[net-next-2.6.git] / drivers / oprofile / oprofile_perf.c

/*
 * Copyright 2010 ARM Ltd.
 *
 * Perf-events backend for OProfile.
 */
#include <linux/perf_event.h>
#include <linux/platform_device.h>
#include <linux/oprofile.h>
#include <linux/slab.h>

/*
 * Per performance monitor configuration as set via oprofilefs.
 */
struct op_counter_config {
	unsigned long count;
	unsigned long enabled;
	unsigned long event;
	unsigned long unit_mask;
	unsigned long kernel;
	unsigned long user;
	struct perf_event_attr attr;
};

static int oprofile_perf_enabled;
static DEFINE_MUTEX(oprofile_perf_mutex);

static struct op_counter_config *counter_config;
static struct perf_event **perf_events[nr_cpumask_bits];
static int num_counters;

/*
 * Overflow callback for oprofile.
 */
static void op_overflow_handler(struct perf_event *event, int unused,
			struct perf_sample_data *data, struct pt_regs *regs)
{
	int id;
	u32 cpu = smp_processor_id();

	for (id = 0; id < num_counters; ++id)
		if (perf_events[cpu][id] == event)
			break;

	if (id != num_counters)
		oprofile_add_sample(regs, id);
	else
		pr_warning("oprofile: ignoring spurious overflow "
				"on cpu %u\n", cpu);
}

/*
 * Called by oprofile_perf_setup to create perf attributes to mirror the oprofile
 * settings in counter_config. Attributes are created as `pinned' events and
 * so are permanently scheduled on the PMU.
 */
static void op_perf_setup(void)
{
	int i;
	u32 size = sizeof(struct perf_event_attr);
	struct perf_event_attr *attr;

	for (i = 0; i < num_counters; ++i) {
		attr = &counter_config[i].attr;
		memset(attr, 0, size);
		attr->type		= PERF_TYPE_RAW;
		attr->size		= size;
		attr->config		= counter_config[i].event;
		attr->sample_period	= counter_config[i].count;
		attr->pinned		= 1;
	}
}

static int op_create_counter(int cpu, int event)
{
	struct perf_event *pevent;

	if (!counter_config[event].enabled || perf_events[cpu][event])
		return 0;

	pevent = perf_event_create_kernel_counter(&counter_config[event].attr,
						  cpu, NULL,
						  op_overflow_handler);

	if (IS_ERR(pevent))
		return PTR_ERR(pevent);

	if (pevent->state != PERF_EVENT_STATE_ACTIVE) {
		perf_event_release_kernel(pevent);
		pr_warning("oprofile: failed to enable event %d "
				"on CPU %d\n", event, cpu);
		return -EBUSY;
	}

	perf_events[cpu][event] = pevent;

	return 0;
}

static void op_destroy_counter(int cpu, int event)
{
	struct perf_event *pevent = perf_events[cpu][event];

	if (pevent) {
		perf_event_release_kernel(pevent);
		perf_events[cpu][event] = NULL;
	}
}

/*
 * Called by oprofile_perf_start to create active perf events based on the
 * perviously configured attributes.
 */
static int op_perf_start(void)
{
	int cpu, event, ret = 0;

	for_each_online_cpu(cpu) {
		for (event = 0; event < num_counters; ++event) {
			ret = op_create_counter(cpu, event);
			if (ret)
				return ret;
		}
	}

	return ret;
}

/*
 * Called by oprofile_perf_stop at the end of a profiling run.
 */
static void op_perf_stop(void)
{
	int cpu, event;

	for_each_online_cpu(cpu)
		for (event = 0; event < num_counters; ++event)
			op_destroy_counter(cpu, event);
}

static int oprofile_perf_create_files(struct super_block *sb, struct dentry *root)
{
	unsigned int i;

	for (i = 0; i < num_counters; i++) {
		struct dentry *dir;
		char buf[4];

		snprintf(buf, sizeof buf, "%d", i);
		dir = oprofilefs_mkdir(sb, root, buf);
		oprofilefs_create_ulong(sb, dir, "enabled", &counter_config[i].enabled);
		oprofilefs_create_ulong(sb, dir, "event", &counter_config[i].event);
		oprofilefs_create_ulong(sb, dir, "count", &counter_config[i].count);
		oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask);
		oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel);
		oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user);
	}

	return 0;
}

static int oprofile_perf_setup(void)
{
	spin_lock(&oprofilefs_lock);
	op_perf_setup();
	spin_unlock(&oprofilefs_lock);
	return 0;
}

static int oprofile_perf_start(void)
{
	int ret = -EBUSY;

	mutex_lock(&oprofile_perf_mutex);
	if (!oprofile_perf_enabled) {
		ret = 0;
		op_perf_start();
		oprofile_perf_enabled = 1;
	}
	mutex_unlock(&oprofile_perf_mutex);
	return ret;
}

static void oprofile_perf_stop(void)
{
	mutex_lock(&oprofile_perf_mutex);
	if (oprofile_perf_enabled)
		op_perf_stop();
	oprofile_perf_enabled = 0;
	mutex_unlock(&oprofile_perf_mutex);
}

#ifdef CONFIG_PM

static int oprofile_perf_suspend(struct platform_device *dev, pm_message_t state)
{
	mutex_lock(&oprofile_perf_mutex);
	if (oprofile_perf_enabled)
		op_perf_stop();
	mutex_unlock(&oprofile_perf_mutex);
	return 0;
}

static int oprofile_perf_resume(struct platform_device *dev)
{
	mutex_lock(&oprofile_perf_mutex);
	if (oprofile_perf_enabled && op_perf_start())
		oprofile_perf_enabled = 0;
	mutex_unlock(&oprofile_perf_mutex);
	return 0;
}

static struct platform_driver oprofile_driver = {
	.driver		= {
		.name		= "oprofile-perf",
	},
	.resume		= oprofile_perf_resume,
	.suspend	= oprofile_perf_suspend,
};

static struct platform_device *oprofile_pdev;

static int __init init_driverfs(void)
{
	int ret;

	ret = platform_driver_register(&oprofile_driver);
	if (ret)
		return ret;

	oprofile_pdev =	platform_device_register_simple(
				oprofile_driver.driver.name, 0, NULL, 0);
	if (IS_ERR(oprofile_pdev)) {
		ret = PTR_ERR(oprofile_pdev);
		platform_driver_unregister(&oprofile_driver);
	}

	return ret;
}

static void exit_driverfs(void)
{
	platform_device_unregister(oprofile_pdev);
	platform_driver_unregister(&oprofile_driver);
}

#else

static inline int  init_driverfs(void) { return 0; }
static inline void exit_driverfs(void) { }

#endif /* CONFIG_PM */

void oprofile_perf_exit(void)
{
	int cpu, id;
	struct perf_event *event;

	for_each_possible_cpu(cpu) {
		for (id = 0; id < num_counters; ++id) {
			event = perf_events[cpu][id];
			if (event)
				perf_event_release_kernel(event);
		}

		kfree(perf_events[cpu]);
	}

	kfree(counter_config);
	exit_driverfs();
}

int __init oprofile_perf_init(struct oprofile_operations *ops)
{
	int cpu, ret = 0;

	ret = init_driverfs();
	if (ret)
		return ret;

	memset(&perf_events, 0, sizeof(perf_events));

	num_counters = perf_num_counters();
	if (num_counters <= 0) {
		pr_info("oprofile: no performance counters\n");
		ret = -ENODEV;
		goto out;
	}

	counter_config = kcalloc(num_counters,
			sizeof(struct op_counter_config), GFP_KERNEL);

	if (!counter_config) {
		pr_info("oprofile: failed to allocate %d "
				"counters\n", num_counters);
		ret = -ENOMEM;
		num_counters = 0;
		goto out;
	}

	for_each_possible_cpu(cpu) {
		perf_events[cpu] = kcalloc(num_counters,
				sizeof(struct perf_event *), GFP_KERNEL);
		if (!perf_events[cpu]) {
			pr_info("oprofile: failed to allocate %d perf events "
					"for cpu %d\n", num_counters, cpu);
			ret = -ENOMEM;
			goto out;
		}
	}

	ops->create_files	= oprofile_perf_create_files;
	ops->setup		= oprofile_perf_setup;
	ops->start		= oprofile_perf_start;
	ops->stop		= oprofile_perf_stop;
	ops->shutdown		= oprofile_perf_stop;
	ops->cpu_type		= op_name_from_perf_id();

	if (!ops->cpu_type)
		ret = -ENODEV;
	else
		pr_info("oprofile: using %s\n", ops->cpu_type);

out:
	if (ret)
		oprofile_perf_exit();

	return ret;
}
Commit	Line	Data
3d90a007 MF	1	/*
	2	* Copyright 2010 ARM Ltd.
	3	*
	4	* Perf-events backend for OProfile.
	5	*/
	6	#include <linux/perf_event.h>
277dd984	7	#include <linux/platform_device.h>
3d90a007 MF	8	#include <linux/oprofile.h>
	9	#include <linux/slab.h>
	10
	11	/*
	12	* Per performance monitor configuration as set via oprofilefs.
	13	*/
	14	struct op_counter_config {
	15	unsigned long count;
	16	unsigned long enabled;
	17	unsigned long event;
	18	unsigned long unit_mask;
	19	unsigned long kernel;
	20	unsigned long user;
	21	struct perf_event_attr attr;
	22	};
	23
	24	static int oprofile_perf_enabled;
	25	static DEFINE_MUTEX(oprofile_perf_mutex);
	26
	27	static struct op_counter_config *counter_config;
	28	static struct perf_event **perf_events[nr_cpumask_bits];
	29	static int num_counters;
	30
	31	/*
	32	* Overflow callback for oprofile.
	33	*/
	34	static void op_overflow_handler(struct perf_event *event, int unused,
	35	struct perf_sample_data data, struct pt_regs regs)
	36	{
	37	int id;
	38	u32 cpu = smp_processor_id();
	39
	40	for (id = 0; id < num_counters; ++id)
	41	if (perf_events[cpu][id] == event)
	42	break;
	43
	44	if (id != num_counters)
	45	oprofile_add_sample(regs, id);
	46	else
	47	pr_warning("oprofile: ignoring spurious overflow "
	48	"on cpu %u\n", cpu);
	49	}
	50
	51	/*
	52	* Called by oprofile_perf_setup to create perf attributes to mirror the oprofile
	53	* settings in counter_config. Attributes are created as `pinned' events and
	54	* so are permanently scheduled on the PMU.
	55	*/
	56	static void op_perf_setup(void)
	57	{
	58	int i;
	59	u32 size = sizeof(struct perf_event_attr);
	60	struct perf_event_attr *attr;
	61
	62	for (i = 0; i < num_counters; ++i) {
	63	attr = &counter_config[i].attr;
	64	memset(attr, 0, size);
	65	attr->type = PERF_TYPE_RAW;
	66	attr->size = size;
	67	attr->config = counter_config[i].event;
	68	attr->sample_period = counter_config[i].count;
	69	attr->pinned = 1;
	70	}
	71	}
72
73	static int op_create_counter(int cpu, int event)
74	{
3d90a007 MF	75	struct perf_event *pevent;
3d90a007 MF	76
2bcb2b64 RR	77	if (!counter_config[event].enabled \|\| perf_events[cpu][event])
2bcb2b64 RR	78	return 0;
3d90a007 MF	79
3d90a007 MF	80	pevent = perf_event_create_kernel_counter(&counter_config[event].attr,
6268464b	81	cpu, NULL,
3d90a007 MF	82	op_overflow_handler);
3d90a007 MF	83
2bcb2b64 RR	84	if (IS_ERR(pevent))
	85	return PTR_ERR(pevent);
	86
	87	if (pevent->state != PERF_EVENT_STATE_ACTIVE) {
81771974	88	perf_event_release_kernel(pevent);
3d90a007 MF	89	pr_warning("oprofile: failed to enable event %d "
3d90a007 MF	90	"on CPU %d\n", event, cpu);
2bcb2b64	91	return -EBUSY;
3d90a007 MF	92	}
3d90a007 MF	93
2bcb2b64 RR	94	perf_events[cpu][event] = pevent;
	95
	96	return 0;
3d90a007 MF	97	}
	98
	99	static void op_destroy_counter(int cpu, int event)
	100	{
	101	struct perf_event *pevent = perf_events[cpu][event];
	102
	103	if (pevent) {
	104	perf_event_release_kernel(pevent);
	105	perf_events[cpu][event] = NULL;
	106	}
	107	}
	108
	109	/*
	110	* Called by oprofile_perf_start to create active perf events based on the
	111	* perviously configured attributes.
	112	*/
	113	static int op_perf_start(void)
	114	{
	115	int cpu, event, ret = 0;
	116
	117	for_each_online_cpu(cpu) {
	118	for (event = 0; event < num_counters; ++event) {
	119	ret = op_create_counter(cpu, event);
	120	if (ret)
9c91283a	121	return ret;
3d90a007 MF	122	}
	123	}
	124
3d90a007 MF	125	return ret;
	126	}
	127
	128	/*
	129	* Called by oprofile_perf_stop at the end of a profiling run.
	130	*/
	131	static void op_perf_stop(void)
	132	{
	133	int cpu, event;
	134
	135	for_each_online_cpu(cpu)
	136	for (event = 0; event < num_counters; ++event)
	137	op_destroy_counter(cpu, event);
	138	}
	139
	140	static int oprofile_perf_create_files(struct super_block sb, struct dentry root)
	141	{
	142	unsigned int i;
	143
	144	for (i = 0; i < num_counters; i++) {
	145	struct dentry *dir;
	146	char buf[4];
	147
	148	snprintf(buf, sizeof buf, "%d", i);
	149	dir = oprofilefs_mkdir(sb, root, buf);
	150	oprofilefs_create_ulong(sb, dir, "enabled", &counter_config[i].enabled);
	151	oprofilefs_create_ulong(sb, dir, "event", &counter_config[i].event);
	152	oprofilefs_create_ulong(sb, dir, "count", &counter_config[i].count);
	153	oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask);
	154	oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel);
	155	oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user);
	156	}
	157
	158	return 0;
	159	}
	160
	161	static int oprofile_perf_setup(void)
	162	{
	163	spin_lock(&oprofilefs_lock);
	164	op_perf_setup();
	165	spin_unlock(&oprofilefs_lock);
	166	return 0;
	167	}
	168
	169	static int oprofile_perf_start(void)
	170	{
	171	int ret = -EBUSY;
	172
	173	mutex_lock(&oprofile_perf_mutex);
	174	if (!oprofile_perf_enabled) {
	175	ret = 0;
	176	op_perf_start();
	177	oprofile_perf_enabled = 1;
	178	}
	179	mutex_unlock(&oprofile_perf_mutex);
	180	return ret;
	181	}
	182
	183	static void oprofile_perf_stop(void)
	184	{
	185	mutex_lock(&oprofile_perf_mutex);
	186	if (oprofile_perf_enabled)
	187	op_perf_stop();
	188	oprofile_perf_enabled = 0;
189	mutex_unlock(&oprofile_perf_mutex);
190	}
191
192	#ifdef CONFIG_PM
cd254f29	193
3d90a007 MF	194	static int oprofile_perf_suspend(struct platform_device *dev, pm_message_t state)
	195	{
	196	mutex_lock(&oprofile_perf_mutex);
	197	if (oprofile_perf_enabled)
	198	op_perf_stop();
	199	mutex_unlock(&oprofile_perf_mutex);
	200	return 0;
	201	}
	202
	203	static int oprofile_perf_resume(struct platform_device *dev)
	204	{
	205	mutex_lock(&oprofile_perf_mutex);
	206	if (oprofile_perf_enabled && op_perf_start())
	207	oprofile_perf_enabled = 0;
	208	mutex_unlock(&oprofile_perf_mutex);
	209	return 0;
	210	}
	211
	212	static struct platform_driver oprofile_driver = {
	213	.driver = {
	214	.name = "oprofile-perf",
	215	},
	216	.resume = oprofile_perf_resume,
	217	.suspend = oprofile_perf_suspend,
	218	};
	219
	220	static struct platform_device *oprofile_pdev;
	221
	222	static int __init init_driverfs(void)
	223	{
	224	int ret;
	225
	226	ret = platform_driver_register(&oprofile_driver);
	227	if (ret)
9c91283a	228	return ret;
3d90a007 MF	229
	230	oprofile_pdev = platform_device_register_simple(
	231	oprofile_driver.driver.name, 0, NULL, 0);
	232	if (IS_ERR(oprofile_pdev)) {
	233	ret = PTR_ERR(oprofile_pdev);
	234	platform_driver_unregister(&oprofile_driver);
	235	}
	236
3d90a007 MF	237	return ret;
	238	}
	239
b3b3a9b6	240	static void exit_driverfs(void)
3d90a007 MF	241	{
	242	platform_device_unregister(oprofile_pdev);
	243	platform_driver_unregister(&oprofile_driver);
	244	}
cd254f29	245
3d90a007	246	#else
cd254f29 RR	247
	248	static inline int init_driverfs(void) { return 0; }
	249	static inline void exit_driverfs(void) { }
	250
3d90a007 MF	251	#endif /* CONFIG_PM */
3d90a007 MF	252
e9677b3c RR	253	void oprofile_perf_exit(void)
	254	{
	255	int cpu, id;
	256	struct perf_event *event;
	257
	258	for_each_possible_cpu(cpu) {
	259	for (id = 0; id < num_counters; ++id) {
	260	event = perf_events[cpu][id];
	261	if (event)
	262	perf_event_release_kernel(event);
	263	}
	264
	265	kfree(perf_events[cpu]);
	266	}
	267
	268	kfree(counter_config);
	269	exit_driverfs();
	270	}
	271
3d90a007 MF	272	int __init oprofile_perf_init(struct oprofile_operations *ops)
	273	{
	274	int cpu, ret = 0;
	275
e9677b3c RR	276	ret = init_driverfs();
	277	if (ret)
	278	return ret;
	279
3d90a007 MF	280	memset(&perf_events, 0, sizeof(perf_events));
	281
	282	num_counters = perf_num_counters();
	283	if (num_counters <= 0) {
	284	pr_info("oprofile: no performance counters\n");
	285	ret = -ENODEV;
	286	goto out;
	287	}
	288
	289	counter_config = kcalloc(num_counters,
	290	sizeof(struct op_counter_config), GFP_KERNEL);
	291
	292	if (!counter_config) {
	293	pr_info("oprofile: failed to allocate %d "
	294	"counters\n", num_counters);
	295	ret = -ENOMEM;
e9677b3c	296	num_counters = 0;
3d90a007 MF	297	goto out;
	298	}
	299
3d90a007 MF	300	for_each_possible_cpu(cpu) {
	301	perf_events[cpu] = kcalloc(num_counters,
	302	sizeof(struct perf_event *), GFP_KERNEL);
	303	if (!perf_events[cpu]) {
	304	pr_info("oprofile: failed to allocate %d perf events "
	305	"for cpu %d\n", num_counters, cpu);
	306	ret = -ENOMEM;
	307	goto out;
	308	}
	309	}
	310
	311	ops->create_files = oprofile_perf_create_files;
	312	ops->setup = oprofile_perf_setup;
	313	ops->start = oprofile_perf_start;
	314	ops->stop = oprofile_perf_stop;
	315	ops->shutdown = oprofile_perf_stop;
	316	ops->cpu_type = op_name_from_perf_id();
	317
	318	if (!ops->cpu_type)
	319	ret = -ENODEV;
	320	else
	321	pr_info("oprofile: using %s\n", ops->cpu_type);
	322
	323	out:
e9677b3c RR	324	if (ret)
e9677b3c RR	325	oprofile_perf_exit();
3d90a007 MF	326
	327	return ret;
	328	}