[PATCH 0/4] Enable MSI support for KVM VT-d

[PATCH 0/4] Enable MSI support for KVM VT-d

[Sheng Yang]

Hi, Avi

This patchset enable MSI support for KVM VT-d.

And here are only kernel space ones. The third patch would go to also goto x86
upstream.

The userspace code would looks like this:

assigned_irq_data.guest_msi_addr = *(uint32_t *)(d->msi_cap + 4);
assigned_irq_data.guest_msi_data = *(uint16_t *)(d->msi_cap + 8);
assigned_irq_data.flags |= KVM_DEV_IRQ_ASSIGN_ENABLE_MSI;
r = kvm_assign_irq(kvm_context, &assigned_irq_data);

I've test the patchset with some userspace hack, it works well.

Thanks!
--
regards
Yang, Sheng

Re: [PATCH 0/4] Enable MSI support for KVM VT-d

[Avi Kivity]

Sheng Yang wrote:
> Hi, Avi
>
> This patchset enable MSI support for KVM VT-d.
>
> And here are only kernel space ones. The third patch would go to also goto x86
> upstream.
>
> The userspace code would looks like this:
>
> assigned_irq_data.guest_msi_addr = *(uint32_t *)(d->msi_cap + 4);
> assigned_irq_data.guest_msi_data = *(uint16_t *)(d->msi_cap + 8);
> assigned_irq_data.flags |= KVM_DEV_IRQ_ASSIGN_ENABLE_MSI;
> r = kvm_assign_irq(kvm_context, &assigned_irq_data);
>
> I've test the patchset with some userspace hack, it works well.
>
>   

Can you resend this patch with all the updates, as well as the userspace 
changes?

-- 
error compiling committee.c: too many arguments to function

RE: [PATCH 0/4] Enable MSI support for KVM VT-d

[Zhang, Xiantao]

Avi Kivity wrote:
> Sheng Yang wrote:
>> Hi, Avi
>> 
>> This patchset enable MSI support for KVM VT-d.
>> 
>> And here are only kernel space ones. The third patch would go to
>> also goto x86 upstream. 
>> 
>> The userspace code would looks like this:
>> 
>> assigned_irq_data.guest_msi_addr = *(uint32_t *)(d->msi_cap + 4);
>> assigned_irq_data.guest_msi_data = *(uint16_t *)(d->msi_cap + 8);
>> assigned_irq_data.flags |= KVM_DEV_IRQ_ASSIGN_ENABLE_MSI;
>> r = kvm_assign_irq(kvm_context, &assigned_irq_data);
>> 
>> I've test the patchset with some userspace hack, it works well.
>> 
>> 
> 
> Can you resend this patch with all the updates, as well as the
> userspace changes?

Maybe Sheng need to make it work on kvm/ia64, and at least the changes
can't break ia64 side. 
Xiantao


> --
> error compiling committee.c: too many arguments to function

Re: [PATCH 0/4] Enable MSI support for KVM VT-d

[Yang, Sheng]

On Monday 20 October 2008 12:09:20 Zhang, Xiantao wrote:
> Avi Kivity wrote:
> > Sheng Yang wrote:
> >> Hi, Avi
> >>
> >> This patchset enable MSI support for KVM VT-d.
> >>
> >> And here are only kernel space ones. The third patch would go to
> >> also goto x86 upstream.
> >>
> >> The userspace code would looks like this:
> >>
> >> assigned_irq_data.guest_msi_addr = *(uint32_t *)(d->msi_cap + 4);
> >> assigned_irq_data.guest_msi_data = *(uint16_t *)(d->msi_cap + 8);
> >> assigned_irq_data.flags |= KVM_DEV_IRQ_ASSIGN_ENABLE_MSI;
> >> r = kvm_assign_irq(kvm_context, &assigned_irq_data);
> >>
> >> I've test the patchset with some userspace hack, it works well.
> >
> > Can you resend this patch with all the updates, as well as the
> > userspace changes?
>
> Maybe Sheng need to make it work on kvm/ia64, and at least the changes
> can't break ia64 side. Xiantao
>
Yes, I would ensure that won't break ia64. I would repost the patchset soon. 

But Avi, for the userspace, the Amit's patch still not checked in, so I 
haven't written a complete version because of lacking code base. I only got a 
experiment patch by hand, expose MSI cap to guest and enable MSI when guest 
wrote MSI enable bit. Well, I prefer to give you a complete version after 
Amit's patch is there. (I will conclude the change of userspace in the first 
mail).

--
regards
Yang, Sheng