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

Pull btrfs fixes from Chris Mason:
 "Filipe is doing a careful pass through fsync problems, and these are
  the fixes so far.  I'll have one more for rc6 that we're still
  testing.

  My big commit is fixing up some inode hash races that Al Viro found
  (thanks Al)"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: use insert_inode_locked4 for inode creation
  Btrfs: fix fsync data loss after a ranged fsync
  Btrfs: kfree()ing ERR_PTRs
  Btrfs: fix crash while doing a ranged fsync
  Btrfs: fix corruption after write/fsync failure + fsync + log recovery
  Btrfs: fix autodefrag with compression

1 files changed, 61 insertions, 16 deletions

diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 7e0e6e3029dd..d296efe2d3e7 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -94,8 +94,10 @@
 #define LOG_WALK_REPLAY_ALL 3
 
 static int btrfs_log_inode(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, struct inode *inode,
-			     int inode_only);
+			   struct btrfs_root *root, struct inode *inode,
+			   int inode_only,
+			   const loff_t start,
+			   const loff_t end);
 static int link_to_fixup_dir(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid);
@@ -3858,8 +3860,10 @@ process:
  * This handles both files and directories.
  */
 static int btrfs_log_inode(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, struct inode *inode,
-			     int inode_only)
+			   struct btrfs_root *root, struct inode *inode,
+			   int inode_only,
+			   const loff_t start,
+			   const loff_t end)
 {
 	struct btrfs_path *path;
 	struct btrfs_path *dst_path;
@@ -3876,6 +3880,7 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 	int ins_nr;
 	bool fast_search = false;
 	u64 ino = btrfs_ino(inode);
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -4049,13 +4054,35 @@ log_extents:
 			goto out_unlock;
 		}
 	} else if (inode_only == LOG_INODE_ALL) {
-		struct extent_map_tree *tree = &BTRFS_I(inode)->extent_tree;
 		struct extent_map *em, *n;
 
-		write_lock(&tree->lock);
-		list_for_each_entry_safe(em, n, &tree->modified_extents, list)
-			list_del_init(&em->list);
-		write_unlock(&tree->lock);
+		write_lock(&em_tree->lock);
+		/*
+		 * We can't just remove every em if we're called for a ranged
+		 * fsync - that is, one that doesn't cover the whole possible
+		 * file range (0 to LLONG_MAX). This is because we can have
+		 * em's that fall outside the range we're logging and therefore
+		 * their ordered operations haven't completed yet
+		 * (btrfs_finish_ordered_io() not invoked yet). This means we
+		 * didn't get their respective file extent item in the fs/subvol
+		 * tree yet, and need to let the next fast fsync (one which
+		 * consults the list of modified extent maps) find the em so
+		 * that it logs a matching file extent item and waits for the
+		 * respective ordered operation to complete (if it's still
+		 * running).
+		 *
+		 * Removing every em outside the range we're logging would make
+		 * the next fast fsync not log their matching file extent items,
+		 * therefore making us lose data after a log replay.
+		 */
+		list_for_each_entry_safe(em, n, &em_tree->modified_extents,
+					 list) {
+			const u64 mod_end = em->mod_start + em->mod_len - 1;
+
+			if (em->mod_start >= start && mod_end <= end)
+				list_del_init(&em->list);
+		}
+		write_unlock(&em_tree->lock);
 	}
 
 	if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) {
@@ -4065,8 +4092,19 @@ log_extents:
 			goto out_unlock;
 		}
 	}
-	BTRFS_I(inode)->logged_trans = trans->transid;
-	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->last_sub_trans;
+
+	write_lock(&em_tree->lock);
+	/*
+	 * If we're doing a ranged fsync and there are still modified extents
+	 * in the list, we must run on the next fsync call as it might cover
+	 * those extents (a full fsync or an fsync for other range).
+	 */
+	if (list_empty(&em_tree->modified_extents)) {
+		BTRFS_I(inode)->logged_trans = trans->transid;
+		BTRFS_I(inode)->last_log_commit =
+			BTRFS_I(inode)->last_sub_trans;
+	}
+	write_unlock(&em_tree->lock);
 out_unlock:
 	if (unlikely(err))
 		btrfs_put_logged_extents(&logged_list);
@@ -4161,7 +4199,10 @@ out:
  */
 static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 			    	  struct btrfs_root *root, struct inode *inode,
-			    	  struct dentry *parent, int exists_only,
+				  struct dentry *parent,
+				  const loff_t start,
+				  const loff_t end,
+				  int exists_only,
 				  struct btrfs_log_ctx *ctx)
 {
 	int inode_only = exists_only ? LOG_INODE_EXISTS : LOG_INODE_ALL;
@@ -4207,7 +4248,7 @@ static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto end_no_trans;
 
-	ret = btrfs_log_inode(trans, root, inode, inode_only);
+	ret = btrfs_log_inode(trans, root, inode, inode_only, start, end);
 	if (ret)
 		goto end_trans;
 
@@ -4235,7 +4276,8 @@ static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 
 		if (BTRFS_I(inode)->generation >
 		    root->fs_info->last_trans_committed) {
-			ret = btrfs_log_inode(trans, root, inode, inode_only);
+			ret = btrfs_log_inode(trans, root, inode, inode_only,
+					      0, LLONG_MAX);
 			if (ret)
 				goto end_trans;
 		}
@@ -4269,13 +4311,15 @@ end_no_trans:
  */
 int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, struct dentry *dentry,
+			  const loff_t start,
+			  const loff_t end,
 			  struct btrfs_log_ctx *ctx)
 {
 	struct dentry *parent = dget_parent(dentry);
 	int ret;
 
 	ret = btrfs_log_inode_parent(trans, root, dentry->d_inode, parent,
-				     0, ctx);
+				     start, end, 0, ctx);
 	dput(parent);
 
 	return ret;
@@ -4512,6 +4556,7 @@ int btrfs_log_new_name(struct btrfs_trans_handle *trans,
 		    root->fs_info->last_trans_committed))
 		return 0;
 
-	return btrfs_log_inode_parent(trans, root, inode, parent, 1, NULL);
+	return btrfs_log_inode_parent(trans, root, inode, parent, 0,
+				      LLONG_MAX, 1, NULL);
 }