Hi, there's now a patch in the upstream bug report <https://bugzilla.kernel.org/show_bug.cgi?id=86231> (and soon to be submitted upstream) that seems to fix the issue on my machine. It turns out that this is probably not a regression in the kernel, but introduced by the fact that systemd restores the rfkill state fairly early during boot - which uncovered a race condition in the driver. Kind regards Ralf