[RFC V3 2/6] KVM: x86: exchange info about lazy_tscdeadline with msr

[Wang Jianchao]

The lazy tsc deadline hasn't work in this version but just tranmit
the physical addrss from gust to host sdie.
 - Add data structure in both guest and host side
 - If feature is enabled, set msr of lazy tscdeadline when guest
   cpu is initialized and clear it when cpu is offlined.
 - Add msr set/get emulation code in host side.

Signed-off-by: Li Shujin <arkinjob@xxxxxxxxxxx>
Signed-off-by: Wang Jianchao <jianchwa@xxxxxxxxxxx>
---
 arch/x86/include/asm/kvm_host.h |  6 ++++++
 arch/x86/kernel/kvm.c           | 13 +++++++++++++
 arch/x86/kvm/x86.c              | 13 +++++++++++++
 3 files changed, 32 insertions(+)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 28bd383..439e4a7 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -713,6 +713,10 @@ struct kvm_queued_exception {
 	bool has_payload;
 };
 
+struct kvm_host_lazy_tscdeadline {
+	u64 msr_val;
+};
+
 struct kvm_vcpu_arch {
 	/*
 	 * rip and regs accesses must go through
@@ -944,6 +948,8 @@ struct kvm_vcpu_arch {
 		struct gfn_to_hva_cache data;
 	} pv_eoi;
 
+	struct kvm_host_lazy_tscdeadline lazy_tscdeadline;
+
 	u64 msr_kvm_poll_control;
 
 	/* set at EPT violation at this point */
diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
index 1cceac5..91eb333 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -67,6 +67,7 @@ early_param("no-steal-acc", parse_no_stealacc);
 
 static DEFINE_PER_CPU_DECRYPTED(struct kvm_vcpu_pv_apf_data, apf_reason) __aligned(64);
 DEFINE_PER_CPU_DECRYPTED(struct kvm_steal_time, steal_time) __aligned(64) __visible;
+DEFINE_PER_CPU_DECRYPTED(struct kvm_lazy_tscdeadline, kvm_lazy_tscdeadline) __aligned(64) __visible;
 static int has_steal_clock = 0;
 
 static int has_guest_poll = 0;
@@ -379,6 +380,16 @@ static void kvm_guest_cpu_init(void)
 
 	if (has_steal_clock)
 		kvm_register_steal_time();
+
+	if (kvm_para_has_feature(KVM_FEATURE_LAZY_TSCDEADLINE)) {
+		struct kvm_lazy_tscdeadline *ptr = this_cpu_ptr(&kvm_lazy_tscdeadline);
+		unsigned long pa;
+
+		BUILD_BUG_ON(__alignof__(kvm_lazy_tscdeadline) < 4);
+		memset(ptr, 0, sizeof(*ptr));
+		pa = slow_virt_to_phys(ptr) | KVM_MSR_ENABLED;
+		wrmsrl(MSR_KVM_LAZY_TSCDEADLINE, pa);
+	}
 }
 
 static void kvm_pv_disable_apf(void)
@@ -452,6 +463,8 @@ static void kvm_guest_cpu_offline(bool shutdown)
 	if (kvm_para_has_feature(KVM_FEATURE_MIGRATION_CONTROL))
 		wrmsrl(MSR_KVM_MIGRATION_CONTROL, 0);
 	kvm_pv_disable_apf();
+	if (kvm_para_has_feature(KVM_FEATURE_LAZY_TSCDEADLINE))
+		wrmsrl(MSR_KVM_LAZY_TSCDEADLINE, 0);
 	if (!shutdown)
 		apf_task_wake_all();
 	kvmclock_disable();
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index a6b9bea..dbbae8f 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -1551,6 +1551,7 @@ static const u32 emulated_msrs_all[] = {
 
 	MSR_K7_HWCR,
 	MSR_KVM_POLL_CONTROL,
+	MSR_KVM_LAZY_TSCDEADLINE,
 };
 
 static u32 emulated_msrs[ARRAY_SIZE(emulated_msrs_all)];
@@ -3873,7 +3874,13 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 
 		vcpu->arch.msr_kvm_poll_control = data;
 		break;
+	case MSR_KVM_LAZY_TSCDEADLINE:
+		if (!guest_pv_has(vcpu, KVM_FEATURE_LAZY_TSCDEADLINE))
+			return 1;
+
+		vcpu->arch.lazy_tscdeadline.msr_val = data;
 
+		break;
 	case MSR_IA32_MCG_CTL:
 	case MSR_IA32_MCG_STATUS:
 	case MSR_IA32_MC0_CTL ... MSR_IA32_MCx_CTL(KVM_MAX_MCE_BANKS) - 1:
@@ -4229,6 +4236,12 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 
 		msr_info->data = vcpu->arch.msr_kvm_poll_control;
 		break;
+	case MSR_KVM_LAZY_TSCDEADLINE:
+		if (!guest_pv_has(vcpu, KVM_FEATURE_LAZY_TSCDEADLINE))
+			return 1;
+
+		msr_info->data = vcpu->arch.lazy_tscdeadline.msr_val;
+		break;
 	case MSR_IA32_P5_MC_ADDR:
 	case MSR_IA32_P5_MC_TYPE:
 	case MSR_IA32_MCG_CAP:
-- 
2.7.4