KVM: x86: Rework user space NMI injection as KVM_CAP_USER_NMI

There is no point in doing the ready_for_nmi_injection/
request_nmi_window dance with user space. First, we don't do this for
in-kernel irqchip anyway, while the code path is the same as for user
space irqchip mode. And second, there is nothing to loose if a pending
NMI is overwritten by another one (in contrast to IRQs where we have to
save the number). Actually, there is even the risk of raising spurious
NMIs this way because the reason for the held-back NMI might already be
handled while processing the first one.

Therefore this patch creates a simplified user space NMI injection
interface, exporting it under KVM_CAP_USER_NMI and dropping the old
KVM_CAP_NMI capability. And this time we also take care to provide the
interface only on archs supporting NMIs via KVM (right now only x86).
Signed-off-by: NJan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: NAvi Kivity <avi@redhat.com>

arch/x86/kvm/vmx.c

@@ -2498,15 +2498,13 @@ static void do_interrupt_requests(struct kvm_vcpu *vcpu,
 	}
 	if (vcpu->arch.nmi_injected) {
 		vmx_inject_nmi(vcpu);
-		if (vcpu->arch.nmi_pending || kvm_run->request_nmi_window)
+		if (vcpu->arch.nmi_pending)
 			enable_nmi_window(vcpu);
 		else if (vcpu->arch.irq_summary
 			 || kvm_run->request_interrupt_window)
 			enable_irq_window(vcpu);
 		return;
 	}
-	if (!vcpu->arch.nmi_window_open || kvm_run->request_nmi_window)
-		enable_nmi_window(vcpu);
 
 	if (vcpu->arch.interrupt_window_open) {
 		if (vcpu->arch.irq_summary && !vcpu->arch.interrupt.pending)
@@ -3040,14 +3038,6 @@ static int handle_nmi_window(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 	vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control);
 	++vcpu->stat.nmi_window_exits;
 
-	/*
-	 * If the user space waits to inject a NMI, exit as soon as possible
-	 */
-	if (kvm_run->request_nmi_window && !vcpu->arch.nmi_pending) {
-		kvm_run->exit_reason = KVM_EXIT_NMI_WINDOW_OPEN;
-		return 0;
-	}
-
 	return 1;
 }
 
@@ -3162,7 +3152,7 @@ static int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 			vmx->soft_vnmi_blocked = 0;
 			vcpu->arch.nmi_window_open = 1;
 		} else if (vmx->vnmi_blocked_time > 1000000000LL &&
-		    (kvm_run->request_nmi_window || vcpu->arch.nmi_pending)) {
+			   vcpu->arch.nmi_pending) {
 			/*
 			 * This CPU don't support us in finding the end of an
 			 * NMI-blocked window if the guest runs with IRQs
@@ -3175,16 +3165,6 @@ static int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 			vmx->soft_vnmi_blocked = 0;
 			vmx->vcpu.arch.nmi_window_open = 1;
 		}
-
-		/*
-		 * If the user space waits to inject an NNI, exit ASAP
-		 */
-		if (vcpu->arch.nmi_window_open && kvm_run->request_nmi_window
-		    && !vcpu->arch.nmi_pending) {
-			kvm_run->exit_reason = KVM_EXIT_NMI_WINDOW_OPEN;
-			++vcpu->stat.nmi_window_exits;
-			return 0;
-		}
 	}
 
 	if (exit_reason < kvm_vmx_max_exit_handlers

arch/x86/kvm/x86.c

@@ -2887,37 +2887,18 @@ static int dm_request_for_irq_injection(struct kvm_vcpu *vcpu,
 		(kvm_x86_ops->get_rflags(vcpu) & X86_EFLAGS_IF));
 }
 
-/*
- * Check if userspace requested a NMI window, and that the NMI window
- * is open.
- *
- * No need to exit to userspace if we already have a NMI queued.
- */
-static int dm_request_for_nmi_injection(struct kvm_vcpu *vcpu,
-					struct kvm_run *kvm_run)
-{
-	return (!vcpu->arch.nmi_pending &&
-		kvm_run->request_nmi_window &&
-		vcpu->arch.nmi_window_open);
-}
-
 static void post_kvm_run_save(struct kvm_vcpu *vcpu,
 			      struct kvm_run *kvm_run)
 {
 	kvm_run->if_flag = (kvm_x86_ops->get_rflags(vcpu) & X86_EFLAGS_IF) != 0;
 	kvm_run->cr8 = kvm_get_cr8(vcpu);
 	kvm_run->apic_base = kvm_get_apic_base(vcpu);
-	if (irqchip_in_kernel(vcpu->kvm)) {
+	if (irqchip_in_kernel(vcpu->kvm))
 		kvm_run->ready_for_interrupt_injection = 1;
-		kvm_run->ready_for_nmi_injection = 1;
-	} else {
+	else
 		kvm_run->ready_for_interrupt_injection =
 					(vcpu->arch.interrupt_window_open &&
 					 vcpu->arch.irq_summary == 0);
-		kvm_run->ready_for_nmi_injection =
-					(vcpu->arch.nmi_window_open &&
-					 vcpu->arch.nmi_pending == 0);
-	}
 }
 
 static void vapic_enter(struct kvm_vcpu *vcpu)
@@ -3093,11 +3074,6 @@ static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 		}
 
 		if (r > 0) {
-			if (dm_request_for_nmi_injection(vcpu, kvm_run)) {
-				r = -EINTR;
-				kvm_run->exit_reason = KVM_EXIT_NMI;
-				++vcpu->stat.request_nmi_exits;
-			}
 			if (dm_request_for_irq_injection(vcpu, kvm_run)) {
 				r = -EINTR;
 				kvm_run->exit_reason = KVM_EXIT_INTR;

include/linux/kvm.h

@@ -84,21 +84,18 @@ struct kvm_irqchip {
 #define KVM_EXIT_S390_RESET       14
 #define KVM_EXIT_DCR              15
 #define KVM_EXIT_NMI              16
-#define KVM_EXIT_NMI_WINDOW_OPEN  17
 
 /* for KVM_RUN, returned by mmap(vcpu_fd, offset=0) */
 struct kvm_run {
 	/* in */
 	__u8 request_interrupt_window;
-	__u8 request_nmi_window;
-	__u8 padding1[6];
+	__u8 padding1[7];
 
 	/* out */
 	__u32 exit_reason;
 	__u8 ready_for_interrupt_injection;
 	__u8 if_flag;
-	__u8 ready_for_nmi_injection;
-	__u8 padding2;
+	__u8 padding2[2];
 
 	/* in (pre_kvm_run), out (post_kvm_run) */
 	__u64 cr8;
@@ -391,12 +388,14 @@ struct kvm_trace_rec {
 #define KVM_CAP_DEVICE_ASSIGNMENT 17
 #endif
 #define KVM_CAP_IOMMU 18
-#define KVM_CAP_NMI 19
 #if defined(CONFIG_X86)
 #define KVM_CAP_DEVICE_MSI 20
 #endif
 /* Bug in KVM_SET_USER_MEMORY_REGION fixed: */
 #define KVM_CAP_DESTROY_MEMORY_REGION_WORKS 21
+#if defined(CONFIG_X86)
+#define KVM_CAP_USER_NMI 22
+#endif
 
 /*
  * ioctls for VM fds