KVM: x86: Use static calls to reduce kvm_pmu_ops overhead

Use static calls to improve kvm_pmu_ops performance, following the same
pattern and naming scheme used by kvm-x86-ops.h.

Here are the worst fenced_rdtsc() cycles numbers for the kvm_pmu_ops
functions that is most often called (up to 7 digits of calls) when running
a single perf test case in a guest on an ICX 2.70GHz host (mitigations=on):

		|	legacy	|	static call
------------------------------------------------------------
.pmc_idx_to_pmc	|	1304840	|	994872 (+23%)
.pmc_is_enabled	|	978670	|	1011750 (-3%)
.msr_idx_to_pmc	|	47828	|	41690 (+12%)
.is_valid_msr	|	28786	|	30108 (-4%)

Signed-off-by: Like Xu <likexu@tencent.com>
[sean: Handle static call updates in pmu.c, tweak changelog]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220329235054.3534728-5-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 34 insertions, 22 deletions

diff --git a/arch/x86/kvm/pmu.c b/arch/x86/kvm/pmu.c
index bb0b1ad0fda5..618f529f1c4d 100644
--- a/arch/x86/kvm/pmu.c
+++ b/arch/x86/kvm/pmu.c
@@ -51,14 +51,28 @@
 
 static struct kvm_pmu_ops kvm_pmu_ops __read_mostly;
 
+#define KVM_X86_PMU_OP(func)					     \
+	DEFINE_STATIC_CALL_NULL(kvm_x86_pmu_##func,			     \
+				*(((struct kvm_pmu_ops *)0)->func));
+#define KVM_X86_PMU_OP_OPTIONAL KVM_X86_PMU_OP
+#include <asm/kvm-x86-pmu-ops.h>
+
 void kvm_pmu_ops_update(const struct kvm_pmu_ops *pmu_ops)
 {
 	memcpy(&kvm_pmu_ops, pmu_ops, sizeof(kvm_pmu_ops));
+
+#define __KVM_X86_PMU_OP(func) \
+	static_call_update(kvm_x86_pmu_##func, kvm_pmu_ops.func);
+#define KVM_X86_PMU_OP(func) \
+	WARN_ON(!kvm_pmu_ops.func); __KVM_X86_PMU_OP(func)
+#define KVM_X86_PMU_OP_OPTIONAL __KVM_X86_PMU_OP
+#include <asm/kvm-x86-pmu-ops.h>
+#undef __KVM_X86_PMU_OP
 }
 
 static inline bool pmc_is_enabled(struct kvm_pmc *pmc)
 {
-	return kvm_pmu_ops.pmc_is_enabled(pmc);
+	return static_call(kvm_x86_pmu_pmc_is_enabled)(pmc);
 }
 
 static void kvm_pmi_trigger_fn(struct irq_work *irq_work)
@@ -225,7 +239,7 @@ void reprogram_gp_counter(struct kvm_pmc *pmc, u64 eventsel)
 			  ARCH_PERFMON_EVENTSEL_CMASK |
 			  HSW_IN_TX |
 			  HSW_IN_TX_CHECKPOINTED))) {
-		config = kvm_pmu_ops.pmc_perf_hw_id(pmc);
+		config = static_call(kvm_x86_pmu_pmc_perf_hw_id)(pmc);
 		if (config != PERF_COUNT_HW_MAX)
 			type = PERF_TYPE_HARDWARE;
 	}
@@ -275,7 +289,7 @@ void reprogram_fixed_counter(struct kvm_pmc *pmc, u8 ctrl, int idx)
 
 	pmc->current_config = (u64)ctrl;
 	pmc_reprogram_counter(pmc, PERF_TYPE_HARDWARE,
-			      kvm_pmu_ops.pmc_perf_hw_id(pmc),
+			      static_call(kvm_x86_pmu_pmc_perf_hw_id)(pmc),
 			      !(en_field & 0x2), /* exclude user */
 			      !(en_field & 0x1), /* exclude kernel */
 			      pmi);
@@ -284,7 +298,7 @@ EXPORT_SYMBOL_GPL(reprogram_fixed_counter);
 
 void reprogram_counter(struct kvm_pmu *pmu, int pmc_idx)
 {
-	struct kvm_pmc *pmc = kvm_pmu_ops.pmc_idx_to_pmc(pmu, pmc_idx);
+	struct kvm_pmc *pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, pmc_idx);
 
 	if (!pmc)
 		return;
@@ -306,7 +320,7 @@ void kvm_pmu_handle_event(struct kvm_vcpu *vcpu)
 	int bit;
 
 	for_each_set_bit(bit, pmu->reprogram_pmi, X86_PMC_IDX_MAX) {
-		struct kvm_pmc *pmc = kvm_pmu_ops.pmc_idx_to_pmc(pmu, bit);
+		struct kvm_pmc *pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, bit);
 
 		if (unlikely(!pmc || !pmc->perf_event)) {
 			clear_bit(bit, pmu->reprogram_pmi);
@@ -328,7 +342,7 @@ void kvm_pmu_handle_event(struct kvm_vcpu *vcpu)
 /* check if idx is a valid index to access PMU */
 bool kvm_pmu_is_valid_rdpmc_ecx(struct kvm_vcpu *vcpu, unsigned int idx)
 {
-	return kvm_pmu_ops.is_valid_rdpmc_ecx(vcpu, idx);
+	return static_call(kvm_x86_pmu_is_valid_rdpmc_ecx)(vcpu, idx);
 }
 
 bool is_vmware_backdoor_pmc(u32 pmc_idx)
@@ -378,7 +392,7 @@ int kvm_pmu_rdpmc(struct kvm_vcpu *vcpu, unsigned idx, u64 *data)
 	if (is_vmware_backdoor_pmc(idx))
 		return kvm_pmu_rdpmc_vmware(vcpu, idx, data);
 
-	pmc = kvm_pmu_ops.rdpmc_ecx_to_pmc(vcpu, idx, &mask);
+	pmc = static_call(kvm_x86_pmu_rdpmc_ecx_to_pmc)(vcpu, idx, &mask);
 	if (!pmc)
 		return 1;
 
@@ -394,22 +408,21 @@ int kvm_pmu_rdpmc(struct kvm_vcpu *vcpu, unsigned idx, u64 *data)
 void kvm_pmu_deliver_pmi(struct kvm_vcpu *vcpu)
 {
 	if (lapic_in_kernel(vcpu)) {
-		if (kvm_pmu_ops.deliver_pmi)
-			kvm_pmu_ops.deliver_pmi(vcpu);
+		static_call_cond(kvm_x86_pmu_deliver_pmi)(vcpu);
 		kvm_apic_local_deliver(vcpu->arch.apic, APIC_LVTPC);
 	}
 }
 
 bool kvm_pmu_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr)
 {
-	return kvm_pmu_ops.msr_idx_to_pmc(vcpu, msr) ||
-		kvm_pmu_ops.is_valid_msr(vcpu, msr);
+	return static_call(kvm_x86_pmu_msr_idx_to_pmc)(vcpu, msr) ||
+		static_call(kvm_x86_pmu_is_valid_msr)(vcpu, msr);
 }
 
 static void kvm_pmu_mark_pmc_in_use(struct kvm_vcpu *vcpu, u32 msr)
 {
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
-	struct kvm_pmc *pmc = kvm_pmu_ops.msr_idx_to_pmc(vcpu, msr);
+	struct kvm_pmc *pmc = static_call(kvm_x86_pmu_msr_idx_to_pmc)(vcpu, msr);
 
 	if (pmc)
 		__set_bit(pmc->idx, pmu->pmc_in_use);
@@ -417,13 +430,13 @@ static void kvm_pmu_mark_pmc_in_use(struct kvm_vcpu *vcpu, u32 msr)
 
 int kvm_pmu_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 {
-	return kvm_pmu_ops.get_msr(vcpu, msr_info);
+	return static_call(kvm_x86_pmu_get_msr)(vcpu, msr_info);
 }
 
 int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 {
 	kvm_pmu_mark_pmc_in_use(vcpu, msr_info->index);
-	return kvm_pmu_ops.set_msr(vcpu, msr_info);
+	return static_call(kvm_x86_pmu_set_msr)(vcpu, msr_info);
 }
 
 /* refresh PMU settings. This function generally is called when underlying
@@ -432,7 +445,7 @@ int kvm_pmu_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
  */
 void kvm_pmu_refresh(struct kvm_vcpu *vcpu)
 {
-	kvm_pmu_ops.refresh(vcpu);
+	static_call(kvm_x86_pmu_refresh)(vcpu);
 }
 
 void kvm_pmu_reset(struct kvm_vcpu *vcpu)
@@ -440,7 +453,7 @@ void kvm_pmu_reset(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 
 	irq_work_sync(&pmu->irq_work);
-	kvm_pmu_ops.reset(vcpu);
+	static_call(kvm_x86_pmu_reset)(vcpu);
 }
 
 void kvm_pmu_init(struct kvm_vcpu *vcpu)
@@ -448,7 +461,7 @@ void kvm_pmu_init(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 
 	memset(pmu, 0, sizeof(*pmu));
-	kvm_pmu_ops.init(vcpu);
+	static_call(kvm_x86_pmu_init)(vcpu);
 	init_irq_work(&pmu->irq_work, kvm_pmi_trigger_fn);
 	pmu->event_count = 0;
 	pmu->need_cleanup = false;
@@ -480,14 +493,13 @@ void kvm_pmu_cleanup(struct kvm_vcpu *vcpu)
 		      pmu->pmc_in_use, X86_PMC_IDX_MAX);
 
 	for_each_set_bit(i, bitmask, X86_PMC_IDX_MAX) {
-		pmc = kvm_pmu_ops.pmc_idx_to_pmc(pmu, i);
+		pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, i);
 
 		if (pmc && pmc->perf_event && !pmc_speculative_in_use(pmc))
 			pmc_stop_counter(pmc);
 	}
 
-	if (kvm_pmu_ops.cleanup)
-		kvm_pmu_ops.cleanup(vcpu);
+	static_call_cond(kvm_x86_pmu_cleanup)(vcpu);
 
 	bitmap_zero(pmu->pmc_in_use, X86_PMC_IDX_MAX);
 }
@@ -517,7 +529,7 @@ static inline bool eventsel_match_perf_hw_id(struct kvm_pmc *pmc,
 	unsigned int config;
 
 	pmc->eventsel &= (ARCH_PERFMON_EVENTSEL_EVENT | ARCH_PERFMON_EVENTSEL_UMASK);
-	config = kvm_pmu_ops.pmc_perf_hw_id(pmc);
+	config = static_call(kvm_x86_pmu_pmc_perf_hw_id)(pmc);
 	pmc->eventsel = old_eventsel;
 	return config == perf_hw_id;
 }
@@ -545,7 +557,7 @@ void kvm_pmu_trigger_event(struct kvm_vcpu *vcpu, u64 perf_hw_id)
 	int i;
 
 	for_each_set_bit(i, pmu->all_valid_pmc_idx, X86_PMC_IDX_MAX) {
-		pmc = kvm_pmu_ops.pmc_idx_to_pmc(pmu, i);
+		pmc = static_call(kvm_x86_pmu_pmc_idx_to_pmc)(pmu, i);
 
 		if (!pmc || !pmc_is_enabled(pmc) || !pmc_speculative_in_use(pmc))
 			continue;