]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
arm64: pmu: add support for interrupt-affinity property
authorWill Deacon <will.deacon@arm.com>
Fri, 6 Mar 2015 11:54:10 +0000 (11:54 +0000)
committerWill Deacon <will.deacon@arm.com>
Tue, 24 Mar 2015 15:09:47 +0000 (15:09 +0000)
Historically, the PMU devicetree bindings have expected SPIs to be
listed in order of *logical* CPU number. This is problematic for
bootloaders, especially when the boot CPU (logical ID 0) isn't listed
first in the devicetree.

This patch adds a new optional property, interrupt-affinity, to the
PMU node which allows the interrupt affinity to be described using
a list of phandled to CPU nodes, with each entry in the list
corresponding to the SPI at the same index in the interrupts property.

Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
arch/arm64/include/asm/pmu.h
arch/arm64/kernel/perf_event.c

index e6f087806aaf11d65732aca8ffb221b291a53c85..b7710a59672c0355b8a6096937bdbc3c703f1cdf 100644 (file)
@@ -44,6 +44,7 @@ struct pmu_hw_events {
 struct arm_pmu {
        struct pmu              pmu;
        cpumask_t               active_irqs;
+       int                     *irq_affinity;
        const char              *name;
        irqreturn_t             (*handle_irq)(int irq_num, void *dev);
        void                    (*enable)(struct hw_perf_event *evt, int idx);
index 68a74151fa6cf84002cd3be98a0fc29f67a40b02..195991dadc3772c67a084396d28aa2a7c8b2019e 100644 (file)
 #include <linux/irq.h>
 #include <linux/kernel.h>
 #include <linux/export.h>
+#include <linux/of.h>
 #include <linux/perf_event.h>
 #include <linux/platform_device.h>
+#include <linux/slab.h>
 #include <linux/spinlock.h>
 #include <linux/uaccess.h>
 
@@ -405,7 +407,12 @@ armpmu_release_hardware(struct arm_pmu *armpmu)
                free_percpu_irq(irq, &cpu_hw_events);
        } else {
                for (i = 0; i < irqs; ++i) {
-                       if (!cpumask_test_and_clear_cpu(i, &armpmu->active_irqs))
+                       int cpu = i;
+
+                       if (armpmu->irq_affinity)
+                               cpu = armpmu->irq_affinity[i];
+
+                       if (!cpumask_test_and_clear_cpu(cpu, &armpmu->active_irqs))
                                continue;
                        irq = platform_get_irq(pmu_device, i);
                        if (irq > 0)
@@ -459,19 +466,24 @@ armpmu_reserve_hardware(struct arm_pmu *armpmu)
                on_each_cpu(armpmu_enable_percpu_irq, &irq, 1);
        } else {
                for (i = 0; i < irqs; ++i) {
+                       int cpu = i;
+
                        err = 0;
                        irq = platform_get_irq(pmu_device, i);
                        if (irq <= 0)
                                continue;
 
+                       if (armpmu->irq_affinity)
+                               cpu = armpmu->irq_affinity[i];
+
                        /*
                         * If we have a single PMU interrupt that we can't shift,
                         * assume that we're running on a uniprocessor machine and
                         * continue. Otherwise, continue without this interrupt.
                         */
-                       if (irq_set_affinity(irq, cpumask_of(i)) && irqs > 1) {
+                       if (irq_set_affinity(irq, cpumask_of(cpu)) && irqs > 1) {
                                pr_warning("unable to set irq affinity (irq=%d, cpu=%u)\n",
-                                               irq, i);
+                                               irq, cpu);
                                continue;
                        }
 
@@ -485,7 +497,7 @@ armpmu_reserve_hardware(struct arm_pmu *armpmu)
                                return err;
                        }
 
-                       cpumask_set_cpu(i, &armpmu->active_irqs);
+                       cpumask_set_cpu(cpu, &armpmu->active_irqs);
                }
        }
 
@@ -1298,9 +1310,46 @@ static const struct of_device_id armpmu_of_device_ids[] = {
 
 static int armpmu_device_probe(struct platform_device *pdev)
 {
+       int i, *irqs;
+
        if (!cpu_pmu)
                return -ENODEV;
 
+       irqs = kcalloc(pdev->num_resources, sizeof(*irqs), GFP_KERNEL);
+       if (!irqs)
+               return -ENOMEM;
+
+       for (i = 0; i < pdev->num_resources; ++i) {
+               struct device_node *dn;
+               int cpu;
+
+               dn = of_parse_phandle(pdev->dev.of_node, "interrupt-affinity",
+                                     i);
+               if (!dn) {
+                       pr_warn("Failed to parse %s/interrupt-affinity[%d]\n",
+                               of_node_full_name(dn), i);
+                       break;
+               }
+
+               for_each_possible_cpu(cpu)
+                       if (arch_find_n_match_cpu_physical_id(dn, cpu, NULL))
+                               break;
+
+               of_node_put(dn);
+               if (cpu >= nr_cpu_ids) {
+                       pr_warn("Failed to find logical CPU for %s\n",
+                               dn->name);
+                       break;
+               }
+
+               irqs[i] = cpu;
+       }
+
+       if (i == pdev->num_resources)
+               cpu_pmu->irq_affinity = irqs;
+       else
+               kfree(irqs);
+
        cpu_pmu->plat_device = pdev;
        return 0;
 }