write_unwritten(ip, start, length);
+		error = xfs_iomap_write_unwritten(ip, start, length, false);
 		if (error)
 			goto out_drop_iolock;
 	}