1 files changed, 91 insertions, 41 deletions

diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 5146387b416b..f0cf8ce26f01 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -138,6 +138,25 @@ static void wait_log_commit(struct btrfs_root *root, int transid);
  * and once to do all the other items.
  */
 
+static struct inode *btrfs_iget_logging(u64 objectid, struct btrfs_root *root)
+{
+	unsigned int nofs_flag;
+	struct inode *inode;
+
+	/*
+	 * We're holding a transaction handle whether we are logging or
+	 * replaying a log tree, so we must make sure NOFS semantics apply
+	 * because btrfs_alloc_inode() may be triggered and it uses GFP_KERNEL
+	 * to allocate an inode, which can recurse back into the filesystem and
+	 * attempt a transaction commit, resulting in a deadlock.
+	 */
+	nofs_flag = memalloc_nofs_save();
+	inode = btrfs_iget(objectid, root);
+	memalloc_nofs_restore(nofs_flag);
+
+	return inode;
+}
+
 /*
  * start a sub transaction and setup the log tree
  * this increments the log tree writer count to make the people
@@ -600,7 +619,7 @@ static noinline struct inode *read_one_inode(struct btrfs_root *root,
 {
 	struct inode *inode;
 
-	inode = btrfs_iget(root->fs_info->sb, objectid, root);
+	inode = btrfs_iget_logging(objectid, root);
 	if (IS_ERR(inode))
 		inode = NULL;
 	return inode;
@@ -1625,7 +1644,8 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
 		if (ret)
 			goto out;
 	}
-	BTRFS_I(inode)->index_cnt = (u64)-1;
+	if (S_ISDIR(inode->i_mode))
+		BTRFS_I(inode)->index_cnt = (u64)-1;
 
 	if (inode->i_nlink == 0) {
 		if (S_ISDIR(inode->i_mode)) {
@@ -2820,7 +2840,7 @@ static void wait_for_writer(struct btrfs_root *root)
 	finish_wait(&root->log_writer_wait, &wait);
 }
 
-void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx, struct inode *inode)
+void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx, struct btrfs_inode *inode)
 {
 	ctx->log_ret = 0;
 	ctx->log_transid = 0;
@@ -2839,7 +2859,7 @@ void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx, struct inode *inode)
 
 void btrfs_init_log_ctx_scratch_eb(struct btrfs_log_ctx *ctx)
 {
-	struct btrfs_inode *inode = BTRFS_I(ctx->inode);
+	struct btrfs_inode *inode = ctx->inode;
 
 	if (!test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags) &&
 	    !test_bit(BTRFS_INODE_COPY_EVERYTHING, &inode->runtime_flags))
@@ -2857,7 +2877,7 @@ void btrfs_release_log_ctx_extents(struct btrfs_log_ctx *ctx)
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_ordered_extent *tmp;
 
-	ASSERT(inode_is_locked(ctx->inode));
+	ASSERT(inode_is_locked(&ctx->inode->vfs_inode));
 
 	list_for_each_entry_safe(ordered, tmp, &ctx->ordered_extents, log_list) {
 		list_del_init(&ordered->log_list);
@@ -4234,8 +4254,10 @@ static int log_inode_item(struct btrfs_trans_handle *trans,
 			  struct btrfs_inode *inode, bool inode_item_dropped)
 {
 	struct btrfs_inode_item *inode_item;
+	struct btrfs_key key;
 	int ret;
 
+	btrfs_get_inode_key(inode, &key);
 	/*
 	 * If we are doing a fast fsync and the inode was logged before in the
 	 * current transaction, then we know the inode was previously logged and
@@ -4247,7 +4269,7 @@ static int log_inode_item(struct btrfs_trans_handle *trans,
 	 * already exists can also result in unnecessarily splitting a leaf.
 	 */
 	if (!inode_item_dropped && inode->logged_trans == trans->transid) {
-		ret = btrfs_search_slot(trans, log, &inode->location, path, 0, 1);
+		ret = btrfs_search_slot(trans, log, &key, path, 0, 1);
 		ASSERT(ret <= 0);
 		if (ret > 0)
 			ret = -ENOENT;
@@ -4261,7 +4283,7 @@ static int log_inode_item(struct btrfs_trans_handle *trans,
 		 * the inode, we set BTRFS_INODE_NEEDS_FULL_SYNC on its runtime
 		 * flags and set ->logged_trans to 0.
 		 */
-		ret = btrfs_insert_empty_item(trans, log, path, &inode->location,
+		ret = btrfs_insert_empty_item(trans, log, path, &key,
 					      sizeof(*inode_item));
 		ASSERT(ret != -EEXIST);
 	}
@@ -4575,6 +4597,7 @@ static int log_extent_csums(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_root *csum_root;
+	u64 block_start;
 	u64 csum_offset;
 	u64 csum_len;
 	u64 mod_start = em->start;
@@ -4584,7 +4607,7 @@ static int log_extent_csums(struct btrfs_trans_handle *trans,
 
 	if (inode->flags & BTRFS_INODE_NODATASUM ||
 	    (em->flags & EXTENT_FLAG_PREALLOC) ||
-	    em->block_start == EXTENT_MAP_HOLE)
+	    em->disk_bytenr == EXTENT_MAP_HOLE)
 		return 0;
 
 	list_for_each_entry(ordered, &ctx->ordered_extents, log_list) {
@@ -4648,17 +4671,18 @@ static int log_extent_csums(struct btrfs_trans_handle *trans,
 	/* If we're compressed we have to save the entire range of csums. */
 	if (extent_map_is_compressed(em)) {
 		csum_offset = 0;
-		csum_len = max(em->block_len, em->orig_block_len);
+		csum_len = em->disk_num_bytes;
 	} else {
 		csum_offset = mod_start - em->start;
 		csum_len = mod_len;
 	}
 
 	/* block start is already adjusted for the file extent offset. */
-	csum_root = btrfs_csum_root(trans->fs_info, em->block_start);
-	ret = btrfs_lookup_csums_list(csum_root, em->block_start + csum_offset,
-				      em->block_start + csum_offset +
-				      csum_len - 1, &ordered_sums, false);
+	block_start = extent_map_block_start(em);
+	csum_root = btrfs_csum_root(trans->fs_info, block_start);
+	ret = btrfs_lookup_csums_list(csum_root, block_start + csum_offset,
+				      block_start + csum_offset + csum_len - 1,
+				      &ordered_sums, false);
 	if (ret < 0)
 		return ret;
 	ret = 0;
@@ -4688,7 +4712,8 @@ static int log_one_extent(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 	struct btrfs_key key;
 	enum btrfs_compression_type compress_type;
-	u64 extent_offset = em->start - em->orig_start;
+	u64 extent_offset = em->offset;
+	u64 block_start = extent_map_block_start(em);
 	u64 block_len;
 	int ret;
 
@@ -4698,14 +4723,13 @@ static int log_one_extent(struct btrfs_trans_handle *trans,
 	else
 		btrfs_set_stack_file_extent_type(&fi, BTRFS_FILE_EXTENT_REG);
 
-	block_len = max(em->block_len, em->orig_block_len);
+	block_len = em->disk_num_bytes;
 	compress_type = extent_map_compression(em);
 	if (compress_type != BTRFS_COMPRESS_NONE) {
-		btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start);
+		btrfs_set_stack_file_extent_disk_bytenr(&fi, block_start);
 		btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
-	} else if (em->block_start < EXTENT_MAP_LAST_BYTE) {
-		btrfs_set_stack_file_extent_disk_bytenr(&fi, em->block_start -
-							extent_offset);
+	} else if (em->disk_bytenr < EXTENT_MAP_LAST_BYTE) {
+		btrfs_set_stack_file_extent_disk_bytenr(&fi, block_start - extent_offset);
 		btrfs_set_stack_file_extent_disk_num_bytes(&fi, block_len);
 	}
 
@@ -4860,18 +4884,23 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans,
 			path->slots[0]++;
 			continue;
 		}
-		if (!dropped_extents) {
-			/*
-			 * Avoid logging extent items logged in past fsync calls
-			 * and leading to duplicate keys in the log tree.
-			 */
+		/*
+		 * Avoid overlapping items in the log tree. The first time we
+		 * get here, get rid of everything from a past fsync. After
+		 * that, if the current extent starts before the end of the last
+		 * extent we copied, truncate the last one. This can happen if
+		 * an ordered extent completion modifies the subvolume tree
+		 * while btrfs_next_leaf() has the tree unlocked.
+		 */
+		if (!dropped_extents || key.offset < truncate_offset) {
 			ret = truncate_inode_items(trans, root->log_root, inode,
-						   truncate_offset,
+						   min(key.offset, truncate_offset),
 						   BTRFS_EXTENT_DATA_KEY);
 			if (ret)
 				goto out;
 			dropped_extents = true;
 		}
+		truncate_offset = btrfs_file_extent_end(path);
 		if (ins_nr == 0)
 			start_slot = slot;
 		ins_nr++;
@@ -5433,7 +5462,6 @@ static int log_new_dir_dentries(struct btrfs_trans_handle *trans,
 				struct btrfs_log_ctx *ctx)
 {
 	struct btrfs_root *root = start_inode->root;
-	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_path *path;
 	LIST_HEAD(dir_list);
 	struct btrfs_dir_list *dir_elem;
@@ -5494,7 +5522,7 @@ again:
 				continue;
 
 			btrfs_release_path(path);
-			di_inode = btrfs_iget(fs_info->sb, di_key.objectid, root);
+			di_inode = btrfs_iget_logging(di_key.objectid, root);
 			if (IS_ERR(di_inode)) {
 				ret = PTR_ERR(di_inode);
 				goto out;
@@ -5554,7 +5582,7 @@ again:
 		btrfs_add_delayed_iput(curr_inode);
 		curr_inode = NULL;
 
-		vfs_inode = btrfs_iget(fs_info->sb, ino, root);
+		vfs_inode = btrfs_iget_logging(ino, root);
 		if (IS_ERR(vfs_inode)) {
 			ret = PTR_ERR(vfs_inode);
 			break;
@@ -5649,7 +5677,7 @@ static int add_conflicting_inode(struct btrfs_trans_handle *trans,
 	if (ctx->num_conflict_inodes >= MAX_CONFLICT_INODES)
 		return BTRFS_LOG_FORCE_COMMIT;
 
-	inode = btrfs_iget(root->fs_info->sb, ino, root);
+	inode = btrfs_iget_logging(ino, root);
 	/*
 	 * If the other inode that had a conflicting dir entry was deleted in
 	 * the current transaction then we either:
@@ -5750,7 +5778,6 @@ static int log_conflicting_inodes(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root,
 				  struct btrfs_log_ctx *ctx)
 {
-	struct btrfs_fs_info *fs_info = root->fs_info;
 	int ret = 0;
 
 	/*
@@ -5781,7 +5808,7 @@ static int log_conflicting_inodes(struct btrfs_trans_handle *trans,
 		list_del(&curr->list);
 		kfree(curr);
 
-		inode = btrfs_iget(fs_info->sb, ino, root);
+		inode = btrfs_iget_logging(ino, root);
 		/*
 		 * If the other inode that had a conflicting dir entry was
 		 * deleted in the current transaction, we need to log its parent
@@ -5792,7 +5819,7 @@ static int log_conflicting_inodes(struct btrfs_trans_handle *trans,
 			if (ret != -ENOENT)
 				break;
 
-			inode = btrfs_iget(fs_info->sb, parent, root);
+			inode = btrfs_iget_logging(parent, root);
 			if (IS_ERR(inode)) {
 				ret = PTR_ERR(inode);
 				break;
@@ -5905,7 +5932,7 @@ again:
 			if (ret < 0) {
 				return ret;
 			} else if (ret > 0 &&
-				   other_ino != btrfs_ino(BTRFS_I(ctx->inode))) {
+				   other_ino != btrfs_ino(ctx->inode)) {
 				if (ins_nr > 0) {
 					ins_nr++;
 				} else {
@@ -6314,7 +6341,6 @@ static int log_new_delayed_dentries(struct btrfs_trans_handle *trans,
 				    struct btrfs_log_ctx *ctx)
 {
 	const bool orig_log_new_dentries = ctx->log_new_dentries;
-	struct btrfs_fs_info *fs_info = trans->fs_info;
 	struct btrfs_delayed_item *item;
 	int ret = 0;
 
@@ -6340,7 +6366,7 @@ static int log_new_delayed_dentries(struct btrfs_trans_handle *trans,
 		if (key.type == BTRFS_ROOT_ITEM_KEY)
 			continue;
 
-		di_inode = btrfs_iget(fs_info->sb, key.objectid, inode->root);
+		di_inode = btrfs_iget_logging(key.objectid, inode->root);
 		if (IS_ERR(di_inode)) {
 			ret = PTR_ERR(di_inode);
 			break;
@@ -6724,7 +6750,6 @@ static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
 				 struct btrfs_inode *inode,
 				 struct btrfs_log_ctx *ctx)
 {
-	struct btrfs_fs_info *fs_info = trans->fs_info;
 	int ret;
 	struct btrfs_path *path;
 	struct btrfs_key key;
@@ -6789,8 +6814,7 @@ static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
 				cur_offset = item_size;
 			}
 
-			dir_inode = btrfs_iget(fs_info->sb, inode_key.objectid,
-					       root);
+			dir_inode = btrfs_iget_logging(inode_key.objectid, root);
 			/*
 			 * If the parent inode was deleted, return an error to
 			 * fallback to a transaction commit. This is to prevent
@@ -6852,7 +6876,6 @@ static int log_new_ancestors(struct btrfs_trans_handle *trans,
 	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
 
 	while (true) {
-		struct btrfs_fs_info *fs_info = root->fs_info;
 		struct extent_buffer *leaf;
 		int slot;
 		struct btrfs_key search_key;
@@ -6867,7 +6890,7 @@ static int log_new_ancestors(struct btrfs_trans_handle *trans,
 		search_key.objectid = found_key.offset;
 		search_key.type = BTRFS_INODE_ITEM_KEY;
 		search_key.offset = 0;
-		inode = btrfs_iget(fs_info->sb, ino, root);
+		inode = btrfs_iget_logging(ino, root);
 		if (IS_ERR(inode))
 			return PTR_ERR(inode);
 
@@ -7056,6 +7079,15 @@ static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 	}
 
 	/*
+	 * If we're logging an inode from a subvolume created in the current
+	 * transaction we must force a commit since the root is not persisted.
+	 */
+	if (btrfs_root_generation(&root->root_item) == trans->transid) {
+		ret = BTRFS_LOG_FORCE_COMMIT;
+		goto end_no_trans;
+	}
+
+	/*
 	 * Skip already logged inodes or inodes corresponding to tmpfiles
 	 * (since logging them is pointless, a link count of 0 means they
 	 * will never be accessible).
@@ -7436,6 +7468,24 @@ void btrfs_record_snapshot_destroy(struct btrfs_trans_handle *trans,
 }
 
 /*
+ * Call this when creating a subvolume in a directory.
+ * Because we don't commit a transaction when creating a subvolume, we can't
+ * allow the directory pointing to the subvolume to be logged with an entry that
+ * points to an unpersisted root if we are still in the transaction used to
+ * create the subvolume, so make any attempt to log the directory to result in a
+ * full log sync.
+ * Also we don't need to worry with renames, since btrfs_rename() marks the log
+ * for full commit when renaming a subvolume.
+ */
+void btrfs_record_new_subvolume(const struct btrfs_trans_handle *trans,
+				struct btrfs_inode *dir)
+{
+	mutex_lock(&dir->log_mutex);
+	dir->last_unlink_trans = trans->transid;
+	mutex_unlock(&dir->log_mutex);
+}
+
+/*
  * Update the log after adding a new name for an inode.
  *
  * @trans:              Transaction handle.
@@ -7567,7 +7617,7 @@ void btrfs_log_new_name(struct btrfs_trans_handle *trans,
 			goto out;
 	}
 
-	btrfs_init_log_ctx(&ctx, &inode->vfs_inode);
+	btrfs_init_log_ctx(&ctx, inode);
 	ctx.logging_new_name = true;
 	btrfs_init_log_ctx_scratch_eb(&ctx);
 	/*