Doesn't support arm64

[Ming Lei]

Without attached patch, we can't create VM on arm64 with KVM.

[Robie Basak (racb)]

Thank you for work and for the patch!

I'm keen to make sure that we don't diverge from Linaro and libvirt upstream on armhf/arm64 enablement. I know that this is an active area of work in Linaro at the moment. I believe Clark Laughlin is working on it?

Clark, how does this fit in with your plans, please? Or perhaps you could review and/or use it?

I just want to avoid putting something into Ubuntu that then conflicts with your work.

Marking Wishlist as arm64 support would be a new feature.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Robie Basak (racb)]

I've just spoken to Ming about this. He wrote this patch as a workaround, and isn't sure that it's suitable to land for the general case. So I'd like to hold from landing this in Ubuntu or from sending it upstream until it has been reviewed for suitability.

[Robie Basak (racb)]

Clark just pointed me to http://libvirt.org/git/?p=libvirt.git;a=commit;h=27e32e0f3d4d5840cee7e361bdff94c68ae181e2 which is committed upstream.

[Ming Lei (tom-leiming)]

Looks the upstreamed commit mentioned in #5 includes most of the patch we are using, and that is great
really, :-)

But maybe below part is needed too:

diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
index e964c75..c947e2e 100644
--- a/src/qemu/qemu_domain.c
+++ b/src/qemu/qemu_domain.c
@@ -793,10 +794,14 @@ qemuDomainDefaultNetModel(const virDomainDef *def)
         def->os.arch == VIR_ARCH_S390X)
         return "virtio";

- if (def->os.arch == VIR_ARCH_ARMV7L) {
+ if (def->os.arch == VIR_ARCH_ARMV7L ||
+ def->os.arch == VIR_ARCH_AARCH64) {
         if (STREQ(def->os.machine, "versatilepb"))
             return "smc91c111";

+ if (strstr(def->os.machine, "virt"))
+ return "virtio";
+
         /* Incomplete. vexpress (and a few others) use this, but not all
          * arm boards */
         return "lan9118";

[Serge Hallyn (serge-hallyn)]

Quoting Ming Lei (<email address hidden>):
> Looks the upstreamed commit mentioned in #5 includes most of the patch we are using, and that is great
> really, :-)
>
> But maybe below part is needed too:
>
> diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
> index e964c75..c947e2e 100644
> --- a/src/qemu/qemu_domain.c
> +++ b/src/qemu/qemu_domain.c
> @@ -793,10 +794,14 @@ qemuDomainDefaultNetModel(const virDomainDef *def)
> def->os.arch == VIR_ARCH_S390X)
> return "virtio";
>
> - if (def->os.arch == VIR_ARCH_ARMV7L) {
> + if (def->os.arch == VIR_ARCH_ARMV7L ||
> + def->os.arch == VIR_ARCH_AARCH64) {
> if (STREQ(def->os.machine, "versatilepb"))
> return "smc91c111";
>
> + if (strstr(def->os.machine, "virt"))
> + return "virtio";
> +
> /* Incomplete. vexpress (and a few others) use this, but not all
> * arm boards */
> return "lan9118";

Can you create a single patch with what remains of what you need? Also,
if you look at the changelog for the upstreamed patch they had to do
some work to the testcase to be accepted - does what remains of your
patch need any testcase tweaking?

[Ming Lei (tom-leiming)]

On Thu, Jan 9, 2014 at 1:39 AM, Serge Hallyn
> Can you create a single patch with what remains of what you need? Also,

Yes, I will do it.

> if you look at the changelog for the upstreamed patch they had to do
> some work to the testcase to be accepted - does what remains of your
> patch need any testcase tweaking?

Sorry, I didn't understand this stuff, and I just remember without the patch,
we can't bring up a VM on arm64 with qemu...

Thanks for your comment.

Thanks
--
Ming Lei