perf_event_cpu.c 8.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
/*
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 *
 * Copyright (C) 2012 ARM Limited
 *
 * Author: Will Deacon <will.deacon@arm.com>
 */
#define pr_fmt(fmt) "CPU PMU: " fmt

#include <linux/bitmap.h>
#include <linux/export.h>
#include <linux/kernel.h>
#include <linux/of.h>
#include <linux/platform_device.h>
26
#include <linux/slab.h>
27
#include <linux/spinlock.h>
28 29
#include <linux/irq.h>
#include <linux/irqdesc.h>
30 31 32 33 34 35 36 37

#include <asm/cputype.h>
#include <asm/irq_regs.h>
#include <asm/pmu.h>

/* Set at runtime when we know what CPU type we are. */
static struct arm_pmu *cpu_pmu;

38
static DEFINE_PER_CPU(struct arm_pmu *, percpu_pmu);
39 40 41 42 43 44 45 46 47 48 49
static DEFINE_PER_CPU(struct pmu_hw_events, cpu_hw_events);

/*
 * Despite the names, these two functions are CPU-specific and are used
 * by the OProfile/perf code.
 */
const char *perf_pmu_name(void)
{
	if (!cpu_pmu)
		return NULL;

50
	return cpu_pmu->name;
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69
}
EXPORT_SYMBOL_GPL(perf_pmu_name);

int perf_num_counters(void)
{
	int max_events = 0;

	if (cpu_pmu != NULL)
		max_events = cpu_pmu->num_events;

	return max_events;
}
EXPORT_SYMBOL_GPL(perf_num_counters);

/* Include the PMU-specific implementations. */
#include "perf_event_xscale.c"
#include "perf_event_v6.c"
#include "perf_event_v7.c"

70 71
static void cpu_pmu_enable_percpu_irq(void *data)
{
72
	int irq = *(int *)data;
73 74 75 76 77 78

	enable_percpu_irq(irq, IRQ_TYPE_NONE);
}

static void cpu_pmu_disable_percpu_irq(void *data)
{
79
	int irq = *(int *)data;
80 81 82 83

	disable_percpu_irq(irq);
}

84
static void cpu_pmu_free_irq(struct arm_pmu *cpu_pmu)
85 86 87 88 89 90
{
	int i, irq, irqs;
	struct platform_device *pmu_device = cpu_pmu->plat_device;

	irqs = min(pmu_device->num_resources, num_possible_cpus());

91 92
	irq = platform_get_irq(pmu_device, 0);
	if (irq >= 0 && irq_is_percpu(irq)) {
93
		on_each_cpu(cpu_pmu_disable_percpu_irq, &irq, 1);
94 95 96 97 98 99 100 101 102
		free_percpu_irq(irq, &percpu_pmu);
	} else {
		for (i = 0; i < irqs; ++i) {
			if (!cpumask_test_and_clear_cpu(i, &cpu_pmu->active_irqs))
				continue;
			irq = platform_get_irq(pmu_device, i);
			if (irq >= 0)
				free_irq(irq, cpu_pmu);
		}
103 104 105
	}
}

106
static int cpu_pmu_request_irq(struct arm_pmu *cpu_pmu, irq_handler_t handler)
107 108 109 110 111 112 113 114 115
{
	int i, err, irq, irqs;
	struct platform_device *pmu_device = cpu_pmu->plat_device;

	if (!pmu_device)
		return -ENODEV;

	irqs = min(pmu_device->num_resources, num_possible_cpus());
	if (irqs < 1) {
116
		pr_warn_once("perf/ARM: No irqs for PMU defined, sampling events not supported\n");
117
		return 0;
118 119
	}

120 121 122
	irq = platform_get_irq(pmu_device, 0);
	if (irq >= 0 && irq_is_percpu(irq)) {
		err = request_percpu_irq(irq, handler, "arm-pmu", &percpu_pmu);
123 124 125 126 127
		if (err) {
			pr_err("unable to request IRQ%d for ARM PMU counters\n",
				irq);
			return err;
		}
128
		on_each_cpu(cpu_pmu_enable_percpu_irq, &irq, 1);
129 130 131 132 133 134 135 136 137 138 139 140 141
	} else {
		for (i = 0; i < irqs; ++i) {
			err = 0;
			irq = platform_get_irq(pmu_device, i);
			if (irq < 0)
				continue;

			/*
			 * 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) {
142 143
				pr_warn("unable to set irq affinity (irq=%d, cpu=%u)\n",
					irq, i);
144 145 146 147 148 149 150 151 152 153 154 155 156 157
				continue;
			}

			err = request_irq(irq, handler,
					  IRQF_NOBALANCING | IRQF_NO_THREAD, "arm-pmu",
					  cpu_pmu);
			if (err) {
				pr_err("unable to request IRQ%d for ARM PMU counters\n",
					irq);
				return err;
			}

			cpumask_set_cpu(i, &cpu_pmu->active_irqs);
		}
158 159 160 161 162
	}

	return 0;
}

163
static void cpu_pmu_init(struct arm_pmu *cpu_pmu)
164 165 166 167 168
{
	int cpu;
	for_each_possible_cpu(cpu) {
		struct pmu_hw_events *events = &per_cpu(cpu_hw_events, cpu);
		raw_spin_lock_init(&events->pmu_lock);
169
		per_cpu(percpu_pmu, cpu) = cpu_pmu;
170
	}
171

M
Mark Rutland 已提交
172
	cpu_pmu->hw_events	= &cpu_hw_events;
173 174
	cpu_pmu->request_irq	= cpu_pmu_request_irq;
	cpu_pmu->free_irq	= cpu_pmu_free_irq;
175 176

	/* Ensure the PMU has sane values out of reset. */
177
	if (cpu_pmu->reset)
178
		on_each_cpu(cpu_pmu->reset, cpu_pmu, 1);
179 180 181 182

	/* If no interrupts available, set the corresponding capability flag */
	if (!platform_get_irq(cpu_pmu->plat_device, 0))
		cpu_pmu->pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT;
183 184 185 186 187 188 189 190
}

/*
 * PMU hardware loses all context when a CPU goes offline.
 * When a CPU is hotplugged back in, since some hardware registers are
 * UNKNOWN at reset, the PMU must be explicitly reset to avoid reading
 * junk values out of them.
 */
191 192
static int cpu_pmu_notify(struct notifier_block *b, unsigned long action,
			  void *hcpu)
193 194 195 196 197
{
	if ((action & ~CPU_TASKS_FROZEN) != CPU_STARTING)
		return NOTIFY_DONE;

	if (cpu_pmu && cpu_pmu->reset)
198
		cpu_pmu->reset(cpu_pmu);
199 200
	else
		return NOTIFY_DONE;
201 202 203 204

	return NOTIFY_OK;
}

205
static struct notifier_block cpu_pmu_hotplug_notifier = {
206 207 208 209 210 211
	.notifier_call = cpu_pmu_notify,
};

/*
 * PMU platform driver and devicetree bindings.
 */
212
static struct of_device_id cpu_pmu_of_device_ids[] = {
213
	{.compatible = "arm,cortex-a17-pmu",	.data = armv7_a17_pmu_init},
214
	{.compatible = "arm,cortex-a15-pmu",	.data = armv7_a15_pmu_init},
215
	{.compatible = "arm,cortex-a12-pmu",	.data = armv7_a12_pmu_init},
216 217 218 219 220
	{.compatible = "arm,cortex-a9-pmu",	.data = armv7_a9_pmu_init},
	{.compatible = "arm,cortex-a8-pmu",	.data = armv7_a8_pmu_init},
	{.compatible = "arm,cortex-a7-pmu",	.data = armv7_a7_pmu_init},
	{.compatible = "arm,cortex-a5-pmu",	.data = armv7_a5_pmu_init},
	{.compatible = "arm,arm11mpcore-pmu",	.data = armv6mpcore_pmu_init},
M
Mark Rutland 已提交
221 222
	{.compatible = "arm,arm1176-pmu",	.data = armv6_1176_pmu_init},
	{.compatible = "arm,arm1136-pmu",	.data = armv6_1136_pmu_init},
223
	{.compatible = "qcom,krait-pmu",	.data = krait_pmu_init},
224 225 226
	{},
};

227
static struct platform_device_id cpu_pmu_plat_device_ids[] = {
228
	{.name = "arm-pmu"},
229 230 231
	{.name = "armv6-pmu"},
	{.name = "armv7-pmu"},
	{.name = "xscale-pmu"},
232 233 234
	{},
};

235 236 237 238 239 240 241 242 243 244 245 246
static const struct pmu_probe_info pmu_probe_table[] = {
	ARM_PMU_PROBE(ARM_CPU_PART_ARM1136, armv6_1136_pmu_init),
	ARM_PMU_PROBE(ARM_CPU_PART_ARM1156, armv6_1156_pmu_init),
	ARM_PMU_PROBE(ARM_CPU_PART_ARM1176, armv6_1176_pmu_init),
	ARM_PMU_PROBE(ARM_CPU_PART_ARM11MPCORE, armv6mpcore_pmu_init),
	ARM_PMU_PROBE(ARM_CPU_PART_CORTEX_A8, armv7_a8_pmu_init),
	ARM_PMU_PROBE(ARM_CPU_PART_CORTEX_A9, armv7_a9_pmu_init),
	XSCALE_PMU_PROBE(ARM_CPU_XSCALE_ARCH_V1, xscale1pmu_init),
	XSCALE_PMU_PROBE(ARM_CPU_XSCALE_ARCH_V2, xscale2pmu_init),
	{ /* sentinel value */ }
};

247 248 249
/*
 * CPU PMU identification and probing.
 */
250
static int probe_current_pmu(struct arm_pmu *pmu)
251 252
{
	int cpu = get_cpu();
253
	unsigned int cpuid = read_cpuid_id();
254
	int ret = -ENODEV;
255
	const struct pmu_probe_info *info;
256 257 258

	pr_info("probing PMU on CPU %d\n", cpu);

259 260 261 262
	for (info = pmu_probe_table; info->init != NULL; info++) {
		if ((cpuid & info->mask) != info->cpuid)
			continue;
		ret = info->init(pmu);
263
		break;
264 265 266
	}

	put_cpu();
267
	return ret;
268 269
}

270
static int cpu_pmu_device_probe(struct platform_device *pdev)
271 272
{
	const struct of_device_id *of_id;
273
	const int (*init_fn)(struct arm_pmu *);
274
	struct device_node *node = pdev->dev.of_node;
275 276
	struct arm_pmu *pmu;
	int ret = -ENODEV;
277 278

	if (cpu_pmu) {
279
		pr_info("attempt to register multiple PMU devices!\n");
280 281 282
		return -ENOSPC;
	}

283 284
	pmu = kzalloc(sizeof(struct arm_pmu), GFP_KERNEL);
	if (!pmu) {
285
		pr_info("failed to allocate PMU device!\n");
286 287 288
		return -ENOMEM;
	}

289 290 291
	cpu_pmu = pmu;
	cpu_pmu->plat_device = pdev;

292 293
	if (node && (of_id = of_match_node(cpu_pmu_of_device_ids, pdev->dev.of_node))) {
		init_fn = of_id->data;
294
		ret = init_fn(pmu);
295
	} else {
296
		ret = probe_current_pmu(pmu);
297 298
	}

299
	if (ret) {
300
		pr_info("failed to probe PMU!\n");
301
		goto out_free;
302
	}
303 304

	cpu_pmu_init(cpu_pmu);
305
	ret = armpmu_register(cpu_pmu, -1);
306

307 308 309 310
	if (!ret)
		return 0;

out_free:
311
	pr_info("failed to register PMU devices!\n");
312 313
	kfree(pmu);
	return ret;
314 315 316 317 318 319 320 321 322 323 324 325 326 327
}

static struct platform_driver cpu_pmu_driver = {
	.driver		= {
		.name	= "arm-pmu",
		.pm	= &armpmu_dev_pm_ops,
		.of_match_table = cpu_pmu_of_device_ids,
	},
	.probe		= cpu_pmu_device_probe,
	.id_table	= cpu_pmu_plat_device_ids,
};

static int __init register_pmu_driver(void)
{
328 329 330 331 332 333 334 335 336 337 338
	int err;

	err = register_cpu_notifier(&cpu_pmu_hotplug_notifier);
	if (err)
		return err;

	err = platform_driver_register(&cpu_pmu_driver);
	if (err)
		unregister_cpu_notifier(&cpu_pmu_hotplug_notifier);

	return err;
339 340
}
device_initcall(register_pmu_driver);