Paul Mackerras's on February 18, 2019 9:06 am: > On Sat, Oct 13, 2018 at 10:04:09PM +1000, Nicholas Piggin wrote: >> Reimplement Book3S idle code in C, moving POWER7/8/9 implementation >> speific HV idle code to the powernv platform code. >> > > [...] > >> @@ -2760,21 +2744,47 @@ BEGIN_FTR_SECTION >> li r4, LPCR_PECE_HVEE@higher >> sldi r4, r4, 32 >> or r5, r5, r4 >> -END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300) >> +FTR_SECTION_ELSE >> + li r3, PNV_THREAD_NAP >> +ALT_FTR_SECTION_END_IFSET(CPU_FTR_ARCH_300) >> mtspr SPRN_LPCR,r5 >> isync >> - li r0, 0 >> - std r0, HSTATE_SCRATCH0(r13) >> - ptesync >> - ld r0, HSTATE_SCRATCH0(r13) >> -1: cmpd r0, r0 >> - bne 1b >> + >> + mr r0, r1 >> + ld r1, PACAEMERGSP(r13) >> + subi r1, r1, STACK_FRAME_OVERHEAD >> + std r0, 0(r1) >> + ld r0, PACAR1(r13) >> + std r0, 8(r1) > > This bit seems wrong to me. If this is a secondary thread on POWER8, > we were already on the emergency stack, and now we've reset r1 back to > the top of the emergency stack and we're overwriting it.
I'll have to find some time to take another look at this stuff. The KVM stuff was a bit hasty. > I wonder why you didn't see secondary threads going off into lala land > in your tests? It must have been because I wasn't testing the guest SMT properly because I did get it to break trivially sometime after posting this patch out. So we were on the emergency stack here, that should make things easier, that may be what's wrong. Thanks, Nick
