Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs

Pull btrfs updates from Chris Mason:
 "I held off on my rc5 pull because I hit an oops during log recovery
  after a crash.  I wanted to make sure it wasn't a regression because
  we have some logging fixes in here.

  It turns out that a commit during the merge window just made it much
  more likely to trigger directory logging instead of full commits,
  which exposed an old bug.

  The new backref walking code got some additional fixes.  This should
  be the final set of them.

  Josef fixed up a corner where our O_DIRECT writes and buffered reads
  could expose old file contents (not stale, just not the most recent).
  He and Liu Bo fixed crashes during tree log recover as well.

  Ilya fixed errors while we resume disk balancing operations on
  readonly mounts."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: run delayed directory updates during log replay
  Btrfs: hold a ref on the inode during writepages
  Btrfs: fix tree log remove space corner case
  Btrfs: fix wrong check during log recovery
  Btrfs: use _IOR for BTRFS_IOC_SUBVOL_GETFLAGS
  Btrfs: resume balance on rw (re)mounts properly
  Btrfs: restore restriper state on all mounts
  Btrfs: fix dio write vs buffered read race
  Btrfs: don't count I/O statistic read errors for missing devices
  Btrfs: resolve tree mod log locking issue in btrfs_next_leaf
  Btrfs: fix tree mod log rewind of ADD operations
  Btrfs: leave critical region in btrfs_find_all_roots as soon as possible
  Btrfs: always put insert_ptr modifications into the tree mod log
  Btrfs: fix tree mod log for root replacements at leaf level
  Btrfs: support root level changes in __resolve_indirect_ref
  Btrfs: avoid waiting for delayed refs when we must not

1 files changed, 51 insertions, 6 deletions

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index d8bb0dbc494..a7d1921ac76 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3754,7 +3754,7 @@ void btrfs_evict_inode(struct inode *inode)
 	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
 	if (root->fs_info->log_root_recovering) {
-		BUG_ON(!test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+		BUG_ON(test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
 				 &BTRFS_I(inode)->runtime_flags));
 		goto no_delete;
 	}
@@ -5876,8 +5876,17 @@ map:
 	bh_result->b_size = len;
 	bh_result->b_bdev = em->bdev;
 	set_buffer_mapped(bh_result);
-	if (create && !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
-		set_buffer_new(bh_result);
+	if (create) {
+		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
+			set_buffer_new(bh_result);
+
+		/*
+		 * Need to update the i_size under the extent lock so buffered
+		 * readers will get the updated i_size when we unlock.
+		 */
+		if (start + len > i_size_read(inode))
+			i_size_write(inode, start + len);
+	}
 
 	free_extent_map(em);
 
@@ -6360,12 +6369,48 @@ static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
 		 */
 		ordered = btrfs_lookup_ordered_range(inode, lockstart,
 						     lockend - lockstart + 1);
-		if (!ordered)
+
+		/*
+		 * We need to make sure there are no buffered pages in this
+		 * range either, we could have raced between the invalidate in
+		 * generic_file_direct_write and locking the extent.  The
+		 * invalidate needs to happen so that reads after a write do not
+		 * get stale data.
+		 */
+		if (!ordered && (!writing ||
+		    !test_range_bit(&BTRFS_I(inode)->io_tree,
+				    lockstart, lockend, EXTENT_UPTODATE, 0,
+				    cached_state)))
 			break;
+
 		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
 				     &cached_state, GFP_NOFS);
-		btrfs_start_ordered_extent(inode, ordered, 1);
-		btrfs_put_ordered_extent(ordered);
+
+		if (ordered) {
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+		} else {
+			/* Screw you mmap */
+			ret = filemap_write_and_wait_range(file->f_mapping,
+							   lockstart,
+							   lockend);
+			if (ret)
+				goto out;
+
+			/*
+			 * If we found a page that couldn't be invalidated just
+			 * fall back to buffered.
+			 */
+			ret = invalidate_inode_pages2_range(file->f_mapping,
+					lockstart >> PAGE_CACHE_SHIFT,
+					lockend >> PAGE_CACHE_SHIFT);
+			if (ret) {
+				if (ret == -EBUSY)
+					ret = 0;
+				goto out;
+			}
+		}
+
 		cond_resched();
 	}