KVM: x86: Don't update RIP or do single-step on faulting emulation

Don't advance RIP or inject a single-step #DB if emulation signals a
fault.  This logic applies to all state updates that are conditional on
clean retirement of the emulation instruction, e.g. updating RFLAGS was
previously handled by commit 38827db ("KVM: x86: Do not update
EFLAGS on faulting emulation").

Not advancing RIP is likely a nop, i.e. ctxt->eip isn't updated with
ctxt->_eip until emulation "retires" anyways.  Skipping #DB injection
fixes a bug reported by Andy Lutomirski where a #UD on SYSCALL due to
invalid state with EFLAGS.TF=1 would loop indefinitely due to emulation
overwriting the #UD with #DB and thus restarting the bad SYSCALL over
and over.

Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: stable@vger.kernel.org
Reported-by: Andy Lutomirski <luto@kernel.org>
Fixes: 663f4c6 ("KVM: x86: handle singlestep during emulation")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>

arch/x86/kvm/x86.c

@@ -6594,12 +6594,13 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu,
 		unsigned long rflags = kvm_x86_ops->get_rflags(vcpu);
 		toggle_interruptibility(vcpu, ctxt->interruptibility);
 		vcpu->arch.emulate_regs_need_sync_to_vcpu = false;
-		kvm_rip_write(vcpu, ctxt->eip);
-		if (r == EMULATE_DONE && ctxt->tf)
-			kvm_vcpu_do_singlestep(vcpu, &r);
 		if (!ctxt->have_exception ||
-		    exception_type(ctxt->exception.vector) == EXCPT_TRAP)
+		    exception_type(ctxt->exception.vector) == EXCPT_TRAP) {
+			kvm_rip_write(vcpu, ctxt->eip);
+			if (r == EMULATE_DONE && ctxt->tf)
+				kvm_vcpu_do_singlestep(vcpu, &r);
 			__kvm_set_rflags(vcpu, ctxt->eflags);
+		}
 
 		/*
 		 * For STI, interrupts are shadowed; so KVM_REQ_EVENT will