npiggin

This is a deadlock due to a synchronous smp_call_function being called from the timer interrupt: [12532.259260] [c000003fe566b320] [c0000000001d5340] smp_call_function_single+0x110/0x180 (unreliable) [12532.259263] [c000003fe566b390] [c0000000001d55e0] smp_call_function_any+0x180/0x250 [12532.259265] [c000003fe566b3f0] [c000000000acd3e8] gpstate_timer_handler+0x1e8/0x580 [12532.259268] [c000003fe566b4a0]...

I proposed a patch to process opal events directly by kopald, and to enable and disable interrupts and also cond_resched between processing each event. It might improve this situation. https://marc.info/?l=linuxppc-embedded&m=150902327030882&w=2

No I haven't had a chance to yet, I've been caught up doing P9 work at the moment.

I brought the patch up to date and resubmitted it, but... I don't think it solves the root cause of the problem, there must be a long running OPAL event...

You are reading the ISA correctly, these MSR bits are reserved. It predates my involvement with OpenPOWER but the code or comment may have inadvertently come from some legacy code...