Re: [PATCH 09/10] KVM: Enable MSI for device assignment

[Avi Kivity <avi@redhat.com>]

Sheng Yang wrote:
> We enable guest MSI and host MSI support in this patch. The userspace want to
> enable MSI should set KVM_DEV_IRQ_ASSIGN_ENABLE_MSI in the assigned_irq's
> flag. Function would return -ENOTTY if can't enable MSI, userspace shouldn't
> set MSI Enable bit when KVM_ASSIGN_IRQ return -ENOTTY with
> KVM_DEV_IRQ_ASSIGN_ENABLE_MSI.
>   

Need a KVM_CAP_MSI so that userspace can tell before trying.

> +	if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_GUEST_INTX)
> +		kvm_set_irq(assigned_dev->kvm,
> +			    assigned_dev->irq_source_id,
> +			    assigned_dev->guest_irq, 1);
> +	else if (assigned_dev->irq_requested_type &
> +				KVM_ASSIGNED_DEV_GUEST_MSI) {
> +		assigned_device_msi_dispatch(assigned_dev);
> +		enable_irq(assigned_dev->host_irq);
> +	}
>   

Hmm.  This means the only path that can trigger msi is device 
assignment.  But we also want virtio-pci to use msi.

How about modifying kvm_set_irq() to dispatch either msi or normal 
irqs?  We'd need a 'struct kvm_guest_irq' if we can't define the msi 
using the bits available in gsi (and also define a new KVM_IRQ_LINE... I 
really hope we can fit the msi info in gsi).

The device-assignment-independent parts should be in a separate patch.

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