Re: [PATCH] KVM: x86: don't hold kvm->lock in KVM_SET_GSI_ROUTING

["Radim Krčmář" <rkrcmar@redhat.com>]

2017-04-26 17:21+0200, David Hildenbrand:
> On 26.04.2017 16:27, Radim Krčmář wrote:
> > 2017-04-26 10:40+0200, David Hildenbrand:
> >> On 25.04.2017 21:59, Radim Krčmář wrote:
> >>> 2017-04-25 21:03+0200, David Hildenbrand:
> >>>> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> >>>> @@ -504,6 +504,7 @@ void vcpu_put(struct kvm_vcpu *vcpu);
> >>>>  #ifdef __KVM_HAVE_IOAPIC
> >>>>  void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm);
> >>>>  void kvm_arch_post_irq_routing_update(struct kvm *kvm);
> >>>> +bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
> >>>
> >>> (A nitpick: it might be useful even without __KVM_HAVE_IOAPIC so weak
> >>>  linking would probably be cleaner for a slow path.)
> >>>
> >>>>  #else
> >>>>  static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
> >>>>  {
> >>>> @@ -511,6 +512,10 @@ static inline void kvm_arch_post_irq_ack_notifier_list_update(struct kvm *kvm)
> >>>>  static inline void kvm_arch_post_irq_routing_update(struct kvm *kvm)
> >>>>  {
> >>>>  }
> >>>> +static bool kvm_arch_can_set_irq_routing(struct kvm *kvm)
> >>>> +{
> >>>> +	return true;
> >>>> +}
> >>>>  #endif
> >>>>  
> >>
> >> Makes sense, shall I resend or can you fix that up when applying?
> > 
> > Please send a v2, it's going to be automatically compile-tested on all
> > arches that way. :)
> > 
> 
> Using a weak symbol for the first time... Is the following what you had in mind?
> 
> ---- snip ----
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 397b7b5..b725d4e 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -1009,6 +1009,7 @@ static inline int mmu_notifier_retry(struct kvm *kvm, unsigned long mmu_seq)
>  #define KVM_MAX_IRQ_ROUTES 1024
>  #endif
>  
> +bool kvm_arch_can_set_irq_routing(struct kvm *kvm);
>  int kvm_set_irq_routing(struct kvm *kvm,
>                         const struct kvm_irq_routing_entry *entries,
>                         unsigned nr,
> diff --git a/virt/kvm/irqchip.c b/virt/kvm/irqchip.c
> index cc30d01..a080659 100644
> --- a/virt/kvm/irqchip.c
> +++ b/virt/kvm/irqchip.c
> @@ -172,6 +172,11 @@ void __attribute__((weak)) kvm_arch_irq_routing_update(struct kvm *kvm)
>  {
>  }
>  
> +bool __attribute__((weak)) kvm_arch_can_set_irq_routing(struct kvm *kvm)

Yes, exactly what I had in mind.
(I'd just write "__weak" to make it shorter. :])

> +{
> +       return true;
> +}
> +
>  int kvm_set_irq_routing(struct kvm *kvm,
>                         const struct kvm_irq_routing_entry *ue,
>                         unsigned nr,
> ---- snip ----
> 
> -- 
> 
> Thanks,
> 
> David