Merge tag 'for-6.11-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - extend tree-checker verification of directory item type

 - fix regression in page/folio and extent state tracking in xarray, the
   dirty status can get out of sync and can cause problems e.g. a hang

 - in send, detect last extent and allow to clone it instead of sending
   it as write, reduces amount of data transferred in the stream

 - fix checking extent references when cleaning deleted subvolumes

 - fix one more case in the extent map shrinker, let it run only in the
   kswapd context so it does not cause latency spikes during other
   operations

* tag 'for-6.11-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix invalid mapping of extent xarray state
  btrfs: send: allow cloning non-aligned extent if it ends at i_size
  btrfs: only run the extent map shrinker from kswapd tasks
  btrfs: tree-checker: reject BTRFS_FT_UNKNOWN dir type
  btrfs: check delayed refs when we're checking if a ref exists

Author: torvalds

fs/btrfs/delayed-ref.c

@@ -1134,6 +1134,73 @@ btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs, u64 byt
 	return find_ref_head(delayed_refs, bytenr, false);
 }
 
+static int find_comp(struct btrfs_delayed_ref_node *entry, u64 root, u64 parent)
+{
+	int type = parent ? BTRFS_SHARED_BLOCK_REF_KEY : BTRFS_TREE_BLOCK_REF_KEY;
+
+	if (type < entry->type)
+		return -1;
+	if (type > entry->type)
+		return 1;
+
+	if (type == BTRFS_TREE_BLOCK_REF_KEY) {
+		if (root < entry->ref_root)
+			return -1;
+		if (root > entry->ref_root)
+			return 1;
+	} else {
+		if (parent < entry->parent)
+			return -1;
+		if (parent > entry->parent)
+			return 1;
+	}
+	return 0;
+}
+
+/*
+ * Check to see if a given root/parent reference is attached to the head.  This
+ * only checks for BTRFS_ADD_DELAYED_REF references that match, as that
+ * indicates the reference exists for the given root or parent.  This is for
+ * tree blocks only.
+ *
+ * @head: the head of the bytenr we're searching.
+ * @root: the root objectid of the reference if it is a normal reference.
+ * @parent: the parent if this is a shared backref.
+ */
+bool btrfs_find_delayed_tree_ref(struct btrfs_delayed_ref_head *head,
+				 u64 root, u64 parent)
+{
+	struct rb_node *node;
+	bool found = false;
+
+	lockdep_assert_held(&head->mutex);
+
+	spin_lock(&head->lock);
+	node = head->ref_tree.rb_root.rb_node;
+	while (node) {
+		struct btrfs_delayed_ref_node *entry;
+		int ret;
+
+		entry = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
+		ret = find_comp(entry, root, parent);
+		if (ret < 0) {
+			node = node->rb_left;
+		} else if (ret > 0) {
+			node = node->rb_right;
+		} else {
+			/*
+			 * We only want to count ADD actions, as drops mean the
+			 * ref doesn't exist.
+			 */
+			if (entry->action == BTRFS_ADD_DELAYED_REF)
+				found = true;
+			break;
+		}
+	}
+	spin_unlock(&head->lock);
+	return found;
+}
+
 void __cold btrfs_delayed_ref_exit(void)
 {
 	kmem_cache_destroy(btrfs_delayed_ref_head_cachep);

fs/btrfs/delayed-ref.h

@@ -389,6 +389,8 @@ void btrfs_dec_delayed_refs_rsv_bg_updates(struct btrfs_fs_info *fs_info);
 int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
 				  enum btrfs_reserve_flush_enum flush);
 bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info);
+bool btrfs_find_delayed_tree_ref(struct btrfs_delayed_ref_head *head,
+				 u64 root, u64 parent);
 
 static inline u64 btrfs_delayed_ref_owner(struct btrfs_delayed_ref_node *node)
 {

fs/btrfs/extent-tree.c

@@ -5472,23 +5472,62 @@ static int check_ref_exists(struct btrfs_trans_handle *trans,
 			    struct btrfs_root *root, u64 bytenr, u64 parent,
 			    int level)
 {
+	struct btrfs_delayed_ref_root *delayed_refs;
+	struct btrfs_delayed_ref_head *head;
 	struct btrfs_path *path;
 	struct btrfs_extent_inline_ref *iref;
 	int ret;
+	bool exists = false;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-
+again:
 	ret = lookup_extent_backref(trans, path, &iref, bytenr,
 				    root->fs_info->nodesize, parent,
 				    btrfs_root_id(root), level, 0);
+	if (ret != -ENOENT) {
+		/*
+		 * If we get 0 then we found our reference, return 1, else
+		 * return the error if it's not -ENOENT;
+		 */
+		btrfs_free_path(path);
+		return (ret < 0 ) ? ret : 1;
+	}
+
+	/*
+	 * We could have a delayed ref with this reference, so look it up while
+	 * we're holding the path open to make sure we don't race with the
+	 * delayed ref running.
+	 */
+	delayed_refs = &trans->transaction->delayed_refs;
+	spin_lock(&delayed_refs->lock);
+	head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
+	if (!head)
+		goto out;
+	if (!mutex_trylock(&head->mutex)) {
+		/*
+		 * We're contended, means that the delayed ref is running, get a
+		 * reference and wait for the ref head to be complete and then
+		 * try again.
+		 */
+		refcount_inc(&head->refs);
+		spin_unlock(&delayed_refs->lock);
+
+		btrfs_release_path(path);
+
+		mutex_lock(&head->mutex);
+		mutex_unlock(&head->mutex);
+		btrfs_put_delayed_ref_head(head);
+		goto again;
+	}
+
+	exists = btrfs_find_delayed_tree_ref(head, root->root_key.objectid, parent);
+	mutex_unlock(&head->mutex);
+out:
+	spin_unlock(&delayed_refs->lock);
 	btrfs_free_path(path);
-	if (ret == -ENOENT)
-		return 0;
-	if (ret < 0)
-		return ret;
-	return 1;
+	return exists ? 1 : 0;
 }
 
 /*

fs/btrfs/extent_io.c

@@ -1496,20 +1496,20 @@ static noinline_for_stack int __extent_writepage_io(struct btrfs_inode *inode,
 		free_extent_map(em);
 		em = NULL;
 
+		/*
+		 * Although the PageDirty bit might be cleared before entering
+		 * this function, subpage dirty bit is not cleared.
+		 * So clear subpage dirty bit here so next time we won't submit
+		 * page for range already written to disk.
+		 */
+		btrfs_folio_clear_dirty(fs_info, page_folio(page), cur, iosize);
 		btrfs_set_range_writeback(inode, cur, cur + iosize - 1);
 		if (!PageWriteback(page)) {
 			btrfs_err(inode->root->fs_info,
 				   "page %lu not writeback, cur %llu end %llu",
 			       page->index, cur, end);
 		}
 
-		/*
-		 * Although the PageDirty bit is cleared before entering this
-		 * function, subpage dirty bit is not cleared.
-		 * So clear subpage dirty bit here so next time we won't submit
-		 * page for range already written to disk.
-		 */
-		btrfs_folio_clear_dirty(fs_info, page_folio(page), cur, iosize);
 
 		submit_extent_page(bio_ctrl, disk_bytenr, page, iosize,
 				   cur - page_offset(page));

fs/btrfs/extent_map.c

@@ -1147,8 +1147,7 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_c
 		return 0;
 
 	/*
-	 * We want to be fast because we can be called from any path trying to
-	 * allocate memory, so if the lock is busy we don't want to spend time
+	 * We want to be fast so if the lock is busy we don't want to spend time
 	 * waiting for it - either some task is about to do IO for the inode or
 	 * we may have another task shrinking extent maps, here in this code, so
 	 * skip this inode.
@@ -1191,9 +1190,7 @@ static long btrfs_scan_inode(struct btrfs_inode *inode, struct btrfs_em_shrink_c
 		/*
 		 * Stop if we need to reschedule or there's contention on the
 		 * lock. This is to avoid slowing other tasks trying to take the
-		 * lock and because the shrinker might be called during a memory
-		 * allocation path and we want to avoid taking a very long time
-		 * and slowing down all sorts of tasks.
+		 * lock.
 		 */
 		if (need_resched() || rwlock_needbreak(&tree->lock))
 			break;
@@ -1222,12 +1219,7 @@ static long btrfs_scan_root(struct btrfs_root *root, struct btrfs_em_shrink_ctx
 		if (ctx->scanned >= ctx->nr_to_scan)
 			break;
 
-		/*
-		 * We may be called from memory allocation paths, so we don't
-		 * want to take too much time and slowdown tasks.
-		 */
-		if (need_resched())
-			break;
+		cond_resched();
 
 		inode = btrfs_find_first_inode(root, min_ino);
 	}
@@ -1285,14 +1277,12 @@ long btrfs_free_extent_maps(struct btrfs_fs_info *fs_info, long nr_to_scan)
 							   ctx.last_ino);
 	}
 
-	/*
-	 * We may be called from memory allocation paths, so we don't want to
-	 * take too much time and slowdown tasks, so stop if we need reschedule.
-	 */
-	while (ctx.scanned < ctx.nr_to_scan && !need_resched()) {
+	while (ctx.scanned < ctx.nr_to_scan) {
 		struct btrfs_root *root;
 		unsigned long count;
 
+		cond_resched();
+
 		spin_lock(&fs_info->fs_roots_radix_lock);
 		count = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
 					       (void **)&root,

fs/btrfs/send.c

@@ -6157,25 +6157,51 @@ static int send_write_or_clone(struct send_ctx *sctx,
 	u64 offset = key->offset;
 	u64 end;
 	u64 bs = sctx->send_root->fs_info->sectorsize;
+	struct btrfs_file_extent_item *ei;
+	u64 disk_byte;
+	u64 data_offset;
+	u64 num_bytes;
+	struct btrfs_inode_info info = { 0 };
 
 	end = min_t(u64, btrfs_file_extent_end(path), sctx->cur_inode_size);
 	if (offset >= end)
 		return 0;
 
-	if (clone_root && IS_ALIGNED(end, bs)) {
-		struct btrfs_file_extent_item *ei;
-		u64 disk_byte;
-		u64 data_offset;
+	num_bytes = end - offset;
 
-		ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
-				    struct btrfs_file_extent_item);
-		disk_byte = btrfs_file_extent_disk_bytenr(path->nodes[0], ei);
-		data_offset = btrfs_file_extent_offset(path->nodes[0], ei);
-		ret = clone_range(sctx, path, clone_root, disk_byte,
-				  data_offset, offset, end - offset);
-	} else {
-		ret = send_extent_data(sctx, path, offset, end - offset);
-	}
+	if (!clone_root)
+		goto write_data;
+
+	if (IS_ALIGNED(end, bs))
+		goto clone_data;
+
+	/*
+	 * If the extent end is not aligned, we can clone if the extent ends at
+	 * the i_size of the inode and the clone range ends at the i_size of the
+	 * source inode, otherwise the clone operation fails with -EINVAL.
+	 */
+	if (end != sctx->cur_inode_size)
+		goto write_data;
+
+	ret = get_inode_info(clone_root->root, clone_root->ino, &info);
+	if (ret < 0)
+		return ret;
+
+	if (clone_root->offset + num_bytes == info.size)
+		goto clone_data;
+
+write_data:
+	ret = send_extent_data(sctx, path, offset, num_bytes);
+	sctx->cur_inode_next_write_offset = end;
+	return ret;
+
+clone_data:
+	ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			    struct btrfs_file_extent_item);
+	disk_byte = btrfs_file_extent_disk_bytenr(path->nodes[0], ei);
+	data_offset = btrfs_file_extent_offset(path->nodes[0], ei);
+	ret = clone_range(sctx, path, clone_root, disk_byte, data_offset, offset,
+			  num_bytes);
 	sctx->cur_inode_next_write_offset = end;
 	return ret;
 }

fs/btrfs/super.c

@@ -28,6 +28,7 @@
 #include <linux/btrfs.h>
 #include <linux/security.h>
 #include <linux/fs_parser.h>
+#include <linux/swap.h>
 #include "messages.h"
 #include "delayed-inode.h"
 #include "ctree.h"
@@ -2409,6 +2410,15 @@ static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_cont
 	const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 
+	/*
+	 * We may be called from any task trying to allocate memory and we don't
+	 * want to slow it down with scanning and dropping extent maps. It would
+	 * also cause heavy lock contention if many tasks concurrently enter
+	 * here. Therefore only allow kswapd tasks to scan and drop extent maps.
+	 */
+	if (!current_is_kswapd())
+		return 0;
+
 	return btrfs_free_extent_maps(fs_info, nr_to_scan);
 }
 

fs/btrfs/tree-checker.c

@@ -569,9 +569,10 @@ static int check_dir_item(struct extent_buffer *leaf,
 
 		/* dir type check */
 		dir_type = btrfs_dir_ftype(leaf, di);
-		if (unlikely(dir_type >= BTRFS_FT_MAX)) {
+		if (unlikely(dir_type <= BTRFS_FT_UNKNOWN ||
+			     dir_type >= BTRFS_FT_MAX)) {
 			dir_item_err(leaf, slot,
-			"invalid dir item type, have %u expect [0, %u)",
+			"invalid dir item type, have %u expect (0, %u)",
 				dir_type, BTRFS_FT_MAX);
 			return -EUCLEAN;
 		}