Re: [qemu-mainline test] 156257: regressions - FAIL

[Anthony PERARD <anthony.perard@xxxxxxxxxx>]

On Wed, Oct 28, 2020 at 02:49:15PM +0000, Julien Grall wrote:
> (+ Anthony and Stefano,)
> 
> Hi Jason,
> 
> On 28/10/2020 13:37, Jason Andryuk wrote:
> > On Tue, Oct 27, 2020 at 5:23 PM osstest service owner
> > <osstest-admin@xxxxxxxxxxxxxx> wrote:
> > > 
> > > flight 156257 qemu-mainline real [real]
> > > flight 156266 qemu-mainline real-retest [real]
> > > http://logs.test-lab.xenproject.org/osstest/logs/156257/
> > > http://logs.test-lab.xenproject.org/osstest/logs/156266/
> > > 
> > > Regressions :-(
> > > 
> > > Tests which did not succeed and are blocking,
> > > including tests which could not be run:
> > >   test-arm64-arm64-libvirt-xsm 14 guest-start              fail REGR. vs. 
> > > 152631
> > 
> > QEMU doesn't start with "qemu-system-i386: -xen-domid 1: Option not
> > supported for this target"
> > 
> > This happens if CONFIG_XEN isn't set.
> > 
> > QEMU is built with:
> >                    host CPU: aarch64
> >             host endianness: little
> >                 target list: i386-softmmu
> > 
> > commit 8a19980e3fc4 "configure: move accelerator logic to meson"
> > introduced this logic:
> > +accelerator_targets = { 'CONFIG_KVM': kvm_targets }
> > +if cpu in ['x86', 'x86_64']
> > +  accelerator_targets += {
> > +    'CONFIG_HAX': ['i386-softmmu', 'x86_64-softmmu'],
> > +    'CONFIG_XEN': ['i386-softmmu', 'x86_64-softmmu'],
> > +    'CONFIG_HVF': ['x86_64-softmmu'],
> > +    'CONFIG_WHPX': ['i386-softmmu', 'x86_64-softmmu'],
> > +  }
> > +endif
> 
> I always wondered when this would come to bite us :). I am surprised it took
> so long.
> 
> > 
> > I guess something like this would fix it:
> > if cpu in ['aarch64', 'arm']
> >    accelerator_targets += { 'CONFIG_XEN': ['i386-softmmu'], }
> > endif
> 
> Per the logic above, I think this correct. @Stefano, @Anthony, can you have
> a look?

Yes, that would probably do the trick and restrict to x86 and arm rather
than any host with xen support. I think this would need a comment
explaining why we enable xen for i386 target on arm (or at least a
comment that we use i386 target on arm).

Thanks for the investigation.

Cheers,

-- 
Anthony PERARD