[RESEND PATCH 1/4] perf/x86/intel/cstate: Cleanup duplicate attr_groups

[RESEND PATCH 1/4] perf/x86/intel/cstate: Cleanup duplicate attr_groups

[kan.liang]

From: Kan Liang <kan.liang@linux.intel.com>

The events of the cstate_core and cstate_pkg PMU have the same format.
They both need to create a "events" group (with empty attrs). The
attr_groups can be shared.

Remove the dedicated attr_groups for each cstate PMU. Use the shared
cstate_attr_groups to replace.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
---
 arch/x86/events/intel/cstate.c | 44 +++++++++-------------------------
 1 file changed, 11 insertions(+), 33 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index cbeb6d2bf5b4..693bdcd92e8c 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -189,20 +189,20 @@ static struct attribute *attrs_empty[] = {
  * "events" group (with empty attrs) before updating
  * it with detected events.
  */
-static struct attribute_group core_events_attr_group = {
+static struct attribute_group cstate_events_attr_group = {
 	.name = "events",
 	.attrs = attrs_empty,
 };
 
-DEFINE_CSTATE_FORMAT_ATTR(core_event, event, "config:0-63");
-static struct attribute *core_format_attrs[] = {
-	&format_attr_core_event.attr,
+DEFINE_CSTATE_FORMAT_ATTR(cstate_event, event, "config:0-63");
+static struct attribute *cstate_format_attrs[] = {
+	&format_attr_cstate_event.attr,
 	NULL,
 };
 
-static struct attribute_group core_format_attr_group = {
+static struct attribute_group cstate_format_attr_group = {
 	.name = "format",
-	.attrs = core_format_attrs,
+	.attrs = cstate_format_attrs,
 };
 
 static cpumask_t cstate_core_cpu_mask;
@@ -217,9 +217,9 @@ static struct attribute_group cpumask_attr_group = {
 	.attrs = cstate_cpumask_attrs,
 };
 
-static const struct attribute_group *core_attr_groups[] = {
-	&core_events_attr_group,
-	&core_format_attr_group,
+static const struct attribute_group *cstate_attr_groups[] = {
+	&cstate_events_attr_group,
+	&cstate_format_attr_group,
 	&cpumask_attr_group,
 	NULL,
 };
@@ -268,30 +268,8 @@ static struct perf_msr pkg_msr[] = {
 	[PERF_CSTATE_PKG_C10_RES] = { MSR_PKG_C10_RESIDENCY,	&group_cstate_pkg_c10,	test_msr },
 };
 
-static struct attribute_group pkg_events_attr_group = {
-	.name = "events",
-	.attrs = attrs_empty,
-};
-
-DEFINE_CSTATE_FORMAT_ATTR(pkg_event, event, "config:0-63");
-static struct attribute *pkg_format_attrs[] = {
-	&format_attr_pkg_event.attr,
-	NULL,
-};
-static struct attribute_group pkg_format_attr_group = {
-	.name = "format",
-	.attrs = pkg_format_attrs,
-};
-
 static cpumask_t cstate_pkg_cpu_mask;
 
-static const struct attribute_group *pkg_attr_groups[] = {
-	&pkg_events_attr_group,
-	&pkg_format_attr_group,
-	&cpumask_attr_group,
-	NULL,
-};
-
 static ssize_t cstate_get_attr_cpumask(struct device *dev,
 				       struct device_attribute *attr,
 				       char *buf)
@@ -478,7 +456,7 @@ static const struct attribute_group *pkg_attr_update[] = {
 };
 
 static struct pmu cstate_core_pmu = {
-	.attr_groups	= core_attr_groups,
+	.attr_groups	= cstate_attr_groups,
 	.attr_update	= core_attr_update,
 	.name		= "cstate_core",
 	.task_ctx_nr	= perf_invalid_context,
@@ -493,7 +471,7 @@ static struct pmu cstate_core_pmu = {
 };
 
 static struct pmu cstate_pkg_pmu = {
-	.attr_groups	= pkg_attr_groups,
+	.attr_groups	= cstate_attr_groups,
 	.attr_update	= pkg_attr_update,
 	.name		= "cstate_pkg",
 	.task_ctx_nr	= perf_invalid_context,
-- 
2.35.1

[RESEND PATCH 2/4] x86/smp: Export symbol cpu_clustergroup_mask

[kan.liang]

From: Kan Liang <kan.liang@linux.intel.com>

Intel cstate PMU driver will invoke the topology_cluster_cpumask() to
retrieve the CPU mask of a cluster. A modpost error is triggered since
the symbol cpu_clustergroup_mask is not exported.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
---
 arch/x86/kernel/smpboot.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 48e040618731..376387ca6be5 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -747,6 +747,7 @@ const struct cpumask *cpu_clustergroup_mask(int cpu)
 {
 	return cpu_l2c_shared_mask(cpu);
 }
+EXPORT_SYMBOL(cpu_clustergroup_mask);
 
 static void impress_friends(void)
 {
-- 
2.35.1

Re: [RESEND PATCH 2/4] x86/smp: Export symbol cpu_clustergroup_mask

[Peter Zijlstra]

On Thu, Nov 16, 2023 at 06:22:43AM -0800, kan.liang@linux.intel.com wrote:
> From: Kan Liang <kan.liang@linux.intel.com>
> 
> Intel cstate PMU driver will invoke the topology_cluster_cpumask() to
> retrieve the CPU mask of a cluster. A modpost error is triggered since
> the symbol cpu_clustergroup_mask is not exported.
> 
> Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
> ---
>  arch/x86/kernel/smpboot.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
> index 48e040618731..376387ca6be5 100644
> --- a/arch/x86/kernel/smpboot.c
> +++ b/arch/x86/kernel/smpboot.c
> @@ -747,6 +747,7 @@ const struct cpumask *cpu_clustergroup_mask(int cpu)
>  {
>  	return cpu_l2c_shared_mask(cpu);
>  }
> +EXPORT_SYMBOL(cpu_clustergroup_mask);

I made that _GPL.

Queued the lot.

Re: [RESEND PATCH 2/4] x86/smp: Export symbol cpu_clustergroup_mask

[Liang, Kan]

On 2023-11-16 9:40 a.m., Peter Zijlstra wrote:
> On Thu, Nov 16, 2023 at 06:22:43AM -0800, kan.liang@linux.intel.com wrote:
>> From: Kan Liang <kan.liang@linux.intel.com>
>>
>> Intel cstate PMU driver will invoke the topology_cluster_cpumask() to
>> retrieve the CPU mask of a cluster. A modpost error is triggered since
>> the symbol cpu_clustergroup_mask is not exported.
>>
>> Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
>> ---
>>  arch/x86/kernel/smpboot.c | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
>> index 48e040618731..376387ca6be5 100644
>> --- a/arch/x86/kernel/smpboot.c
>> +++ b/arch/x86/kernel/smpboot.c
>> @@ -747,6 +747,7 @@ const struct cpumask *cpu_clustergroup_mask(int cpu)
>>  {
>>  	return cpu_l2c_shared_mask(cpu);
>>  }
>> +EXPORT_SYMBOL(cpu_clustergroup_mask);
> 
> I made that _GPL.

Sure.
> 
> Queued the lot.

Thanks!

Kan

[tip: perf/core] x86/smp: Export symbol cpu_clustergroup_mask()

[tip-bot2 for Kan Liang]

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     c3dd1995620cdcd65cf4944c4164b0dbc16e557c
Gitweb:        https://git.kernel.org/tip/c3dd1995620cdcd65cf4944c4164b0dbc16e557c
Author:        Kan Liang <kan.liang@linux.intel.com>
AuthorDate:    Thu, 16 Nov 2023 06:22:43 -08:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 17 Nov 2023 10:54:52 +01:00

x86/smp: Export symbol cpu_clustergroup_mask()

Intel cstate PMU driver will invoke the topology_cluster_cpumask() to
retrieve the CPU mask of a cluster. A modpost error is triggered since
the symbol cpu_clustergroup_mask is not exported.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20231116142245.1233485-2-kan.liang@linux.intel.com
---
 arch/x86/kernel/smpboot.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2cc2aa1..3f57ce6 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -757,6 +757,7 @@ const struct cpumask *cpu_clustergroup_mask(int cpu)
 {
 	return cpu_l2c_shared_mask(cpu);
 }
+EXPORT_SYMBOL_GPL(cpu_clustergroup_mask);
 
 static void impress_friends(void)
 {

[RESEND PATCH 3/4] perf/x86/intel/cstate: Add Sierra Forest support

[kan.liang]

From: Kan Liang <kan.liang@linux.intel.com>

A new module C6 Residency Counter is introduced in the Sierra Forest.
The scope of the new counter is module (A cluster of cores shared L2
cache). Create a brand new cstate_module PMU to profile the new counter.
The only differences between the new cstate_module PMU and the existing
cstate PMU are the scope and events.

Regarding the choice of the new cstate_module PMU name, the current
naming rule of a cstate PMU is "cstate_" + the scope of the PMU. The
scope of the PMU is the cores shared L2. On SRF, Intel calls it
"module", while the internal Linux sched code calls it "cluster". The
"cstate_module" is used as the new PMU name, because
- The Cstate PMU driver is a Intel specific driver. It doesn't impact
  other ARCHs. The name makes it consistent with the documentation.
- The "cluster" mainly be used by the scheduler developer, while the
  user of cstate PMU is more likely a researcher reading HW docs and
  optimizing power.
- In the Intel's SDM, the "cluster" has a different meaning/scope for
  topology. Using it will mislead the end users.

Besides the module C6, the core C1/C6 and pkg C6 residency counters are
supported in the Sierra Forest as well.

Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
---
 arch/x86/events/intel/cstate.c | 113 +++++++++++++++++++++++++++++++--
 1 file changed, 109 insertions(+), 4 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index 693bdcd92e8c..4a46ef315284 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -41,7 +41,7 @@
  *	MSR_CORE_C1_RES: CORE C1 Residency Counter
  *			 perf code: 0x00
  *			 Available model: SLM,AMT,GLM,CNL,ICX,TNT,ADL,RPL
- *					  MTL
+ *					  MTL,SRF
  *			 Scope: Core (each processor core has a MSR)
  *	MSR_CORE_C3_RESIDENCY: CORE C3 Residency Counter
  *			       perf code: 0x01
@@ -52,7 +52,7 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
  *			       Scope: Core
  *	MSR_CORE_C7_RESIDENCY: CORE C7 Residency Counter
  *			       perf code: 0x03
@@ -75,7 +75,7 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
  *			       Scope: Package (physical package)
  *	MSR_PKG_C7_RESIDENCY:  Package C7 Residency Counter.
  *			       perf code: 0x03
@@ -97,6 +97,10 @@
  *			       Available model: HSW ULT,KBL,GLM,CNL,CML,ICL,TGL,
  *						TNT,RKL,ADL,RPL,MTL
  *			       Scope: Package (physical package)
+ *	MSR_MODULE_C6_RES_MS:  Module C6 Residency Counter.
+ *			       perf code: 0x00
+ *			       Available model: SRF
+ *			       Scope: A cluster of cores shared L2 cache
  *
  */
 
@@ -130,6 +134,7 @@ static ssize_t cstate_get_attr_cpumask(struct device *dev,
 struct cstate_model {
 	unsigned long		core_events;
 	unsigned long		pkg_events;
+	unsigned long		module_events;
 	unsigned long		quirks;
 };
 
@@ -270,6 +275,28 @@ static struct perf_msr pkg_msr[] = {
 
 static cpumask_t cstate_pkg_cpu_mask;
 
+/* cstate_module PMU */
+static struct pmu cstate_module_pmu;
+static bool has_cstate_module;
+
+enum perf_cstate_module_events {
+	PERF_CSTATE_MODULE_C6_RES = 0,
+
+	PERF_CSTATE_MODULE_EVENT_MAX,
+};
+
+PMU_EVENT_ATTR_STRING(c6-residency, attr_cstate_module_c6, "event=0x00");
+
+static unsigned long module_msr_mask;
+
+PMU_EVENT_GROUP(events, cstate_module_c6);
+
+static struct perf_msr module_msr[] = {
+	[PERF_CSTATE_MODULE_C6_RES]  = { MSR_MODULE_C6_RES_MS,	&group_cstate_module_c6,	test_msr },
+};
+
+static cpumask_t cstate_module_cpu_mask;
+
 static ssize_t cstate_get_attr_cpumask(struct device *dev,
 				       struct device_attribute *attr,
 				       char *buf)
@@ -280,6 +307,8 @@ static ssize_t cstate_get_attr_cpumask(struct device *dev,
 		return cpumap_print_to_pagebuf(true, buf, &cstate_core_cpu_mask);
 	else if (pmu == &cstate_pkg_pmu)
 		return cpumap_print_to_pagebuf(true, buf, &cstate_pkg_cpu_mask);
+	else if (pmu == &cstate_module_pmu)
+		return cpumap_print_to_pagebuf(true, buf, &cstate_module_cpu_mask);
 	else
 		return 0;
 }
@@ -320,6 +349,15 @@ static int cstate_pmu_event_init(struct perf_event *event)
 		event->hw.event_base = pkg_msr[cfg].msr;
 		cpu = cpumask_any_and(&cstate_pkg_cpu_mask,
 				      topology_die_cpumask(event->cpu));
+	} else if (event->pmu == &cstate_module_pmu) {
+		if (cfg >= PERF_CSTATE_MODULE_EVENT_MAX)
+			return -EINVAL;
+		cfg = array_index_nospec((unsigned long)cfg, PERF_CSTATE_MODULE_EVENT_MAX);
+		if (!(module_msr_mask & (1 << cfg)))
+			return -EINVAL;
+		event->hw.event_base = module_msr[cfg].msr;
+		cpu = cpumask_any_and(&cstate_module_cpu_mask,
+				      topology_cluster_cpumask(event->cpu));
 	} else {
 		return -ENOENT;
 	}
@@ -407,6 +445,17 @@ static int cstate_cpu_exit(unsigned int cpu)
 			perf_pmu_migrate_context(&cstate_pkg_pmu, cpu, target);
 		}
 	}
+
+	if (has_cstate_module &&
+	    cpumask_test_and_clear_cpu(cpu, &cstate_module_cpu_mask)) {
+
+		target = cpumask_any_but(topology_cluster_cpumask(cpu), cpu);
+		/* Migrate events if there is a valid target */
+		if (target < nr_cpu_ids) {
+			cpumask_set_cpu(target, &cstate_module_cpu_mask);
+			perf_pmu_migrate_context(&cstate_module_pmu, cpu, target);
+		}
+	}
 	return 0;
 }
 
@@ -433,6 +482,15 @@ static int cstate_cpu_init(unsigned int cpu)
 	if (has_cstate_pkg && target >= nr_cpu_ids)
 		cpumask_set_cpu(cpu, &cstate_pkg_cpu_mask);
 
+	/*
+	 * If this is the first online thread of that cluster, set it
+	 * in the cluster cpu mask as the designated reader.
+	 */
+	target = cpumask_any_and(&cstate_module_cpu_mask,
+				 topology_cluster_cpumask(cpu));
+	if (has_cstate_module && target >= nr_cpu_ids)
+		cpumask_set_cpu(cpu, &cstate_module_cpu_mask);
+
 	return 0;
 }
 
@@ -455,6 +513,11 @@ static const struct attribute_group *pkg_attr_update[] = {
 	NULL,
 };
 
+static const struct attribute_group *module_attr_update[] = {
+	&group_cstate_module_c6,
+	NULL
+};
+
 static struct pmu cstate_core_pmu = {
 	.attr_groups	= cstate_attr_groups,
 	.attr_update	= core_attr_update,
@@ -485,6 +548,21 @@ static struct pmu cstate_pkg_pmu = {
 	.module		= THIS_MODULE,
 };
 
+static struct pmu cstate_module_pmu = {
+	.attr_groups	= cstate_attr_groups,
+	.attr_update	= module_attr_update,
+	.name		= "cstate_module",
+	.task_ctx_nr	= perf_invalid_context,
+	.event_init	= cstate_pmu_event_init,
+	.add		= cstate_pmu_event_add,
+	.del		= cstate_pmu_event_del,
+	.start		= cstate_pmu_event_start,
+	.stop		= cstate_pmu_event_stop,
+	.read		= cstate_pmu_event_update,
+	.capabilities	= PERF_PMU_CAP_NO_INTERRUPT | PERF_PMU_CAP_NO_EXCLUDE,
+	.module		= THIS_MODULE,
+};
+
 static const struct cstate_model nhm_cstates __initconst = {
 	.core_events		= BIT(PERF_CSTATE_CORE_C3_RES) |
 				  BIT(PERF_CSTATE_CORE_C6_RES),
@@ -599,6 +677,15 @@ static const struct cstate_model glm_cstates __initconst = {
 				  BIT(PERF_CSTATE_PKG_C10_RES),
 };
 
+static const struct cstate_model srf_cstates __initconst = {
+	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
+				  BIT(PERF_CSTATE_CORE_C6_RES),
+
+	.pkg_events		= BIT(PERF_CSTATE_PKG_C6_RES),
+
+	.module_events		= BIT(PERF_CSTATE_MODULE_C6_RES),
+};
+
 
 static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(NEHALEM,		&nhm_cstates),
@@ -651,6 +738,7 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT_L,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_GRACEMONT,	&adl_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT_X,	&srf_cstates),
 
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_L,		&icl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE,		&icl_cstates),
@@ -692,10 +780,14 @@ static int __init cstate_probe(const struct cstate_model *cm)
 	pkg_msr_mask = perf_msr_probe(pkg_msr, PERF_CSTATE_PKG_EVENT_MAX,
 				      true, (void *) &cm->pkg_events);
 
+	module_msr_mask = perf_msr_probe(module_msr, PERF_CSTATE_MODULE_EVENT_MAX,
+				      true, (void *) &cm->module_events);
+
 	has_cstate_core = !!core_msr_mask;
 	has_cstate_pkg  = !!pkg_msr_mask;
+	has_cstate_module  = !!module_msr_mask;
 
-	return (has_cstate_core || has_cstate_pkg) ? 0 : -ENODEV;
+	return (has_cstate_core || has_cstate_pkg || has_cstate_module) ? 0 : -ENODEV;
 }
 
 static inline void cstate_cleanup(void)
@@ -708,6 +800,9 @@ static inline void cstate_cleanup(void)
 
 	if (has_cstate_pkg)
 		perf_pmu_unregister(&cstate_pkg_pmu);
+
+	if (has_cstate_module)
+		perf_pmu_unregister(&cstate_module_pmu);
 }
 
 static int __init cstate_init(void)
@@ -744,6 +839,16 @@ static int __init cstate_init(void)
 			return err;
 		}
 	}
+
+	if (has_cstate_module) {
+		err = perf_pmu_register(&cstate_module_pmu, cstate_module_pmu.name, -1);
+		if (err) {
+			has_cstate_module = false;
+			pr_info("Failed to register cstate cluster pmu\n");
+			cstate_cleanup();
+			return err;
+		}
+	}
 	return 0;
 }
 
-- 
2.35.1

[tip: perf/core] perf/x86/intel/cstate: Add Sierra Forest support

[tip-bot2 for Kan Liang]

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     3877d55a0db2688c2e4ab8a319614a0c81f8e2d2
Gitweb:        https://git.kernel.org/tip/3877d55a0db2688c2e4ab8a319614a0c81f8e2d2
Author:        Kan Liang <kan.liang@linux.intel.com>
AuthorDate:    Thu, 16 Nov 2023 06:22:44 -08:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 17 Nov 2023 10:54:53 +01:00

perf/x86/intel/cstate: Add Sierra Forest support

A new module C6 Residency Counter is introduced in the Sierra Forest.
The scope of the new counter is module (A cluster of cores shared L2
cache). Create a brand new cstate_module PMU to profile the new counter.
The only differences between the new cstate_module PMU and the existing
cstate PMU are the scope and events.

Regarding the choice of the new cstate_module PMU name, the current
naming rule of a cstate PMU is "cstate_" + the scope of the PMU. The
scope of the PMU is the cores shared L2. On SRF, Intel calls it
"module", while the internal Linux sched code calls it "cluster". The
"cstate_module" is used as the new PMU name, because
- The Cstate PMU driver is a Intel specific driver. It doesn't impact
  other ARCHs. The name makes it consistent with the documentation.
- The "cluster" mainly be used by the scheduler developer, while the
  user of cstate PMU is more likely a researcher reading HW docs and
  optimizing power.
- In the Intel's SDM, the "cluster" has a different meaning/scope for
  topology. Using it will mislead the end users.

Besides the module C6, the core C1/C6 and pkg C6 residency counters are
supported in the Sierra Forest as well.

Suggested-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20231116142245.1233485-3-kan.liang@linux.intel.com
---
 arch/x86/events/intel/cstate.c | 113 ++++++++++++++++++++++++++++++--
 1 file changed, 109 insertions(+), 4 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index 693bdcd..4a46ef3 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -41,7 +41,7 @@
  *	MSR_CORE_C1_RES: CORE C1 Residency Counter
  *			 perf code: 0x00
  *			 Available model: SLM,AMT,GLM,CNL,ICX,TNT,ADL,RPL
- *					  MTL
+ *					  MTL,SRF
  *			 Scope: Core (each processor core has a MSR)
  *	MSR_CORE_C3_RESIDENCY: CORE C3 Residency Counter
  *			       perf code: 0x01
@@ -52,7 +52,7 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
  *			       Scope: Core
  *	MSR_CORE_C7_RESIDENCY: CORE C7 Residency Counter
  *			       perf code: 0x03
@@ -75,7 +75,7 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
  *			       Scope: Package (physical package)
  *	MSR_PKG_C7_RESIDENCY:  Package C7 Residency Counter.
  *			       perf code: 0x03
@@ -97,6 +97,10 @@
  *			       Available model: HSW ULT,KBL,GLM,CNL,CML,ICL,TGL,
  *						TNT,RKL,ADL,RPL,MTL
  *			       Scope: Package (physical package)
+ *	MSR_MODULE_C6_RES_MS:  Module C6 Residency Counter.
+ *			       perf code: 0x00
+ *			       Available model: SRF
+ *			       Scope: A cluster of cores shared L2 cache
  *
  */
 
@@ -130,6 +134,7 @@ static ssize_t cstate_get_attr_cpumask(struct device *dev,
 struct cstate_model {
 	unsigned long		core_events;
 	unsigned long		pkg_events;
+	unsigned long		module_events;
 	unsigned long		quirks;
 };
 
@@ -270,6 +275,28 @@ static struct perf_msr pkg_msr[] = {
 
 static cpumask_t cstate_pkg_cpu_mask;
 
+/* cstate_module PMU */
+static struct pmu cstate_module_pmu;
+static bool has_cstate_module;
+
+enum perf_cstate_module_events {
+	PERF_CSTATE_MODULE_C6_RES = 0,
+
+	PERF_CSTATE_MODULE_EVENT_MAX,
+};
+
+PMU_EVENT_ATTR_STRING(c6-residency, attr_cstate_module_c6, "event=0x00");
+
+static unsigned long module_msr_mask;
+
+PMU_EVENT_GROUP(events, cstate_module_c6);
+
+static struct perf_msr module_msr[] = {
+	[PERF_CSTATE_MODULE_C6_RES]  = { MSR_MODULE_C6_RES_MS,	&group_cstate_module_c6,	test_msr },
+};
+
+static cpumask_t cstate_module_cpu_mask;
+
 static ssize_t cstate_get_attr_cpumask(struct device *dev,
 				       struct device_attribute *attr,
 				       char *buf)
@@ -280,6 +307,8 @@ static ssize_t cstate_get_attr_cpumask(struct device *dev,
 		return cpumap_print_to_pagebuf(true, buf, &cstate_core_cpu_mask);
 	else if (pmu == &cstate_pkg_pmu)
 		return cpumap_print_to_pagebuf(true, buf, &cstate_pkg_cpu_mask);
+	else if (pmu == &cstate_module_pmu)
+		return cpumap_print_to_pagebuf(true, buf, &cstate_module_cpu_mask);
 	else
 		return 0;
 }
@@ -320,6 +349,15 @@ static int cstate_pmu_event_init(struct perf_event *event)
 		event->hw.event_base = pkg_msr[cfg].msr;
 		cpu = cpumask_any_and(&cstate_pkg_cpu_mask,
 				      topology_die_cpumask(event->cpu));
+	} else if (event->pmu == &cstate_module_pmu) {
+		if (cfg >= PERF_CSTATE_MODULE_EVENT_MAX)
+			return -EINVAL;
+		cfg = array_index_nospec((unsigned long)cfg, PERF_CSTATE_MODULE_EVENT_MAX);
+		if (!(module_msr_mask & (1 << cfg)))
+			return -EINVAL;
+		event->hw.event_base = module_msr[cfg].msr;
+		cpu = cpumask_any_and(&cstate_module_cpu_mask,
+				      topology_cluster_cpumask(event->cpu));
 	} else {
 		return -ENOENT;
 	}
@@ -407,6 +445,17 @@ static int cstate_cpu_exit(unsigned int cpu)
 			perf_pmu_migrate_context(&cstate_pkg_pmu, cpu, target);
 		}
 	}
+
+	if (has_cstate_module &&
+	    cpumask_test_and_clear_cpu(cpu, &cstate_module_cpu_mask)) {
+
+		target = cpumask_any_but(topology_cluster_cpumask(cpu), cpu);
+		/* Migrate events if there is a valid target */
+		if (target < nr_cpu_ids) {
+			cpumask_set_cpu(target, &cstate_module_cpu_mask);
+			perf_pmu_migrate_context(&cstate_module_pmu, cpu, target);
+		}
+	}
 	return 0;
 }
 
@@ -433,6 +482,15 @@ static int cstate_cpu_init(unsigned int cpu)
 	if (has_cstate_pkg && target >= nr_cpu_ids)
 		cpumask_set_cpu(cpu, &cstate_pkg_cpu_mask);
 
+	/*
+	 * If this is the first online thread of that cluster, set it
+	 * in the cluster cpu mask as the designated reader.
+	 */
+	target = cpumask_any_and(&cstate_module_cpu_mask,
+				 topology_cluster_cpumask(cpu));
+	if (has_cstate_module && target >= nr_cpu_ids)
+		cpumask_set_cpu(cpu, &cstate_module_cpu_mask);
+
 	return 0;
 }
 
@@ -455,6 +513,11 @@ static const struct attribute_group *pkg_attr_update[] = {
 	NULL,
 };
 
+static const struct attribute_group *module_attr_update[] = {
+	&group_cstate_module_c6,
+	NULL
+};
+
 static struct pmu cstate_core_pmu = {
 	.attr_groups	= cstate_attr_groups,
 	.attr_update	= core_attr_update,
@@ -485,6 +548,21 @@ static struct pmu cstate_pkg_pmu = {
 	.module		= THIS_MODULE,
 };
 
+static struct pmu cstate_module_pmu = {
+	.attr_groups	= cstate_attr_groups,
+	.attr_update	= module_attr_update,
+	.name		= "cstate_module",
+	.task_ctx_nr	= perf_invalid_context,
+	.event_init	= cstate_pmu_event_init,
+	.add		= cstate_pmu_event_add,
+	.del		= cstate_pmu_event_del,
+	.start		= cstate_pmu_event_start,
+	.stop		= cstate_pmu_event_stop,
+	.read		= cstate_pmu_event_update,
+	.capabilities	= PERF_PMU_CAP_NO_INTERRUPT | PERF_PMU_CAP_NO_EXCLUDE,
+	.module		= THIS_MODULE,
+};
+
 static const struct cstate_model nhm_cstates __initconst = {
 	.core_events		= BIT(PERF_CSTATE_CORE_C3_RES) |
 				  BIT(PERF_CSTATE_CORE_C6_RES),
@@ -599,6 +677,15 @@ static const struct cstate_model glm_cstates __initconst = {
 				  BIT(PERF_CSTATE_PKG_C10_RES),
 };
 
+static const struct cstate_model srf_cstates __initconst = {
+	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
+				  BIT(PERF_CSTATE_CORE_C6_RES),
+
+	.pkg_events		= BIT(PERF_CSTATE_PKG_C6_RES),
+
+	.module_events		= BIT(PERF_CSTATE_MODULE_C6_RES),
+};
+
 
 static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(NEHALEM,		&nhm_cstates),
@@ -651,6 +738,7 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT_L,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_GRACEMONT,	&adl_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT_X,	&srf_cstates),
 
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_L,		&icl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE,		&icl_cstates),
@@ -692,10 +780,14 @@ static int __init cstate_probe(const struct cstate_model *cm)
 	pkg_msr_mask = perf_msr_probe(pkg_msr, PERF_CSTATE_PKG_EVENT_MAX,
 				      true, (void *) &cm->pkg_events);
 
+	module_msr_mask = perf_msr_probe(module_msr, PERF_CSTATE_MODULE_EVENT_MAX,
+				      true, (void *) &cm->module_events);
+
 	has_cstate_core = !!core_msr_mask;
 	has_cstate_pkg  = !!pkg_msr_mask;
+	has_cstate_module  = !!module_msr_mask;
 
-	return (has_cstate_core || has_cstate_pkg) ? 0 : -ENODEV;
+	return (has_cstate_core || has_cstate_pkg || has_cstate_module) ? 0 : -ENODEV;
 }
 
 static inline void cstate_cleanup(void)
@@ -708,6 +800,9 @@ static inline void cstate_cleanup(void)
 
 	if (has_cstate_pkg)
 		perf_pmu_unregister(&cstate_pkg_pmu);
+
+	if (has_cstate_module)
+		perf_pmu_unregister(&cstate_module_pmu);
 }
 
 static int __init cstate_init(void)
@@ -744,6 +839,16 @@ static int __init cstate_init(void)
 			return err;
 		}
 	}
+
+	if (has_cstate_module) {
+		err = perf_pmu_register(&cstate_module_pmu, cstate_module_pmu.name, -1);
+		if (err) {
+			has_cstate_module = false;
+			pr_info("Failed to register cstate cluster pmu\n");
+			cstate_cleanup();
+			return err;
+		}
+	}
 	return 0;
 }
 

[RESEND PATCH 4/4] perf/x86/intel/cstate: Add Grand Ridge support

[kan.liang]

From: Kan Liang <kan.liang@linux.intel.com>

The same as the Sierra Forest, the Grand Ridge supports core C1/C6 and
module C6. But it doesn't support pkg C6 residency counter.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
---
 arch/x86/events/intel/cstate.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index 4a46ef315284..4b50a3a9818a 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -41,7 +41,7 @@
  *	MSR_CORE_C1_RES: CORE C1 Residency Counter
  *			 perf code: 0x00
  *			 Available model: SLM,AMT,GLM,CNL,ICX,TNT,ADL,RPL
- *					  MTL,SRF
+ *					  MTL,SRF,GRR
  *			 Scope: Core (each processor core has a MSR)
  *	MSR_CORE_C3_RESIDENCY: CORE C3 Residency Counter
  *			       perf code: 0x01
@@ -52,7 +52,8 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF,
+ *						GRR
  *			       Scope: Core
  *	MSR_CORE_C7_RESIDENCY: CORE C7 Residency Counter
  *			       perf code: 0x03
@@ -99,7 +100,7 @@
  *			       Scope: Package (physical package)
  *	MSR_MODULE_C6_RES_MS:  Module C6 Residency Counter.
  *			       perf code: 0x00
- *			       Available model: SRF
+ *			       Available model: SRF,GRR
  *			       Scope: A cluster of cores shared L2 cache
  *
  */
@@ -677,6 +678,13 @@ static const struct cstate_model glm_cstates __initconst = {
 				  BIT(PERF_CSTATE_PKG_C10_RES),
 };
 
+static const struct cstate_model grr_cstates __initconst = {
+	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
+				  BIT(PERF_CSTATE_CORE_C6_RES),
+
+	.module_events		= BIT(PERF_CSTATE_MODULE_C6_RES),
+};
+
 static const struct cstate_model srf_cstates __initconst = {
 	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
 				  BIT(PERF_CSTATE_CORE_C6_RES),
@@ -739,6 +747,7 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT_L,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_GRACEMONT,	&adl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT_X,	&srf_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT,	&grr_cstates),
 
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_L,		&icl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE,		&icl_cstates),
-- 
2.35.1

[tip: perf/core] perf/x86/intel/cstate: Add Grand Ridge support

[tip-bot2 for Kan Liang]

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     bbb968696d0f3442ab823598def3b756cf4735c6
Gitweb:        https://git.kernel.org/tip/bbb968696d0f3442ab823598def3b756cf4735c6
Author:        Kan Liang <kan.liang@linux.intel.com>
AuthorDate:    Thu, 16 Nov 2023 06:22:45 -08:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 17 Nov 2023 10:54:53 +01:00

perf/x86/intel/cstate: Add Grand Ridge support

The same as the Sierra Forest, the Grand Ridge supports core C1/C6 and
module C6. But it doesn't support pkg C6 residency counter.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20231116142245.1233485-4-kan.liang@linux.intel.com
---
 arch/x86/events/intel/cstate.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index 4a46ef3..4b50a3a 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -41,7 +41,7 @@
  *	MSR_CORE_C1_RES: CORE C1 Residency Counter
  *			 perf code: 0x00
  *			 Available model: SLM,AMT,GLM,CNL,ICX,TNT,ADL,RPL
- *					  MTL,SRF
+ *					  MTL,SRF,GRR
  *			 Scope: Core (each processor core has a MSR)
  *	MSR_CORE_C3_RESIDENCY: CORE C3 Residency Counter
  *			       perf code: 0x01
@@ -52,7 +52,8 @@
  *			       perf code: 0x02
  *			       Available model: SLM,AMT,NHM,WSM,SNB,IVB,HSW,BDW,
  *						SKL,KNL,GLM,CNL,KBL,CML,ICL,ICX,
- *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF
+ *						TGL,TNT,RKL,ADL,RPL,SPR,MTL,SRF,
+ *						GRR
  *			       Scope: Core
  *	MSR_CORE_C7_RESIDENCY: CORE C7 Residency Counter
  *			       perf code: 0x03
@@ -99,7 +100,7 @@
  *			       Scope: Package (physical package)
  *	MSR_MODULE_C6_RES_MS:  Module C6 Residency Counter.
  *			       perf code: 0x00
- *			       Available model: SRF
+ *			       Available model: SRF,GRR
  *			       Scope: A cluster of cores shared L2 cache
  *
  */
@@ -677,6 +678,13 @@ static const struct cstate_model glm_cstates __initconst = {
 				  BIT(PERF_CSTATE_PKG_C10_RES),
 };
 
+static const struct cstate_model grr_cstates __initconst = {
+	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
+				  BIT(PERF_CSTATE_CORE_C6_RES),
+
+	.module_events		= BIT(PERF_CSTATE_MODULE_C6_RES),
+};
+
 static const struct cstate_model srf_cstates __initconst = {
 	.core_events		= BIT(PERF_CSTATE_CORE_C1_RES) |
 				  BIT(PERF_CSTATE_CORE_C6_RES),
@@ -739,6 +747,7 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_TREMONT_L,	&glm_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_GRACEMONT,	&adl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT_X,	&srf_cstates),
+	X86_MATCH_INTEL_FAM6_MODEL(ATOM_CRESTMONT,	&grr_cstates),
 
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE_L,		&icl_cstates),
 	X86_MATCH_INTEL_FAM6_MODEL(ICELAKE,		&icl_cstates),

[tip: perf/core] perf/x86/intel/cstate: Cleanup duplicate attr_groups

[tip-bot2 for Kan Liang]

The following commit has been merged into the perf/core branch of tip:

Commit-ID:     243218ca93037631f0224fdbefea045912cb761a
Gitweb:        https://git.kernel.org/tip/243218ca93037631f0224fdbefea045912cb761a
Author:        Kan Liang <kan.liang@linux.intel.com>
AuthorDate:    Thu, 16 Nov 2023 06:22:42 -08:00
Committer:     Peter Zijlstra <peterz@infradead.org>
CommitterDate: Fri, 17 Nov 2023 10:54:52 +01:00

perf/x86/intel/cstate: Cleanup duplicate attr_groups

The events of the cstate_core and cstate_pkg PMU have the same format.
They both need to create a "events" group (with empty attrs). The
attr_groups can be shared.

Remove the dedicated attr_groups for each cstate PMU. Use the shared
cstate_attr_groups to replace.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20231116142245.1233485-1-kan.liang@linux.intel.com
---
 arch/x86/events/intel/cstate.c | 44 ++++++++-------------------------
 1 file changed, 11 insertions(+), 33 deletions(-)

diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
index cbeb6d2..693bdcd 100644
--- a/arch/x86/events/intel/cstate.c
+++ b/arch/x86/events/intel/cstate.c
@@ -189,20 +189,20 @@ static struct attribute *attrs_empty[] = {
  * "events" group (with empty attrs) before updating
  * it with detected events.
  */
-static struct attribute_group core_events_attr_group = {
+static struct attribute_group cstate_events_attr_group = {
 	.name = "events",
 	.attrs = attrs_empty,
 };
 
-DEFINE_CSTATE_FORMAT_ATTR(core_event, event, "config:0-63");
-static struct attribute *core_format_attrs[] = {
-	&format_attr_core_event.attr,
+DEFINE_CSTATE_FORMAT_ATTR(cstate_event, event, "config:0-63");
+static struct attribute *cstate_format_attrs[] = {
+	&format_attr_cstate_event.attr,
 	NULL,
 };
 
-static struct attribute_group core_format_attr_group = {
+static struct attribute_group cstate_format_attr_group = {
 	.name = "format",
-	.attrs = core_format_attrs,
+	.attrs = cstate_format_attrs,
 };
 
 static cpumask_t cstate_core_cpu_mask;
@@ -217,9 +217,9 @@ static struct attribute_group cpumask_attr_group = {
 	.attrs = cstate_cpumask_attrs,
 };
 
-static const struct attribute_group *core_attr_groups[] = {
-	&core_events_attr_group,
-	&core_format_attr_group,
+static const struct attribute_group *cstate_attr_groups[] = {
+	&cstate_events_attr_group,
+	&cstate_format_attr_group,
 	&cpumask_attr_group,
 	NULL,
 };
@@ -268,30 +268,8 @@ static struct perf_msr pkg_msr[] = {
 	[PERF_CSTATE_PKG_C10_RES] = { MSR_PKG_C10_RESIDENCY,	&group_cstate_pkg_c10,	test_msr },
 };
 
-static struct attribute_group pkg_events_attr_group = {
-	.name = "events",
-	.attrs = attrs_empty,
-};
-
-DEFINE_CSTATE_FORMAT_ATTR(pkg_event, event, "config:0-63");
-static struct attribute *pkg_format_attrs[] = {
-	&format_attr_pkg_event.attr,
-	NULL,
-};
-static struct attribute_group pkg_format_attr_group = {
-	.name = "format",
-	.attrs = pkg_format_attrs,
-};
-
 static cpumask_t cstate_pkg_cpu_mask;
 
-static const struct attribute_group *pkg_attr_groups[] = {
-	&pkg_events_attr_group,
-	&pkg_format_attr_group,
-	&cpumask_attr_group,
-	NULL,
-};
-
 static ssize_t cstate_get_attr_cpumask(struct device *dev,
 				       struct device_attribute *attr,
 				       char *buf)
@@ -478,7 +456,7 @@ static const struct attribute_group *pkg_attr_update[] = {
 };
 
 static struct pmu cstate_core_pmu = {
-	.attr_groups	= core_attr_groups,
+	.attr_groups	= cstate_attr_groups,
 	.attr_update	= core_attr_update,
 	.name		= "cstate_core",
 	.task_ctx_nr	= perf_invalid_context,
@@ -493,7 +471,7 @@ static struct pmu cstate_core_pmu = {
 };
 
 static struct pmu cstate_pkg_pmu = {
-	.attr_groups	= pkg_attr_groups,
+	.attr_groups	= cstate_attr_groups,
 	.attr_update	= pkg_attr_update,
 	.name		= "cstate_pkg",
 	.task_ctx_nr	= perf_invalid_context,