Messages in this thread | | | From | Vince Weaver <> | Date | Tue, 17 Mar 2015 11:00:44 -0400 (EDT) | Subject | Re: perf: fuzzer causes lockup in x86_pmu_event_init() |
| |
On Tue, 17 Mar 2015, Jiri Olsa wrote:
> > I'll start my test-machine fuzzing with this plus the infinite-interrupt > > fix and see what happens. The machines do eventually trigger the issue > > somwhat consistently but it can take up to a week before they hit it > > so it's a tough one to test for. > > I guess no news is good news in here.. ;-)
yes, with both the irq-storm patch plus your patch I've been fuzzing for over a week with no crashes. (I'm only using your patch for the lockup problem, not Peter's).
There are other minor warnings that pop up as per usual, and if you fuzz with paranoid set to 1 you quickly crash with the uncore issue I reported a while back, but otherwise things seem really solid with the two patches I mentioned.
Of course that's what I always say just before things start crashing again, although right now I had to stop the fuzzing temporarily as somehow the fuzzer is leaking children for reasons I don't understand and I'm trying to debug.
Vince
| |