Re: perf: fuzzer: spinlock hang

[Vince Weaver]

On Wed, 17 Aug 2016, Peter Zijlstra wrote:

> On Tue, Aug 09, 2016 at 11:51:48AM -0400, Vince Weaver wrote:
> > 
> > So still haswell/4.8-rc1, after 6 hours of fuzzing this is where the 
> > system finally locked up permanently.
> > 
> > It looks like some sort of spinlock error.
> > The logs continued being spammed with this stuff overnight until I 
> > finally put it out of its misery, but I think just the beginning here
> > is the relevant part.
> 
> That sadly looks vaguely familiar... I think its the point where I gave
> up chasing bugs last year, because I was going soft in the head staring
> at this stuff for several months non-stop.
> 
> It taking several hours to reproduce didn't help.
> 
> I suspect its something iffy with the Intel driver, because I could
> never trigger it on AMD, but of course, no guarantees ...

I'm working on getting an AMD and an ARM box fuzzing along (in addition 
to the Haswell and Core2 box), maybe that will help narrow down if it's 
an Intel issue or not.

If you're lucky I'll run out of time once classes start up next week and 
you won't have any more fuzz reports until January (well, probably 
November, as I'll need some material for the fuzzing microconference at 
Linux Plumbers).

Vince