bcachefs: Fix lock ordering under traverse_all()

traverse_all() traverses btree paths in sorted order, so it should never
see transaction restarts due to lock ordering violations. But some code
in __bch2_btree_path_upgrade(), while necessary when not running under
traverse_all(), was causing some confusing lock ordering violations -
disabling this code under traverse_all() will let us put in some more
assertions.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This commit is contained in:
Kent Overstreet 2022-03-06 21:17:43 -05:00 committed by Kent Overstreet
parent a897ef6827
commit 61a66469a4

View File

@ -473,14 +473,15 @@ bool __bch2_btree_path_upgrade(struct btree_trans *trans,
* before interior nodes - now that's handled by
* bch2_btree_path_traverse_all().
*/
trans_for_each_path(trans, linked)
if (linked != path &&
linked->cached == path->cached &&
linked->btree_id == path->btree_id &&
linked->locks_want < new_locks_want) {
linked->locks_want = new_locks_want;
btree_path_get_locks(trans, linked, true);
}
if (!path->cached && !trans->in_traverse_all)
trans_for_each_path(trans, linked)
if (linked != path &&
linked->cached == path->cached &&
linked->btree_id == path->btree_id &&
linked->locks_want < new_locks_want) {
linked->locks_want = new_locks_want;
btree_path_get_locks(trans, linked, true);
}
return false;
}