Re: [PATCH v2 5/6] KVM: x86: prevent setup of invalid routes

From: Paolo Bonzini
Date: Tue Jan 10 2017 - 04:43:21 EST




On 10/01/2017 06:26, Peter Xu wrote:
> On Fri, Dec 16, 2016 at 04:10:05PM +0100, Radim KrÄmÃÅ wrote:
>> The check in kvm_set_pic_irq() and kvm_set_ioapic_irq() was just a
>> temporary measure until the code improved enough for us to do this.
>>
>> This changes APIC in a case when KVM_SET_GSI_ROUTING is called to set up pic
>> and ioapic routes before KVM_CREATE_IRQCHIP. Those rules would get overwritten
>> by KVM_CREATE_IRQCHIP at best, so it is pointless to allow it. Userspaces
>> hopefully noticed that things don't work if they do that and don't do that.
>>
>> Reviewed-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>> Signed-off-by: Radim KrÄmÃÅ <rkrcmar@xxxxxxxxxx>
>
> Since we are at here, do we need to protect KVM_SET_GSI_ROUTING in
> general as well to make sure kernel APIC is there?

Skipping the check is harmless. I agree that it should have been always
done like you suggest, but right now it may break something.

Paolo

>
> ---------8<---------
>
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index 482612b..31141a7 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -3057,6 +3057,10 @@ static long kvm_vm_ioctl(struct file *filp,
> struct kvm_irq_routing_entry *entries = NULL;
>
> r = -EFAULT;
> +
> + if (!irqchip_in_kernel(kvm))
> + goto out;
> +
> if (copy_from_user(&routing, argp, sizeof(routing)))
> goto out;
> r = -EINVAL;
>
> --------->8---------
>
> Thanks,
>
> -- peterx
>