Skip to contentSkip to conversation

KVM: Only use posted interrupts for Fixed/LowPrio MSIs

PATCH Superseded Alexander Graf 2 participants 8 messages Sep 4, 2019

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover KVM: Only use posted interrupts for Fixed/LowPrio MSIs

The MSI-X descriptor has a "delivery mode" field which can be set to various different targets, such as "Fixed" (default), SMI, NMI or INIT.

Usually when we pass devices into guests, we only ever see this MSI-X descriptor configured as Fixed, so nobody realized that the other modes were broken when using posted interrupts.

With posted interrupts, we end up configuring these special modes just the same as a Fixed interrupt. That means instead of generating an SMI, we inject a normal GSI into the guest.

Of course, that if completely broken. These two patches attempt to fix the situation for x86 systems. If anyone has a great idea how to generalize the filtering though, I'm all ears.

Alex

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

Alexander Graf (2):
KVM: VMX: Disable posted interrupts for odd IRQs
KVM: SVM: Disable posted interrupts for odd IRQs

 arch/x86/kvm/svm.c     | 16 ++++++++++++++++
 arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
 2 files changed, 38 insertions(+)
Signature · 12 lines
2.17.1




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
#1 lore raw

PATCH v2 1/2 KVM: VMX: Disable posted interrupts for odd IRQs

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the VMX PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
 arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
arch/x86/kvm/vmx/vmx.c +22 -0
index 570a233e272b..8029fe658c30 100644
@@ -7401,6 +7401,28 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 			continue;
 		}
 
+		switch (irq.delivery_mode) {
+		case dest_Fixed:
+		case dest_LowestPrio:
+			break;
+		default:
+			/*
+			 * For non-trivial interrupt events, we need to go
+			 * through the full KVM IRQ code, so refuse to take
+			 * any direct PI assignments here.
+			 */
+
+			ret = irq_set_vcpu_affinity(host_irq, NULL);
+			if (ret < 0) {
+				printk(KERN_INFO
+				    "non-std IRQ failed to recover, irq: %u\n",
+				    host_irq);
+				goto out;
+			}
+
+			continue;
+		}
+
 		vcpu_info.pi_desc_addr = __pa(vcpu_to_pi_desc(vcpu));
 		vcpu_info.vector = irq.vector;
 
Signature · 12 lines
2.17.1




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
#2 lore raw

Re: [PATCH v2 1/2] KVM: VMX: Disable posted interrupts for odd IRQs

Quoted 36 lines · Alexander Graf

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the VMX PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
 arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 570a233e272b..8029fe658c30 100644
@@ -7401,6 +7401,28 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 			continue;
 		}
 

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7413

Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +10 -0
+		switch (irq.delivery_mode) {
+		case dest_Fixed:
+		case dest_LowestPrio:
+			break;
+		default:
+			/*
+			 * For non-trivial interrupt events, we need to go
+			 * through the full KVM IRQ code, so refuse to take
+			 * any direct PI assignments here.
+			 */

IMO, a beefy comment is unnecessary, anyone that is digging through this code has hopefully read the PI spec or at least understands the basic concepts. I.e. it should be obvious that PI can't be used for SMI, etc...

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7422

Quoted 9 lines
arch/x86/kvm/vmx/vmx.c +9 -0
+			ret = irq_set_vcpu_affinity(host_irq, NULL);
+			if (ret < 0) {
+				printk(KERN_INFO
+				    "non-std IRQ failed to recover, irq: %u\n",
+				    host_irq);
+				goto out;
+			}
+
+			continue;

Using a switch to filter out two types is a bit of overkill. It also probably makes sense to perform the deliver_mode checks before calling kvm_intr_is_single_vcpu(). Why not simply something like this? The existing comment and error message are even generic enough to keep as is.

arch/x86/kvm/vmx/vmx.c +3 -1
index c030c96fc81a..e0111a271a5e 100644
@@ -7372,7 +7372,9 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
                 */
 
                kvm_set_msi_irq(kvm, e, &irq);
-               if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+               if ((irq.delivery_mode != dest_Fixed &&
+                    irq.delivery_mode != dest_LowestPrio) ||
+                   !kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
                        /*
                         * Make sure the IRTE is in remapped mode if
                         * we don't handle it in posted mode.
Quoted 21 lines
arch/x86/kvm/vmx/vmx.c +2 -0
+		}
+
 		vcpu_info.pi_desc_addr = __pa(vcpu_to_pi_desc(vcpu));
 		vcpu_info.vector = irq.vector;
 
Signature · 12 lines
2.17.1




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
#3 lore raw

Re: [PATCH v2 1/2] KVM: VMX: Disable posted interrupts for odd IRQs

Quoted 52 lines · Sean Christopherson
Quoted 36 lines · Alexander Graf

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the VMX PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
  arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
  1 file changed, 22 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 570a233e272b..8029fe658c30 100644
@@ -7401,6 +7401,28 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
  			continue;
  		}
 

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7413

Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +10 -0
+		switch (irq.delivery_mode) {
+		case dest_Fixed:
+		case dest_LowestPrio:
+			break;
+		default:
+			/*
+			 * For non-trivial interrupt events, we need to go
+			 * through the full KVM IRQ code, so refuse to take
+			 * any direct PI assignments here.
+			 */

IMO, a beefy comment is unnecessary, anyone that is digging through this code has hopefully read the PI spec or at least understands the basic concepts. I.e. it should be obvious that PI can't be used for SMI, etc...

Quoted 11 lines

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7422

Quoted 9 lines
arch/x86/kvm/vmx/vmx.c +9 -0
+			ret = irq_set_vcpu_affinity(host_irq, NULL);
+			if (ret < 0) {
+				printk(KERN_INFO
+				    "non-std IRQ failed to recover, irq: %u\n",
+				    host_irq);
+				goto out;
+			}
+
+			continue;

Using a switch to filter out two types is a bit of overkill. It also

The switch should compile into the same as the if() below, it's just a matter of being more verbose in code.

Quoted 3 lines

probably makes sense to perform the deliver_mode checks before calling kvm_intr_is_single_vcpu(). Why not simply something like this? The existing comment and error message are even generic enough to keep as is.

Ok, so how about this, even though it goes against Liran's comment on the combined debug print?

If you think it's reasonable despite the broken formatting, I'll be happy to fold the patches and submit as v3.

Alex

arch/x86/include/asm/kvm_host.h +0 -0
index 44a5ce57a905..55f68fb0d791 100644
@@ -1581,6 +1581,12 @@ bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct kvm_lapic_irq *irq,
 void kvm_set_msi_irq(struct kvm *kvm, struct 

kvm_kernel_irq_routing_entry *e,
struct kvm_lapic_irq *irq);

+static inline bool kvm_irq_is_generic(struct kvm_lapic_irq *irq) +{
+ return (irq->delivery_mode == dest_Fixed ||
+ irq->delivery_mode == dest_LowestPrio);
+}
+
static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) {
if (kvm_x86_ops->vcpu_blocking)

arch/x86/kvm/svm.c +3 -1
index 1f220a85514f..34cc59518cbb 100644
@@ -5260,7 +5260,8 @@ get_pi_vcpu_info(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
 
 	kvm_set_msi_irq(kvm, e, &irq);
 
-	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+	    !kvm_irq_is_generic(&irq)) {
 		pr_debug("SVM: %s: use legacy intr remap mode for irq %u\n",
 			 __func__, irq.vector);
 		return -1;
@@ -5314,6 +5315,7 @@ static int svm_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 		 * 1. When cannot target interrupt to a specific vcpu.
 		 * 2. Unsetting posted interrupt.
 		 * 3. APIC virtialization is disabled for the vcpu.
+		 * 4. IRQ has extended delivery mode (SMI, INIT, etc)
 		 */
 		if (!get_pi_vcpu_info(kvm, e, &vcpu_info, &svm) && set &&
 		    kvm_vcpu_apicv_active(&svm->vcpu)) {
arch/x86/kvm/vmx/vmx.c +5 -1
index 570a233e272b..69f53809c7bb 100644
@@ -7382,10 +7382,14 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 		 * irqbalance to make the interrupts single-CPU.
 		 *
 		 * We will support full lowest-priority interrupt later.
+		 *
+		 * In addition, we can only inject generic interrupts using
+		 * the PI mechanism, refuse to route others through it.
 		 */
 
 		kvm_set_msi_irq(kvm, e, &irq);
-		if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+		if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+		    !kvm_irq_is_generic(&irq)) {
 			/*
 			 * Make sure the IRTE is in remapped mode if
 			 * we don't handle it in posted mode.

Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879

#4 lore raw

Re: [PATCH v2 1/2] KVM: VMX: Disable posted interrupts for odd IRQs

Quoted 69 lines · Alexander Graf
Quoted 51 lines · Sean Christopherson
Quoted 35 lines · Alexander Graf

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the VMX PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
 arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 570a233e272b..8029fe658c30 100644
@@ -7401,6 +7401,28 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 			continue;
 		}

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7412

Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +10 -0
+		switch (irq.delivery_mode) {
+		case dest_Fixed:
+		case dest_LowestPrio:
+			break;
+		default:
+			/*
+			 * For non-trivial interrupt events, we need to go
+			 * through the full KVM IRQ code, so refuse to take
+			 * any direct PI assignments here.
+			 */

IMO, a beefy comment is unnecessary, anyone that is digging through this code has hopefully read the PI spec or at least understands the basic concepts. I.e. it should be obvious that PI can't be used for SMI, etc...

Quoted 11 lines

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7421

Quoted 9 lines
arch/x86/kvm/vmx/vmx.c +9 -0
+			ret = irq_set_vcpu_affinity(host_irq, NULL);
+			if (ret < 0) {
+				printk(KERN_INFO
+				    "non-std IRQ failed to recover, irq: %u\n",
+				    host_irq);
+				goto out;
+			}
+
+			continue;

Using a switch to filter out two types is a bit of overkill. It also

The switch should compile into the same as the if() below, it's just a matter of being more verbose in code.

Quoted 6 lines
Quoted 3 lines

probably makes sense to perform the deliver_mode checks before calling kvm_intr_is_single_vcpu(). Why not simply something like this? The existing comment and error message are even generic enough to keep as is.

Ok, so how about this, even though it goes against Liran's comment on the combined debug print?

I missed that comment.

How often do we expect irq_set_vcpu_affinity() to fail? If it's frequent enough that the debug message matters, maybe it should be a tracepoint.

Quoted 32 lines

If you think it's reasonable despite the broken formatting, I'll be happy to fold the patches and submit as v3.

Alex

arch/x86/include/asm/kvm_host.h +6 -0
index 44a5ce57a905..55f68fb0d791 100644
@@ -1581,6 +1581,12 @@ bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct
 kvm_lapic_irq *irq,
 void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry
 *e,
 		     struct kvm_lapic_irq *irq);
 
+static inline bool kvm_irq_is_generic(struct kvm_lapic_irq *irq)
+{
+	return (irq->delivery_mode == dest_Fixed ||
+		irq->delivery_mode == dest_LowestPrio);
+}
+
 static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)

{

arch/x86/include/asm/kvm_host.h +0 -0
 	if (kvm_x86_ops->vcpu_blocking)
arch/x86/kvm/svm.c +0 -0
index 1f220a85514f..34cc59518cbb 100644

arch/x86/kvm/svm.c get_pi_vcpu_info(struct kvm *kvm, struct · line 5265

Quoted 7 lines
arch/x86/kvm/svm.c +2 -1
@@ -5260,7 +5260,8 @@ get_pi_vcpu_info(struct kvm *kvm, struct
 kvm_kernel_irq_routing_entry *e,
 
 	kvm_set_msi_irq(kvm, e, &irq);
 
-	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+	    !kvm_irq_is_generic(&irq)) {

I've never heard/seen the term generic used to describe x86 interrupts. Maybe kvm_irq_is_intr() or kvm_irq_is_vectored_intr()?

arch/x86/kvm/svm.c static int svm_update_pi_irte(struct kvm *kvm, · line 5319

Quoted 9 lines
arch/x86/kvm/svm.c +0 -0
 		pr_debug("SVM: %s: use legacy intr remap mode for irq %u\n",
 			 __func__, irq.vector);
arch/x86/kvm/svm.c +1 -0
 		return -1;
@@ -5314,6 +5315,7 @@ static int svm_update_pi_irte(struct kvm *kvm,
 unsigned int host_irq,
 		 * 1. When cannot target interrupt to a specific vcpu.
 		 * 2. Unsetting posted interrupt.
 		 * 3. APIC virtialization is disabled for the vcpu.
+		 * 4. IRQ has extended delivery mode (SMI, INIT, etc)

Similarly, 'extended delivery mode' isn't really a thing, it's simply the delivery mode.

4. IRQ is not a vectored interrupt.

Quoted 37 lines
arch/x86/kvm/svm.c +0 -0
 		 */
 		if (!get_pi_vcpu_info(kvm, e, &vcpu_info, &svm) && set &&
arch/x86/kvm/svm.c +0 -0
 		    kvm_vcpu_apicv_active(&svm->vcpu)) {
arch/x86/kvm/vmx/vmx.c +5 -1
index 570a233e272b..69f53809c7bb 100644
@@ -7382,10 +7382,14 @@ static int vmx_update_pi_irte(struct kvm *kvm,
 unsigned int host_irq,
 		 * irqbalance to make the interrupts single-CPU.
 		 *
 		 * We will support full lowest-priority interrupt later.
+		 *
+		 * In addition, we can only inject generic interrupts using
+		 * the PI mechanism, refuse to route others through it.
 		 */
 
 		kvm_set_msi_irq(kvm, e, &irq);
-		if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+		if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+		    !kvm_irq_is_generic(&irq)) {
 			/*
 			 * Make sure the IRTE is in remapped mode if

* we don't handle it in posted mode.

Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879

#5 lore raw

Re: [PATCH v2 1/2] KVM: VMX: Disable posted interrupts for odd IRQs

Quoted 74 lines · Sean Christopherson
Quoted 73 lines · Alexander Graf
Quoted 51 lines · Sean Christopherson
Quoted 35 lines · Alexander Graf

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the VMX PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
  arch/x86/kvm/vmx/vmx.c | 22 ++++++++++++++++++++++
  1 file changed, 22 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 570a233e272b..8029fe658c30 100644
@@ -7401,6 +7401,28 @@ static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
  			continue;
  		}

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7412

Quoted 10 lines
arch/x86/kvm/vmx/vmx.c +10 -0
+		switch (irq.delivery_mode) {
+		case dest_Fixed:
+		case dest_LowestPrio:
+			break;
+		default:
+			/*
+			 * For non-trivial interrupt events, we need to go
+			 * through the full KVM IRQ code, so refuse to take
+			 * any direct PI assignments here.
+			 */

IMO, a beefy comment is unnecessary, anyone that is digging through this code has hopefully read the PI spec or at least understands the basic concepts. I.e. it should be obvious that PI can't be used for SMI, etc...

Quoted 11 lines

arch/x86/kvm/vmx/vmx.c static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, · line 7421

Quoted 9 lines
arch/x86/kvm/vmx/vmx.c +9 -0
+			ret = irq_set_vcpu_affinity(host_irq, NULL);
+			if (ret < 0) {
+				printk(KERN_INFO
+				    "non-std IRQ failed to recover, irq: %u\n",
+				    host_irq);
+				goto out;
+			}
+
+			continue;

Using a switch to filter out two types is a bit of overkill. It also

The switch should compile into the same as the if() below, it's just a matter of being more verbose in code.

Quoted 3 lines

probably makes sense to perform the deliver_mode checks before calling kvm_intr_is_single_vcpu(). Why not simply something like this? The existing comment and error message are even generic enough to keep as is.

Quoted 7 lines
Quoted 2 lines

Ok, so how about this, even though it goes against Liran's comment on the combined debug print?

I missed that comment.

How often do we expect irq_set_vcpu_affinity() to fail? If it's frequent enough that the debug message matters, maybe it should be a tracepoint.

I don't expect to ever hit that debug print, so I don't think it matters really.

Quoted 35 lines
Quoted 35 lines

If you think it's reasonable despite the broken formatting, I'll be happy to fold the patches and submit as v3.

Alex

arch/x86/include/asm/kvm_host.h +6 -0
index 44a5ce57a905..55f68fb0d791 100644
@@ -1581,6 +1581,12 @@ bool kvm_intr_is_single_vcpu(struct kvm *kvm, struct
 kvm_lapic_irq *irq,
  void kvm_set_msi_irq(struct kvm *kvm, struct kvm_kernel_irq_routing_entry
 *e,
  		     struct kvm_lapic_irq *irq);
 
+static inline bool kvm_irq_is_generic(struct kvm_lapic_irq *irq)
+{
+	return (irq->delivery_mode == dest_Fixed ||
+		irq->delivery_mode == dest_LowestPrio);
+}
+
  static inline void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu)
  {
  	if (kvm_x86_ops->vcpu_blocking)
arch/x86/kvm/svm.c +0 -0
index 1f220a85514f..34cc59518cbb 100644
@@ -5260,7 +5260,8 @@ get_pi_vcpu_info(struct kvm *kvm, struct
 kvm_kernel_irq_routing_entry *e,
 
Quoted 8 lines

arch/x86/kvm/svm.c get_pi_vcpu_info(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e, · line 5264

Quoted 5 lines
arch/x86/kvm/svm.c +2 -1
  	kvm_set_msi_irq(kvm, e, &irq);
 
-	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+	    !kvm_irq_is_generic(&irq)) {

I've never heard/seen the term generic used to describe x86 interrupts. Maybe kvm_irq_is_intr() or kvm_irq_is_vectored_intr()?

I was trying to come up with any name that describes "interrupt that we can post". If "intr" is that, I'll be happy to take it. Vectored_intr sounds even worse IMHO :).

Quoted 13 lines

arch/x86/kvm/svm.c static int svm_update_pi_irte(struct kvm *kvm, · line 5319

Quoted 10 lines
arch/x86/kvm/svm.c +0 -0
  		pr_debug("SVM: %s: use legacy intr remap mode for irq %u\n",
  			 __func__, irq.vector);

return -1;

arch/x86/kvm/svm.c +1 -0
@@ -5314,6 +5315,7 @@ static int svm_update_pi_irte(struct kvm *kvm,
 unsigned int host_irq,
  		 * 1. When cannot target interrupt to a specific vcpu.
  		 * 2. Unsetting posted interrupt.
  		 * 3. APIC virtialization is disabled for the vcpu.
+		 * 4. IRQ has extended delivery mode (SMI, INIT, etc)

Similarly, 'extended delivery mode' isn't really a thing, it's simply the delivery mode.

s/extended/incompatible/ maybe?

Alex

Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879

#6 lore raw

Re: [PATCH v2 1/2] KVM: VMX: Disable posted interrupts for odd IRQs

Quoted 11 lines · Alexander Graf
Quoted 8 lines · Sean Christopherson

arch/x86/kvm/svm.c get_pi_vcpu_info(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e, · line 5264

Quoted 4 lines · Alexander Graf
arch/x86/kvm/svm.c +2 -1
-	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu)) {
+	if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) ||
+	    !kvm_irq_is_generic(&irq)) {

I've never heard/seen the term generic used to describe x86 interrupts. Maybe kvm_irq_is_intr() or kvm_irq_is_vectored_intr()?

Quoted 3 lines

I was trying to come up with any name that describes "interrupt that we can post". If "intr" is that, I'll be happy to take it. Vectored_intr sounds even worse IMHO :).

kvm_irq_is_intr() is fine by me if it's clear to everyone else. Alternatively, we could be more literal, e.g. kvm_irq_is_postable().

Quoted 15 lines
Quoted 14 lines

arch/x86/kvm/svm.c static int svm_update_pi_irte(struct kvm *kvm, · line 5319

Quoted 11 lines
arch/x86/kvm/svm.c +1 -0
 		pr_debug("SVM: %s: use legacy intr remap mode for irq %u\n",
 			 __func__, irq.vector);
 		return -1;
@@ -5314,6 +5315,7 @@ static int svm_update_pi_irte(struct kvm *kvm,
 unsigned int host_irq,
 		 * 1. When cannot target interrupt to a specific vcpu.
 		 * 2. Unsetting posted interrupt.
 		 * 3. APIC virtialization is disabled for the vcpu.
+		 * 4. IRQ has extended delivery mode (SMI, INIT, etc)

Similarly, 'extended delivery mode' isn't really a thing, it's simply the delivery mode.

Quoted 1 line

s/extended/incompatible/ maybe?

Ya, much better.

#7 lore raw

PATCH v2 2/2 KVM: SVM: Disable posted interrupts for odd IRQs

In reply to Alexander Graf

We can easily route hardware interrupts directly into VM context when they target the "Fixed" or "LowPriority" delivery modes.

However, on modes such as "SMI" or "Init", we need to go via KVM code to actually put the vCPU into a different mode of operation, so we can not post the interrupt

Add code in the SVM PI logic to explicitly refuse to establish posted mappings for advanced IRQ deliver modes. This reflects the logic in __apic_accept_irq() which also only ever passes Fixed and LowPriority interrupts as posted interrupts into the guest.

This fixes a bug I have with code which configures real hardware to inject virtual SMIs into my guest.

---

v1 -> v2:

- Make error message more unique
- Update commit message to point to __apic_accept_irq()

---
 arch/x86/kvm/svm.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
arch/x86/kvm/svm.c +16 -0
index 1f220a85514f..b86b45b85da8 100644
@@ -5266,6 +5266,21 @@ get_pi_vcpu_info(struct kvm *kvm, struct kvm_kernel_irq_routing_entry *e,
 		return -1;
 	}
 
+	switch (irq.delivery_mode) {
+	case dest_Fixed:
+	case dest_LowestPrio:
+		break;
+	default:
+		/*
+		 * For non-trivial interrupt events, we need to go
+		 * through the full KVM IRQ code, so refuse to take
+		 * any direct PI assignments here.
+		 */
+		pr_debug("SVM: %s: use legacy intr mode for non-std irq %u\n",
+			 __func__, irq.vector);
+		return -1;
+	}
+
 	pr_debug("SVM: %s: use GA mode for irq %u\n", __func__,
 		 irq.vector);
 	*svm = to_svm(vcpu);
@@ -5314,6 +5329,7 @@ static int svm_update_pi_irte(struct kvm *kvm, unsigned int host_irq,
 		 * 1. When cannot target interrupt to a specific vcpu.
 		 * 2. Unsetting posted interrupt.
 		 * 3. APIC virtialization is disabled for the vcpu.
+		 * 4. IRQ has extended delivery mode (SMI, INIT, etc)
 		 */
 		if (!get_pi_vcpu_info(kvm, e, &vcpu_info, &svm) && set &&
 		    kvm_vcpu_apicv_active(&svm->vcpu)) {
Signature · 12 lines
2.17.1




Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879