Skip to contentSkip to conversation

KVM vPMU code refine

PATCH In review Dapeng Mi 2 participants 12 messages Jun 19, 2024 – Jun 26, 2024

Subsystems (4)
lore
Share & tools
lore

Dapeng Mi

author
#0 lore raw

cover KVM vPMU code refine

This small patchset refines KVM vPMU code and relevant selftests. Patch 1/2 defines new macro KVM_PMC_MAX_GENERIC to avoid the Intel specific macro KVM_INTEL_PMC_MAX_GENERIC to be used in vPMU x86 common code. Patch 2/2 reduces the verbosity of "Random seed" messages to avoid the hugh number of messages to flood the regular output of selftests.

No logic change in this patchset and Kselftests/KUT PMU related tests passed with this patchset.

Dapeng Mi (2):
KVM: x86/pmu: Define KVM_PMC_MAX_GENERIC for platform independence
selftests: kvm: Reduce verbosity of "Random seed" messages

 arch/x86/include/asm/kvm_host.h            | 9 +++++----
 arch/x86/kvm/svm/pmu.c                     | 2 +-
 arch/x86/kvm/vmx/pmu_intel.c               | 2 ++
 tools/testing/selftests/kvm/lib/kvm_util.c | 2 +-
 4 files changed, 9 insertions(+), 6 deletions(-)

base-commit: 0ce958282e66b3d1882e2bb2f503a5e2cebcc3ef

Signature · 1 line
2.34.1
author
#1 lore raw

PATCH 1/2 KVM: x86/pmu: Define KVM_PMC_MAX_GENERIC for platform independence

The existing macro, KVM_INTEL_PMC_MAX_GENERIC, ambiguously represents the maximum supported General Purpose (GP) counter number for both Intel and AMD platforms. This could lead to issues if AMD begins to support more GP counters than Intel.

To resolve this, a new platform-independent macro, KVM_PMC_MAX_GENERIC, is introduced to represent the maximum GP counter number across all x86 platforms.

No logic changes are introduced in this patch.

---
 arch/x86/include/asm/kvm_host.h | 9 +++++----
 arch/x86/kvm/svm/pmu.c          | 2 +-
 arch/x86/kvm/vmx/pmu_intel.c    | 2 ++
 3 files changed, 8 insertions(+), 5 deletions(-)
arch/x86/include/asm/kvm_host.h +5 -4
index 57440bda4dc4..18137be6504a 100644
@@ -534,11 +534,12 @@ struct kvm_pmc {
 
 /* More counters may conflict with other existing Architectural MSRs */
 #define KVM_INTEL_PMC_MAX_GENERIC	8
-#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
-#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
+#define KVM_AMD_PMC_MAX_GENERIC	6
+#define KVM_PMC_MAX_GENERIC		KVM_INTEL_PMC_MAX_GENERIC
+#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_PMC_MAX_GENERIC - 1)
+#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_PMC_MAX_GENERIC - 1)
 #define KVM_PMC_MAX_FIXED	3
 #define MSR_ARCH_PERFMON_FIXED_CTR_MAX	(MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
-#define KVM_AMD_PMC_MAX_GENERIC	6
 
 struct kvm_pmu {
 	u8 version;
@@ -554,7 +555,7 @@ struct kvm_pmu {
 	u64 global_status_rsvd;
 	u64 reserved_bits;
 	u64 raw_event_mask;
-	struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
+	struct kvm_pmc gp_counters[KVM_PMC_MAX_GENERIC];
 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
 
 	/*
arch/x86/kvm/svm/pmu.c +1 -1
index 6e908bdc3310..2fca247798eb 100644
@@ -218,7 +218,7 @@ static void amd_pmu_init(struct kvm_vcpu *vcpu)
 	int i;
 
 	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > AMD64_NUM_COUNTERS_CORE);
-	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > INTEL_PMC_MAX_GENERIC);
+	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
 
 	for (i = 0; i < KVM_AMD_PMC_MAX_GENERIC ; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
arch/x86/kvm/vmx/pmu_intel.c +2 -0
index fb5cbd6cbeff..a4b0bee04596 100644
@@ -570,6 +570,8 @@ static void intel_pmu_init(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
 
+	BUILD_BUG_ON(KVM_INTEL_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
+
 	for (i = 0; i < KVM_INTEL_PMC_MAX_GENERIC; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
 		pmu->gp_counters[i].vcpu = vcpu;
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH 1/2] KVM: x86/pmu: Define KVM_PMC_MAX_GENERIC for platform independence

Quoted 23 lines · Dapeng Mi

The existing macro, KVM_INTEL_PMC_MAX_GENERIC, ambiguously represents the maximum supported General Purpose (GP) counter number for both Intel and AMD platforms. This could lead to issues if AMD begins to support more GP counters than Intel.

To resolve this, a new platform-independent macro, KVM_PMC_MAX_GENERIC, is introduced to represent the maximum GP counter number across all x86 platforms.

No logic changes are introduced in this patch.

---
 arch/x86/include/asm/kvm_host.h | 9 +++++----
 arch/x86/kvm/svm/pmu.c          | 2 +-
 arch/x86/kvm/vmx/pmu_intel.c    | 2 ++
 3 files changed, 8 insertions(+), 5 deletions(-)
arch/x86/include/asm/kvm_host.h +0 -0
index 57440bda4dc4..18137be6504a 100644

arch/x86/include/asm/kvm_host.h struct kvm_pmc { · line 538

Quoted 7 lines
arch/x86/include/asm/kvm_host.h +2 -2
@@ -534,11 +534,12 @@ struct kvm_pmc {
 
 /* More counters may conflict with other existing Architectural MSRs */
 #define KVM_INTEL_PMC_MAX_GENERIC	8
-#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
-#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
+#define KVM_AMD_PMC_MAX_GENERIC	6
+#define KVM_PMC_MAX_GENERIC		KVM_INTEL_PMC_MAX_GENERIC

Since we're changing the macro, maybe take the opportunity to use a better name? E.g. KVM_MAX_NR_GP_COUNTERS? And then in a follow-up patch, give fixed counters the same treatment, e.g. KVM_MAX_NR_FIXED_COUNTERS. Or maybe KVM_MAX_NR_GP_PMCS and KVM_MAX_NR_FIXED_PMCS?

arch/x86/include/asm/kvm_host.h struct kvm_pmc { · line 540

Quoted 2 lines
arch/x86/include/asm/kvm_host.h +2 -0
+#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_PMC_MAX_GENERIC - 1)
+#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_PMC_MAX_GENERIC - 1)

And I'm very, very tempted to say we should simply delete these two, along with MSR_ARCH_PERFMON_FIXED_CTR_MAX, and just open code the "end" MSR in the one user. Especially since "KVM" doesn't appear anyone in the name, i.e. because the names misrepresent KVM's semi-arbitrary max as the *architectural* max.

arch/x86/kvm/x86.c +6 -3
index 6ad19d913d31..547dfe40d017 100644
@@ -7432,17 +7432,20 @@ static void kvm_probe_msr_to_save(u32 msr_index)
                     intel_pt_validate_hw_cap(PT_CAP_num_address_ranges) * 2))
                        return;
                break;
-       case MSR_ARCH_PERFMON_PERFCTR0 ... MSR_ARCH_PERFMON_PERFCTR_MAX:
+       case MSR_ARCH_PERFMON_PERFCTR0 ...
+            MSR_ARCH_PERFMON_PERFCTR0 + KVM_MAX_NR_GP_COUNTERS - 1:
                if (msr_index - MSR_ARCH_PERFMON_PERFCTR0 >=
                    kvm_pmu_cap.num_counters_gp)
                        return;
                break;
-       case MSR_ARCH_PERFMON_EVENTSEL0 ... MSR_ARCH_PERFMON_EVENTSEL_MAX:
+       case MSR_ARCH_PERFMON_EVENTSEL0 ...
+            MSR_ARCH_PERFMON_EVENTSEL0 + KVM_MAX_NR_GP_COUNTERS - 1:
                if (msr_index - MSR_ARCH_PERFMON_EVENTSEL0 >=
                    kvm_pmu_cap.num_counters_gp)
                        return;
                break;
-       case MSR_ARCH_PERFMON_FIXED_CTR0 ... MSR_ARCH_PERFMON_FIXED_CTR_MAX:
+       case MSR_ARCH_PERFMON_FIXED_CTR0 ...
+            MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_MAR_NR_FIXED_COUNTERS - 1:
                if (msr_index - MSR_ARCH_PERFMON_FIXED_CTR0 >=
                    kvm_pmu_cap.num_counters_fixed)
                        return;
Quoted 33 lines
arch/x86/include/asm/kvm_host.h +1 -2
 #define KVM_PMC_MAX_FIXED	3
 #define MSR_ARCH_PERFMON_FIXED_CTR_MAX	(MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
-#define KVM_AMD_PMC_MAX_GENERIC	6
 
 struct kvm_pmu {
 	u8 version;
@@ -554,7 +555,7 @@ struct kvm_pmu {
 	u64 global_status_rsvd;
 	u64 reserved_bits;
 	u64 raw_event_mask;
-	struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
+	struct kvm_pmc gp_counters[KVM_PMC_MAX_GENERIC];
 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
 
 	/*
arch/x86/kvm/svm/pmu.c +1 -1
index 6e908bdc3310..2fca247798eb 100644
@@ -218,7 +218,7 @@ static void amd_pmu_init(struct kvm_vcpu *vcpu)
 	int i;
 
 	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > AMD64_NUM_COUNTERS_CORE);
-	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > INTEL_PMC_MAX_GENERIC);
+	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
 
 	for (i = 0; i < KVM_AMD_PMC_MAX_GENERIC ; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
arch/x86/kvm/vmx/pmu_intel.c +0 -0
index fb5cbd6cbeff..a4b0bee04596 100644

arch/x86/kvm/vmx/pmu_intel.c static void intel_pmu_init(struct kvm_vcpu *vcpu) · line 573

Quoted 4 lines
arch/x86/kvm/vmx/pmu_intel.c +1 -0
@@ -570,6 +570,8 @@ static void intel_pmu_init(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
 
+	BUILD_BUG_ON(KVM_INTEL_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);

Rather than BUILD_BUG_ON() for both Intel and AMD, can't we just do?

#define KVM_MAX_NR_GP_COUNTERS max(KVM_INTEL_PMC_MAX_GENERIC, KVM_AMD_PMC_MAX_GENERIC)

Quoted 7 lines
arch/x86/kvm/vmx/pmu_intel.c +1 -0
+
 	for (i = 0; i < KVM_INTEL_PMC_MAX_GENERIC; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
 		pmu->gp_counters[i].vcpu = vcpu;
Signature · 1 line
2.34.1
author
#3 lore raw

Re: [PATCH 1/2] KVM: x86/pmu: Define KVM_PMC_MAX_GENERIC for platform independence

Quoted 25 lines · Sean Christopherson
Quoted 24 lines · Dapeng Mi

The existing macro, KVM_INTEL_PMC_MAX_GENERIC, ambiguously represents the maximum supported General Purpose (GP) counter number for both Intel and AMD platforms. This could lead to issues if AMD begins to support more GP counters than Intel.

To resolve this, a new platform-independent macro, KVM_PMC_MAX_GENERIC, is introduced to represent the maximum GP counter number across all x86 platforms.

No logic changes are introduced in this patch.

---
 arch/x86/include/asm/kvm_host.h | 9 +++++----
 arch/x86/kvm/svm/pmu.c          | 2 +-
 arch/x86/kvm/vmx/pmu_intel.c    | 2 ++
 3 files changed, 8 insertions(+), 5 deletions(-)
arch/x86/include/asm/kvm_host.h +0 -0
index 57440bda4dc4..18137be6504a 100644
@@ -534,11 +534,12 @@ struct kvm_pmc {
 
Quoted 10 lines

arch/x86/include/asm/kvm_host.h struct kvm_pmc { · line 538

Quoted 6 lines
arch/x86/include/asm/kvm_host.h +2 -2
 /* More counters may conflict with other existing Architectural MSRs */
 #define KVM_INTEL_PMC_MAX_GENERIC	8
-#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
-#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
+#define KVM_AMD_PMC_MAX_GENERIC	6
+#define KVM_PMC_MAX_GENERIC		KVM_INTEL_PMC_MAX_GENERIC

Since we're changing the macro, maybe take the opportunity to use a better name? E.g. KVM_MAX_NR_GP_COUNTERS? And then in a follow-up patch, give fixed counters the same treatment, e.g. KVM_MAX_NR_FIXED_COUNTERS. Or maybe KVM_MAX_NR_GP_PMCS and KVM_MAX_NR_FIXED_PMCS?

Yeah, would change to KVM_MAX_NR_GP_COUNTERS and KVM_MAX_NR_FIXED_COUNTERS in next version.

Quoted 7 lines

arch/x86/include/asm/kvm_host.h struct kvm_pmc { · line 540

Quoted 3 lines
arch/x86/include/asm/kvm_host.h +2 -0
+#define MSR_ARCH_PERFMON_PERFCTR_MAX	(MSR_ARCH_PERFMON_PERFCTR0 + KVM_PMC_MAX_GENERIC - 1)
+#define MSR_ARCH_PERFMON_EVENTSEL_MAX	(MSR_ARCH_PERFMON_EVENTSEL0 + KVM_PMC_MAX_GENERIC - 1)

And I'm very, very tempted to say we should simply delete these two, along with MSR_ARCH_PERFMON_FIXED_CTR_MAX, and just open code the "end" MSR in the one user. Especially since "KVM" doesn't appear anyone in the name, i.e. because the names misrepresent KVM's semi-arbitrary max as the *architectural* max.

Agree. MSR_ARCH_PERFMON_PERFCTR_MAX indeed brings some misleading which make users think it's a HW's limitation instead of KVM's limitation.

Quoted 19 lines
arch/x86/kvm/x86.c +4 -2
index 6ad19d913d31..547dfe40d017 100644
@@ -7432,17 +7432,20 @@ static void kvm_probe_msr_to_save(u32 msr_index)
                     intel_pt_validate_hw_cap(PT_CAP_num_address_ranges) * 2))
                        return;
                break;
-       case MSR_ARCH_PERFMON_PERFCTR0 ... MSR_ARCH_PERFMON_PERFCTR_MAX:
+       case MSR_ARCH_PERFMON_PERFCTR0 ...
+            MSR_ARCH_PERFMON_PERFCTR0 + KVM_MAX_NR_GP_COUNTERS - 1:
                if (msr_index - MSR_ARCH_PERFMON_PERFCTR0 >=
                    kvm_pmu_cap.num_counters_gp)
                        return;
                break;
-       case MSR_ARCH_PERFMON_EVENTSEL0 ... MSR_ARCH_PERFMON_EVENTSEL_MAX:
+       case MSR_ARCH_PERFMON_EVENTSEL0 ...
+            MSR_ARCH_PERFMON_EVENTSEL0 + KVM_MAX_NR_GP_COUNTERS - 1:

arch/x86/kvm/x86.c static void kvm_probe_msr_to_save(u32 msr_index) · line 7451

Quoted 10 lines
arch/x86/kvm/x86.c +2 -1
                if (msr_index - MSR_ARCH_PERFMON_EVENTSEL0 >=
                    kvm_pmu_cap.num_counters_gp)
                        return;
                break;
-       case MSR_ARCH_PERFMON_FIXED_CTR0 ... MSR_ARCH_PERFMON_FIXED_CTR_MAX:
+       case MSR_ARCH_PERFMON_FIXED_CTR0 ...
+            MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_MAR_NR_FIXED_COUNTERS - 1:
                if (msr_index - MSR_ARCH_PERFMON_FIXED_CTR0 >=
                    kvm_pmu_cap.num_counters_fixed)
                        return;

Thanks.

Quoted 37 lines
Quoted 37 lines
arch/x86/include/asm/kvm_host.h +1 -2
 #define KVM_PMC_MAX_FIXED	3
 #define MSR_ARCH_PERFMON_FIXED_CTR_MAX	(MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
-#define KVM_AMD_PMC_MAX_GENERIC	6
 
 struct kvm_pmu {
 	u8 version;
@@ -554,7 +555,7 @@ struct kvm_pmu {
 	u64 global_status_rsvd;
 	u64 reserved_bits;
 	u64 raw_event_mask;
-	struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
+	struct kvm_pmc gp_counters[KVM_PMC_MAX_GENERIC];
 	struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
 
 	/*
arch/x86/kvm/svm/pmu.c +1 -1
index 6e908bdc3310..2fca247798eb 100644
@@ -218,7 +218,7 @@ static void amd_pmu_init(struct kvm_vcpu *vcpu)
 	int i;
 
 	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > AMD64_NUM_COUNTERS_CORE);
-	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > INTEL_PMC_MAX_GENERIC);
+	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
 
 	for (i = 0; i < KVM_AMD_PMC_MAX_GENERIC ; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
arch/x86/kvm/vmx/pmu_intel.c +0 -0
index fb5cbd6cbeff..a4b0bee04596 100644
@@ -570,6 +570,8 @@ static void intel_pmu_init(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
 
Quoted 4 lines

arch/x86/kvm/vmx/pmu_intel.c static void intel_pmu_init(struct kvm_vcpu *vcpu) · line 573

Quoted 1 line
arch/x86/kvm/vmx/pmu_intel.c +1 -0
+	BUILD_BUG_ON(KVM_INTEL_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);

Rather than BUILD_BUG_ON() for both Intel and AMD, can't we just do?

#define KVM_MAX_NR_GP_COUNTERS max(KVM_INTEL_PMC_MAX_GENERIC, KVM_AMD_PMC_MAX_GENERIC)

Actually I tried this, but compiler would report the below error since KVM_PMC_MAX_GENERIC would used to define the array gp_counters[KVM_PMC_MAX_GENERIC];

./include/linux/minmax.h:48:50: error: braced-group within expression allowed only inside a function

Quoted 8 lines
Quoted 8 lines
arch/x86/kvm/vmx/pmu_intel.c +1 -0
+
 	for (i = 0; i < KVM_INTEL_PMC_MAX_GENERIC; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
 		pmu->gp_counters[i].vcpu = vcpu;
Signature · 1 line
2.34.1
#4 lore raw

Re: [PATCH 1/2] KVM: x86/pmu: Define KVM_PMC_MAX_GENERIC for platform independence

Quoted 24 lines · Dapeng Mi
Quoted 24 lines
Quoted 18 lines
arch/x86/kvm/svm/pmu.c +1 -1
index 6e908bdc3310..2fca247798eb 100644
@@ -218,7 +218,7 @@ static void amd_pmu_init(struct kvm_vcpu *vcpu)
 	int i;
 
 	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > AMD64_NUM_COUNTERS_CORE);
-	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > INTEL_PMC_MAX_GENERIC);
+	BUILD_BUG_ON(KVM_AMD_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);
 
 	for (i = 0; i < KVM_AMD_PMC_MAX_GENERIC ; i++) {
 		pmu->gp_counters[i].type = KVM_PMC_GP;
arch/x86/kvm/vmx/pmu_intel.c +0 -0
index fb5cbd6cbeff..a4b0bee04596 100644

arch/x86/kvm/vmx/pmu_intel.c static void intel_pmu_init(struct kvm_vcpu *vcpu) · line 573

Quoted 4 lines
arch/x86/kvm/vmx/pmu_intel.c +1 -0
@@ -570,6 +570,8 @@ static void intel_pmu_init(struct kvm_vcpu *vcpu)
 	struct kvm_pmu *pmu = vcpu_to_pmu(vcpu);
 	struct lbr_desc *lbr_desc = vcpu_to_lbr_desc(vcpu);
 
+	BUILD_BUG_ON(KVM_INTEL_PMC_MAX_GENERIC > KVM_PMC_MAX_GENERIC);

Rather than BUILD_BUG_ON() for both Intel and AMD, can't we just do?

Quoted 8 lines
Quoted 1 line

#define KVM_MAX_NR_GP_COUNTERS max(KVM_INTEL_PMC_MAX_GENERIC, KVM_AMD_PMC_MAX_GENERIC)

Actually I tried this, but compiler would report the below error since KVM_PMC_MAX_GENERIC would used to define the array gp_counters[KVM_PMC_MAX_GENERIC];

./include/linux/minmax.h:48:50: error: braced-group within expression allowed only inside a function

Oh, right, the min/max macros are super fancy to deal with types. How about this (getting there over 2-3 patches)?

I don't love the "#define MAX", but I don't see a better option. I suppose maybe we should use __MAX or KVM_MAX since kvm_host.h is likely included outside of KVM?

arch/x86/include/asm/kvm_host.h +13 -8
index 5c0415899a07..ce0c9191eb85 100644
@@ -532,13 +532,18 @@ struct kvm_pmc {
        u64 current_config;
 };
 
+#define MAX(a, b) ((a) >= (b) ? (a) : (b))
+
 /* More counters may conflict with other existing Architectural MSRs */
-#define KVM_INTEL_PMC_MAX_GENERIC      8
-#define MSR_ARCH_PERFMON_PERFCTR_MAX   (MSR_ARCH_PERFMON_PERFCTR0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
-#define MSR_ARCH_PERFMON_EVENTSEL_MAX  (MSR_ARCH_PERFMON_EVENTSEL0 + KVM_INTEL_PMC_MAX_GENERIC - 1)
-#define KVM_PMC_MAX_FIXED      3
-#define MSR_ARCH_PERFMON_FIXED_CTR_MAX (MSR_ARCH_PERFMON_FIXED_CTR0 + KVM_PMC_MAX_FIXED - 1)
-#define KVM_AMD_PMC_MAX_GENERIC        6
+#define KVM_MAX_NR_INTEL_GP_COUNTERS   8
+#define KVM_MAX_NR_AMD_GP_COUNTERS     6
+#define KVM_MAX_NR_GP_COUNTERS         MAX(KVM_MAX_NR_INTEL_GP_COUNTERS, \
+                                           KVM_MAX_NR_AMD_GP_COUNTERS)
+
+#define KVM_MAX_NR_INTEL_FIXED_COUTNERS        3
+#define KVM_MAX_NR_AMD_FIXED_COUTNERS  0
+#define KVM_MAX_NR_FIXED_COUNTERS      MAX(KVM_MAX_NR_INTEL_FIXED_COUTNERS, \
+                                           KVM_MAX_NR_AMD_FIXED_COUTNERS)
 
 struct kvm_pmu {
        u8 version;
@@ -554,8 +559,8 @@ struct kvm_pmu {
        u64 global_status_rsvd;
        u64 reserved_bits;
        u64 raw_event_mask;
-       struct kvm_pmc gp_counters[KVM_INTEL_PMC_MAX_GENERIC];
-       struct kvm_pmc fixed_counters[KVM_PMC_MAX_FIXED];
+       struct kvm_pmc gp_counters[KVM_MAX_NR_GP_COUNTERS];
+       struct kvm_pmc fixed_counters[KVM_MAX_NR_FIXED_COUNTERS];
 
        /*
         * Overlay the bitmap with a 64-bit atomic so that all bits can be
author
#5 lore raw

PATCH 2/2 selftests: kvm: Reduce verbosity of "Random seed" messages

In reply to Dapeng Mi

Huge number of "Random seed:" messages are printed when running pmu_counters_test. It leads to the regular test output is totally flooded by these over-verbose messages.

Downgrade "Random seed" message printing level to debug and prevent it to be printed in normal case.

---
 tools/testing/selftests/kvm/lib/kvm_util.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/testing/selftests/kvm/lib/kvm_util.c +1 -1
index ad00e4761886..8568c7d619c3 100644
@@ -434,7 +434,7 @@ struct kvm_vm *__vm_create(struct vm_shape shape, uint32_t nr_runnable_vcpus,
 	slot0 = memslot2region(vm, 0);
 	ucall_init(vm, slot0->region.guest_phys_addr + slot0->region.memory_size);
 
-	pr_info("Random seed: 0x%x\n", guest_random_seed);
+	pr_debug("Random seed: 0x%x\n", guest_random_seed);
 	guest_rng = new_guest_random_state(guest_random_seed);
 	sync_global_to_guest(vm, guest_rng);
 
Signature · 1 line
2.34.1
#6 lore raw

Re: [PATCH 2/2] selftests: kvm: Reduce verbosity of "Random seed" messages

Quoted 6 lines · Dapeng Mi

Huge number of "Random seed:" messages are printed when running pmu_counters_test. It leads to the regular test output is totally flooded by these over-verbose messages.

Downgrade "Random seed" message printing level to debug and prevent it to be printed in normal case.

I completely agree this is annoying, but the whole point of printing the seed is so that the seed is automatically captured if a test fails, e.g. so that the failure can be reproduced if it is dependent on some random decision.

Rather than simply hiding the message, what if print the seed if and only if it changes?

--
From: Sean Christopherson <[email protected]>
Date: Thu, 20 Jun 2024 10:29:53 -0700
Subject: [PATCH] KVM: selftests: Print the seed for the guest pRNG iff it has
changed

Print the guest's random seed during VM creation if and only if the seed has changed since the seed was last printed. The vast majority of tests, if not all tests at this point, set the seed during test initialization and never change the seed, i.e. printing it every time a VM is created is useless noise.

Snapshot and print the seed during early selftest init to play nice with tests that use the kselftests harness, at the cost of printing an unused seed for tests that change the seed during test-specific initialization, e.g. dirty_log_perf_test. The kselftests harness runs each testcase in a separate process that is forked from the original process before creating each testcase's VM, i.e. waiting until first VM creation will result in the seed being printed by each testcase despite it never changing. And long term, the hope/goal is that setting the seed will be handled by the core framework, i.e. that the dirty_log_perf_test wart will naturally go away.

---
 tools/testing/selftests/kvm/lib/kvm_util.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
tools/testing/selftests/kvm/lib/kvm_util.c +7 -2
index ad00e4761886..56b170b725b3 100644
@@ -21,6 +21,7 @@
 
 uint32_t guest_random_seed;
 struct guest_random_state guest_rng;
+static uint32_t last_guest_seed;
 
 static int vcpu_mmap_sz(void);
 
@@ -434,7 +435,10 @@ struct kvm_vm *__vm_create(struct vm_shape shape, uint32_t nr_runnable_vcpus,
 	slot0 = memslot2region(vm, 0);
 	ucall_init(vm, slot0->region.guest_phys_addr + slot0->region.memory_size);
 
-	pr_info("Random seed: 0x%x\n", guest_random_seed);
+	if (guest_random_seed != last_guest_seed) {
+		pr_info("Random seed: 0x%x\n", guest_random_seed);
+		last_guest_seed = guest_random_seed;
+	}
 	guest_rng = new_guest_random_state(guest_random_seed);
 	sync_global_to_guest(vm, guest_rng);
 
@@ -2319,7 +2323,8 @@ void __attribute((constructor)) kvm_selftest_init(void)
 	/* Tell stdout not to buffer its content. */
 	setbuf(stdout, NULL);
 
-	guest_random_seed = random();
+	guest_random_seed = last_guest_seed = random();
+	pr_info("Random seed: 0x%x\n", guest_random_seed);
 
 	kvm_selftest_arch_init();
 }

base-commit: c81b138d5075c6f5ba3419ac1d2a2e7047719c14

author
#7 lore raw

Re: [PATCH 2/2] selftests: kvm: Reduce verbosity of "Random seed" messages

Quoted 5 lines · Sean Christopherson
Quoted 4 lines · Dapeng Mi

Huge number of "Random seed:" messages are printed when running pmu_counters_test. It leads to the regular test output is totally flooded by these over-verbose messages.

Quoted 8 lines
Quoted 2 lines

Downgrade "Random seed" message printing level to debug and prevent it to be printed in normal case.

I completely agree this is annoying, but the whole point of printing the seed is so that the seed is automatically captured if a test fails, e.g. so that the failure can be reproduced if it is dependent on some random decision.

Rather than simply hiding the message, what if print the seed if and only if it changes?

Yeah, it's indeed better.

Quoted 6 lines

--
From: Sean Christopherson <[email protected]>
Date: Thu, 20 Jun 2024 10:29:53 -0700
Subject: [PATCH] KVM: selftests: Print the seed for the guest pRNG iff it has
changed

s/iff/if/

Quoted 61 lines

Print the guest's random seed during VM creation if and only if the seed has changed since the seed was last printed. The vast majority of tests, if not all tests at this point, set the seed during test initialization and never change the seed, i.e. printing it every time a VM is created is useless noise.

Snapshot and print the seed during early selftest init to play nice with tests that use the kselftests harness, at the cost of printing an unused seed for tests that change the seed during test-specific initialization, e.g. dirty_log_perf_test. The kselftests harness runs each testcase in a separate process that is forked from the original process before creating each testcase's VM, i.e. waiting until first VM creation will result in the seed being printed by each testcase despite it never changing. And long term, the hope/goal is that setting the seed will be handled by the core framework, i.e. that the dirty_log_perf_test wart will naturally go away.

---
 tools/testing/selftests/kvm/lib/kvm_util.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
tools/testing/selftests/kvm/lib/kvm_util.c +7 -2
index ad00e4761886..56b170b725b3 100644
@@ -21,6 +21,7 @@
 
 uint32_t guest_random_seed;
 struct guest_random_state guest_rng;
+static uint32_t last_guest_seed;
 
 static int vcpu_mmap_sz(void);
 
@@ -434,7 +435,10 @@ struct kvm_vm *__vm_create(struct vm_shape shape, uint32_t nr_runnable_vcpus,
 	slot0 = memslot2region(vm, 0);
 	ucall_init(vm, slot0->region.guest_phys_addr + slot0->region.memory_size);
 
-	pr_info("Random seed: 0x%x\n", guest_random_seed);
+	if (guest_random_seed != last_guest_seed) {
+		pr_info("Random seed: 0x%x\n", guest_random_seed);
+		last_guest_seed = guest_random_seed;
+	}
 	guest_rng = new_guest_random_state(guest_random_seed);
 	sync_global_to_guest(vm, guest_rng);
 
@@ -2319,7 +2323,8 @@ void __attribute((constructor)) kvm_selftest_init(void)
 	/* Tell stdout not to buffer its content. */
 	setbuf(stdout, NULL);
 
-	guest_random_seed = random();
+	guest_random_seed = last_guest_seed = random();
+	pr_info("Random seed: 0x%x\n", guest_random_seed);
 
 	kvm_selftest_arch_init();
 }

base-commit: c81b138d5075c6f5ba3419ac1d2a2e7047719c14

#8 lore raw

Re: [PATCH 2/2] selftests: kvm: Reduce verbosity of "Random seed" messages

Quoted 7 lines · Dapeng Mi
Quoted 5 lines

--
From: Sean Christopherson <[email protected]>
Date: Thu, 20 Jun 2024 10:29:53 -0700
Subject: [PATCH] KVM: selftests: Print the seed for the guest pRNG iff it has
changed

s/iff/if/

"iff" is shorthand for "if and only if". I try to write out the full "if and only if" when possible, but use "iff" in shortlogs when I want to squeeze in more words.

author
#9 lore raw

Re: [PATCH 2/2] selftests: kvm: Reduce verbosity of "Random seed" messages

Quoted 9 lines · Sean Christopherson
Quoted 6 lines · Dapeng Mi
Quoted 5 lines

--
From: Sean Christopherson <[email protected]>
Date: Thu, 20 Jun 2024 10:29:53 -0700
Subject: [PATCH] KVM: selftests: Print the seed for the guest pRNG iff it has
changed

s/iff/if/

"iff" is shorthand for "if and only if". I try to write out the full "if and only if" when possible, but use "iff" in shortlogs when I want to squeeze in more words.

Good to know. Thanks.

#10 lore raw
In reply to Dapeng Mi
Quoted 5 lines · Dapeng Mi

This small patchset refines KVM vPMU code and relevant selftests. Patch 1/2 defines new macro KVM_PMC_MAX_GENERIC to avoid the Intel specific macro KVM_INTEL_PMC_MAX_GENERIC to be used in vPMU x86 common code. Patch 2/2 reduces the verbosity of "Random seed" messages to avoid the hugh number of messages to flood the regular output of selftests.

In the future, please post these as separate patches, they are completely unrelated.

author
#11 lore raw
Quoted 7 lines · Sean Christopherson
Quoted 5 lines · Dapeng Mi

This small patchset refines KVM vPMU code and relevant selftests. Patch 1/2 defines new macro KVM_PMC_MAX_GENERIC to avoid the Intel specific macro KVM_INTEL_PMC_MAX_GENERIC to be used in vPMU x86 common code. Patch 2/2 reduces the verbosity of "Random seed" messages to avoid the hugh number of messages to flood the regular output of selftests.

In the future, please post these as separate patches, they are completely unrelated.

Sure. Thanks.