On Wed, Dec 07, 2011 at 12:33:05PM +0200, Gleb Natapov wrote: > On Wed, Dec 07, 2011 at 02:31:11PM +0400, Cyrill Gorcunov wrote: > > On Wed, Dec 07, 2011 at 12:21:52PM +0200, Gleb Natapov wrote: > > > On Tue, Dec 06, 2011 at 10:42:55PM +0200, Sasha Levin wrote: > > > > +static void handle_nmi(int fd, u32 type, u32 len, u8 *msg) > > > > +{ > > > > + u32 vcpu = *(u32 *)msg; > > > > + > > > > + ioctl(kvm_cpus[vcpu]->vcpu_fd, KVM_NMI); > > > > > > You need to check that vcpu apic's LINT1 is configured to receive > > > NMI (and not masked obviously) before injecting NMI. > > > > > > > I've been configuring mptable to have lint1 as nmi receiver, > > so it should remain so I suppose (if only we've not masked it > > somewhere else ;) > > > > That's up to the guest. mptable is just a hint to an OS on how things is > wired in HW. > Yup, but while we support linux kernels only it should be fine. Still of course on long term we need a check. Cyrill -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html