Merge git://oss.sgi.com:8090/xfs/xfs-2.6

* git://oss.sgi.com:8090/xfs/xfs-2.6:
  [XFS] Add lockdep support for XFS
  [XFS] Fix race in xfs_write() b/w dmapi callout and direct I/O checks.
  [XFS] Get rid of redundant "required" in msg.
  [XFS] Export via a function xfs_buftarg_list for use by kdb/xfsidbg.
  [XFS] Remove unused ilen variable and references.
  [XFS] Fix to prevent the notorious 'NULL files' problem after a crash.
  [XFS] Fix race condition in xfs_write().
  [XFS] Fix uquota and oquota enforcement problems.
  [XFS] propogate return codes from flush routines
  [XFS] Fix quotaon syscall failures for group enforcement requests.
  [XFS] Invalidate quotacheck when mounting without a quota type.
  [XFS] reducing the number of random number functions.
  [XFS] remove more misc. unused args
  [XFS] the "aendp" arg to xfs_dir2_data_freescan is always NULL, remove it.
  [XFS] The last argument "lsn" of xfs_trans_commit() is always called with

8 files changed, 212 insertions, 90 deletions

diff --git a/fs/xfs/linux-2.6/mrlock.h b/fs/xfs/linux-2.6/mrlock.h
index af168a1a98c..c110bb00266 100644
--- a/fs/xfs/linux-2.6/mrlock.h
+++ b/fs/xfs/linux-2.6/mrlock.h
@@ -43,6 +43,18 @@ static inline void mrupdate(mrlock_t *mrp)
 	mrp->mr_writer = 1;
 }
 
+static inline void mraccess_nested(mrlock_t *mrp, int subclass)
+{
+	down_read_nested(&mrp->mr_lock, subclass);
+}
+
+static inline void mrupdate_nested(mrlock_t *mrp, int subclass)
+{
+	down_write_nested(&mrp->mr_lock, subclass);
+	mrp->mr_writer = 1;
+}
+
+
 static inline int mrtryaccess(mrlock_t *mrp)
 {
 	return down_read_trylock(&mrp->mr_lock);
diff --git a/fs/xfs/linux-2.6/xfs_aops.c b/fs/xfs/linux-2.6/xfs_aops.c
index 143ffc851c9..4475588e973 100644
--- a/fs/xfs/linux-2.6/xfs_aops.c
+++ b/fs/xfs/linux-2.6/xfs_aops.c
@@ -141,9 +141,46 @@ xfs_destroy_ioend(
 }
 
 /*
+ * Update on-disk file size now that data has been written to disk.
+ * The current in-memory file size is i_size.  If a write is beyond
+ * eof io_new_size will be the intended file size until i_size is
+ * updated.  If this write does not extend all the way to the valid
+ * file size then restrict this update to the end of the write.
+ */
+STATIC void
+xfs_setfilesize(
+	xfs_ioend_t		*ioend)
+{
+	xfs_inode_t		*ip;
+	xfs_fsize_t		isize;
+	xfs_fsize_t		bsize;
+
+	ip = xfs_vtoi(ioend->io_vnode);
+
+	ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFREG);
+	ASSERT(ioend->io_type != IOMAP_READ);
+
+	if (unlikely(ioend->io_error))
+		return;
+
+	bsize = ioend->io_offset + ioend->io_size;
+
+	xfs_ilock(ip, XFS_ILOCK_EXCL);
+
+	isize = MAX(ip->i_size, ip->i_iocore.io_new_size);
+	isize = MIN(isize, bsize);
+
+	if (ip->i_d.di_size < isize) {
+		ip->i_d.di_size = isize;
+		ip->i_update_core = 1;
+		ip->i_update_size = 1;
+	}
+
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+}
+
+/*
  * Buffered IO write completion for delayed allocate extents.
- * TODO: Update ondisk isize now that we know the file data
- * has been flushed (i.e. the notorious "NULL file" problem).
  */
 STATIC void
 xfs_end_bio_delalloc(
@@ -152,6 +189,7 @@ xfs_end_bio_delalloc(
 	xfs_ioend_t		*ioend =
 		container_of(work, xfs_ioend_t, io_work);
 
+	xfs_setfilesize(ioend);
 	xfs_destroy_ioend(ioend);
 }
 
@@ -165,6 +203,7 @@ xfs_end_bio_written(
 	xfs_ioend_t		*ioend =
 		container_of(work, xfs_ioend_t, io_work);
 
+	xfs_setfilesize(ioend);
 	xfs_destroy_ioend(ioend);
 }
 
@@ -184,8 +223,23 @@ xfs_end_bio_unwritten(
 	xfs_off_t		offset = ioend->io_offset;
 	size_t			size = ioend->io_size;
 
-	if (likely(!ioend->io_error))
+	if (likely(!ioend->io_error)) {
 		bhv_vop_bmap(vp, offset, size, BMAPI_UNWRITTEN, NULL, NULL);
+		xfs_setfilesize(ioend);
+	}
+	xfs_destroy_ioend(ioend);
+}
+
+/*
+ * IO read completion for regular, written extents.
+ */
+STATIC void
+xfs_end_bio_read(
+	struct work_struct	*work)
+{
+	xfs_ioend_t		*ioend =
+		container_of(work, xfs_ioend_t, io_work);
+
 	xfs_destroy_ioend(ioend);
 }
 
@@ -224,6 +278,8 @@ xfs_alloc_ioend(
 		INIT_WORK(&ioend->io_work, xfs_end_bio_unwritten);
 	else if (type == IOMAP_DELAY)
 		INIT_WORK(&ioend->io_work, xfs_end_bio_delalloc);
+	else if (type == IOMAP_READ)
+		INIT_WORK(&ioend->io_work, xfs_end_bio_read);
 	else
 		INIT_WORK(&ioend->io_work, xfs_end_bio_written);
 
@@ -913,7 +969,7 @@ xfs_page_state_convert(
 	bh = head = page_buffers(page);
 	offset = page_offset(page);
 	flags = -1;
-	type = 0;
+	type = IOMAP_READ;
 
 	/* TODO: cleanup count and page_dirty */
 
@@ -999,7 +1055,7 @@ xfs_page_state_convert(
 			 * That means it must already have extents allocated
 			 * underneath it. Map the extent by reading it.
 			 */
-			if (!iomap_valid || type != 0) {
+			if (!iomap_valid || type != IOMAP_READ) {
 				flags = BMAPI_READ;
 				size = xfs_probe_cluster(inode, page, bh,
 								head, 1);
@@ -1010,7 +1066,7 @@ xfs_page_state_convert(
 				iomap_valid = xfs_iomap_valid(&iomap, offset);
 			}
 
-			type = 0;
+			type = IOMAP_READ;
 			if (!test_and_set_bit(BH_Lock, &bh->b_state)) {
 				ASSERT(buffer_mapped(bh));
 				if (iomap_valid)
@@ -1356,12 +1412,21 @@ xfs_end_io_direct(
 	 * completion handler in the future, in which case all this can
 	 * go away.
 	 */
-	if (private && size > 0) {
-		ioend->io_offset = offset;
-		ioend->io_size = size;
+	ioend->io_offset = offset;
+	ioend->io_size = size;
+	if (ioend->io_type == IOMAP_READ) {
+		xfs_finish_ioend(ioend);
+	} else if (private && size > 0) {
 		xfs_finish_ioend(ioend);
 	} else {
-		xfs_destroy_ioend(ioend);
+		/*
+		 * A direct I/O write ioend starts it's life in unwritten
+		 * state in case they map an unwritten extent.  This write
+		 * didn't map an unwritten extent so switch it's completion
+		 * handler.
+		 */
+		INIT_WORK(&ioend->io_work, xfs_end_bio_written);
+		xfs_finish_ioend(ioend);
 	}
 
 	/*
@@ -1392,15 +1457,15 @@ xfs_vm_direct_IO(
 	if (error)
 		return -error;
 
-	iocb->private = xfs_alloc_ioend(inode, IOMAP_UNWRITTEN);
-
 	if (rw == WRITE) {
+		iocb->private = xfs_alloc_ioend(inode, IOMAP_UNWRITTEN);
 		ret = blockdev_direct_IO_own_locking(rw, iocb, inode,
 			iomap.iomap_target->bt_bdev,
 			iov, offset, nr_segs,
 			xfs_get_blocks_direct,
 			xfs_end_io_direct);
 	} else {
+		iocb->private = xfs_alloc_ioend(inode, IOMAP_READ);
 		ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
 			iomap.iomap_target->bt_bdev,
 			iov, offset, nr_segs,
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index 69e9e80735d..fe4f66a5af1 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -1426,7 +1426,7 @@ xfs_free_bufhash(
 /*
  *	buftarg list for delwrite queue processing
  */
-LIST_HEAD(xfs_buftarg_list);
+static LIST_HEAD(xfs_buftarg_list);
 static DEFINE_SPINLOCK(xfs_buftarg_lock);
 
 STATIC void
@@ -1867,3 +1867,11 @@ xfs_buf_terminate(void)
 	ktrace_free(xfs_buf_trace_buf);
 #endif
 }
+
+#ifdef CONFIG_KDB_MODULES
+struct list_head *
+xfs_get_buftarg_list(void)
+{
+	return &xfs_buftarg_list;
+}
+#endif
diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h
index 9e8ef8fef39..b6241f6201a 100644
--- a/fs/xfs/linux-2.6/xfs_buf.h
+++ b/fs/xfs/linux-2.6/xfs_buf.h
@@ -411,6 +411,9 @@ extern void xfs_free_buftarg(xfs_buftarg_t *, int);
 extern void xfs_wait_buftarg(xfs_buftarg_t *);
 extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int, unsigned int);
 extern int xfs_flush_buftarg(xfs_buftarg_t *, int);
+#ifdef CONFIG_KDB_MODULES
+extern struct list_head *xfs_get_buftarg_list(void);
+#endif
 
 #define xfs_getsize_buftarg(buftarg)	block_size((buftarg)->bt_bdev)
 #define xfs_readonly_buftarg(buftarg)	bdev_read_only((buftarg)->bt_bdev)
diff --git a/fs/xfs/linux-2.6/xfs_fs_subr.c b/fs/xfs/linux-2.6/xfs_fs_subr.c
index dc0562828e7..2eb87cd082a 100644
--- a/fs/xfs/linux-2.6/xfs_fs_subr.c
+++ b/fs/xfs/linux-2.6/xfs_fs_subr.c
@@ -35,7 +35,7 @@ fs_tosspages(
 		truncate_inode_pages(ip->i_mapping, first);
 }
 
-void
+int
 fs_flushinval_pages(
 	bhv_desc_t	*bdp,
 	xfs_off_t	first,
@@ -44,13 +44,16 @@ fs_flushinval_pages(
 {
 	bhv_vnode_t	*vp = BHV_TO_VNODE(bdp);
 	struct inode	*ip = vn_to_inode(vp);
+	int		ret = 0;
 
 	if (VN_CACHED(vp)) {
 		if (VN_TRUNC(vp))
 			VUNTRUNCATE(vp);
-		filemap_write_and_wait(ip->i_mapping);
-		truncate_inode_pages(ip->i_mapping, first);
+		ret = filemap_write_and_wait(ip->i_mapping);
+		if (!ret)
+			truncate_inode_pages(ip->i_mapping, first);
 	}
+	return ret;
 }
 
 int
@@ -63,14 +66,18 @@ fs_flush_pages(
 {
 	bhv_vnode_t	*vp = BHV_TO_VNODE(bdp);
 	struct inode	*ip = vn_to_inode(vp);
+	int		ret = 0;
+	int		ret2;
 
 	if (VN_DIRTY(vp)) {
 		if (VN_TRUNC(vp))
 			VUNTRUNCATE(vp);
-		filemap_fdatawrite(ip->i_mapping);
+		ret = filemap_fdatawrite(ip->i_mapping);
 		if (flags & XFS_B_ASYNC)
-			return 0;
-		filemap_fdatawait(ip->i_mapping);
+			return ret;
+		ret2 = filemap_fdatawait(ip->i_mapping);
+		if (!ret)
+			ret = ret2;
 	}
-	return 0;
+	return ret;
 }
diff --git a/fs/xfs/linux-2.6/xfs_fs_subr.h b/fs/xfs/linux-2.6/xfs_fs_subr.h
index aee9ccdd18f..c1b53118a30 100644
--- a/fs/xfs/linux-2.6/xfs_fs_subr.h
+++ b/fs/xfs/linux-2.6/xfs_fs_subr.h
@@ -23,7 +23,7 @@ extern int  fs_noerr(void);
 extern int  fs_nosys(void);
 extern void fs_noval(void);
 extern void fs_tosspages(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
-extern void fs_flushinval_pages(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
+extern int  fs_flushinval_pages(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
 extern int  fs_flush_pages(bhv_desc_t *, xfs_off_t, xfs_off_t, uint64_t, int);
 
 #endif	/* __XFS_FS_SUBR_H__ */
diff --git a/fs/xfs/linux-2.6/xfs_lrw.c b/fs/xfs/linux-2.6/xfs_lrw.c
index 558076dd075..86fb671a8bc 100644
--- a/fs/xfs/linux-2.6/xfs_lrw.c
+++ b/fs/xfs/linux-2.6/xfs_lrw.c
@@ -191,7 +191,7 @@ xfs_read(
 	struct file		*file = iocb->ki_filp;
 	struct inode		*inode = file->f_mapping->host;
 	size_t			size = 0;
-	ssize_t			ret;
+	ssize_t			ret = 0;
 	xfs_fsize_t		n;
 	xfs_inode_t		*ip;
 	xfs_mount_t		*mp;
@@ -224,7 +224,7 @@ xfs_read(
 				mp->m_rtdev_targp : mp->m_ddev_targp;
 		if ((*offset & target->bt_smask) ||
 		    (size & target->bt_smask)) {
-			if (*offset == ip->i_d.di_size) {
+			if (*offset == ip->i_size) {
 				return (0);
 			}
 			return -XFS_ERROR(EINVAL);
@@ -263,9 +263,13 @@ xfs_read(
 
 	if (unlikely(ioflags & IO_ISDIRECT)) {
 		if (VN_CACHED(vp))
-			bhv_vop_flushinval_pages(vp, ctooff(offtoct(*offset)),
+			ret = bhv_vop_flushinval_pages(vp, ctooff(offtoct(*offset)),
 						 -1, FI_REMAPF_LOCKED);
 		mutex_unlock(&inode->i_mutex);
+		if (ret) {
+			xfs_iunlock(ip, XFS_IOLOCK_SHARED);
+			return ret;
+		}
 	}
 
 	xfs_rw_enter_trace(XFS_READ_ENTER, &ip->i_iocore,
@@ -383,9 +387,10 @@ xfs_splice_write(
 {
 	xfs_inode_t		*ip = XFS_BHVTOI(bdp);
 	xfs_mount_t		*mp = ip->i_mount;
+	xfs_iocore_t		*io = &ip->i_iocore;
 	ssize_t			ret;
 	struct inode		*inode = outfilp->f_mapping->host;
-	xfs_fsize_t		isize;
+	xfs_fsize_t		isize, new_size;
 
 	XFS_STATS_INC(xs_write_calls);
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount))
@@ -406,6 +411,14 @@ xfs_splice_write(
 			return -error;
 		}
 	}
+
+	new_size = *ppos + count;
+
+	xfs_ilock(ip, XFS_ILOCK_EXCL);
+	if (new_size > ip->i_size)
+		io->io_new_size = new_size;
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
 	xfs_rw_enter_trace(XFS_SPLICE_WRITE_ENTER, &ip->i_iocore,
 			   pipe, count, *ppos, ioflags);
 	ret = generic_file_splice_write(pipe, outfilp, ppos, count, flags);
@@ -416,14 +429,18 @@ xfs_splice_write(
 	if (unlikely(ret < 0 && ret != -EFAULT && *ppos > isize))
 		*ppos = isize;
 
-	if (*ppos > ip->i_d.di_size) {
+	if (*ppos > ip->i_size) {
 		xfs_ilock(ip, XFS_ILOCK_EXCL);
-		if (*ppos > ip->i_d.di_size) {
-			ip->i_d.di_size = *ppos;
-			i_size_write(inode, *ppos);
-			ip->i_update_core = 1;
-			ip->i_update_size = 1;
-		}
+		if (*ppos > ip->i_size)
+			ip->i_size = *ppos;
+		xfs_iunlock(ip, XFS_ILOCK_EXCL);
+	}
+
+	if (io->io_new_size) {
+		xfs_ilock(ip, XFS_ILOCK_EXCL);
+		io->io_new_size = 0;
+		if (ip->i_d.di_size > ip->i_size)
+			ip->i_d.di_size = ip->i_size;
 		xfs_iunlock(ip, XFS_ILOCK_EXCL);
 	}
 	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
@@ -644,7 +661,7 @@ xfs_write(
 	bhv_vrwlock_t		locktype;
 	size_t			ocount = 0, count;
 	loff_t			pos;
-	int			need_i_mutex = 1, need_flush = 0;
+	int			need_i_mutex;
 
 	XFS_STATS_INC(xs_write_calls);
 
@@ -669,39 +686,20 @@ xfs_write(
 	if (XFS_FORCED_SHUTDOWN(mp))
 		return -EIO;
 
-	if (ioflags & IO_ISDIRECT) {
-		xfs_buftarg_t	*target =
-			(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
-				mp->m_rtdev_targp : mp->m_ddev_targp;
-
-		if ((pos & target->bt_smask) || (count & target->bt_smask))
-			return XFS_ERROR(-EINVAL);
-
-		if (!VN_CACHED(vp) && pos < i_size_read(inode))
-			need_i_mutex = 0;
-
-		if (VN_CACHED(vp))
-			need_flush = 1;
-	}
-
 relock:
-	if (need_i_mutex) {
+	if (ioflags & IO_ISDIRECT) {
+		iolock = XFS_IOLOCK_SHARED;
+		locktype = VRWLOCK_WRITE_DIRECT;
+		need_i_mutex = 0;
+	} else {
 		iolock = XFS_IOLOCK_EXCL;
 		locktype = VRWLOCK_WRITE;
-
+		need_i_mutex = 1;
 		mutex_lock(&inode->i_mutex);
-	} else {
-		iolock = XFS_IOLOCK_SHARED;
-		locktype = VRWLOCK_WRITE_DIRECT;
 	}
 
 	xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
 
-	isize = i_size_read(inode);
-
-	if (file->f_flags & O_APPEND)
-		*offset = isize;
-
 start:
 	error = -generic_write_checks(file, &pos, &count,
 					S_ISBLK(inode->i_mode));
@@ -710,13 +708,8 @@ start:
 		goto out_unlock_mutex;
 	}
 
-	new_size = pos + count;
-	if (new_size > isize)
-		io->io_new_size = new_size;
-
 	if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
 	    !(ioflags & IO_INVIS) && !eventsent)) {
-		loff_t		savedsize = pos;
 		int		dmflags = FILP_DELAY_FLAG(file);
 
 		if (need_i_mutex)
@@ -727,8 +720,7 @@ start:
 				      pos, count,
 				      dmflags, &locktype);
 		if (error) {
-			xfs_iunlock(xip, iolock);
-			goto out_unlock_mutex;
+			goto out_unlock_internal;
 		}
 		xfs_ilock(xip, XFS_ILOCK_EXCL);
 		eventsent = 1;
@@ -740,12 +732,35 @@ start:
 		 * event prevents another call to XFS_SEND_DATA, which is
 		 * what allows the size to change in the first place.
 		 */
-		if ((file->f_flags & O_APPEND) && savedsize != isize) {
-			pos = isize = xip->i_d.di_size;
+		if ((file->f_flags & O_APPEND) && pos != xip->i_size)
+			goto start;
+	}
+
+	if (ioflags & IO_ISDIRECT) {
+		xfs_buftarg_t	*target =
+			(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
+				mp->m_rtdev_targp : mp->m_ddev_targp;
+
+		if ((pos & target->bt_smask) || (count & target->bt_smask)) {
+			xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
+			return XFS_ERROR(-EINVAL);
+		}
+
+		if (!need_i_mutex && (VN_CACHED(vp) || pos > xip->i_size)) {
+			xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
+			iolock = XFS_IOLOCK_EXCL;
+			locktype = VRWLOCK_WRITE;
+			need_i_mutex = 1;
+			mutex_lock(&inode->i_mutex);
+			xfs_ilock(xip, XFS_ILOCK_EXCL|iolock);
 			goto start;
 		}
 	}
 
+	new_size = pos + count;
+	if (new_size > xip->i_size)
+		io->io_new_size = new_size;
+
 	if (likely(!(ioflags & IO_INVIS))) {
 		file_update_time(file);
 		xfs_ichgtime_fast(xip, inode,
@@ -761,11 +776,11 @@ start:
 	 * to zero it out up to the new size.
 	 */
 
-	if (pos > isize) {
-		error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, pos, isize);
+	if (pos > xip->i_size) {
+		error = xfs_zero_eof(BHV_TO_VNODE(bdp), io, pos, xip->i_size);
 		if (error) {
-			xfs_iunlock(xip, XFS_ILOCK_EXCL|iolock);
-			goto out_unlock_mutex;
+			xfs_iunlock(xip, XFS_ILOCK_EXCL);
+			goto out_unlock_internal;
 		}
 	}
 	xfs_iunlock(xip, XFS_ILOCK_EXCL);
@@ -785,8 +800,7 @@ start:
 		if (likely(!error))
 			error = -remove_suid(file->f_path.dentry);
 		if (unlikely(error)) {
-			xfs_iunlock(xip, iolock);
-			goto out_unlock_mutex;
+			goto out_unlock_internal;
 		}
 	}
 
@@ -795,11 +809,14 @@ retry:
 	current->backing_dev_info = mapping->backing_dev_info;
 
 	if ((ioflags & IO_ISDIRECT)) {
-		if (need_flush) {
+		if (VN_CACHED(vp)) {
+			WARN_ON(need_i_mutex == 0);
 			xfs_inval_cached_trace(io, pos, -1,
 					ctooff(offtoct(pos)), -1);
-			bhv_vop_flushinval_pages(vp, ctooff(offtoct(pos)),
+			error = bhv_vop_flushinval_pages(vp, ctooff(offtoct(pos)),
 					-1, FI_REMAPF_LOCKED);
+			if (error)
+				goto out_unlock_internal;
 		}
 
 		if (need_i_mutex) {
@@ -827,7 +844,6 @@ retry:
 			pos += ret;
 			count -= ret;
 
-			need_i_mutex = 1;
 			ioflags &= ~IO_ISDIRECT;
 			xfs_iunlock(xip, iolock);
 			goto relock;
@@ -854,12 +870,12 @@ retry:
 		error = XFS_SEND_NAMESP(xip->i_mount, DM_EVENT_NOSPACE, vp,
 				DM_RIGHT_NULL, vp, DM_RIGHT_NULL, NULL, NULL,
 				0, 0, 0); /* Delay flag intentionally  unused */
-		if (error)
-			goto out_nounlocks;
 		if (need_i_mutex)
 			mutex_lock(&inode->i_mutex);
 		xfs_rwlock(bdp, locktype);
-		pos = xip->i_d.di_size;
+		if (error)
+			goto out_unlock_internal;
+		pos = xip->i_size;
 		ret = 0;
 		goto retry;
 	}
@@ -868,14 +884,10 @@ retry:
 	if (unlikely(ret < 0 && ret != -EFAULT && *offset > isize))
 		*offset = isize;
 
-	if (*offset > xip->i_d.di_size) {
+	if (*offset > xip->i_size) {
 		xfs_ilock(xip, XFS_ILOCK_EXCL);
-		if (*offset > xip->i_d.di_size) {
-			xip->i_d.di_size = *offset;
-			i_size_write(inode, *offset);
-			xip->i_update_core = 1;
-			xip->i_update_size = 1;
-		}
+		if (*offset > xip->i_size)
+			xip->i_size = *offset;
 		xfs_iunlock(xip, XFS_ILOCK_EXCL);
 	}
 
@@ -897,16 +909,31 @@ retry:
 
 		error = sync_page_range(inode, mapping, pos, ret);
 		if (!error)
-			error = ret;
-		return error;
+			error = -ret;
+		if (need_i_mutex)
+			mutex_lock(&inode->i_mutex);
+		xfs_rwlock(bdp, locktype);
 	}
 
  out_unlock_internal:
+	if (io->io_new_size) {
+		xfs_ilock(xip, XFS_ILOCK_EXCL);
+		io->io_new_size = 0;
+		/*
+		 * If this was a direct or synchronous I/O that failed (such
+		 * as ENOSPC) then part of the I/O may have been written to
+		 * disk before the error occured.  In this case the on-disk
+		 * file size may have been adjusted beyond the in-memory file
+		 * size and now needs to be truncated back.
+		 */
+		if (xip->i_d.di_size > xip->i_size)
+			xip->i_d.di_size = xip->i_size;
+		xfs_iunlock(xip, XFS_ILOCK_EXCL);
+	}
 	xfs_rwunlock(bdp, locktype);
  out_unlock_mutex:
 	if (need_i_mutex)
 		mutex_unlock(&inode->i_mutex);
- out_nounlocks:
 	return -error;
 }
 
diff --git a/fs/xfs/linux-2.6/xfs_vnode.h b/fs/xfs/linux-2.6/xfs_vnode.h
index b76118cf489..d1b2d01843d 100644
--- a/fs/xfs/linux-2.6/xfs_vnode.h
+++ b/fs/xfs/linux-2.6/xfs_vnode.h
@@ -194,7 +194,7 @@ typedef	int	(*vop_attr_list_t)(bhv_desc_t *, char *, int, int,
 typedef void	(*vop_link_removed_t)(bhv_desc_t *, bhv_vnode_t *, int);
 typedef void	(*vop_vnode_change_t)(bhv_desc_t *, bhv_vchange_t, __psint_t);
 typedef void	(*vop_ptossvp_t)(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
-typedef void	(*vop_pflushinvalvp_t)(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
+typedef int	(*vop_pflushinvalvp_t)(bhv_desc_t *, xfs_off_t, xfs_off_t, int);
 typedef int	(*vop_pflushvp_t)(bhv_desc_t *, xfs_off_t, xfs_off_t,
 				uint64_t, int);
 typedef int	(*vop_iflush_t)(bhv_desc_t *, int);