perf: arm_pmu: Request specific affinities for per CPU NMIs/interrupts

Let the PMU driver request both NMIs and normal interrupts with an affinity mask
matching the PMU affinity.

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Will Deacon <will@kernel.org>
Link: https://patch.msgid.link/20251020122944.3074811-19-maz@kernel.org

Author: willdeacon

drivers/perf/arm_pmu.c

@@ -26,7 +26,8 @@
 
 #include <asm/irq_regs.h>
 
-static int armpmu_count_irq_users(const int irq);
+static int armpmu_count_irq_users(const struct cpumask *affinity,
+				  const int irq);
 
 struct pmu_irq_ops {
 	void (*enable_pmuirq)(unsigned int irq);
@@ -64,7 +65,9 @@ static void armpmu_enable_percpu_pmuirq(unsigned int irq)
 static void armpmu_free_percpu_pmuirq(unsigned int irq, int cpu,
 				   void __percpu *devid)
 {
-	if (armpmu_count_irq_users(irq) == 1)
+	struct arm_pmu *armpmu = *per_cpu_ptr((void * __percpu *)devid, cpu);
+
+	if (armpmu_count_irq_users(&armpmu->supported_cpus, irq) == 1)
 		free_percpu_irq(irq, devid);
 }
 
@@ -89,7 +92,9 @@ static void armpmu_disable_percpu_pmunmi(unsigned int irq)
 static void armpmu_free_percpu_pmunmi(unsigned int irq, int cpu,
 				      void __percpu *devid)
 {
-	if (armpmu_count_irq_users(irq) == 1)
+	struct arm_pmu *armpmu = *per_cpu_ptr((void * __percpu *)devid, cpu);
+
+	if (armpmu_count_irq_users(&armpmu->supported_cpus, irq) == 1)
 		free_percpu_nmi(irq, devid);
 }
 
@@ -580,24 +585,25 @@ static const struct attribute_group armpmu_common_attr_group = {
 	.attrs = armpmu_common_attrs,
 };
 
-static int armpmu_count_irq_users(const int irq)
+static int armpmu_count_irq_users(const struct cpumask *affinity, const int irq)
 {
 	int cpu, count = 0;
 
-	for_each_possible_cpu(cpu) {
+	for_each_cpu(cpu, affinity) {
 		if (per_cpu(cpu_irq, cpu) == irq)
 			count++;
 	}
 
 	return count;
 }
 
-static const struct pmu_irq_ops *armpmu_find_irq_ops(int irq)
+static const struct pmu_irq_ops *
+armpmu_find_irq_ops(const struct cpumask *affinity, int irq)
 {
 	const struct pmu_irq_ops *ops = NULL;
 	int cpu;
 
-	for_each_possible_cpu(cpu) {
+	for_each_cpu(cpu, affinity) {
 		if (per_cpu(cpu_irq, cpu) != irq)
 			continue;
 
@@ -609,22 +615,25 @@ static const struct pmu_irq_ops *armpmu_find_irq_ops(int irq)
 	return ops;
 }
 
-void armpmu_free_irq(int irq, int cpu)
+void armpmu_free_irq(struct arm_pmu * __percpu *armpmu, int irq, int cpu)
 {
 	if (per_cpu(cpu_irq, cpu) == 0)
 		return;
 	if (WARN_ON(irq != per_cpu(cpu_irq, cpu)))
 		return;
 
-	per_cpu(cpu_irq_ops, cpu)->free_pmuirq(irq, cpu, &cpu_armpmu);
+	per_cpu(cpu_irq_ops, cpu)->free_pmuirq(irq, cpu, armpmu);
 
 	per_cpu(cpu_irq, cpu) = 0;
 	per_cpu(cpu_irq_ops, cpu) = NULL;
 }
 
-int armpmu_request_irq(int irq, int cpu)
+int armpmu_request_irq(struct arm_pmu * __percpu *pcpu_armpmu, int irq, int cpu)
 {
 	int err = 0;
+	struct arm_pmu **armpmu = per_cpu_ptr(pcpu_armpmu, cpu);
+	const struct cpumask *affinity = *armpmu ? &(*armpmu)->supported_cpus :
+						   cpu_possible_mask; /* ACPI */
 	const irq_handler_t handler = armpmu_dispatch_irq;
 	const struct pmu_irq_ops *irq_ops;
 
@@ -646,33 +655,32 @@ int armpmu_request_irq(int irq, int cpu)
 			    IRQF_NOBALANCING | IRQF_NO_AUTOEN |
 			    IRQF_NO_THREAD;
 
-		err = request_nmi(irq, handler, irq_flags, "arm-pmu",
-				  per_cpu_ptr(&cpu_armpmu, cpu));
+		err = request_nmi(irq, handler, irq_flags, "arm-pmu", armpmu);
 
 		/* If cannot get an NMI, get a normal interrupt */
 		if (err) {
 			err = request_irq(irq, handler, irq_flags, "arm-pmu",
-					  per_cpu_ptr(&cpu_armpmu, cpu));
+					  armpmu);
 			irq_ops = &pmuirq_ops;
 		} else {
 			has_nmi = true;
 			irq_ops = &pmunmi_ops;
 		}
-	} else if (armpmu_count_irq_users(irq) == 0) {
-		err = request_percpu_nmi(irq, handler, "arm-pmu", NULL, &cpu_armpmu);
+	} else if (armpmu_count_irq_users(affinity, irq) == 0) {
+		err = request_percpu_nmi(irq, handler, "arm-pmu", affinity, pcpu_armpmu);
 
 		/* If cannot get an NMI, get a normal interrupt */
 		if (err) {
-			err = request_percpu_irq(irq, handler, "arm-pmu",
-						 &cpu_armpmu);
+			err = request_percpu_irq_affinity(irq, handler, "arm-pmu",
+							  affinity, pcpu_armpmu);
 			irq_ops = &percpu_pmuirq_ops;
 		} else {
 			has_nmi = true;
 			irq_ops = &percpu_pmunmi_ops;
 		}
 	} else {
 		/* Per cpudevid irq was already requested by another CPU */
-		irq_ops = armpmu_find_irq_ops(irq);
+		irq_ops = armpmu_find_irq_ops(affinity, irq);
 
 		if (WARN_ON(!irq_ops))
 			err = -EINVAL;

drivers/perf/arm_pmu_acpi.c

@@ -218,7 +218,7 @@ static int arm_pmu_acpi_parse_irqs(void)
 		 * them with their PMUs.
 		 */
 		per_cpu(pmu_irqs, cpu) = irq;
-		err = armpmu_request_irq(irq, cpu);
+		err = armpmu_request_irq(&probed_pmus, irq, cpu);
 		if (err)
 			goto out_err;
 	}

drivers/perf/arm_pmu_platform.c

@@ -165,7 +165,7 @@ static int armpmu_request_irqs(struct arm_pmu *armpmu)
 		if (!irq)
 			continue;
 
-		err = armpmu_request_irq(irq, cpu);
+		err = armpmu_request_irq(&hw_events->percpu_pmu, irq, cpu);
 		if (err)
 			break;
 	}
@@ -181,7 +181,7 @@ static void armpmu_free_irqs(struct arm_pmu *armpmu)
 	for_each_cpu(cpu, &armpmu->supported_cpus) {
 		int irq = per_cpu(hw_events->irq, cpu);
 
-		armpmu_free_irq(irq, cpu);
+		armpmu_free_irq(&hw_events->percpu_pmu, irq, cpu);
 	}
 }
 

include/linux/perf/arm_pmu.h

@@ -190,8 +190,8 @@ bool arm_pmu_irq_is_nmi(void);
 struct arm_pmu *armpmu_alloc(void);
 void armpmu_free(struct arm_pmu *pmu);
 int armpmu_register(struct arm_pmu *pmu);
-int armpmu_request_irq(int irq, int cpu);
-void armpmu_free_irq(int irq, int cpu);
+int armpmu_request_irq(struct arm_pmu * __percpu *armpmu, int irq, int cpu);
+void armpmu_free_irq(struct arm_pmu * __percpu *armpmu, int irq, int cpu);
 
 #define ARMV8_PMU_PDEV_NAME "armv8-pmu"