Unless it's a regression, it's rather a thing to be reported to the upstream. Could you try to report it the upstream, e.g. linux-wireless ML?