2017-11-21 17:30+0200, Liran Alon: > An alternative could have been done to return -EBUSY in this case. > For now, we decided to just silently override exception and warn on > such an attempt. > > Signed-off-by: Liran Alon <liran.alon@xxxxxxxxxx> > Reviewed-by: Nikita Leshenko <nikita.leshchenko@xxxxxxxxxx> > Reviewed-by: Krish Sadhukhan <krish.sadhukhan@xxxxxxxxxx> > Signed-off-by: Krish Sadhukhan <krish.sadhukhan@xxxxxxxxxx> > --- > arch/x86/kvm/x86.c | 13 +++++++++++++ > 1 file changed, 13 insertions(+) > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c > index 1490da89de4b..c8cec7c39c1c 100644 > --- a/arch/x86/kvm/x86.c > +++ b/arch/x86/kvm/x86.c > @@ -3153,12 +3153,25 @@ static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu, > return -EINVAL; > > process_nmi(vcpu); > + > + /* > + * Warn if userspace is overriding existing > + * injected exception > + */ > + WARN_ON_ONCE(vcpu->arch.exception.injected && > + events->exception.injected); I think that overwriting the injected exception/interrupt is a perfectly valid operation -- userspace could have rolled back the state to a time of the previous injection. Syzkaller would complain sooner or later and I don't see it as a useful printk, so dropping this patch would be preferred, thanks.
