Merge tag 'for-6.6-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fix from David Sterba:
 "One more fix for a problem with snapshot of a newly created subvolume
  that can lead to inconsistent data under some circumstances. Kernel
  6.5 added a performance optimization to skip transaction commit for
  subvolume creation but this could end up with newer data on disk but
  not linked to other structures.

  The fix itself is an added condition, the rest of the patch is a
  parameter added to several functions"

* tag 'for-6.6-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix unwritten extent buffer after snapshotting a new subvolume

5 files changed, 33 insertions, 15 deletions

diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
index b7d54efb4728..a4a809efc92f 100644
--- a/fs/btrfs/backref.c
+++ b/fs/btrfs/backref.c
@@ -3196,12 +3196,14 @@ static int handle_direct_tree_backref(struct btrfs_backref_cache *cache,
  * We still need to do a tree search to find out the parents. This is for
  * TREE_BLOCK_REF backref (keyed or inlined).
  *
+ * @trans:	Transaction handle.
  * @ref_key:	The same as @ref_key in  handle_direct_tree_backref()
  * @tree_key:	The first key of this tree block.
  * @path:	A clean (released) path, to avoid allocating path every time
  *		the function get called.
  */
-static int handle_indirect_tree_backref(struct btrfs_backref_cache *cache,
+static int handle_indirect_tree_backref(struct btrfs_trans_handle *trans,
+					struct btrfs_backref_cache *cache,
 					struct btrfs_path *path,
 					struct btrfs_key *ref_key,
 					struct btrfs_key *tree_key,
@@ -3315,7 +3317,7 @@ static int handle_indirect_tree_backref(struct btrfs_backref_cache *cache,
 			 * If we know the block isn't shared we can avoid
 			 * checking its backrefs.
 			 */
-			if (btrfs_block_can_be_shared(root, eb))
+			if (btrfs_block_can_be_shared(trans, root, eb))
 				upper->checked = 0;
 			else
 				upper->checked = 1;
@@ -3363,11 +3365,13 @@ out:
  *	 links aren't yet bi-directional. Needs to finish such links.
  *	 Use btrfs_backref_finish_upper_links() to finish such linkage.
  *
+ * @trans:	Transaction handle.
  * @path:	Released path for indirect tree backref lookup
  * @iter:	Released backref iter for extent tree search
  * @node_key:	The first key of the tree block
  */
-int btrfs_backref_add_tree_node(struct btrfs_backref_cache *cache,
+int btrfs_backref_add_tree_node(struct btrfs_trans_handle *trans,
+				struct btrfs_backref_cache *cache,
 				struct btrfs_path *path,
 				struct btrfs_backref_iter *iter,
 				struct btrfs_key *node_key,
@@ -3467,8 +3471,8 @@ int btrfs_backref_add_tree_node(struct btrfs_backref_cache *cache,
 			 * offset means the root objectid. We need to search
 			 * the tree to get its parent bytenr.
 			 */
-			ret = handle_indirect_tree_backref(cache, path, &key, node_key,
-							   cur);
+			ret = handle_indirect_tree_backref(trans, cache, path,
+							   &key, node_key, cur);
 			if (ret < 0)
 				goto out;
 		}
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
index 1616e3e3f1e4..71d535e03dca 100644
--- a/fs/btrfs/backref.h
+++ b/fs/btrfs/backref.h
@@ -540,7 +540,8 @@ static inline void btrfs_backref_panic(struct btrfs_fs_info *fs_info,
 		    bytenr);
 }
 
-int btrfs_backref_add_tree_node(struct btrfs_backref_cache *cache,
+int btrfs_backref_add_tree_node(struct btrfs_trans_handle *trans,
+				struct btrfs_backref_cache *cache,
 				struct btrfs_path *path,
 				struct btrfs_backref_iter *iter,
 				struct btrfs_key *node_key,
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index da519c1b6ad0..617d4827eec2 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -367,7 +367,8 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 /*
  * check if the tree block can be shared by multiple trees
  */
-int btrfs_block_can_be_shared(struct btrfs_root *root,
+int btrfs_block_can_be_shared(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
 			      struct extent_buffer *buf)
 {
 	/*
@@ -376,11 +377,21 @@ int btrfs_block_can_be_shared(struct btrfs_root *root,
 	 * not allocated by tree relocation, we know the block is not shared.
 	 */
 	if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state) &&
-	    buf != root->node && buf != root->commit_root &&
+	    buf != root->node &&
 	    (btrfs_header_generation(buf) <=
 	     btrfs_root_last_snapshot(&root->root_item) ||
-	     btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
-		return 1;
+	     btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC))) {
+		if (buf != root->commit_root)
+			return 1;
+		/*
+		 * An extent buffer that used to be the commit root may still be
+		 * shared because the tree height may have increased and it
+		 * became a child of a higher level root. This can happen when
+		 * snapshotting a subvolume created in the current transaction.
+		 */
+		if (btrfs_header_generation(buf) == trans->transid)
+			return 1;
+	}
 
 	return 0;
 }
@@ -415,7 +426,7 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 	 * are only allowed for blocks use full backrefs.
 	 */
 
-	if (btrfs_block_can_be_shared(root, buf)) {
+	if (btrfs_block_can_be_shared(trans, root, buf)) {
 		ret = btrfs_lookup_extent_info(trans, fs_info, buf->start,
 					       btrfs_header_level(buf), 1,
 					       &refs, &flags);
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 9419f4e37a58..ff40acd63a37 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -540,7 +540,8 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
 		      struct extent_buffer *buf,
 		      struct extent_buffer **cow_ret, u64 new_root_objectid);
-int btrfs_block_can_be_shared(struct btrfs_root *root,
+int btrfs_block_can_be_shared(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
 			      struct extent_buffer *buf);
 int btrfs_del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		  struct btrfs_path *path, int level, int slot);
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index 9951a0caf5bb..c6d4bb8cbe29 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -466,6 +466,7 @@ static bool handle_useless_nodes(struct reloc_control *rc,
  * cached.
  */
 static noinline_for_stack struct btrfs_backref_node *build_backref_tree(
+			struct btrfs_trans_handle *trans,
 			struct reloc_control *rc, struct btrfs_key *node_key,
 			int level, u64 bytenr)
 {
@@ -499,8 +500,8 @@ static noinline_for_stack struct btrfs_backref_node *build_backref_tree(
 
 	/* Breadth-first search to build backref cache */
 	do {
-		ret = btrfs_backref_add_tree_node(cache, path, iter, node_key,
-						  cur);
+		ret = btrfs_backref_add_tree_node(trans, cache, path, iter,
+						  node_key, cur);
 		if (ret < 0) {
 			err = ret;
 			goto out;
@@ -2803,7 +2804,7 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans,
 
 	/* Do tree relocation */
 	rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
-		node = build_backref_tree(rc, &block->key,
+		node = build_backref_tree(trans, rc, &block->key,
 					  block->level, block->bytenr);
 		if (IS_ERR(node)) {
 			err = PTR_ERR(node);