Test 73 Sig_trap fails on arm64 (was Re: [PATCH] perf test: Test 73 Sig_trap fails on s390)

Leo Yan leo.yan at linaro.org
Wed Feb 16 05:13:06 PST 2022


On Wed, Feb 16, 2022 at 12:54:16PM +0100, Dmitry Vyukov wrote:
> On Wed, 16 Feb 2022 at 12:47, John Garry <john.garry at huawei.com> wrote:

[...]

> > > Signals make this messy, as the step logic will step_into_  the signal
> > > handler -- we have to do this, otherwise we would miss break/watchpoints
> > > triggered by the signal handler if we had disabled them for the step.
> > > However, it means that when we return back from the signal handler we will
> > > run back into the break/watchpoint which we initially stepped over. When
> > > perf uses SIGTRAP to notify userspace that we hit a break/watchpoint,
> > > then we'll get stuck because we'll step into the handler every time.
> > >
> > > Hopefully that clears things up a bit. Ideally, the kernel wouldn't
> > > pretend to handle this stepping at all for arm64 as it adds a bunch of
> > > complexity, overhead to our context-switch and I don't think the current
> > > behaviour is particularly useful.
> > >
> >
> > Right, so what I am hearing altogether is that for now we should just
> > skip this test.
> >
> > And since the kernel does not seem to advertise this capability we need
> > to disable for specific architectures.
> 
> It does and fwiw I am just trying to use it. Things work only on x86 so far.

So here we have agreement to disable the cases for Arm64/Arm.

John, since you put much efforts to follow up the issue, I'd like to
leave decision to you if you want to proceed for patches?  Otherwise,
I will send patches to disable cases in perf.

Thanks,
Leo



More information about the linux-arm-kernel mailing list