[linux-2.6-block.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"

#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;
}

static 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_event_attr *attr = &evsel->attr;
	int track = !evsel->idx; /* only the first counter needs these */

	attr->sample_id_all = opts->sample_id_all_avail ? 1 : 0;
	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 (opts->system_wide)
		attr->sample_type	|= PERF_SAMPLE_CPU;

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

	if (opts->sample_id_all_avail &&
	    (opts->sample_time || opts->system_wide ||
	     !opts->no_inherit || opts->cpu_list))
		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;
	}

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

	if (opts->target_pid == -1 && opts->target_tid == -1 && !opts->system_wide) {
		attr->disabled = 1;
		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;

	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 unsuported 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;
	}

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