1 files changed, 112 insertions, 28 deletions

diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 179a383a4aaa..98c397eb054c 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2729,7 +2729,8 @@ out:
  * in the tree of log roots
  */
 static int update_log_root(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *log)
+			   struct btrfs_root *log,
+			   struct btrfs_root_item *root_item)
 {
 	struct btrfs_fs_info *fs_info = log->fs_info;
 	int ret;
@@ -2737,10 +2738,10 @@ static int update_log_root(struct btrfs_trans_handle *trans,
 	if (log->log_transid == 1) {
 		/* insert root item on the first sync */
 		ret = btrfs_insert_root(trans, fs_info->log_root_tree,
-				&log->root_key, &log->root_item);
+				&log->root_key, root_item);
 	} else {
 		ret = btrfs_update_root(trans, fs_info->log_root_tree,
-				&log->root_key, &log->root_item);
+				&log->root_key, root_item);
 	}
 	return ret;
 }
@@ -2836,6 +2837,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_root *log = root->log_root;
 	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
+	struct btrfs_root_item new_root_item;
 	int log_transid = 0;
 	struct btrfs_log_ctx root_log_ctx;
 	struct blk_plug plug;
@@ -2901,7 +2903,21 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 
+	/*
+	 * We _must_ update under the root->log_mutex in order to make sure we
+	 * have a consistent view of the log root we are trying to commit at
+	 * this moment.
+	 *
+	 * We _must_ copy this into a local copy, because we are not holding the
+	 * log_root_tree->log_mutex yet.  This is important because when we
+	 * commit the log_root_tree we must have a consistent view of the
+	 * log_root_tree when we update the super block to point at the
+	 * log_root_tree bytenr.  If we update the log_root_tree here we'll race
+	 * with the commit and possibly point at the new block which we may not
+	 * have written out.
+	 */
 	btrfs_set_root_node(&log->root_item, log->node);
+	memcpy(&new_root_item, &log->root_item, sizeof(new_root_item));
 
 	root->log_transid++;
 	log->log_transid = root->log_transid;
@@ -2925,9 +2941,15 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 
 	mutex_unlock(&log_root_tree->log_mutex);
 
-	ret = update_log_root(trans, log);
-
 	mutex_lock(&log_root_tree->log_mutex);
+
+	/*
+	 * Now we are safe to update the log_root_tree because we're under the
+	 * log_mutex, and we're a current writer so we're holding the commit
+	 * open until we drop the log_mutex.
+	 */
+	ret = update_log_root(trans, log, &new_root_item);
+
 	if (atomic_dec_and_test(&log_root_tree->log_writers)) {
 		/*
 		 * Implicit memory barrier after atomic_dec_and_test
@@ -3149,6 +3171,30 @@ int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
 }
 
 /*
+ * Check if an inode was logged in the current transaction. We can't always rely
+ * on an inode's logged_trans value, because it's an in-memory only field and
+ * therefore not persisted. This means that its value is lost if the inode gets
+ * evicted and loaded again from disk (in which case it has a value of 0, and
+ * certainly it is smaller then any possible transaction ID), when that happens
+ * the full_sync flag is set in the inode's runtime flags, so on that case we
+ * assume eviction happened and ignore the logged_trans value, assuming the
+ * worst case, that the inode was logged before in the current transaction.
+ */
+static bool inode_logged(struct btrfs_trans_handle *trans,
+			 struct btrfs_inode *inode)
+{
+	if (inode->logged_trans == trans->transid)
+		return true;
+
+	if (inode->last_trans == trans->transid &&
+	    test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags) &&
+	    !test_bit(BTRFS_FS_LOG_RECOVERING, &trans->fs_info->flags))
+		return true;
+
+	return false;
+}
+
+/*
  * If both a file and directory are logged, and unlinks or renames are
  * mixed in, we have a few interesting corners:
  *
@@ -3182,7 +3228,7 @@ int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 	int bytes_del = 0;
 	u64 dir_ino = btrfs_ino(dir);
 
-	if (dir->logged_trans < trans->transid)
+	if (!inode_logged(trans, dir))
 		return 0;
 
 	ret = join_running_log_trans(root);
@@ -3287,7 +3333,7 @@ int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
 	u64 index;
 	int ret;
 
-	if (inode->logged_trans < trans->transid)
+	if (!inode_logged(trans, inode))
 		return 0;
 
 	ret = join_running_log_trans(root);
@@ -3422,9 +3468,16 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 	}
 	btrfs_release_path(path);
 
-	/* find the first key from this transaction again */
+	/*
+	 * Find the first key from this transaction again.  See the note for
+	 * log_new_dir_dentries, if we're logging a directory recursively we
+	 * won't be holding its i_mutex, which means we can modify the directory
+	 * while we're logging it.  If we remove an entry between our first
+	 * search and this search we'll not find the key again and can just
+	 * bail.
+	 */
 	ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
-	if (WARN_ON(ret != 0))
+	if (ret != 0)
 		goto done;
 
 	/*
@@ -4005,6 +4058,7 @@ fill_holes:
 							       *last_extent, 0,
 							       0, len, 0, len,
 							       0, 0, 0);
+				*last_extent += len;
 			}
 		}
 	}
@@ -4501,6 +4555,19 @@ static int logged_inode_size(struct btrfs_root *log, struct btrfs_inode *inode,
 		item = btrfs_item_ptr(path->nodes[0], path->slots[0],
 				      struct btrfs_inode_item);
 		*size_ret = btrfs_inode_size(path->nodes[0], item);
+		/*
+		 * If the in-memory inode's i_size is smaller then the inode
+		 * size stored in the btree, return the inode's i_size, so
+		 * that we get a correct inode size after replaying the log
+		 * when before a power failure we had a shrinking truncate
+		 * followed by addition of a new name (rename / new hard link).
+		 * Otherwise return the inode size from the btree, to avoid
+		 * data loss when replaying a log due to previously doing a
+		 * write that expands the inode's size and logging a new name
+		 * immediately after.
+		 */
+		if (*size_ret > inode->vfs_inode.i_size)
+			*size_ret = inode->vfs_inode.i_size;
 	}
 
 	btrfs_release_path(path);
@@ -5073,7 +5140,7 @@ again:
 						BTRFS_I(other_inode),
 						LOG_OTHER_INODE, 0, LLONG_MAX,
 						ctx);
-				iput(other_inode);
+				btrfs_add_delayed_iput(other_inode);
 				if (err)
 					goto out_unlock;
 				else
@@ -5241,9 +5308,19 @@ log_extents:
 		}
 	}
 
+	/*
+	 * Don't update last_log_commit if we logged that an inode exists after
+	 * it was loaded to memory (full_sync bit set).
+	 * This is to prevent data loss when we do a write to the inode, then
+	 * the inode gets evicted after all delalloc was flushed, then we log
+	 * it exists (due to a rename for example) and then fsync it. This last
+	 * fsync would do nothing (not logging the extents previously written).
+	 */
 	spin_lock(&inode->lock);
 	inode->logged_trans = trans->transid;
-	inode->last_log_commit = inode->last_sub_trans;
+	if (inode_only != LOG_INODE_EXISTS ||
+	    !test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags))
+		inode->last_log_commit = inode->last_sub_trans;
 	spin_unlock(&inode->lock);
 out_unlock:
 	if (unlikely(err))
@@ -5307,7 +5384,6 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 {
 	int ret = 0;
 	struct dentry *old_parent = NULL;
-	struct btrfs_inode *orig_inode = inode;
 
 	/*
 	 * for regular files, if its inode is already on disk, we don't
@@ -5327,16 +5403,6 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 	}
 
 	while (1) {
-		/*
-		 * If we are logging a directory then we start with our inode,
-		 * not our parent's inode, so we need to skip setting the
-		 * logged_trans so that further down in the log code we don't
-		 * think this inode has already been logged.
-		 */
-		if (inode != orig_inode)
-			inode->logged_trans = trans->transid;
-		smp_mb();
-
 		if (btrfs_must_commit_transaction(trans, inode)) {
 			ret = 1;
 			break;
@@ -5491,7 +5557,7 @@ process_leaf:
 			}
 
 			if (btrfs_inode_in_log(BTRFS_I(di_inode), trans->transid)) {
-				iput(di_inode);
+				btrfs_add_delayed_iput(di_inode);
 				break;
 			}
 
@@ -5503,7 +5569,7 @@ process_leaf:
 			if (!ret &&
 			    btrfs_must_commit_transaction(trans, BTRFS_I(di_inode)))
 				ret = 1;
-			iput(di_inode);
+			btrfs_add_delayed_iput(di_inode);
 			if (ret)
 				goto next_dir_inode;
 			if (ctx->log_new_dentries) {
@@ -5650,7 +5716,7 @@ static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
 			if (!ret && ctx && ctx->log_new_dentries)
 				ret = log_new_dir_dentries(trans, root,
 						   BTRFS_I(dir_inode), ctx);
-			iput(dir_inode);
+			btrfs_add_delayed_iput(dir_inode);
 			if (ret)
 				goto out;
 		}
@@ -5952,9 +6018,28 @@ again:
 		wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key);
 		if (IS_ERR(wc.replay_dest)) {
 			ret = PTR_ERR(wc.replay_dest);
+
+			/*
+			 * We didn't find the subvol, likely because it was
+			 * deleted.  This is ok, simply skip this log and go to
+			 * the next one.
+			 *
+			 * We need to exclude the root because we can't have
+			 * other log replays overwriting this log as we'll read
+			 * it back in a few more times.  This will keep our
+			 * block from being modified, and we'll just bail for
+			 * each subsequent pass.
+			 */
+			if (ret == -ENOENT)
+				ret = btrfs_pin_extent_for_log_replay(fs_info,
+							log->node->start,
+							log->node->len);
 			free_extent_buffer(log->node);
 			free_extent_buffer(log->commit_root);
 			kfree(log);
+
+			if (!ret)
+				goto next;
 			btrfs_handle_fs_error(fs_info, ret,
 				"Couldn't read target root for tree log recovery.");
 			goto error;
@@ -5986,7 +6071,6 @@ again:
 						  &root->highest_objectid);
 		}
 
-		key.offset = found_key.offset - 1;
 		wc.replay_dest->log_root = NULL;
 		free_extent_buffer(log->node);
 		free_extent_buffer(log->commit_root);
@@ -5994,9 +6078,10 @@ again:
 
 		if (ret)
 			goto error;
-
+next:
 		if (found_key.offset == 0)
 			break;
+		key.offset = found_key.offset - 1;
 	}
 	btrfs_release_path(path);
 
@@ -6066,7 +6151,6 @@ void btrfs_record_unlink_dir(struct btrfs_trans_handle *trans,
 	 * if this directory was already logged any new
 	 * names for this file/dir will get recorded
 	 */
-	smp_mb();
 	if (dir->logged_trans == trans->transid)
 		return;