[Bug 198149] [hwpmc] pmcstat -P -t (top mode, process sampling) stops after a while

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=198149

--- Comment #12 from John Baldwin <jhb at FreeBSD.org> ---
Finally snagged the hang with the first hack in place.  By the end of the trace
all the PMCs have super high thresholds again, but here is where it went wrong:

211497   1  305651558868602 MDP:WRI:1: iaf-write cpu=1 ri=2 msr=0x40000002
v=ffffffdaa2a0 iafctrl=0 pmc=ffffffdaa2a0
211490   1  305651558791430 MDP:REA:1: iaf-read cpu=1 ri=2 msr=0x40000002 ->
v=25ec32
211481   1  305651558772962 MDP:WRI:1: iaf-write cpu=1 ri=2 msr=0x40000002
v=ffffffd9db37 iafctrl=0 pmc=ffffffd9db37
211474   1  305651558684430 MDP:REA:1: iaf-read cpu=1 ri=2 msr=0x40000002 ->
v=ffffffffffd8db37
211465   1  305651556116174 MDP:WRI:1: iaf-write cpu=1 ri=2 msr=0x40000002
v=1000000000000 iafctrl=0 pmc=0
211458   1  305651555926610 MDP:REA:1: iaf-read cpu=1 ri=2 msr=0x40000002 ->
v=c220
211449   1  305651555915314 MDP:WRI:1: iaf-write cpu=1 ri=2 msr=0x40000002
v=ffffffff1ef0 iafctrl=0 pmc=ffffffff1ef0
211442   1  305651555647630 MDP:REA:1: iaf-read cpu=1 ri=2 msr=0x40000002 ->
v=97d8
211433   1  305651555563074 MDP:WRI:1: iaf-write cpu=1 ri=2 msr=0x40000002
v=ffffffff438c iafctrl=0 pmc=ffffffff438c

In particular, after writing 0 to the pmc in 211465, things went downhill.  I'm
still looking in the debug logs, but one thing is that we definitely can't put
a value of 0 into a PMC.  We should possibly force it to -1 in this case so
that the interrupt triggers on the next event in that case?

-- 
You are receiving this mail because:
You are the assignee for the bug.