kvm: vmx: Reinstate support for CPUs without virtual NMI

This is more or less a revert of commit 2c82878b ("KVM: VMX: require
virtual NMI support", 2017-03-27); it turns out that Core 2 Duo machines
only had virtual NMIs in some SKUs.

The revert is not trivial because in the meanwhile there have been several
fixes to nested NMI injection.  Therefore, the entire vNMI state is moved
to struct loaded_vmcs.

Another change compared to before the patch is a simplification here:

       if (unlikely(!cpu_has_virtual_nmis() && vmx->soft_vnmi_blocked &&
           !(is_guest_mode(vcpu) && nested_cpu_has_virtual_nmis(
                                       get_vmcs12(vcpu))))) {

The final condition here is always true (because nested_cpu_has_virtual_nmis
is always false) and is removed.

Fixes: 2c82878b
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1490803
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>

arch/x86/kvm/vmx.c

@@ -202,6 +202,10 @@ struct loaded_vmcs {
 	bool nmi_known_unmasked;
 	unsigned long vmcs_host_cr3;	/* May not match real cr3 */
 	unsigned long vmcs_host_cr4;	/* May not match real cr4 */
+	/* Support for vnmi-less CPUs */
+	int soft_vnmi_blocked;
+	ktime_t entry_time;
+	s64 vnmi_blocked_time;
 	struct list_head loaded_vmcss_on_cpu_link;
 };
 
@@ -1291,6 +1295,11 @@ static inline bool cpu_has_vmx_invpcid(void)
 		SECONDARY_EXEC_ENABLE_INVPCID;
 }
 
+static inline bool cpu_has_virtual_nmis(void)
+{
+	return vmcs_config.pin_based_exec_ctrl & PIN_BASED_VIRTUAL_NMIS;
+}
+
 static inline bool cpu_has_vmx_wbinvd_exit(void)
 {
 	return vmcs_config.cpu_based_2nd_exec_ctrl &
@@ -1348,11 +1357,6 @@ static inline bool nested_cpu_has2(struct vmcs12 *vmcs12, u32 bit)
 		(vmcs12->secondary_vm_exec_control & bit);
 }
 
-static inline bool nested_cpu_has_virtual_nmis(struct vmcs12 *vmcs12)
-{
-	return vmcs12->pin_based_vm_exec_control & PIN_BASED_VIRTUAL_NMIS;
-}
-
 static inline bool nested_cpu_has_preemption_timer(struct vmcs12 *vmcs12)
 {
 	return vmcs12->pin_based_vm_exec_control &
@@ -3712,9 +3716,9 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf)
 				&_vmexit_control) < 0)
 		return -EIO;
 
-	min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING |
-		PIN_BASED_VIRTUAL_NMIS;
-	opt = PIN_BASED_POSTED_INTR | PIN_BASED_VMX_PREEMPTION_TIMER;
+	min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING;
+	opt = PIN_BASED_VIRTUAL_NMIS | PIN_BASED_POSTED_INTR |
+		 PIN_BASED_VMX_PREEMPTION_TIMER;
 	if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PINBASED_CTLS,
 				&_pin_based_exec_control) < 0)
 		return -EIO;
@@ -5666,7 +5670,8 @@ static void enable_irq_window(struct kvm_vcpu *vcpu)
 
 static void enable_nmi_window(struct kvm_vcpu *vcpu)
 {
-	if (vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) {
+	if (!cpu_has_virtual_nmis() ||
+	    vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) {
 		enable_irq_window(vcpu);
 		return;
 	}
@@ -5706,6 +5711,19 @@ static void vmx_inject_nmi(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 
+	if (!cpu_has_virtual_nmis()) {
+		/*
+		 * Tracking the NMI-blocked state in software is built upon
+		 * finding the next open IRQ window. This, in turn, depends on
+		 * well-behaving guests: They have to keep IRQs disabled at
+		 * least as long as the NMI handler runs. Otherwise we may
+		 * cause NMI nesting, maybe breaking the guest. But as this is
+		 * highly unlikely, we can live with the residual risk.
+		 */
+		vmx->loaded_vmcs->soft_vnmi_blocked = 1;
+		vmx->loaded_vmcs->vnmi_blocked_time = 0;
+	}
+
 	++vcpu->stat.nmi_injections;
 	vmx->loaded_vmcs->nmi_known_unmasked = false;
 
@@ -5724,6 +5742,8 @@ static bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu)
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	bool masked;
 
+	if (!cpu_has_virtual_nmis())
+		return vmx->loaded_vmcs->soft_vnmi_blocked;
 	if (vmx->loaded_vmcs->nmi_known_unmasked)
 		return false;
 	masked = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_NMI;
@@ -5735,6 +5755,12 @@ static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 
+	if (!cpu_has_virtual_nmis()) {
+		if (vmx->loaded_vmcs->soft_vnmi_blocked != masked) {
+			vmx->loaded_vmcs->soft_vnmi_blocked = masked;
+			vmx->loaded_vmcs->vnmi_blocked_time = 0;
+		}
+	} else {
 		vmx->loaded_vmcs->nmi_known_unmasked = !masked;
 		if (masked)
 			vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO,
@@ -5742,6 +5768,7 @@ static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked)
 		else
 			vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO,
 					GUEST_INTR_STATE_NMI);
+	}
 }
 
 static int vmx_nmi_allowed(struct kvm_vcpu *vcpu)
@@ -5749,6 +5776,10 @@ static int vmx_nmi_allowed(struct kvm_vcpu *vcpu)
 	if (to_vmx(vcpu)->nested.nested_run_pending)
 		return 0;
 
+	if (!cpu_has_virtual_nmis() &&
+	    to_vmx(vcpu)->loaded_vmcs->soft_vnmi_blocked)
+		return 0;
+
 	return	!(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) &
 		  (GUEST_INTR_STATE_MOV_SS | GUEST_INTR_STATE_STI
 		   | GUEST_INTR_STATE_NMI));
@@ -6476,6 +6507,7 @@ static int handle_ept_violation(struct kvm_vcpu *vcpu)
 	 * AAK134, BY25.
 	 */
 	if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) &&
+			cpu_has_virtual_nmis() &&
 			(exit_qualification & INTR_INFO_UNBLOCK_NMI))
 		vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, GUEST_INTR_STATE_NMI);
 
@@ -6962,7 +6994,7 @@ static struct loaded_vmcs *nested_get_current_vmcs02(struct vcpu_vmx *vmx)
 	}
 
 	/* Create a new VMCS */
-	item = kmalloc(sizeof(struct vmcs02_list), GFP_KERNEL);
+	item = kzalloc(sizeof(struct vmcs02_list), GFP_KERNEL);
 	if (!item)
 		return NULL;
 	item->vmcs02.vmcs = alloc_vmcs();
@@ -7979,6 +8011,7 @@ static int handle_pml_full(struct kvm_vcpu *vcpu)
 	 * "blocked by NMI" bit has to be set before next VM entry.
 	 */
 	if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) &&
+			cpu_has_virtual_nmis() &&
 			(exit_qualification & INTR_INFO_UNBLOCK_NMI))
 		vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO,
 				GUEST_INTR_STATE_NMI);
@@ -8823,6 +8856,25 @@ static int vmx_handle_exit(struct kvm_vcpu *vcpu)
 		return 0;
 	}
 
+	if (unlikely(!cpu_has_virtual_nmis() &&
+		     vmx->loaded_vmcs->soft_vnmi_blocked)) {
+		if (vmx_interrupt_allowed(vcpu)) {
+			vmx->loaded_vmcs->soft_vnmi_blocked = 0;
+		} else if (vmx->loaded_vmcs->vnmi_blocked_time > 1000000000LL &&
+			   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
+			 * disabled. So we pull the trigger after 1 s of
+			 * futile waiting, but inform the user about this.
+			 */
+			printk(KERN_WARNING "%s: Breaking out of NMI-blocked "
+			       "state on VCPU %d after 1 s timeout\n",
+			       __func__, vcpu->vcpu_id);
+			vmx->loaded_vmcs->soft_vnmi_blocked = 0;
+		}
+	}
+
 	if (exit_reason < kvm_vmx_max_exit_handlers
 	    && kvm_vmx_exit_handlers[exit_reason])
 		return kvm_vmx_exit_handlers[exit_reason](vcpu);
@@ -9105,6 +9157,7 @@ static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx)
 
 	idtv_info_valid = vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK;
 
+	if (cpu_has_virtual_nmis()) {
 		if (vmx->loaded_vmcs->nmi_known_unmasked)
 			return;
 		/*
@@ -9132,6 +9185,10 @@ static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx)
 			vmx->loaded_vmcs->nmi_known_unmasked =
 				!(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO)
 				  & GUEST_INTR_STATE_NMI);
+	} else if (unlikely(vmx->loaded_vmcs->soft_vnmi_blocked))
+		vmx->loaded_vmcs->vnmi_blocked_time +=
+			ktime_to_ns(ktime_sub(ktime_get(),
+					      vmx->loaded_vmcs->entry_time));
 }
 
 static void __vmx_complete_interrupts(struct kvm_vcpu *vcpu,
@@ -9248,6 +9305,11 @@ static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu)
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	unsigned long debugctlmsr, cr3, cr4;
 
+	/* Record the guest's net vcpu time for enforced NMI injections. */
+	if (unlikely(!cpu_has_virtual_nmis() &&
+		     vmx->loaded_vmcs->soft_vnmi_blocked))
+		vmx->loaded_vmcs->entry_time = ktime_get();
+
 	/* Don't enter VMX if guest state is invalid, let the exit handler
 	   start emulation until we arrive back to a valid state */
 	if (vmx->emulation_required)