Commit 517236cb authored by Kent Overstreet's avatar Kent Overstreet
Browse files

bcachefs: Kill read lock dropping in bch2_btree_node_lock_write_nofail() 



dropping read locks in bch2_btree_node_lock_write_nofail() dates from
before we had the cycle detector; we can now tell the cycle detector
directly when taking a lock may not fail because we can't handle
transaction restarts.

This is needed for adding should_be_locked asserts.

Signed-off-by: default avatarKent Overstreet <kent.overstreet@linux.dev>
parent beccf291

fs/bcachefs/btree_locking.c

+1 −27
Original line number Diff line number Diff line
@@ -440,33 +440,7 @@ void bch2_btree_node_lock_write_nofail(struct btree_trans *trans, 
				       struct btree_path *path,

				       struct btree_bkey_cached_common *b)

{

	struct btree_path *linked;

	unsigned i, iter;

	int ret;



	/*

	 * XXX BIG FAT NOTICE

	 *

	 * Drop all read locks before taking a write lock:

	 *

	 * This is a hack, because bch2_btree_node_lock_write_nofail() is a

	 * hack - but by dropping read locks first, this should never fail, and

	 * we only use this in code paths where whatever read locks we've

	 * already taken are no longer needed:

	 */



	trans_for_each_path(trans, linked, iter) {

		if (!linked->nodes_locked)

			continue;



		for (i = 0; i < BTREE_MAX_DEPTH; i++)

			if (btree_node_read_locked(linked, i)) {

				btree_node_unlock(trans, linked, i);

				btree_path_set_dirty(linked, BTREE_ITER_NEED_RELOCK);

			}

	}



	ret = __btree_node_lock_write(trans, path, b, true);

	int ret = __btree_node_lock_write(trans, path, b, true);

	BUG_ON(ret);

}