Re: [Qemu-devel] [RFC PATCH] target-arm/psci.c: wake up sleeping CPUs (MTTCG)

[Andrew Jones <drjones@redhat.com>]

On Fri, Jun 26, 2015 at 08:06:55AM +0100, Alex Bennée wrote:
> 
> Andrew Jones <drjones@redhat.com> writes:
> 
> > On Wed, Jun 24, 2015 at 08:12:52PM +0100, Peter Maydell wrote:
> >> On 24 June 2015 at 18:18, Alex Bennée <alex.bennee@linaro.org> wrote:
> >> >
> >> > Paolo Bonzini <pbonzini@redhat.com> writes:
> >> >
> >> >> On 24/06/2015 17:34, Alex Bennée wrote:
> >> >>> Testing with Alexander's bare metal syncronisation tests fails in MTTCG
> >> >>> leaving one CPU spinning forever waiting for the second CPU to wake up.
> >> >>> We simply need to poke the halt_cond once we have processed the PSCI
> >> >>> power on call.
> >> >>>
> >> >>> Tested-by: Alex Bennée <alex.bennee@linaro.org>
> >> >>> CC: Alexander Spyridakis <a.spyridakis@virtualopensystems.com>
> >> >>>
> >> >>> ---
> >> >>> TODO
> >> >>>   - exactly how does the vexpress wake up it's sleeping CPUs?
> >> >>> ---
> >> >>>  target-arm/psci.c | 2 ++
> >> >>>  1 file changed, 2 insertions(+)
> >> >>>
> >> >>> diff --git a/target-arm/psci.c b/target-arm/psci.c
> >> >>> index d8fafab..661ff28 100644
> >> >>> --- a/target-arm/psci.c
> >> >>> +++ b/target-arm/psci.c
> >> >>> @@ -196,6 +196,8 @@ void arm_handle_psci_call(ARMCPU *cpu)
> >> >>>          }
> >> >>>          target_cpu_class->set_pc(target_cpu_state, entry);
> >> >>>
> >> >>> +        qemu_cond_signal(target_cpu_state->halt_cond);
> >> >>
> >> >> That's called qemu_cpu_kick(target_cpu_state). :)  The patch should be
> >> >> acceptable now upstream, I think.
> >> >
> >> > Oh so this might well fail in KVM too?
> >> 
> >> In KVM we won't use target-arm/psci.c because PSCI calls
> >> are handled in the kernel.
> >>
> >
> > It's also not valid to use Alexander's test on KVM, as the test
> > framework doesn't enable the mmu, and thus {ldr,str}ex won't work
> > as expected.
> >
> > I guess I need to do a better job at advertising/documenting
> > kvm-unit-tests/arm, as that framework would suit this test just
> > fine. I've attached a patch porting the test over to k-u-t[1].
> > After applying the patch, do
> >
> > ./configure --arch=arm64 --cross-prefix=aarch64-linux-gnu-
> > OR
> > ./configure --arch=arm --cross-prefix=arm-linux-gnu-
> >
> > make

One more step here, that you probably already figured out

export QEMU=<path-to-qemu-you-want-to-test>

> >
> > arm/run arm/vos-spinlock-test.flat -smp 4 # non-atomic locks
> > OR
> > arm/run arm/vos-spinlock-test.flat -smp 4 -append atomic # atomic
> > locks
> 
> Thanks for that. I shall have a play with it today.

Feel free to make suggestions/patches for the framework, and, of course,
to write unit tests :-)

drew

> 
> >
> > Thanks,
> > drew
> >
> > [1] git://git.kernel.org/pub/scm/virt/kvm/kvm-unit-tests.git
> 
> -- 
> Alex Bennée
> 
>