Merge tag 'for-4.21-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux

Pull btrfs updates from David Sterba:
 "New features:

   - swapfile support - after a long time it's here, with some
     limitations where COW design does not work well with the swap
     implementation (nodatacow file, no compression, cannot be
     snapshotted, not possible on multiple devices, ...), as this is the
     most restricted but working setup, we'll try to improve that in the
     future

   - metadata uuid - an optional incompat feature to assign a new
     filesystem UUID without overwriting all metadata blocks, stored
     only in superblock

   - more balance messages are printed to system log, initial is in the
     format of the command line that would be used to start it

  Fixes:

   - tag pages of a snapshot to better separate pages that are involved
     in the snapshot (and need to get synced) from newly dirtied pages
     that could slow down or even livelock the snapshot operation

   - improved check of filesystem id associated with a device during
     scan to detect duplicate devices that could be mixed up during
     mount

   - fix device replace state transitions, eg. when it ends up
     interrupted and reboot tries to restart balance too, or when
     start/cancel ioctls race

   - fix a crash due to a race when quotas are enabled during snapshot
     creation

   - GFP_NOFS/memalloc_nofs_* fixes due to GFP_KERNEL allocations in
     transaction context

   - fix fsync of files with multiple hard links in new directories

   - fix race of send with transaction commits that create snapshots

  Core changes:

   - cleanups:
      * further removals of now-dead fsync code
      * core function for finding free extent has been split and
        provides a base for further cleanups to make the logic more
        understandable
      * removed lot of indirect callbacks for data and metadata inodes
      * simplified refcounting and locking for cloned extent buffers
      * removed redundant function arguments
      * defines converted to enums where appropriate

   - separate reserve for delayed refs from global reserve, update logic
     to do less trickery and ad-hoc heuristics, move out some related
     expensive operations from transaction commit or file truncate

   - dev-replace switched from custom locking scheme to semaphore

   - remove first phase of balance that tried to make some space for the
     relocation by calling shrink and grow, this did not work as
     expected and only introduced more error states due to potential
     resize failures, slightly improves the runtime as the chunks on all
     devices are not needlessly enumerated

   - clone and deduplication now use generic helper that adds a few more
     checks that were missing from the original btrfs implementation of
     the ioctls"

* tag 'for-4.21-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux: (125 commits)
  btrfs: Fix typos in comments and strings
  btrfs: improve error handling of btrfs_add_link
  Btrfs: use generic_remap_file_range_prep() for cloning and deduplication
  btrfs: Refactor main loop in extent_readpages
  btrfs: Remove 1st shrink/grow phase from balance
  Btrfs: send, fix race with transaction commits that create snapshots
  Btrfs: use nofs context when initializing security xattrs to avoid deadlock
  btrfs: run delayed items before dropping the snapshot
  btrfs: catch cow on deleting snapshots
  btrfs: extent-tree: cleanup one-shot usage of @blocksize in do_walk_down
  Btrfs: scrub, move setup of nofs contexts higher in the stack
  btrfs: scrub: move scrub_setup_ctx allocation out of device_list_mutex
  btrfs: scrub: pass fs_info to scrub_setup_ctx
  btrfs: fix truncate throttling
  btrfs: don't run delayed refs in the end transaction logic
  btrfs: rework btrfs_check_space_for_delayed_refs
  btrfs: add new flushing states for the delayed refs rsv
  btrfs: update may_commit_transaction to use the delayed refs rsv
  btrfs: introduce delayed_refs_rsv
  btrfs: only track ref_heads in delayed_ref_updates
  ...

1 files changed, 20 insertions, 24 deletions

diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index a5ce99a6c936..ac232b3d6d7e 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1144,7 +1144,7 @@ next:
 	}
 	btrfs_release_path(path);
 
-	/* look for a conflicing name */
+	/* look for a conflicting name */
 	di = btrfs_lookup_dir_item(trans, root, path, btrfs_ino(dir),
 				   name, namelen, 0);
 	if (di && !IS_ERR(di)) {
@@ -3149,7 +3149,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	mutex_unlock(&log_root_tree->log_mutex);
 
 	/*
-	 * nobody else is going to jump in and write the the ctree
+	 * Nobody else is going to jump in and write the ctree
 	 * super here because the log_commit atomic below is protecting
 	 * us.  We must be called with a transaction handle pinning
 	 * the running transaction open, so a full commit can't hop
@@ -3201,8 +3201,6 @@ static void free_log_tree(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *log)
 {
 	int ret;
-	u64 start;
-	u64 end;
 	struct walk_control wc = {
 		.free = 1,
 		.process_func = process_one_buffer
@@ -3216,18 +3214,8 @@ static void free_log_tree(struct btrfs_trans_handle *trans,
 			btrfs_handle_fs_error(log->fs_info, ret, NULL);
 	}
 
-	while (1) {
-		ret = find_first_extent_bit(&log->dirty_log_pages,
-				0, &start, &end,
-				EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT,
-				NULL);
-		if (ret)
-			break;
-
-		clear_extent_bits(&log->dirty_log_pages, start, end,
-				  EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT);
-	}
-
+	clear_extent_bits(&log->dirty_log_pages, 0, (u64)-1,
+			  EXTENT_DIRTY | EXTENT_NEW | EXTENT_NEED_WAIT);
 	free_extent_buffer(log->node);
 	kfree(log);
 }
@@ -4383,7 +4371,6 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
 	struct extent_map *em, *n;
 	struct list_head extents;
 	struct extent_map_tree *tree = &inode->extent_tree;
-	u64 logged_start, logged_end;
 	u64 test_gen;
 	int ret = 0;
 	int num = 0;
@@ -4392,8 +4379,6 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
 
 	write_lock(&tree->lock);
 	test_gen = root->fs_info->last_trans_committed;
-	logged_start = start;
-	logged_end = end;
 
 	list_for_each_entry_safe(em, n, &tree->modified_extents, list) {
 		/*
@@ -4434,11 +4419,6 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
 		    em->start >= i_size_read(&inode->vfs_inode))
 			continue;
 
-		if (em->start < logged_start)
-			logged_start = em->start;
-		if ((em->start + em->len - 1) > logged_end)
-			logged_end = em->start + em->len - 1;
-
 		/* Need a ref to keep it from getting evicted from cache */
 		refcount_inc(&em->refs);
 		set_bit(EXTENT_FLAG_LOGGING, &em->flags);
@@ -5778,6 +5758,22 @@ static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 			goto end_trans;
 	}
 
+	/*
+	 * If a new hard link was added to the inode in the current transaction
+	 * and its link count is now greater than 1, we need to fallback to a
+	 * transaction commit, otherwise we can end up not logging all its new
+	 * parents for all the hard links. Here just from the dentry used to
+	 * fsync, we can not visit the ancestor inodes for all the other hard
+	 * links to figure out if any is new, so we fallback to a transaction
+	 * commit (instead of adding a lot of complexity of scanning a btree,
+	 * since this scenario is not a common use case).
+	 */
+	if (inode->vfs_inode.i_nlink > 1 &&
+	    inode->last_link_trans > last_committed) {
+		ret = -EMLINK;
+		goto end_trans;
+	}
+
 	while (1) {
 		if (!parent || d_really_is_negative(parent) || sb != parent->d_sb)
 			break;