Re: [kvm-unit-tests PATCH 14/32] powerpc: general interrupt tests

["Nicholas Piggin" <npiggin@gmail.com>]

On Sat Mar 2, 2024 at 12:14 AM AEST, Andrew Jones wrote:
> On Fri, Mar 01, 2024 at 02:57:04PM +0100, Thomas Huth wrote:
> > On 01/03/2024 14.45, Andrew Jones wrote:
> > > On Fri, Mar 01, 2024 at 01:41:22PM +0100, Thomas Huth wrote:
> > > > On 26/02/2024 11.12, Nicholas Piggin wrote:
> > > > > Add basic testing of various kinds of interrupts, machine check,
> > > > > page fault, illegal, decrementer, trace, syscall, etc.
> > > > > 
> > > > > This has a known failure on QEMU TCG pseries machines where MSR[ME]
> > > > > can be incorrectly set to 0.
> > > > 
> > > > Two questions out of curiosity:
> > > > 
> > > > Any chance that this could be fixed easily in QEMU?
> > > > 
> > > > Or is there a way to detect TCG from within the test? (for example, we have
> > > > a host_is_tcg() function for s390x so we can e.g. use report_xfail() for
> > > > tests that are known to fail on TCG there)
> > > 
> > > If there's nothing better, then it should be possible to check the
> > > QEMU_ACCEL environment variable which will be there with the default
> > > environ.
> > 
> > Well, but that's only available from the host side, not within the test
> > (i.e. the guest). So that does not help much with report_xfail...
>
> powerpc has had environment variables in guests since commit f266c3e8ef15
> ("powerpc: enable environ"). QEMU_ACCEL is one of the environment
> variables given to unit tests by default when ENVIRON_DEFAULT is 'yes', as
> is the default set in configure. But...
>
> > I was rather thinking of something like checking the device tree, e.g. for
> > the compatible property in /hypervisor to see whether it's KVM or TCG...?
>
> ...while QEMU_ACCEL will work when the environ is present, DT will always
> be present, so checking the hypervisor node sounds better to me.

Yeah I got that.

One issue with xfail I noted when looking at this earlier, is that it
*always* expects a fail, and fails if it succeeds. So if you fix a QEMU
bug then you introduce a fail to kvm-unit-tests. So we really want a
report_known_bug(cond, "SPRs look sane", "QEMU TCG before v9.0 and
POWER9 DD2.0 is known to fail..."); and that could report a maybe-fail
that doesn't make the test group fail.

Thanks,
Nick