On Mon, Nov 17, 2014 at 10:09:58AM +0000, Marc Zyngier wrote:
> On 17/11/14 09:27, wanghaibin wrote:
> > Maybe inject irq action is not only from userspace(like arch_timer or
> > another in-kernel device in future). So, it maybe better that guarantee
> > the parameter valid before inject irq action.
>
> First, we don't defend ourselves from other in-kernel callers. If they
> wrong, we fix them.
>
> > Signed-off-by: wanghaibin <wanghaibin.wang@xxxxxxxxxx>
> > ---
> > virt/kvm/arm/vgic.c | 10 ++++++++++
> > 1 files changed, 10 insertions(+), 0 deletions(-)
> >
> > diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
> > index 12026e6..52f3bfa 100644
> > --- a/virt/kvm/arm/vgic.c
> > +++ b/virt/kvm/arm/vgic.c
> > @@ -1621,10 +1621,20 @@ static bool vgic_update_irq_pending(struct kvm *kvm, int cpuid,
> > {
> > struct vgic_dist *dist = &kvm->arch.vgic;
> > struct kvm_vcpu *vcpu;
> > + int nrcpus = atomic_read(&kvm->online_vcpus);
> > int edge_triggered, level_triggered;
> > int enabled;
> > bool ret = true;
> >
> > + if ((cpuid >= nrcpus) || (irq_num >= dist->nr_irqs))
> > + return false;
> > +
> > + if (irq_num < VGIC_NR_SGIS)
> > + return false;
> > +
> > + if ((irq_num >= VGIC_NR_PRIVATE_IRQS) && (cpuid != 0))
> > + return false;
> > +
> > spin_lock(&dist->lock);
> >
> > vcpu = kvm_get_vcpu(kvm, cpuid);
> >
>
> The only check that could be useful is the one against dist->nr_irqs.
> The rest of it is completely useless.
>
I think the right thin to do is to change the check in
arch/arm/kvm/arm.c:kvm_vm_ioctl_irq_line() to check against
dist->nr_irqs for KVM_ARM_IRQ_TYPE_SPI and return an error in that case.
Wanghaibin, can you send such a patch instead of this one?
-Christoffer
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
