> Date: Tue, 28 Dec 2021 08:41:16 +0900 > From: Ryo ONODERA <r...@tetera.org> > > panic: mutex_vector_enter,518: uninitialized lock (lock=0xffffd80023501458, > from=ffffffff807c091c) > cpu0: Begin traceback... > vpanic() at netbsd:vpanic+0x156 > panic() at netbsd:panic+0x3c > lockdebug_wantlock() at netbsd:lockdebug_wantlock+0x180 > mutex_enter() at netbsd:mutex_enter+0x213 > intel_rps_mark_interactive() at netbsd:intel_rps_mark_interactive+0x1e
Weird -- can you insert db_stacktrace() into intel_rps_init_early and intel_rps_fini? This looks like i915->gt.rps.power.mutex, judging by the stack trace, but I don't see how we can get here without first initializing the mutex.