[PATCH 0/2] KVM: X86: Add dedicated pCPU hint and per-VM non-HLT-exiting capability

[PATCH 0/2] KVM: X86: Add dedicated pCPU hint and per-VM non-HLT-exiting capability

[Wanpeng Li]

Waiman Long mentioned that:

 Generally speaking, unfair lock performs well for VMs with a small
 number of vCPUs. Native qspinlock may perform better than pvqspinlock
 if there is vCPU pinning and there is no vCPU over-commitment.

This patchset adds a PV_DEDICATED performance hint to allow hypervisor 
admin to choose the qspinlock to be used when a dedicated pCPU is 
available.

In addition, according to the original hlt in VMX non-root mode discussion, 
https://www.spinics.net/lists/kvm/msg152397.html This patchset also 
adds the per-VM non-HLT-exiting capability to further improve performance 
under the dedicated pCPU scenarios.

Wanpeng Li (2):
  KVM: X86: Add dedicated pCPU hint PV_DEDICATED
  KVM: X86: Add per-VM no-HLT-exiting capability

 Documentation/virtual/kvm/api.txt    | 11 +++++++++++
 Documentation/virtual/kvm/cpuid.txt  |  6 ++++++
 arch/x86/include/asm/kvm_host.h      |  2 ++
 arch/x86/include/uapi/asm/kvm_para.h |  1 +
 arch/x86/kernel/kvm.c                |  6 ++++++
 arch/x86/kvm/vmx.c                   | 21 +++++++++++++++++++++
 arch/x86/kvm/x86.c                   |  5 +++++
 arch/x86/kvm/x86.h                   |  5 +++++
 include/uapi/linux/kvm.h             |  1 +
 9 files changed, 58 insertions(+)

-- 
2.7.4

[PATCH 1/2] KVM: X86: Add dedicated pCPU hint PV_DEDICATED

[Wanpeng Li]

From: Wanpeng Li <wanpengli@tencent.com>

Waiman Long mentioned that:

 Generally speaking, unfair lock performs well for VMs with a small
 number of vCPUs. Native qspinlock may perform better than pvqspinlock
 if there is vCPU pinning and there is no vCPU over-commitment.

This patch adds a performance hint to allow hypervisor admin to choose
the qspinlock to be used when a dedicated pCPU is available.

PV_DEDICATED = 1, PV_UNHALT = anything: default is qspinlock
PV_DEDICATED = 0, PV_UNHALT = 1: default is Hybrid PV queued/unfair lock
PV_DEDICATED = 0, PV_UNHALT = 0: default is tas

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
---
 Documentation/virtual/kvm/cpuid.txt  | 6 ++++++
 arch/x86/include/uapi/asm/kvm_para.h | 1 +
 arch/x86/kernel/kvm.c                | 6 ++++++
 3 files changed, 13 insertions(+)

diff --git a/Documentation/virtual/kvm/cpuid.txt b/Documentation/virtual/kvm/cpuid.txt
index 87a7506..c0740b1 100644
--- a/Documentation/virtual/kvm/cpuid.txt
+++ b/Documentation/virtual/kvm/cpuid.txt
@@ -54,6 +54,12 @@ KVM_FEATURE_PV_UNHALT              ||     7 || guest checks this feature bit
                                    ||       || before enabling paravirtualized
                                    ||       || spinlock support.
 ------------------------------------------------------------------------------
+KVM_FEATURE_PV_DEDICATED           ||     8 || guest checks this feature bit
+                                   ||       || to determine if they run on
+                                   ||       || dedicated vCPUs, allowing opti-
+                                   ||       || mizations such as usage of
+                                   ||       || qspinlocks.
+------------------------------------------------------------------------------
 KVM_FEATURE_PV_TLB_FLUSH           ||     9 || guest checks this feature bit
                                    ||       || before enabling paravirtualized
                                    ||       || tlb flush.
diff --git a/arch/x86/include/uapi/asm/kvm_para.h b/arch/x86/include/uapi/asm/kvm_para.h
index 6cfa9c8..9a5ef67 100644
--- a/arch/x86/include/uapi/asm/kvm_para.h
+++ b/arch/x86/include/uapi/asm/kvm_para.h
@@ -25,6 +25,7 @@
 #define KVM_FEATURE_STEAL_TIME		5
 #define KVM_FEATURE_PV_EOI		6
 #define KVM_FEATURE_PV_UNHALT		7
+#define KVM_FEATURE_PV_DEDICATED	8
 #define KVM_FEATURE_PV_TLB_FLUSH	9
 #define KVM_FEATURE_ASYNC_PF_VMEXIT	10
 
diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
index aa2b706..6f0e43f 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -726,6 +726,12 @@ void __init kvm_spinlock_init(void)
 {
 	if (!kvm_para_available())
 		return;
+
+	if (kvm_para_has_feature(KVM_FEATURE_PV_DEDICATED)) {
+		static_branch_disable(&virt_spin_lock_key);
+		return;
+	}
+
 	/* Does host kernel support KVM_FEATURE_PV_UNHALT? */
 	if (!kvm_para_has_feature(KVM_FEATURE_PV_UNHALT))
 		return;
-- 
2.7.4

[PATCH 2/2] KVM: X86: Add per-VM no-HLT-exiting capability

[Wanpeng Li]

From: Wanpeng Li <wanpengli@tencent.com>

If host CPUs are dedicated to a VM, we can avoid VM exits on HLT.
This patch adds the per-VM non-HLT-exiting capability.

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
---
 Documentation/virtual/kvm/api.txt | 11 +++++++++++
 arch/x86/include/asm/kvm_host.h   |  2 ++
 arch/x86/kvm/vmx.c                | 21 +++++++++++++++++++++
 arch/x86/kvm/x86.c                |  5 +++++
 arch/x86/kvm/x86.h                |  5 +++++
 include/uapi/linux/kvm.h          |  1 +
 6 files changed, 45 insertions(+)

diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
index e5f1743..573a3e5 100644
--- a/Documentation/virtual/kvm/api.txt
+++ b/Documentation/virtual/kvm/api.txt
@@ -4222,6 +4222,17 @@ enables QEMU to build error log and branch to guest kernel registered
 machine check handling routine. Without this capability KVM will
 branch to guests' 0x200 interrupt vector.
 
+7.13 KVM_CAP_X86_GUEST_HLT
+
+Architectures: x86
+Parameters: none
+Returns: 0 on success
+
+This capability indicates that a guest using HLT to stop a virtual CPU
+will not cause a VM exit. As such, time spent while a virtual CPU is
+halted in this way will then be accounted for as guest running time on
+the host, KVM_FEATURE_PV_UNHALT should be disabled.
+
 8. Other capabilities.
 ----------------------
 
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index dd6f57a..c566ea0 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -804,6 +804,8 @@ struct kvm_arch {
 
 	gpa_t wall_clock;
 
+	bool hlt_in_guest;
+
 	bool ept_identity_pagetable_done;
 	gpa_t ept_identity_map_addr;
 
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index b1e554a..6cfd8d3 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2478,6 +2478,19 @@ static int nested_vmx_check_exception(struct kvm_vcpu *vcpu, unsigned long *exit
 	return 0;
 }
 
+static void vmx_clear_hlt(struct kvm_vcpu *vcpu)
+{
+	/*
+	 * Ensure that we clear the HLT state in the VMCS.  We don't need to
+	 * explicitly skip the instruction because if the HLT state is set,
+	 * then the instruction is already executing and RIP has already been
+	 * advanced.
+	 */
+	if (kvm_hlt_in_guest(vcpu->kvm) &&
+			vmcs_read32(GUEST_ACTIVITY_STATE) == GUEST_ACTIVITY_HLT)
+		vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE);
+}
+
 static void vmx_queue_exception(struct kvm_vcpu *vcpu)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
@@ -2508,6 +2521,8 @@ static void vmx_queue_exception(struct kvm_vcpu *vcpu)
 		intr_info |= INTR_TYPE_HARD_EXCEPTION;
 
 	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info);
+
+	vmx_clear_hlt(vcpu);
 }
 
 static bool vmx_rdtscp_supported(void)
@@ -5301,6 +5316,8 @@ static u32 vmx_exec_control(struct vcpu_vmx *vmx)
 		exec_control |= CPU_BASED_CR3_STORE_EXITING |
 				CPU_BASED_CR3_LOAD_EXITING  |
 				CPU_BASED_INVLPG_EXITING;
+	if (kvm_hlt_in_guest(vmx->vcpu.kvm))
+		exec_control &= ~CPU_BASED_HLT_EXITING;
 	return exec_control;
 }
 
@@ -5729,6 +5746,8 @@ static void vmx_inject_irq(struct kvm_vcpu *vcpu)
 	} else
 		intr |= INTR_TYPE_EXT_INTR;
 	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr);
+
+	vmx_clear_hlt(vcpu);
 }
 
 static void vmx_inject_nmi(struct kvm_vcpu *vcpu)
@@ -5759,6 +5778,8 @@ static void vmx_inject_nmi(struct kvm_vcpu *vcpu)
 
 	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
 			INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR);
+
+	vmx_clear_hlt(vcpu);
 }
 
 static bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index c13cd14..a508247 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2740,6 +2740,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	case KVM_CAP_SET_BOOT_CPU_ID:
  	case KVM_CAP_SPLIT_IRQCHIP:
 	case KVM_CAP_IMMEDIATE_EXIT:
+	case KVM_CAP_X86_GUEST_HLT:
 		r = 1;
 		break;
 	case KVM_CAP_ADJUST_CLOCK:
@@ -4061,6 +4062,10 @@ static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 
 		r = 0;
 		break;
+	case KVM_CAP_X86_GUEST_HLT:
+		kvm->arch.hlt_in_guest = cap->args[0];
+		r = 0;
+		break;
 	default:
 		r = -EINVAL;
 		break;
diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
index b91215d..96fe84e 100644
--- a/arch/x86/kvm/x86.h
+++ b/arch/x86/kvm/x86.h
@@ -270,4 +270,9 @@ static inline bool kvm_mwait_in_guest(void)
 		!boot_cpu_has_bug(X86_BUG_MONITOR);
 }
 
+static inline bool kvm_hlt_in_guest(struct kvm *kvm)
+{
+	return kvm->arch.hlt_in_guest;
+}
+
 #endif
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index 62c564d..6cbf080 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -932,6 +932,7 @@ struct kvm_ppc_resize_hpt {
 #define KVM_CAP_HYPERV_SYNIC2 148
 #define KVM_CAP_HYPERV_VP_INDEX 149
 #define KVM_CAP_S390_AIS_MIGRATION 150
+#define KVM_CAP_X86_GUEST_HLT 151
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
-- 
2.7.4

Re: [PATCH 2/2] KVM: X86: Add per-VM no-HLT-exiting capability

[Radim Krčmář]

2018-02-01 23:11-0800, Wanpeng Li:
> From: Wanpeng Li <wanpengli@tencent.com>
> 
> If host CPUs are dedicated to a VM, we can avoid VM exits on HLT.
> This patch adds the per-VM non-HLT-exiting capability.
> 
> Cc: Paolo Bonzini <pbonzini@redhat.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
> ---

SMM handling needs more work: I think we should vmx_clear_hlt() upon SMI
and RSM needs to implement auto halt restart (AMD might differ).

And also look if we don't need vmx_clear_hlt() around INIT handling,
thanks.

Re: [PATCH 2/2] KVM: X86: Add per-VM no-HLT-exiting capability

[Wanpeng Li]

2018-02-03 3:02 GMT+08:00 Radim Krčmář <rkrcmar@redhat.com>:
> 2018-02-01 23:11-0800, Wanpeng Li:
>> From: Wanpeng Li <wanpengli@tencent.com>
>>
>> If host CPUs are dedicated to a VM, we can avoid VM exits on HLT.
>> This patch adds the per-VM non-HLT-exiting capability.
>>
>> Cc: Paolo Bonzini <pbonzini@redhat.com>
>> Cc: Radim Krčmář <rkrcmar@redhat.com>
>> Signed-off-by: Wanpeng Li <wanpengli@tencent.com>
>> ---
>
> SMM handling needs more work: I think we should vmx_clear_hlt() upon SMI
> and RSM needs to implement auto halt restart (AMD might differ).

Do it in v2.

>
> And also look if we don't need vmx_clear_hlt() around INIT handling,

You are right, the guest stuck after utilizing system_reset command in
qemu monitor w/o vmx_clear_hlt() around INIT handling, fix it in v2.

Regards,
Wanpeng Li