[PATCH v2] arm/arm64: KVM: VGIC: limit ITARGETSR bits to number of VCPUs

[Marc Zyngier]

On 16/11/16 17:57, Andre Przywara wrote:
> The GICv2 spec says in section 4.3.12 that a "CPU targets field bit that
> corresponds to an unimplemented CPU interface is RAZ/WI."
> Currently we allow the guest to write any value in there and it can
> read that back.
> Mask the written value with the proper CPU mask to be spec compliant.
> 
> Signed-off-by: Andre Przywara <andre.przywara at arm.com>
> ---
> Changes v1 .. v2:
> - use GENMASK() instead of open-coding mask
> - drop explicit 0xff masking, since cpu_mask is stronger anyway
> 
>  virt/kvm/arm/vgic/vgic-mmio-v2.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-mmio-v2.c b/virt/kvm/arm/vgic/vgic-mmio-v2.c
> index b44b359..78e34bc 100644
> --- a/virt/kvm/arm/vgic/vgic-mmio-v2.c
> +++ b/virt/kvm/arm/vgic/vgic-mmio-v2.c
> @@ -129,6 +129,7 @@ static void vgic_mmio_write_target(struct kvm_vcpu *vcpu,
>  				   unsigned long val)
>  {
>  	u32 intid = VGIC_ADDR_TO_INTID(addr, 8);
> +	u8 cpu_mask = GENMASK(atomic_read(&vcpu->kvm->online_vcpus) - 1, 0);
>  	int i;
>  
>  	/* GICD_ITARGETSR[0-7] are read-only */
> @@ -141,7 +142,7 @@ static void vgic_mmio_write_target(struct kvm_vcpu *vcpu,
>  
>  		spin_lock(&irq->irq_lock);
>  
> -		irq->targets = (val >> (i * 8)) & 0xff;
> +		irq->targets = (val >> (i * 8)) & cpu_mask;
>  		target = irq->targets ? __ffs(irq->targets) : 0;
>  		irq->target_vcpu = kvm_get_vcpu(vcpu->kvm, target);
>  
> 

Applied, thanks.

	M.
-- 
Jazz is not dead. It just smells funny...