Michael Hudson-Doyle <michael.hudson@linaro.org> writes:

> Pranavkumar Sawargaonkar <pranavkumar@linaro.org> writes:
>
>> Hi,
>>
>> On 6 June 2014 08:58, Michael Hudson-Doyle <michael.hudson@linaro.org>
>> wrote:
>>> Hi all,
>>>
>>> As a bit of a test for myself, I'm trying to run openstack tests on top
>>> of the mostly-ready PSCI support the virtualization team has been
>>> working on.  I've built myself an Ubuntu kernel with the needed patches
>>> (actually all of kvm/next) and now need a matching qemu.  I fished these
>>> patches:
>>>
>>> http://lists.gnu.org/archive/html/qemu-devel/2014-05/msg04514.html
>>>
>>> out of patchwork and applied them on top of the qemu that's in trusty
>>> but it fails to build.  I guess these patches depends that are in qemu
>>> git but not in 2.0.0?  In any case, before I dive into trying to fix
>>> this myself, does anyone have binaries or even better debs with those
>>> patches in?
>>
>> It should get applied on the latest qemu git code. (git://
>> git.qemu.org/qemu.git)
>
> Thanks.
>
>> Please get the attached binary I build for testing.
>
> Even easier :-)
>
>> Also please make sure your host and guest kernel has PSCI0.2 support
>> (currently it is available in linux-next tree only), if you want to test
>> functions like reset and poweroff of VM.
>
> My goal is to run the openstack tests, several of which fail because of
> the lack of suspend, restart etc.  So thanks for the pointers!
>
> I had built myself a host kernel (utopic kernel + kvm/next) but had
> forgotten to prepare that as the guest kernel.  My host kernel doesn't
> appear to have the right drivers to do SATA on mustang so my forgetting
> the guest hadn't immediately caused problems :-)

So, a week or so later, I have a kernel that should support PSCI (3.15 +
ubuntu sauce + kvm/next) in both host and guest and a
qemu-system-aarch64 that should support PSCI (the binary you gave me).

I think things are working to some extent because running "reboot"
inside a VM works, but "nova suspend" still doesn't work (it just kills
the instance).

Can you suggest how I might go about debugging this?

Cheers,
mwh