KVM: x86: Don't report guest userspace emulation error to userspace
authorNadav Amit <namit@cs.technion.ac.il>
Tue, 16 Sep 2014 23:50:50 +0000 (02:50 +0300)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 24 Sep 2014 12:07:49 +0000 (14:07 +0200)
Commit fc3a9157d314 ("KVM: X86: Don't report L2 emulation failures to
user-space") disabled the reporting of L2 (nested guest) emulation failures to
userspace due to race-condition between a vmexit and the instruction emulator.
The same rational applies also to userspace applications that are permitted by
the guest OS to access MMIO area or perform PIO.

This patch extends the current behavior - of injecting a #UD instead of
reporting it to userspace - also for guest userspace code.

Signed-off-by: Nadav Amit <namit@cs.technion.ac.il>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/x86.c

index 2d7f65daa8d0983538c9fe6d577fcc06937e76c8..e46da50e2cec51d56cbff2199c93a8656798e5cd 100644 (file)
@@ -5000,7 +5000,7 @@ static int handle_emulation_failure(struct kvm_vcpu *vcpu)
 
        ++vcpu->stat.insn_emulation_fail;
        trace_kvm_emulate_insn_failed(vcpu);
-       if (!is_guest_mode(vcpu)) {
+       if (!is_guest_mode(vcpu) && kvm_x86_ops->get_cpl(vcpu) == 0) {
                vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR;
                vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION;
                vcpu->run->internal.ndata = 0;