[mirror_ubuntu-artful-kernel.git] / tools / perf / util / evsel.c

/*
 * Copyright (C) 2011, Red Hat Inc, Arnaldo Carvalho de Melo <acme@redhat.com>
 *
 * Parts came from builtin-{top,stat,record}.c, see those files for further
 * copyright notes.
 *
 * Released under the GPL v2. (and only v2, not any later version)
 */

#include <byteswap.h>
#include "asm/bug.h"
#include "evsel.h"
#include "evlist.h"
#include "util.h"
#include "cpumap.h"
#include "thread_map.h"
#include "target.h"

#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
#define GROUP_FD(group_fd, cpu) (*(int *)xyarray__entry(group_fd, cpu, 0))

int __perf_evsel__sample_size(u64 sample_type)
{
	u64 mask = sample_type & PERF_SAMPLE_MASK;
	int size = 0;
	int i;

	for (i = 0; i < 64; i++) {
		if (mask & (1ULL << i))
			size++;
	}

	size *= sizeof(u64);

	return size;
}

void hists__init(struct hists *hists)
{
	memset(hists, 0, sizeof(*hists));
	hists->entries_in_array[0] = hists->entries_in_array[1] = RB_ROOT;
	hists->entries_in = &hists->entries_in_array[0];
	hists->entries_collapsed = RB_ROOT;
	hists->entries = RB_ROOT;
	pthread_mutex_init(&hists->lock, NULL);
}

void perf_evsel__init(struct perf_evsel *evsel,
		      struct perf_event_attr *attr, int idx)
{
	evsel->idx	   = idx;
	evsel->attr	   = *attr;
	INIT_LIST_HEAD(&evsel->node);
	hists__init(&evsel->hists);
}

struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
{
	struct perf_evsel *evsel = zalloc(sizeof(*evsel));

	if (evsel != NULL)
		perf_evsel__init(evsel, attr, idx);

	return evsel;
}

void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts,
			struct perf_evsel *first)
{
	struct perf_event_attr *attr = &evsel->attr;
	int track = !evsel->idx; /* only the first counter needs these */

	attr->disabled = 1;
	attr->sample_id_all = opts->sample_id_all_missing ? 0 : 1;
	attr->inherit	    = !opts->no_inherit;
	attr->read_format   = PERF_FORMAT_TOTAL_TIME_ENABLED |
			      PERF_FORMAT_TOTAL_TIME_RUNNING |
			      PERF_FORMAT_ID;

	attr->sample_type  |= PERF_SAMPLE_IP | PERF_SAMPLE_TID;

	/*
	 * We default some events to a 1 default interval. But keep
	 * it a weak assumption overridable by the user.
	 */
	if (!attr->sample_period || (opts->user_freq != UINT_MAX &&
				     opts->user_interval != ULLONG_MAX)) {
		if (opts->freq) {
			attr->sample_type	|= PERF_SAMPLE_PERIOD;
			attr->freq		= 1;
			attr->sample_freq	= opts->freq;
		} else {
			attr->sample_period = opts->default_interval;
		}
	}

	if (opts->no_samples)
		attr->sample_freq = 0;

	if (opts->inherit_stat)
		attr->inherit_stat = 1;

	if (opts->sample_address) {
		attr->sample_type	|= PERF_SAMPLE_ADDR;
		attr->mmap_data = track;
	}

	if (opts->call_graph)
		attr->sample_type	|= PERF_SAMPLE_CALLCHAIN;

	if (perf_target__has_cpu(&opts->target))
		attr->sample_type	|= PERF_SAMPLE_CPU;

	if (opts->period)
		attr->sample_type	|= PERF_SAMPLE_PERIOD;

	if (!opts->sample_id_all_missing &&
	    (opts->sample_time || !opts->no_inherit ||
	     perf_target__has_cpu(&opts->target)))
		attr->sample_type	|= PERF_SAMPLE_TIME;

	if (opts->raw_samples) {
		attr->sample_type	|= PERF_SAMPLE_TIME;
		attr->sample_type	|= PERF_SAMPLE_RAW;
		attr->sample_type	|= PERF_SAMPLE_CPU;
	}

	if (opts->no_delay) {
		attr->watermark = 0;
		attr->wakeup_events = 1;
	}
	if (opts->branch_stack) {
		attr->sample_type	|= PERF_SAMPLE_BRANCH_STACK;
		attr->branch_sample_type = opts->branch_stack;
	}

	attr->mmap = track;
	attr->comm = track;

	if (perf_target__none(&opts->target) &&
	    (!opts->group || evsel == first)) {
		attr->enable_on_exec = 1;
	}
}

int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
{
	int cpu, thread;
	evsel->fd = xyarray__new(ncpus, nthreads, sizeof(int));

	if (evsel->fd) {
		for (cpu = 0; cpu < ncpus; cpu++) {
			for (thread = 0; thread < nthreads; thread++) {
				FD(evsel, cpu, thread) = -1;
			}
		}
	}

	return evsel->fd != NULL ? 0 : -ENOMEM;
}

int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads)
{
	evsel->sample_id = xyarray__new(ncpus, nthreads, sizeof(struct perf_sample_id));
	if (evsel->sample_id == NULL)
		return -ENOMEM;

	evsel->id = zalloc(ncpus * nthreads * sizeof(u64));
	if (evsel->id == NULL) {
		xyarray__delete(evsel->sample_id);
		evsel->sample_id = NULL;
		return -ENOMEM;
	}

	return 0;
}

int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus)
{
	evsel->counts = zalloc((sizeof(*evsel->counts) +
				(ncpus * sizeof(struct perf_counts_values))));
	return evsel->counts != NULL ? 0 : -ENOMEM;
}

void perf_evsel__free_fd(struct perf_evsel *evsel)
{
	xyarray__delete(evsel->fd);
	evsel->fd = NULL;
}

void perf_evsel__free_id(struct perf_evsel *evsel)
{
	xyarray__delete(evsel->sample_id);
	evsel->sample_id = NULL;
	free(evsel->id);
	evsel->id = NULL;
}

void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
{
	int cpu, thread;

	for (cpu = 0; cpu < ncpus; cpu++)
		for (thread = 0; thread < nthreads; ++thread) {
			close(FD(evsel, cpu, thread));
			FD(evsel, cpu, thread) = -1;
		}
}

void perf_evsel__exit(struct perf_evsel *evsel)
{
	assert(list_empty(&evsel->node));
	xyarray__delete(evsel->fd);
	xyarray__delete(evsel->sample_id);
	free(evsel->id);
}

void perf_evsel__delete(struct perf_evsel *evsel)
{
	perf_evsel__exit(evsel);
	close_cgroup(evsel->cgrp);
	free(evsel->name);
	free(evsel);
}

int __perf_evsel__read_on_cpu(struct perf_evsel *evsel,
			      int cpu, int thread, bool scale)
{
	struct perf_counts_values count;
	size_t nv = scale ? 3 : 1;

	if (FD(evsel, cpu, thread) < 0)
		return -EINVAL;

	if (evsel->counts == NULL && perf_evsel__alloc_counts(evsel, cpu + 1) < 0)
		return -ENOMEM;

	if (readn(FD(evsel, cpu, thread), &count, nv * sizeof(u64)) < 0)
		return -errno;

	if (scale) {
		if (count.run == 0)
			count.val = 0;
		else if (count.run < count.ena)
			count.val = (u64)((double)count.val * count.ena / count.run + 0.5);
	} else
		count.ena = count.run = 0;

	evsel->counts->cpu[cpu] = count;
	return 0;
}

int __perf_evsel__read(struct perf_evsel *evsel,
		       int ncpus, int nthreads, bool scale)
{
	size_t nv = scale ? 3 : 1;
	int cpu, thread;
	struct perf_counts_values *aggr = &evsel->counts->aggr, count;

	aggr->val = aggr->ena = aggr->run = 0;

	for (cpu = 0; cpu < ncpus; cpu++) {
		for (thread = 0; thread < nthreads; thread++) {
			if (FD(evsel, cpu, thread) < 0)
				continue;

			if (readn(FD(evsel, cpu, thread),
				  &count, nv * sizeof(u64)) < 0)
				return -errno;

			aggr->val += count.val;
			if (scale) {
				aggr->ena += count.ena;
				aggr->run += count.run;
			}
		}
	}

	evsel->counts->scaled = 0;
	if (scale) {
		if (aggr->run == 0) {
			evsel->counts->scaled = -1;
			aggr->val = 0;
			return 0;
		}

		if (aggr->run < aggr->ena) {
			evsel->counts->scaled = 1;
			aggr->val = (u64)((double)aggr->val * aggr->ena / aggr->run + 0.5);
		}
	} else
		aggr->ena = aggr->run = 0;

	return 0;
}

static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
			      struct thread_map *threads, bool group,
			      struct xyarray *group_fds)
{
	int cpu, thread;
	unsigned long flags = 0;
	int pid = -1, err;

	if (evsel->fd == NULL &&
	    perf_evsel__alloc_fd(evsel, cpus->nr, threads->nr) < 0)
		return -ENOMEM;

	if (evsel->cgrp) {
		flags = PERF_FLAG_PID_CGROUP;
		pid = evsel->cgrp->fd;
	}

	for (cpu = 0; cpu < cpus->nr; cpu++) {
		int group_fd = group_fds ? GROUP_FD(group_fds, cpu) : -1;

		for (thread = 0; thread < threads->nr; thread++) {

			if (!evsel->cgrp)
				pid = threads->map[thread];

			FD(evsel, cpu, thread) = sys_perf_event_open(&evsel->attr,
								     pid,
								     cpus->map[cpu],
								     group_fd, flags);
			if (FD(evsel, cpu, thread) < 0) {
				err = -errno;
				goto out_close;
			}

			if (group && group_fd == -1)
				group_fd = FD(evsel, cpu, thread);
		}
	}

	return 0;

out_close:
	do {
		while (--thread >= 0) {
			close(FD(evsel, cpu, thread));
			FD(evsel, cpu, thread) = -1;
		}
		thread = threads->nr;
	} while (--cpu >= 0);
	return err;
}

void perf_evsel__close(struct perf_evsel *evsel, int ncpus, int nthreads)
{
	if (evsel->fd == NULL)
		return;

	perf_evsel__close_fd(evsel, ncpus, nthreads);
	perf_evsel__free_fd(evsel);
	evsel->fd = NULL;
}

static struct {
	struct cpu_map map;
	int cpus[1];
} empty_cpu_map = {
	.map.nr	= 1,
	.cpus	= { -1, },
};

static struct {
	struct thread_map map;
	int threads[1];
} empty_thread_map = {
	.map.nr	 = 1,
	.threads = { -1, },
};

int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
		     struct thread_map *threads, bool group,
		     struct xyarray *group_fd)
{
	if (cpus == NULL) {
		/* Work around old compiler warnings about strict aliasing */
		cpus = &empty_cpu_map.map;
	}

	if (threads == NULL)
		threads = &empty_thread_map.map;

	return __perf_evsel__open(evsel, cpus, threads, group, group_fd);
}

int perf_evsel__open_per_cpu(struct perf_evsel *evsel,
			     struct cpu_map *cpus, bool group,
			     struct xyarray *group_fd)
{
	return __perf_evsel__open(evsel, cpus, &empty_thread_map.map, group,
				  group_fd);
}

int perf_evsel__open_per_thread(struct perf_evsel *evsel,
				struct thread_map *threads, bool group,
				struct xyarray *group_fd)
{
	return __perf_evsel__open(evsel, &empty_cpu_map.map, threads, group,
				  group_fd);
}

static int perf_event__parse_id_sample(const union perf_event *event, u64 type,
				       struct perf_sample *sample)
{
	const u64 *array = event->sample.array;

	array += ((event->header.size -
		   sizeof(event->header)) / sizeof(u64)) - 1;

	if (type & PERF_SAMPLE_CPU) {
		u32 *p = (u32 *)array;
		sample->cpu = *p;
		array--;
	}

	if (type & PERF_SAMPLE_STREAM_ID) {
		sample->stream_id = *array;
		array--;
	}

	if (type & PERF_SAMPLE_ID) {
		sample->id = *array;
		array--;
	}

	if (type & PERF_SAMPLE_TIME) {
		sample->time = *array;
		array--;
	}

	if (type & PERF_SAMPLE_TID) {
		u32 *p = (u32 *)array;
		sample->pid = p[0];
		sample->tid = p[1];
	}

	return 0;
}

static bool sample_overlap(const union perf_event *event,
			   const void *offset, u64 size)
{
	const void *base = event;

	if (offset + size > base + event->header.size)
		return true;

	return false;
}

int perf_event__parse_sample(const union perf_event *event, u64 type,
			     int sample_size, bool sample_id_all,
			     struct perf_sample *data, bool swapped)
{
	const u64 *array;

	/*
	 * used for cross-endian analysis. See git commit 65014ab3
	 * for why this goofiness is needed.
	 */
	union {
		u64 val64;
		u32 val32[2];
	} u;

	memset(data, 0, sizeof(*data));
	data->cpu = data->pid = data->tid = -1;
	data->stream_id = data->id = data->time = -1ULL;
	data->period = 1;

	if (event->header.type != PERF_RECORD_SAMPLE) {
		if (!sample_id_all)
			return 0;
		return perf_event__parse_id_sample(event, type, data);
	}

	array = event->sample.array;

	if (sample_size + sizeof(event->header) > event->header.size)
		return -EFAULT;

	if (type & PERF_SAMPLE_IP) {
		data->ip = event->ip.ip;
		array++;
	}

	if (type & PERF_SAMPLE_TID) {
		u.val64 = *array;
		if (swapped) {
			/* undo swap of u64, then swap on individual u32s */
			u.val64 = bswap_64(u.val64);
			u.val32[0] = bswap_32(u.val32[0]);
			u.val32[1] = bswap_32(u.val32[1]);
		}

		data->pid = u.val32[0];
		data->tid = u.val32[1];
		array++;
	}

	if (type & PERF_SAMPLE_TIME) {
		data->time = *array;
		array++;
	}

	data->addr = 0;
	if (type & PERF_SAMPLE_ADDR) {
		data->addr = *array;
		array++;
	}

	data->id = -1ULL;
	if (type & PERF_SAMPLE_ID) {
		data->id = *array;
		array++;
	}

	if (type & PERF_SAMPLE_STREAM_ID) {
		data->stream_id = *array;
		array++;
	}

	if (type & PERF_SAMPLE_CPU) {

		u.val64 = *array;
		if (swapped) {
			/* undo swap of u64, then swap on individual u32s */
			u.val64 = bswap_64(u.val64);
			u.val32[0] = bswap_32(u.val32[0]);
		}

		data->cpu = u.val32[0];
		array++;
	}

	if (type & PERF_SAMPLE_PERIOD) {
		data->period = *array;
		array++;
	}

	if (type & PERF_SAMPLE_READ) {
		fprintf(stderr, "PERF_SAMPLE_READ is unsupported for now\n");
		return -1;
	}

	if (type & PERF_SAMPLE_CALLCHAIN) {
		if (sample_overlap(event, array, sizeof(data->callchain->nr)))
			return -EFAULT;

		data->callchain = (struct ip_callchain *)array;

		if (sample_overlap(event, array, data->callchain->nr))
			return -EFAULT;

		array += 1 + data->callchain->nr;
	}

	if (type & PERF_SAMPLE_RAW) {
		const u64 *pdata;

		u.val64 = *array;
		if (WARN_ONCE(swapped,
			      "Endianness of raw data not corrected!\n")) {
			/* undo swap of u64, then swap on individual u32s */
			u.val64 = bswap_64(u.val64);
			u.val32[0] = bswap_32(u.val32[0]);
			u.val32[1] = bswap_32(u.val32[1]);
		}

		if (sample_overlap(event, array, sizeof(u32)))
			return -EFAULT;

		data->raw_size = u.val32[0];
		pdata = (void *) array + sizeof(u32);

		if (sample_overlap(event, pdata, data->raw_size))
			return -EFAULT;

		data->raw_data = (void *) pdata;

		array = (void *)array + data->raw_size + sizeof(u32);
	}

	if (type & PERF_SAMPLE_BRANCH_STACK) {
		u64 sz;

		data->branch_stack = (struct branch_stack *)array;
		array++; /* nr */

		sz = data->branch_stack->nr * sizeof(struct branch_entry);
		sz /= sizeof(u64);
		array += sz;
	}
	return 0;
}

int perf_event__synthesize_sample(union perf_event *event, u64 type,
				  const struct perf_sample *sample,
				  bool swapped)
{
	u64 *array;

	/*
	 * used for cross-endian analysis. See git commit 65014ab3
	 * for why this goofiness is needed.
	 */
	union {
		u64 val64;
		u32 val32[2];
	} u;

	array = event->sample.array;

	if (type & PERF_SAMPLE_IP) {
		event->ip.ip = sample->ip;
		array++;
	}

	if (type & PERF_SAMPLE_TID) {
		u.val32[0] = sample->pid;
		u.val32[1] = sample->tid;
		if (swapped) {
			/*
			 * Inverse of what is done in perf_event__parse_sample
			 */
			u.val32[0] = bswap_32(u.val32[0]);
			u.val32[1] = bswap_32(u.val32[1]);
			u.val64 = bswap_64(u.val64);
		}

		*array = u.val64;
		array++;
	}

	if (type & PERF_SAMPLE_TIME) {
		*array = sample->time;
		array++;
	}

	if (type & PERF_SAMPLE_ADDR) {
		*array = sample->addr;
		array++;
	}

	if (type & PERF_SAMPLE_ID) {
		*array = sample->id;
		array++;
	}

	if (type & PERF_SAMPLE_STREAM_ID) {
		*array = sample->stream_id;
		array++;
	}

	if (type & PERF_SAMPLE_CPU) {
		u.val32[0] = sample->cpu;
		if (swapped) {
			/*
			 * Inverse of what is done in perf_event__parse_sample
			 */
			u.val32[0] = bswap_32(u.val32[0]);
			u.val64 = bswap_64(u.val64);
		}
		*array = u.val64;
		array++;
	}

	if (type & PERF_SAMPLE_PERIOD) {
		*array = sample->period;
		array++;
	}

	return 0;
}
Commit	Line	Data
f8a95309 ACM	1	/*
	2	* Copyright (C) 2011, Red Hat Inc, Arnaldo Carvalho de Melo <acme@redhat.com>
	3	*
	4	* Parts came from builtin-{top,stat,record}.c, see those files for further
	5	* copyright notes.
	6	*
	7	* Released under the GPL v2. (and only v2, not any later version)
	8	*/
	9
936be503 DA	10	#include <byteswap.h>
936be503 DA	11	#include "asm/bug.h"
69aad6f1	12	#include "evsel.h"
70082dd9	13	#include "evlist.h"
69aad6f1	14	#include "util.h"
86bd5e86	15	#include "cpumap.h"
fd78260b	16	#include "thread_map.h"
12864b31	17	#include "target.h"
69aad6f1	18
c52b12ed	19	#define FD(e, x, y) ((int )xyarray__entry(e->fd, x, y))
727ab04e	20	#define GROUP_FD(group_fd, cpu) ((int )xyarray__entry(group_fd, cpu, 0))
c52b12ed	21
c2a70653 ACM	22	int __perf_evsel__sample_size(u64 sample_type)
	23	{
	24	u64 mask = sample_type & PERF_SAMPLE_MASK;
	25	int size = 0;
	26	int i;
	27
	28	for (i = 0; i < 64; i++) {
	29	if (mask & (1ULL << i))
	30	size++;
	31	}
	32
	33	size *= sizeof(u64);
	34
	35	return size;
	36	}
	37
4bf9ce1b	38	void hists__init(struct hists *hists)
0e2a5f10 ACM	39	{
	40	memset(hists, 0, sizeof(*hists));
	41	hists->entries_in_array[0] = hists->entries_in_array[1] = RB_ROOT;
	42	hists->entries_in = &hists->entries_in_array[0];
	43	hists->entries_collapsed = RB_ROOT;
	44	hists->entries = RB_ROOT;
	45	pthread_mutex_init(&hists->lock, NULL);
	46	}
	47
ef1d1af2 ACM	48	void perf_evsel__init(struct perf_evsel *evsel,
	49	struct perf_event_attr *attr, int idx)
	50	{
	51	evsel->idx = idx;
	52	evsel->attr = *attr;
	53	INIT_LIST_HEAD(&evsel->node);
1980c2eb	54	hists__init(&evsel->hists);
ef1d1af2 ACM	55	}
ef1d1af2 ACM	56
23a2f3ab	57	struct perf_evsel perf_evsel__new(struct perf_event_attr attr, int idx)
69aad6f1 ACM	58	{
	59	struct perf_evsel evsel = zalloc(sizeof(evsel));
	60
ef1d1af2 ACM	61	if (evsel != NULL)
ef1d1af2 ACM	62	perf_evsel__init(evsel, attr, idx);
69aad6f1 ACM	63
	64	return evsel;
	65	}
	66
5090c6ae NK	67	void perf_evsel__config(struct perf_evsel evsel, struct perf_record_opts opts,
5090c6ae NK	68	struct perf_evsel *first)
0f82ebc4 ACM	69	{
	70	struct perf_event_attr *attr = &evsel->attr;
	71	int track = !evsel->idx; /* only the first counter needs these */
	72
5e1c81d9	73	attr->disabled = 1;
808e1226	74	attr->sample_id_all = opts->sample_id_all_missing ? 0 : 1;
0f82ebc4 ACM	75	attr->inherit = !opts->no_inherit;
	76	attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED \|
	77	PERF_FORMAT_TOTAL_TIME_RUNNING \|
	78	PERF_FORMAT_ID;
	79
	80	attr->sample_type \|= PERF_SAMPLE_IP \| PERF_SAMPLE_TID;
	81
	82	/*
	83	* We default some events to a 1 default interval. But keep
	84	* it a weak assumption overridable by the user.
	85	*/
	86	if (!attr->sample_period \|\| (opts->user_freq != UINT_MAX &&
	87	opts->user_interval != ULLONG_MAX)) {
	88	if (opts->freq) {
	89	attr->sample_type \|= PERF_SAMPLE_PERIOD;
	90	attr->freq = 1;
	91	attr->sample_freq = opts->freq;
	92	} else {
	93	attr->sample_period = opts->default_interval;
	94	}
	95	}
	96
	97	if (opts->no_samples)
	98	attr->sample_freq = 0;
	99
	100	if (opts->inherit_stat)
	101	attr->inherit_stat = 1;
	102
	103	if (opts->sample_address) {
	104	attr->sample_type \|= PERF_SAMPLE_ADDR;
	105	attr->mmap_data = track;
	106	}
	107
	108	if (opts->call_graph)
	109	attr->sample_type \|= PERF_SAMPLE_CALLCHAIN;
	110
e40ee742	111	if (perf_target__has_cpu(&opts->target))
0f82ebc4 ACM	112	attr->sample_type \|= PERF_SAMPLE_CPU;
0f82ebc4 ACM	113
3e76ac78 AV	114	if (opts->period)
	115	attr->sample_type \|= PERF_SAMPLE_PERIOD;
	116
808e1226	117	if (!opts->sample_id_all_missing &&
d67356e7	118	(opts->sample_time \|\| !opts->no_inherit \|\|
aa22dd49	119	perf_target__has_cpu(&opts->target)))
0f82ebc4 ACM	120	attr->sample_type \|= PERF_SAMPLE_TIME;
	121
	122	if (opts->raw_samples) {
	123	attr->sample_type \|= PERF_SAMPLE_TIME;
	124	attr->sample_type \|= PERF_SAMPLE_RAW;
	125	attr->sample_type \|= PERF_SAMPLE_CPU;
	126	}
	127
	128	if (opts->no_delay) {
	129	attr->watermark = 0;
	130	attr->wakeup_events = 1;
	131	}
bdfebd84 RAV	132	if (opts->branch_stack) {
	133	attr->sample_type \|= PERF_SAMPLE_BRANCH_STACK;
	134	attr->branch_sample_type = opts->branch_stack;
	135	}
0f82ebc4 ACM	136
	137	attr->mmap = track;
	138	attr->comm = track;
	139
d67356e7 NK	140	if (perf_target__none(&opts->target) &&
d67356e7 NK	141	(!opts->group \|\| evsel == first)) {
0f82ebc4 ACM	142	attr->enable_on_exec = 1;
	143	}
	144	}
	145
69aad6f1 ACM	146	int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
69aad6f1 ACM	147	{
4af4c955	148	int cpu, thread;
69aad6f1	149	evsel->fd = xyarray__new(ncpus, nthreads, sizeof(int));
4af4c955 DA	150
	151	if (evsel->fd) {
	152	for (cpu = 0; cpu < ncpus; cpu++) {
	153	for (thread = 0; thread < nthreads; thread++) {
	154	FD(evsel, cpu, thread) = -1;
	155	}
	156	}
	157	}
	158
69aad6f1 ACM	159	return evsel->fd != NULL ? 0 : -ENOMEM;
	160	}
	161
70db7533 ACM	162	int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads)
70db7533 ACM	163	{
a91e5431 ACM	164	evsel->sample_id = xyarray__new(ncpus, nthreads, sizeof(struct perf_sample_id));
	165	if (evsel->sample_id == NULL)
	166	return -ENOMEM;
	167
	168	evsel->id = zalloc(ncpus * nthreads * sizeof(u64));
	169	if (evsel->id == NULL) {
	170	xyarray__delete(evsel->sample_id);
	171	evsel->sample_id = NULL;
	172	return -ENOMEM;
	173	}
	174
	175	return 0;
70db7533 ACM	176	}
70db7533 ACM	177
c52b12ed ACM	178	int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus)
	179	{
	180	evsel->counts = zalloc((sizeof(*evsel->counts) +
	181	(ncpus * sizeof(struct perf_counts_values))));
	182	return evsel->counts != NULL ? 0 : -ENOMEM;
	183	}
	184
69aad6f1 ACM	185	void perf_evsel__free_fd(struct perf_evsel *evsel)
	186	{
	187	xyarray__delete(evsel->fd);
	188	evsel->fd = NULL;
	189	}
	190
70db7533 ACM	191	void perf_evsel__free_id(struct perf_evsel *evsel)
70db7533 ACM	192	{
a91e5431 ACM	193	xyarray__delete(evsel->sample_id);
	194	evsel->sample_id = NULL;
	195	free(evsel->id);
70db7533 ACM	196	evsel->id = NULL;
	197	}
	198
c52b12ed ACM	199	void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
	200	{
	201	int cpu, thread;
	202
	203	for (cpu = 0; cpu < ncpus; cpu++)
	204	for (thread = 0; thread < nthreads; ++thread) {
	205	close(FD(evsel, cpu, thread));
	206	FD(evsel, cpu, thread) = -1;
	207	}
	208	}
	209
ef1d1af2	210	void perf_evsel__exit(struct perf_evsel *evsel)
69aad6f1 ACM	211	{
	212	assert(list_empty(&evsel->node));
	213	xyarray__delete(evsel->fd);
a91e5431 ACM	214	xyarray__delete(evsel->sample_id);
a91e5431 ACM	215	free(evsel->id);
ef1d1af2 ACM	216	}
	217
	218	void perf_evsel__delete(struct perf_evsel *evsel)
	219	{
	220	perf_evsel__exit(evsel);
023695d9	221	close_cgroup(evsel->cgrp);
f0c55bcf	222	free(evsel->name);
69aad6f1 ACM	223	free(evsel);
69aad6f1 ACM	224	}
c52b12ed ACM	225
	226	int __perf_evsel__read_on_cpu(struct perf_evsel *evsel,
	227	int cpu, int thread, bool scale)
	228	{
	229	struct perf_counts_values count;
	230	size_t nv = scale ? 3 : 1;
	231
	232	if (FD(evsel, cpu, thread) < 0)
	233	return -EINVAL;
	234
4eed11d5 ACM	235	if (evsel->counts == NULL && perf_evsel__alloc_counts(evsel, cpu + 1) < 0)
	236	return -ENOMEM;
	237
c52b12ed ACM	238	if (readn(FD(evsel, cpu, thread), &count, nv * sizeof(u64)) < 0)
	239	return -errno;
	240
	241	if (scale) {
	242	if (count.run == 0)
	243	count.val = 0;
	244	else if (count.run < count.ena)
	245	count.val = (u64)((double)count.val * count.ena / count.run + 0.5);
	246	} else
	247	count.ena = count.run = 0;
	248
	249	evsel->counts->cpu[cpu] = count;
	250	return 0;
	251	}
	252
	253	int __perf_evsel__read(struct perf_evsel *evsel,
	254	int ncpus, int nthreads, bool scale)
	255	{
	256	size_t nv = scale ? 3 : 1;
	257	int cpu, thread;
	258	struct perf_counts_values *aggr = &evsel->counts->aggr, count;
	259
52bcd994	260	aggr->val = aggr->ena = aggr->run = 0;
c52b12ed ACM	261
	262	for (cpu = 0; cpu < ncpus; cpu++) {
	263	for (thread = 0; thread < nthreads; thread++) {
	264	if (FD(evsel, cpu, thread) < 0)
	265	continue;
	266
	267	if (readn(FD(evsel, cpu, thread),
	268	&count, nv * sizeof(u64)) < 0)
	269	return -errno;
	270
	271	aggr->val += count.val;
	272	if (scale) {
	273	aggr->ena += count.ena;
	274	aggr->run += count.run;
	275	}
	276	}
	277	}
	278
	279	evsel->counts->scaled = 0;
	280	if (scale) {
	281	if (aggr->run == 0) {
	282	evsel->counts->scaled = -1;
	283	aggr->val = 0;
	284	return 0;
	285	}
	286
	287	if (aggr->run < aggr->ena) {
	288	evsel->counts->scaled = 1;
	289	aggr->val = (u64)((double)aggr->val * aggr->ena / aggr->run + 0.5);
	290	}
	291	} else
	292	aggr->ena = aggr->run = 0;
	293
	294	return 0;
	295	}
48290609	296
0252208e	297	static int __perf_evsel__open(struct perf_evsel evsel, struct cpu_map cpus,
727ab04e ACM	298	struct thread_map *threads, bool group,
727ab04e ACM	299	struct xyarray *group_fds)
48290609	300	{
0252208e	301	int cpu, thread;
023695d9	302	unsigned long flags = 0;
727ab04e	303	int pid = -1, err;
48290609	304
0252208e ACM	305	if (evsel->fd == NULL &&
0252208e ACM	306	perf_evsel__alloc_fd(evsel, cpus->nr, threads->nr) < 0)
727ab04e	307	return -ENOMEM;
4eed11d5	308
023695d9 SE	309	if (evsel->cgrp) {
	310	flags = PERF_FLAG_PID_CGROUP;
	311	pid = evsel->cgrp->fd;
	312	}
	313
86bd5e86	314	for (cpu = 0; cpu < cpus->nr; cpu++) {
727ab04e	315	int group_fd = group_fds ? GROUP_FD(group_fds, cpu) : -1;
9d04f178	316
0252208e	317	for (thread = 0; thread < threads->nr; thread++) {
023695d9 SE	318
	319	if (!evsel->cgrp)
	320	pid = threads->map[thread];
	321
0252208e	322	FD(evsel, cpu, thread) = sys_perf_event_open(&evsel->attr,
023695d9	323	pid,
f08199d3	324	cpus->map[cpu],
023695d9	325	group_fd, flags);
727ab04e ACM	326	if (FD(evsel, cpu, thread) < 0) {
727ab04e ACM	327	err = -errno;
0252208e	328	goto out_close;
727ab04e	329	}
f08199d3 ACM	330
	331	if (group && group_fd == -1)
	332	group_fd = FD(evsel, cpu, thread);
0252208e	333	}
48290609 ACM	334	}
	335
	336	return 0;
	337
	338	out_close:
0252208e ACM	339	do {
	340	while (--thread >= 0) {
	341	close(FD(evsel, cpu, thread));
	342	FD(evsel, cpu, thread) = -1;
	343	}
	344	thread = threads->nr;
	345	} while (--cpu >= 0);
727ab04e ACM	346	return err;
	347	}
	348
	349	void perf_evsel__close(struct perf_evsel *evsel, int ncpus, int nthreads)
	350	{
	351	if (evsel->fd == NULL)
	352	return;
	353
	354	perf_evsel__close_fd(evsel, ncpus, nthreads);
	355	perf_evsel__free_fd(evsel);
	356	evsel->fd = NULL;
48290609 ACM	357	}
48290609 ACM	358
0252208e ACM	359	static struct {
	360	struct cpu_map map;
	361	int cpus[1];
	362	} empty_cpu_map = {
	363	.map.nr = 1,
	364	.cpus = { -1, },
	365	};
	366
	367	static struct {
	368	struct thread_map map;
	369	int threads[1];
	370	} empty_thread_map = {
	371	.map.nr = 1,
	372	.threads = { -1, },
	373	};
	374
f08199d3	375	int perf_evsel__open(struct perf_evsel evsel, struct cpu_map cpus,
727ab04e ACM	376	struct thread_map *threads, bool group,
727ab04e ACM	377	struct xyarray *group_fd)
48290609	378	{
0252208e ACM	379	if (cpus == NULL) {
	380	/* Work around old compiler warnings about strict aliasing */
	381	cpus = &empty_cpu_map.map;
48290609 ACM	382	}
48290609 ACM	383
0252208e ACM	384	if (threads == NULL)
0252208e ACM	385	threads = &empty_thread_map.map;
48290609	386
727ab04e	387	return __perf_evsel__open(evsel, cpus, threads, group, group_fd);
48290609 ACM	388	}
48290609 ACM	389
f08199d3	390	int perf_evsel__open_per_cpu(struct perf_evsel *evsel,
727ab04e ACM	391	struct cpu_map *cpus, bool group,
727ab04e ACM	392	struct xyarray *group_fd)
48290609	393	{
727ab04e ACM	394	return __perf_evsel__open(evsel, cpus, &empty_thread_map.map, group,
727ab04e ACM	395	group_fd);
0252208e	396	}
48290609	397
f08199d3	398	int perf_evsel__open_per_thread(struct perf_evsel *evsel,
727ab04e ACM	399	struct thread_map *threads, bool group,
727ab04e ACM	400	struct xyarray *group_fd)
0252208e	401	{
727ab04e ACM	402	return __perf_evsel__open(evsel, &empty_cpu_map.map, threads, group,
727ab04e ACM	403	group_fd);
48290609	404	}
70082dd9	405
8115d60c ACM	406	static int perf_event__parse_id_sample(const union perf_event *event, u64 type,
8115d60c ACM	407	struct perf_sample *sample)
d0dd74e8 ACM	408	{
	409	const u64 *array = event->sample.array;
	410
	411	array += ((event->header.size -
	412	sizeof(event->header)) / sizeof(u64)) - 1;
	413
	414	if (type & PERF_SAMPLE_CPU) {
	415	u32 p = (u32 )array;
	416	sample->cpu = *p;
	417	array--;
	418	}
	419
	420	if (type & PERF_SAMPLE_STREAM_ID) {
	421	sample->stream_id = *array;
	422	array--;
	423	}
	424
	425	if (type & PERF_SAMPLE_ID) {
	426	sample->id = *array;
	427	array--;
	428	}
	429
	430	if (type & PERF_SAMPLE_TIME) {
	431	sample->time = *array;
	432	array--;
	433	}
	434
	435	if (type & PERF_SAMPLE_TID) {
	436	u32 p = (u32 )array;
	437	sample->pid = p[0];
	438	sample->tid = p[1];
	439	}
	440
	441	return 0;
	442	}
	443
98e1da90 FW	444	static bool sample_overlap(const union perf_event *event,
	445	const void *offset, u64 size)
	446	{
	447	const void *base = event;
	448
	449	if (offset + size > base + event->header.size)
	450	return true;
	451
	452	return false;
	453	}
	454
8115d60c	455	int perf_event__parse_sample(const union perf_event *event, u64 type,
a2854124	456	int sample_size, bool sample_id_all,
936be503	457	struct perf_sample *data, bool swapped)
d0dd74e8 ACM	458	{
	459	const u64 *array;
	460
936be503 DA	461	/*
	462	* used for cross-endian analysis. See git commit 65014ab3
	463	* for why this goofiness is needed.
	464	*/
	465	union {
	466	u64 val64;
	467	u32 val32[2];
	468	} u;
	469
f3bda2c9	470	memset(data, 0, sizeof(*data));
d0dd74e8 ACM	471	data->cpu = data->pid = data->tid = -1;
d0dd74e8 ACM	472	data->stream_id = data->id = data->time = -1ULL;
a4a03fc7	473	data->period = 1;
d0dd74e8 ACM	474
	475	if (event->header.type != PERF_RECORD_SAMPLE) {
	476	if (!sample_id_all)
	477	return 0;
8115d60c	478	return perf_event__parse_id_sample(event, type, data);
d0dd74e8 ACM	479	}
	480
	481	array = event->sample.array;
	482
a2854124 FW	483	if (sample_size + sizeof(event->header) > event->header.size)
	484	return -EFAULT;
	485
d0dd74e8 ACM	486	if (type & PERF_SAMPLE_IP) {
	487	data->ip = event->ip.ip;
	488	array++;
	489	}
	490
	491	if (type & PERF_SAMPLE_TID) {
936be503 DA	492	u.val64 = *array;
	493	if (swapped) {
	494	/* undo swap of u64, then swap on individual u32s */
	495	u.val64 = bswap_64(u.val64);
	496	u.val32[0] = bswap_32(u.val32[0]);
	497	u.val32[1] = bswap_32(u.val32[1]);
	498	}
	499
	500	data->pid = u.val32[0];
	501	data->tid = u.val32[1];
d0dd74e8 ACM	502	array++;
	503	}
	504
	505	if (type & PERF_SAMPLE_TIME) {
	506	data->time = *array;
	507	array++;
	508	}
	509
7cec0922	510	data->addr = 0;
d0dd74e8 ACM	511	if (type & PERF_SAMPLE_ADDR) {
	512	data->addr = *array;
	513	array++;
	514	}
	515
	516	data->id = -1ULL;
	517	if (type & PERF_SAMPLE_ID) {
	518	data->id = *array;
	519	array++;
	520	}
	521
	522	if (type & PERF_SAMPLE_STREAM_ID) {
	523	data->stream_id = *array;
	524	array++;
	525	}
	526
	527	if (type & PERF_SAMPLE_CPU) {
936be503 DA	528
	529	u.val64 = *array;
	530	if (swapped) {
	531	/* undo swap of u64, then swap on individual u32s */
	532	u.val64 = bswap_64(u.val64);
	533	u.val32[0] = bswap_32(u.val32[0]);
	534	}
	535
	536	data->cpu = u.val32[0];
d0dd74e8 ACM	537	array++;
	538	}
	539
	540	if (type & PERF_SAMPLE_PERIOD) {
	541	data->period = *array;
	542	array++;
	543	}
	544
	545	if (type & PERF_SAMPLE_READ) {
f9d36996	546	fprintf(stderr, "PERF_SAMPLE_READ is unsupported for now\n");
d0dd74e8 ACM	547	return -1;
	548	}
	549
	550	if (type & PERF_SAMPLE_CALLCHAIN) {
98e1da90 FW	551	if (sample_overlap(event, array, sizeof(data->callchain->nr)))
	552	return -EFAULT;
	553
d0dd74e8	554	data->callchain = (struct ip_callchain *)array;
98e1da90 FW	555
	556	if (sample_overlap(event, array, data->callchain->nr))
	557	return -EFAULT;
	558
d0dd74e8 ACM	559	array += 1 + data->callchain->nr;
	560	}
	561
	562	if (type & PERF_SAMPLE_RAW) {
8e303f20 JO	563	const u64 *pdata;
8e303f20 JO	564
936be503 DA	565	u.val64 = *array;
	566	if (WARN_ONCE(swapped,
	567	"Endianness of raw data not corrected!\n")) {
	568	/* undo swap of u64, then swap on individual u32s */
	569	u.val64 = bswap_64(u.val64);
	570	u.val32[0] = bswap_32(u.val32[0]);
	571	u.val32[1] = bswap_32(u.val32[1]);
	572	}
98e1da90 FW	573
	574	if (sample_overlap(event, array, sizeof(u32)))
	575	return -EFAULT;
	576
936be503	577	data->raw_size = u.val32[0];
8e303f20	578	pdata = (void *) array + sizeof(u32);
98e1da90	579
8e303f20	580	if (sample_overlap(event, pdata, data->raw_size))
98e1da90 FW	581	return -EFAULT;
98e1da90 FW	582
8e303f20	583	data->raw_data = (void *) pdata;
fa30c964 SE	584
fa30c964 SE	585	array = (void *)array + data->raw_size + sizeof(u32);
d0dd74e8 ACM	586	}
d0dd74e8 ACM	587
b5387528 RAV	588	if (type & PERF_SAMPLE_BRANCH_STACK) {
	589	u64 sz;
	590
	591	data->branch_stack = (struct branch_stack *)array;
	592	array++; /* nr */
	593
	594	sz = data->branch_stack->nr * sizeof(struct branch_entry);
	595	sz /= sizeof(u64);
	596	array += sz;
	597	}
d0dd74e8 ACM	598	return 0;
d0dd74e8 ACM	599	}
74eec26f AV	600
	601	int perf_event__synthesize_sample(union perf_event *event, u64 type,
	602	const struct perf_sample *sample,
	603	bool swapped)
	604	{
	605	u64 *array;
	606
	607	/*
	608	* used for cross-endian analysis. See git commit 65014ab3
	609	* for why this goofiness is needed.
	610	*/
	611	union {
	612	u64 val64;
	613	u32 val32[2];
	614	} u;
	615
	616	array = event->sample.array;
	617
	618	if (type & PERF_SAMPLE_IP) {
	619	event->ip.ip = sample->ip;
	620	array++;
	621	}
	622
	623	if (type & PERF_SAMPLE_TID) {
	624	u.val32[0] = sample->pid;
	625	u.val32[1] = sample->tid;
	626	if (swapped) {
	627	/*
	628	* Inverse of what is done in perf_event__parse_sample
	629	*/
	630	u.val32[0] = bswap_32(u.val32[0]);
	631	u.val32[1] = bswap_32(u.val32[1]);
	632	u.val64 = bswap_64(u.val64);
	633	}
	634
	635	*array = u.val64;
	636	array++;
	637	}
	638
	639	if (type & PERF_SAMPLE_TIME) {
	640	*array = sample->time;
	641	array++;
	642	}
	643
	644	if (type & PERF_SAMPLE_ADDR) {
	645	*array = sample->addr;
	646	array++;
	647	}
	648
	649	if (type & PERF_SAMPLE_ID) {
	650	*array = sample->id;
	651	array++;
	652	}
	653
	654	if (type & PERF_SAMPLE_STREAM_ID) {
	655	*array = sample->stream_id;
	656	array++;
	657	}
	658
	659	if (type & PERF_SAMPLE_CPU) {
	660	u.val32[0] = sample->cpu;
	661	if (swapped) {
	662	/*
	663	* Inverse of what is done in perf_event__parse_sample
664	*/
665	u.val32[0] = bswap_32(u.val32[0]);
666	u.val64 = bswap_64(u.val64);
667	}
668	*array = u.val64;
669	array++;
670	}
671
672	if (type & PERF_SAMPLE_PERIOD) {
673	*array = sample->period;
674	array++;
675	}
676
677	return 0;
678	}