06:53 <janneg> Did we break `bt` from the HV working for anyone

06:54 <janneg> it's not working at all for me on the m2 pro, i.e. produces an empty stack trace

06:56 <janneg> do I remember correctly that there was mmio there reads/writes might block?

06:58 <janneg> I have a hung probe for dcpext on the m2 pro (used as model for hdmi out on the m2 ultra)

07:00 <janneg> but unfortunately no idea what's hanging. I don't even reach the dcp probe. the darts are at least in so far ok as I can dump them from the HV

07:01 <janneg> could of course be just deadlock in linux but lockdep output was not helpful

07:09 <janneg> yes, lockdep just complains about cpuidle, warnings go away with deselecting ARM_APPLE_CPUIDLE

08:11 <janneg> huh, backtrace works without cpuidle. it's hanging in dart_map_pages, probably for dart-dcpext0

08:20 <janneg> expected that idle cpus don't show a bt, just bad luck that the first 7 CPUs were all idle. the loop over all cpu's failed over the the first deactivated core on a 10 core m2 pro device

16:01 <marcan> janneg: `bt` won't work for CPUs in deep idle because the stack is in a funny state and there is no LR nor frame pointer after waking up from that

16:02 <marcan> we could at least attempt to detect that (print something if the return address is after a wfi)