Skip to contentSkip to conversation

KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

PATCH Superseded [email protected] 2 participants 8 messages Sep 22, 2020 – Nov 6, 2020

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

From: Yadong Qi <[email protected]>

Background: We have a lightweight HV, it needs INIT-VMExit and SIPI-VMExit to wake-up APs for guests since it do not monitor the Local APIC. But currently virtual wait-for-SIPI(WFS) state is not supported in nVMX, so when running on top of KVM, the L1 HV cannot receive the INIT-VMExit and SIPI-VMExit which cause the L2 guest cannot wake up the APs.

According to Intel SDM Chapter 25.2 Other Causes of VM Exits, SIPIs cause VM exits when a logical processor is in wait-for-SIPI state.

In this patch:
1. introduce SIPI exit reason,
2. introduce wait-for-SIPI state for nVMX,
3. advertise wait-for-SIPI support to guest.

When L1 hypervisor is not monitoring Local APIC, L0 need to emulate INIT-VMExit and SIPI-VMExit to L1 to emulate INIT-SIPI-SIPI for L2. L2 LAPIC write would be traped by L0 Hypervisor(KVM), L0 should emulate the INIT/SIPI vmexit to L1 hypervisor to set proper state for L2's vcpu state.

Handle procdure:
Source vCPU:
L2 write LAPIC.ICR(INIT).
L0 trap LAPIC.ICR write(INIT): inject a latched INIT event to target
vCPU.
Target vCPU:
L0 emulate an INIT VMExit to L1 if is guest mode.
L1 set guest VMCS, guest_activity_state=WAIT_SIPI, vmresume. L0 set vcpu.mp_state to INIT_RECEIVED if (vmcs12.guest_activity_state
== WAIT_SIPI).

Source vCPU:
L2 write LAPIC.ICR(SIPI).
L0 trap LAPIC.ICR write(INIT): inject a latched SIPI event to traget
vCPU.
Target vCPU:
L0 emulate an SIPI VMExit to L1 if (vcpu.mp_state == INIT_RECEIVED). L1 set CS:IP, guest_activity_state=ACTIVE, vmresume.
L0 resume to L2.
L2 start-up.

---
 arch/x86/include/asm/vmx.h      |  1 +
 arch/x86/include/uapi/asm/vmx.h |  2 ++
 arch/x86/kvm/lapic.c            |  5 ++--
 arch/x86/kvm/vmx/nested.c       | 53 ++++++++++++++++++++++++---------
 4 files changed, 45 insertions(+), 16 deletions(-)
arch/x86/include/asm/vmx.h +1 -0
index cd7de4b401fe..bff06dc64c52 100644
@@ -113,6 +113,7 @@
 #define VMX_MISC_PREEMPTION_TIMER_RATE_MASK	0x0000001f
 #define VMX_MISC_SAVE_EFER_LMA			0x00000020
 #define VMX_MISC_ACTIVITY_HLT			0x00000040
+#define VMX_MISC_ACTIVITY_WAIT_SIPI		0x00000100
 #define VMX_MISC_ZERO_LEN_INS			0x40000000
 #define VMX_MISC_MSR_LIST_MULTIPLIER		512
 
arch/x86/include/uapi/asm/vmx.h +2 -0
index b8ff9e8ac0d5..ada955c5ebb6 100644
@@ -32,6 +32,7 @@
 #define EXIT_REASON_EXTERNAL_INTERRUPT  1
 #define EXIT_REASON_TRIPLE_FAULT        2
 #define EXIT_REASON_INIT_SIGNAL			3
+#define EXIT_REASON_SIPI_SIGNAL         4
 
 #define EXIT_REASON_INTERRUPT_WINDOW    7
 #define EXIT_REASON_NMI_WINDOW          8
@@ -94,6 +95,7 @@
 	{ EXIT_REASON_EXTERNAL_INTERRUPT,    "EXTERNAL_INTERRUPT" }, \
 	{ EXIT_REASON_TRIPLE_FAULT,          "TRIPLE_FAULT" }, \
 	{ EXIT_REASON_INIT_SIGNAL,           "INIT_SIGNAL" }, \
+	{ EXIT_REASON_SIPI_SIGNAL,           "SIPI_SIGNAL" }, \
 	{ EXIT_REASON_INTERRUPT_WINDOW,      "INTERRUPT_WINDOW" }, \
 	{ EXIT_REASON_NMI_WINDOW,            "NMI_WINDOW" }, \
 	{ EXIT_REASON_TASK_SWITCH,           "TASK_SWITCH" }, \
arch/x86/kvm/lapic.c +3 -2
index 5ccbee7165a2..d04ac7dc6adf 100644
@@ -2839,7 +2839,7 @@ void kvm_apic_accept_events(struct kvm_vcpu *vcpu)
 
 	/*
 	 * INITs are latched while CPU is in specific states
-	 * (SMM, VMX non-root mode, SVM with GIF=0).
+	 * (SMM, SVM with GIF=0).
 	 * Because a CPU cannot be in these states immediately
 	 * after it has processed an INIT signal (and thus in
 	 * KVM_MP_STATE_INIT_RECEIVED state), just eat SIPIs
@@ -2847,7 +2847,8 @@ void kvm_apic_accept_events(struct kvm_vcpu *vcpu)
 	 */
 	if (kvm_vcpu_latch_init(vcpu)) {
 		WARN_ON_ONCE(vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED);
-		if (test_bit(KVM_APIC_SIPI, &apic->pending_events))
+		if (test_bit(KVM_APIC_SIPI, &apic->pending_events) &&
+		    !is_guest_mode(vcpu))
 			clear_bit(KVM_APIC_SIPI, &apic->pending_events);
 		return;
 	}
arch/x86/kvm/vmx/nested.c +39 -14

79 lines not on this page · raw message

Signature · 1 line
2.25.1
maintainer
#1 lore raw
Quoted 35 lines · On 22/09/20 07:23, [email protected] wrote:

From: Yadong Qi <[email protected]>

Background: We have a lightweight HV, it needs INIT-VMExit and SIPI-VMExit to wake-up APs for guests since it do not monitor the Local APIC. But currently virtual wait-for-SIPI(WFS) state is not supported in nVMX, so when running on top of KVM, the L1 HV cannot receive the INIT-VMExit and SIPI-VMExit which cause the L2 guest cannot wake up the APs.

According to Intel SDM Chapter 25.2 Other Causes of VM Exits, SIPIs cause VM exits when a logical processor is in wait-for-SIPI state.

In this patch:
1. introduce SIPI exit reason,
2. introduce wait-for-SIPI state for nVMX,
3. advertise wait-for-SIPI support to guest.

When L1 hypervisor is not monitoring Local APIC, L0 need to emulate INIT-VMExit and SIPI-VMExit to L1 to emulate INIT-SIPI-SIPI for L2. L2 LAPIC write would be traped by L0 Hypervisor(KVM), L0 should emulate the INIT/SIPI vmexit to L1 hypervisor to set proper state for L2's vcpu state.

Handle procdure:
Source vCPU:
L2 write LAPIC.ICR(INIT).
L0 trap LAPIC.ICR write(INIT): inject a latched INIT event to target
vCPU.
Target vCPU:
L0 emulate an INIT VMExit to L1 if is guest mode.
L1 set guest VMCS, guest_activity_state=WAIT_SIPI, vmresume. L0 set vcpu.mp_state to INIT_RECEIVED if (vmcs12.guest_activity_state
== WAIT_SIPI).

Quoted 9 lines

Source vCPU:
L2 write LAPIC.ICR(SIPI).
L0 trap LAPIC.ICR write(INIT): inject a latched SIPI event to traget
vCPU.
Target vCPU:
L0 emulate an SIPI VMExit to L1 if (vcpu.mp_state == INIT_RECEIVED). L1 set CS:IP, guest_activity_state=ACTIVE, vmresume.
L0 resume to L2.
L2 start-up.

Again, this looks good but it needs testcases.

Thanks,

Paolo

Quoted 151 lines
---
 arch/x86/include/asm/vmx.h      |  1 +
 arch/x86/include/uapi/asm/vmx.h |  2 ++
 arch/x86/kvm/lapic.c            |  5 ++--
 arch/x86/kvm/vmx/nested.c       | 53 ++++++++++++++++++++++++---------
 4 files changed, 45 insertions(+), 16 deletions(-)
arch/x86/include/asm/vmx.h +1 -0
index cd7de4b401fe..bff06dc64c52 100644
@@ -113,6 +113,7 @@
 #define VMX_MISC_PREEMPTION_TIMER_RATE_MASK	0x0000001f
 #define VMX_MISC_SAVE_EFER_LMA			0x00000020
 #define VMX_MISC_ACTIVITY_HLT			0x00000040
+#define VMX_MISC_ACTIVITY_WAIT_SIPI		0x00000100
 #define VMX_MISC_ZERO_LEN_INS			0x40000000
 #define VMX_MISC_MSR_LIST_MULTIPLIER		512
 
arch/x86/include/uapi/asm/vmx.h +2 -0
index b8ff9e8ac0d5..ada955c5ebb6 100644
@@ -32,6 +32,7 @@
 #define EXIT_REASON_EXTERNAL_INTERRUPT  1
 #define EXIT_REASON_TRIPLE_FAULT        2
 #define EXIT_REASON_INIT_SIGNAL			3
+#define EXIT_REASON_SIPI_SIGNAL         4
 
 #define EXIT_REASON_INTERRUPT_WINDOW    7
 #define EXIT_REASON_NMI_WINDOW          8
@@ -94,6 +95,7 @@
 	{ EXIT_REASON_EXTERNAL_INTERRUPT,    "EXTERNAL_INTERRUPT" }, \
 	{ EXIT_REASON_TRIPLE_FAULT,          "TRIPLE_FAULT" }, \
 	{ EXIT_REASON_INIT_SIGNAL,           "INIT_SIGNAL" }, \
+	{ EXIT_REASON_SIPI_SIGNAL,           "SIPI_SIGNAL" }, \
 	{ EXIT_REASON_INTERRUPT_WINDOW,      "INTERRUPT_WINDOW" }, \
 	{ EXIT_REASON_NMI_WINDOW,            "NMI_WINDOW" }, \
 	{ EXIT_REASON_TASK_SWITCH,           "TASK_SWITCH" }, \
arch/x86/kvm/lapic.c +3 -2
index 5ccbee7165a2..d04ac7dc6adf 100644
@@ -2839,7 +2839,7 @@ void kvm_apic_accept_events(struct kvm_vcpu *vcpu)
 
 	/*
 	 * INITs are latched while CPU is in specific states
-	 * (SMM, VMX non-root mode, SVM with GIF=0).
+	 * (SMM, SVM with GIF=0).
 	 * Because a CPU cannot be in these states immediately
 	 * after it has processed an INIT signal (and thus in
 	 * KVM_MP_STATE_INIT_RECEIVED state), just eat SIPIs
@@ -2847,7 +2847,8 @@ void kvm_apic_accept_events(struct kvm_vcpu *vcpu)
 	 */
 	if (kvm_vcpu_latch_init(vcpu)) {
 		WARN_ON_ONCE(vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED);
-		if (test_bit(KVM_APIC_SIPI, &apic->pending_events))
+		if (test_bit(KVM_APIC_SIPI, &apic->pending_events) &&
+		    !is_guest_mode(vcpu))
 			clear_bit(KVM_APIC_SIPI, &apic->pending_events);
 		return;
 	}
arch/x86/kvm/vmx/nested.c +39 -14
index 1bb6b31eb646..fe3bb68df987 100644
@@ -2946,7 +2946,8 @@ static int nested_vmx_check_vmcs_link_ptr(struct kvm_vcpu *vcpu,
 static int nested_check_guest_non_reg_state(struct vmcs12 *vmcs12)
 {
 	if (CC(vmcs12->guest_activity_state != GUEST_ACTIVITY_ACTIVE &&
-	       vmcs12->guest_activity_state != GUEST_ACTIVITY_HLT))
+	       vmcs12->guest_activity_state != GUEST_ACTIVITY_HLT &&
+	       vmcs12->guest_activity_state != GUEST_ACTIVITY_WAIT_SIPI))
 		return -EINVAL;
 
 	return 0;
@@ -3543,19 +3544,29 @@ static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch)
 	 */
 	nested_cache_shadow_vmcs12(vcpu, vmcs12);
 
-	/*
-	 * If we're entering a halted L2 vcpu and the L2 vcpu won't be
-	 * awakened by event injection or by an NMI-window VM-exit or
-	 * by an interrupt-window VM-exit, halt the vcpu.
-	 */
-	if ((vmcs12->guest_activity_state == GUEST_ACTIVITY_HLT) &&
-	    !(vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK) &&
-	    !(vmcs12->cpu_based_vm_exec_control & CPU_BASED_NMI_WINDOW_EXITING) &&
-	    !((vmcs12->cpu_based_vm_exec_control & CPU_BASED_INTR_WINDOW_EXITING) &&
-	      (vmcs12->guest_rflags & X86_EFLAGS_IF))) {
+	switch (vmcs12->guest_activity_state) {
+	case GUEST_ACTIVITY_HLT:
+		/*
+		 * If we're entering a halted L2 vcpu and the L2 vcpu won't be
+		 * awakened by event injection or by an NMI-window VM-exit or
+		 * by an interrupt-window VM-exit, halt the vcpu.
+		 */
+		if (!(vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK) &&
+		    !nested_cpu_has(vmcs12, CPU_BASED_NMI_WINDOW_EXITING) &&
+		    !(nested_cpu_has(vmcs12, CPU_BASED_INTR_WINDOW_EXITING) &&
+		      (vmcs12->guest_rflags & X86_EFLAGS_IF))) {
+			vmx->nested.nested_run_pending = 0;
+			return kvm_vcpu_halt(vcpu);
+		}
+		break;
+	case GUEST_ACTIVITY_WAIT_SIPI:
 		vmx->nested.nested_run_pending = 0;
-		return kvm_vcpu_halt(vcpu);
+		vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
+		break;
+	default:
+		break;
 	}
+
 	return 1;
 
 vmentry_failed:
@@ -3781,7 +3792,20 @@ static int vmx_check_nested_events(struct kvm_vcpu *vcpu)
 			return -EBUSY;
 		nested_vmx_update_pending_dbg(vcpu);
 		clear_bit(KVM_APIC_INIT, &apic->pending_events);
-		nested_vmx_vmexit(vcpu, EXIT_REASON_INIT_SIGNAL, 0, 0);
+		if (vcpu->arch.mp_state != KVM_MP_STATE_INIT_RECEIVED)
+			nested_vmx_vmexit(vcpu, EXIT_REASON_INIT_SIGNAL, 0, 0);
+		return 0;
+	}
+
+	if (lapic_in_kernel(vcpu) &&
+	    test_bit(KVM_APIC_SIPI, &apic->pending_events)) {
+		if (block_nested_events)
+			return -EBUSY;
+
+		clear_bit(KVM_APIC_SIPI, &apic->pending_events);
+		if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED)
+			nested_vmx_vmexit(vcpu, EXIT_REASON_SIPI_SIGNAL, 0,
+						apic->sipi_vector & 0xFFUL);
 		return 0;
 	}
 
@@ -6471,7 +6495,8 @@ void nested_vmx_setup_ctls_msrs(struct nested_vmx_msrs *msrs, u32 ept_caps)
 	msrs->misc_low |=
 		MSR_IA32_VMX_MISC_VMWRITE_SHADOW_RO_FIELDS |
 		VMX_MISC_EMULATED_PREEMPTION_TIMER_RATE |
-		VMX_MISC_ACTIVITY_HLT;
+		VMX_MISC_ACTIVITY_HLT |
+		VMX_MISC_ACTIVITY_WAIT_SIPI;
 	msrs->misc_high = 0;
 
 	/*
author
#2 lore raw
Quoted 1 line

Again, this looks good but it needs testcases.

Yes, the unit test development is WIP.

Best Regard
Yadong

author
#3 lore raw
Quoted 6 lines

Subject: RE: [PATCH] KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

Quoted 1 line

Again, this looks good but it needs testcases.

Yes, the unit test development is WIP.

Hi, Paolo

I have sent out the unit test patch.
https://patchwork.kernel.org/patch/11799305/
Could you help review it? Thanks a lot.

Best Regard
Yadong

maintainer
#4 lore raw
Quoted 4 lines · Yadong
Quoted 4 lines

Subject: RE: [PATCH] KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

Quoted 1 line

Again, this looks good but it needs testcases.

Quoted 8 lines
Quoted 2 lines

Yes, the unit test development is WIP.

Hi, Paolo

I have sent out the unit test patch.
https://patchwork.kernel.org/patch/11799305/
Could you help review it? Thanks a lot.

Yes, I've seen it. Tomorrow. :)

Paolo

author
#5 lore raw
In reply to Paolo Bonzini
Quoted 60 lines

-----Original Message-----
From: Paolo Bonzini <[email protected]>
Sent: Tuesday, September 22, 2020 5:10 PM
To: Qi, Yadong <[email protected]>; [email protected]; linux- [email protected]; [email protected]
Cc: Christopherson, Sean J <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; Gao, Chao <[email protected]>; Tian, Kevin <[email protected]>; Chen, Luhai <[email protected]>; Zhu, Bing <[email protected]>; Wang, Kai Z <[email protected]>
Subject: Re: [PATCH] KVM: x86: emulate wait-for-SIPI and SIPI-VMExit

Quoted 34 lines · On 22/09/20 07:23, [email protected] wrote:

From: Yadong Qi <[email protected]>

Background: We have a lightweight HV, it needs INIT-VMExit and SIPI-VMExit to wake-up APs for guests since it do not monitor the Local APIC. But currently virtual wait-for-SIPI(WFS) state is not supported in nVMX, so when running on top of KVM, the L1 HV cannot receive the INIT-VMExit and SIPI-VMExit which cause the L2 guest cannot wake up the APs.

According to Intel SDM Chapter 25.2 Other Causes of VM Exits, SIPIs cause VM exits when a logical processor is in wait-for-SIPI state.

In this patch:
1. introduce SIPI exit reason,
2. introduce wait-for-SIPI state for nVMX,
3. advertise wait-for-SIPI support to guest.

When L1 hypervisor is not monitoring Local APIC, L0 need to emulate INIT-VMExit and SIPI-VMExit to L1 to emulate INIT-SIPI-SIPI for L2. L2 LAPIC write would be traped by L0 Hypervisor(KVM), L0 should emulate the INIT/SIPI vmexit to L1 hypervisor to set proper state for L2's vcpu state.

Handle procdure:
Source vCPU:
L2 write LAPIC.ICR(INIT).
L0 trap LAPIC.ICR write(INIT): inject a latched INIT event to target
vCPU.
Target vCPU:
L0 emulate an INIT VMExit to L1 if is guest mode.
L1 set guest VMCS, guest_activity_state=WAIT_SIPI, vmresume. L0 set vcpu.mp_state to INIT_RECEIVED if (vmcs12.guest_activity_state
== WAIT_SIPI).

Quoted 9 lines

Source vCPU:
L2 write LAPIC.ICR(SIPI).
L0 trap LAPIC.ICR write(INIT): inject a latched SIPI event to traget
vCPU.
Target vCPU:
L0 emulate an SIPI VMExit to L1 if (vcpu.mp_state == INIT_RECEIVED). L1 set CS:IP, guest_activity_state=ACTIVE, vmresume.
L0 resume to L2.
L2 start-up.

Quoted 2 lines

Again, this looks good but it needs testcases.

Hi, Paolo

I saw you queued the testcase patch: https://patchwork.kernel.org/project/kvm/patch/[email protected]/
Will you also queue this patch? Or there are some additional comments of this patch?

Best Regard
Yadong

maintainer
#6 lore raw
In reply to [email protected]
Quoted 18 lines · On 22/09/20 07:23, [email protected] wrote:

From: Yadong Qi <[email protected]>

Background: We have a lightweight HV, it needs INIT-VMExit and SIPI-VMExit to wake-up APs for guests since it do not monitor the Local APIC. But currently virtual wait-for-SIPI(WFS) state is not supported in nVMX, so when running on top of KVM, the L1 HV cannot receive the INIT-VMExit and SIPI-VMExit which cause the L2 guest cannot wake up the APs.

According to Intel SDM Chapter 25.2 Other Causes of VM Exits, SIPIs cause VM exits when a logical processor is in wait-for-SIPI state.

In this patch:
1. introduce SIPI exit reason,
2. introduce wait-for-SIPI state for nVMX,
3. advertise wait-for-SIPI support to guest.

Quoted 5 lines

When L1 hypervisor is not monitoring Local APIC, L0 need to emulate INIT-VMExit and SIPI-VMExit to L1 to emulate INIT-SIPI-SIPI for L2. L2 LAPIC write would be traped by L0 Hypervisor(KVM), L0 should emulate the INIT/SIPI vmexit to L1 hypervisor to set proper state for L2's vcpu state.

There is a problem in this patch, in that this change is incorrect:

Review void kvm_apic_accept_events(struct kvm_vcpu *vcpu) · line 2854

Quoted 11 lines
(quoted diff) +2 -1
@@ -2847,7 +2847,8 @@ void kvm_apic_accept_events(struct kvm_vcpu *vcpu)
 	 */
 	if (kvm_vcpu_latch_init(vcpu)) {
 		WARN_ON_ONCE(vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED);
-		if (test_bit(KVM_APIC_SIPI, &apic->pending_events))
+		if (test_bit(KVM_APIC_SIPI, &apic->pending_events) &&
+		    !is_guest_mode(vcpu))
 			clear_bit(KVM_APIC_SIPI, &apic->pending_events);
 		return;
 	}

Here you're not trying to process a latched INIT; you just want to delay the processing of the SIPI until check_nested_events.

The change does have a correct part in it. In particular, vmx_apic_init_signal_blocked should have been

arch/x86/kvm/vmx/vmx.c +1 -1
index 47b8357b9751..64339121a4f0 100644
@@ -7558,7 +7558,7 @@ static void enable_smi_window(struct kvm_vcpu *vcpu)
 
 static bool vmx_apic_init_signal_blocked(struct kvm_vcpu *vcpu)
 {
-	return to_vmx(vcpu)->nested.vmxon;
+	return to_vmx(vcpu)->nested.vmxon && !is_guest_mode(vcpu);
 }
 
 static void vmx_migrate_timers(struct kvm_vcpu *vcpu)

to only latch INIT signals in root mode. However, SIPI must be cleared unconditionally on SVM; the "!is_guest_mode" test in that case is incorrect.

The right way to do it is to call check_nested_events from kvm_apic_accept_events. This will cause an INIT or SIPI vmexit, as required. There is some extra complication to read pending_events *before* kvm_apic_accept_events and not steal from the guest any INIT or SIPI that is sent after kvm_apic_accept_events returns.

Thanks to your test case, I will test a patch and send it.

Paolo

author
#7 lore raw
Quoted 40 lines

There is a problem in this patch, in that this change is incorrect:

Review void kvm_apic_accept_events(struct kvm_vcpu

Quoted 2 lines
(quoted diff) +0 -0
@@ -2847,7 +2847,8 @@ void kvm_apic_accept_events(struct kvm_vcpu

*vcpu)

Review void kvm_apic_accept_events(struct kvm_vcpu · line 2849

Quoted 3 lines
(quoted diff) +0 -0
 	 */
 	if (kvm_vcpu_latch_init(vcpu)) {
 		WARN_ON_ONCE(vcpu->arch.mp_state ==

KVM_MP_STATE_INIT_RECEIVED);

Review void kvm_apic_accept_events(struct kvm_vcpu · line 2854

Quoted 6 lines
(quoted diff) +2 -1
-		if (test_bit(KVM_APIC_SIPI, &apic->pending_events))
+		if (test_bit(KVM_APIC_SIPI, &apic->pending_events) &&
+		    !is_guest_mode(vcpu))
 			clear_bit(KVM_APIC_SIPI, &apic->pending_events);
 		return;
 	}

Here you're not trying to process a latched INIT; you just want to delay the processing of the SIPI until check_nested_events.

The change does have a correct part in it. In particular, vmx_apic_init_signal_blocked should have been

arch/x86/kvm/vmx/vmx.c index renamed from arch/x86/kvm/vmx/vmx.c +0 -0

47b8357b9751..64339121a4f0 100644

arch/x86/kvm/vmx/vmx.c +1 -1
@@ -7558,7 +7558,7 @@ static void enable_smi_window(struct kvm_vcpu
 *vcpu)
 
  static bool vmx_apic_init_signal_blocked(struct kvm_vcpu *vcpu)
  {
-	return to_vmx(vcpu)->nested.vmxon;
+	return to_vmx(vcpu)->nested.vmxon && !is_guest_mode(vcpu);
  }
 

static void vmx_migrate_timers(struct kvm_vcpu *vcpu)

to only latch INIT signals in root mode. However, SIPI must be cleared unconditionally on SVM; the "!is_guest_mode" test in that case is incorrect.

Quoted 8 lines

The right way to do it is to call check_nested_events from kvm_apic_accept_events. This will cause an INIT or SIPI vmexit, as required. There is some extra complication to read pending_events
*before* kvm_apic_accept_events and not steal from the guest any INIT or SIPI that is sent after kvm_apic_accept_events returns.

Thanks to your test case, I will test a patch and send it.

Thanks very much, Paolo.
BTW, I noticed another issue in sync_vmcs02_to_vmcs12():
vmcs12->guest_activity_state is not set properly when mp_state is INIT_RECEIVED. I will correct it and send v2 of Patch 2/2.