2011-01-30 05:46:46 -07:00
|
|
|
/*
|
|
|
|
* 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)
|
|
|
|
*/
|
2011-01-11 17:30:02 -07:00
|
|
|
#include <poll.h>
|
2011-01-30 05:46:46 -07:00
|
|
|
#include "cpumap.h"
|
|
|
|
#include "thread_map.h"
|
2011-01-11 15:56:53 -07:00
|
|
|
#include "evlist.h"
|
|
|
|
#include "evsel.h"
|
|
|
|
#include "util.h"
|
|
|
|
|
2011-01-30 05:46:46 -07:00
|
|
|
#include <sys/mman.h>
|
|
|
|
|
2011-01-12 17:39:13 -07:00
|
|
|
#include <linux/bitops.h>
|
|
|
|
#include <linux/hash.h>
|
|
|
|
|
2011-01-30 05:46:46 -07:00
|
|
|
#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
|
|
|
|
#define SID(e, x, y) xyarray__entry(e->id, x, y)
|
|
|
|
|
2011-01-18 16:41:45 -07:00
|
|
|
void perf_evlist__init(struct perf_evlist *evlist)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < PERF_EVLIST__HLIST_SIZE; ++i)
|
|
|
|
INIT_HLIST_HEAD(&evlist->heads[i]);
|
|
|
|
INIT_LIST_HEAD(&evlist->entries);
|
|
|
|
}
|
|
|
|
|
2011-01-11 15:56:53 -07:00
|
|
|
struct perf_evlist *perf_evlist__new(void)
|
|
|
|
{
|
|
|
|
struct perf_evlist *evlist = zalloc(sizeof(*evlist));
|
|
|
|
|
2011-01-18 16:41:45 -07:00
|
|
|
if (evlist != NULL)
|
|
|
|
perf_evlist__init(evlist);
|
2011-01-11 15:56:53 -07:00
|
|
|
|
|
|
|
return evlist;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void perf_evlist__purge(struct perf_evlist *evlist)
|
|
|
|
{
|
|
|
|
struct perf_evsel *pos, *n;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(pos, n, &evlist->entries, node) {
|
|
|
|
list_del_init(&pos->node);
|
|
|
|
perf_evsel__delete(pos);
|
|
|
|
}
|
|
|
|
|
|
|
|
evlist->nr_entries = 0;
|
|
|
|
}
|
|
|
|
|
2011-01-18 16:41:45 -07:00
|
|
|
void perf_evlist__exit(struct perf_evlist *evlist)
|
2011-01-11 15:56:53 -07:00
|
|
|
{
|
2011-01-12 17:39:13 -07:00
|
|
|
free(evlist->mmap);
|
2011-01-11 17:30:02 -07:00
|
|
|
free(evlist->pollfd);
|
2011-01-18 16:41:45 -07:00
|
|
|
evlist->mmap = NULL;
|
|
|
|
evlist->pollfd = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
void perf_evlist__delete(struct perf_evlist *evlist)
|
|
|
|
{
|
|
|
|
perf_evlist__purge(evlist);
|
|
|
|
perf_evlist__exit(evlist);
|
2011-01-11 15:56:53 -07:00
|
|
|
free(evlist);
|
|
|
|
}
|
|
|
|
|
|
|
|
void perf_evlist__add(struct perf_evlist *evlist, struct perf_evsel *entry)
|
|
|
|
{
|
|
|
|
list_add_tail(&entry->node, &evlist->entries);
|
|
|
|
++evlist->nr_entries;
|
|
|
|
}
|
|
|
|
|
|
|
|
int perf_evlist__add_default(struct perf_evlist *evlist)
|
|
|
|
{
|
|
|
|
struct perf_event_attr attr = {
|
|
|
|
.type = PERF_TYPE_HARDWARE,
|
|
|
|
.config = PERF_COUNT_HW_CPU_CYCLES,
|
|
|
|
};
|
|
|
|
struct perf_evsel *evsel = perf_evsel__new(&attr, 0);
|
|
|
|
|
|
|
|
if (evsel == NULL)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
perf_evlist__add(evlist, evsel);
|
|
|
|
return 0;
|
|
|
|
}
|
2011-01-11 17:30:02 -07:00
|
|
|
|
|
|
|
int perf_evlist__alloc_pollfd(struct perf_evlist *evlist, int ncpus, int nthreads)
|
|
|
|
{
|
|
|
|
int nfds = ncpus * nthreads * evlist->nr_entries;
|
|
|
|
evlist->pollfd = malloc(sizeof(struct pollfd) * nfds);
|
|
|
|
return evlist->pollfd != NULL ? 0 : -ENOMEM;
|
|
|
|
}
|
2011-01-12 12:03:24 -07:00
|
|
|
|
|
|
|
void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd)
|
|
|
|
{
|
|
|
|
fcntl(fd, F_SETFL, O_NONBLOCK);
|
|
|
|
evlist->pollfd[evlist->nr_fds].fd = fd;
|
|
|
|
evlist->pollfd[evlist->nr_fds].events = POLLIN;
|
|
|
|
evlist->nr_fds++;
|
|
|
|
}
|
2011-01-12 17:39:13 -07:00
|
|
|
|
2011-01-30 05:46:46 -07:00
|
|
|
static int perf_evlist__id_hash(struct perf_evlist *evlist, struct perf_evsel *evsel,
|
|
|
|
int cpu, int thread, int fd)
|
|
|
|
{
|
|
|
|
struct perf_sample_id *sid;
|
|
|
|
u64 read_data[4] = { 0, };
|
|
|
|
int hash, id_idx = 1; /* The first entry is the counter value */
|
|
|
|
|
|
|
|
if (!(evsel->attr.read_format & PERF_FORMAT_ID) ||
|
|
|
|
read(fd, &read_data, sizeof(read_data)) == -1)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
if (evsel->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED)
|
|
|
|
++id_idx;
|
|
|
|
if (evsel->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING)
|
|
|
|
++id_idx;
|
|
|
|
|
|
|
|
sid = SID(evsel, cpu, thread);
|
|
|
|
sid->id = read_data[id_idx];
|
|
|
|
sid->evsel = evsel;
|
|
|
|
hash = hash_64(sid->id, PERF_EVLIST__HLIST_BITS);
|
|
|
|
hlist_add_head(&sid->node, &evlist->heads[hash]);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-01-12 17:39:13 -07:00
|
|
|
struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id)
|
|
|
|
{
|
|
|
|
struct hlist_head *head;
|
|
|
|
struct hlist_node *pos;
|
|
|
|
struct perf_sample_id *sid;
|
|
|
|
int hash;
|
|
|
|
|
|
|
|
if (evlist->nr_entries == 1)
|
|
|
|
return list_entry(evlist->entries.next, struct perf_evsel, node);
|
|
|
|
|
|
|
|
hash = hash_64(id, PERF_EVLIST__HLIST_BITS);
|
|
|
|
head = &evlist->heads[hash];
|
|
|
|
|
|
|
|
hlist_for_each_entry(sid, pos, head, node)
|
|
|
|
if (sid->id == id)
|
|
|
|
return sid->evsel;
|
|
|
|
return NULL;
|
|
|
|
}
|
2011-01-15 05:40:59 -07:00
|
|
|
|
2011-01-29 09:01:45 -07:00
|
|
|
union perf_event *perf_evlist__read_on_cpu(struct perf_evlist *evlist, int cpu)
|
2011-01-15 05:40:59 -07:00
|
|
|
{
|
|
|
|
/* XXX Move this to perf.c, making it generally available */
|
|
|
|
unsigned int page_size = sysconf(_SC_PAGE_SIZE);
|
|
|
|
struct perf_mmap *md = &evlist->mmap[cpu];
|
|
|
|
unsigned int head = perf_mmap__read_head(md);
|
|
|
|
unsigned int old = md->prev;
|
|
|
|
unsigned char *data = md->base + page_size;
|
2011-01-29 09:01:45 -07:00
|
|
|
union perf_event *event = NULL;
|
2011-01-15 05:40:59 -07:00
|
|
|
|
2011-01-29 04:08:13 -07:00
|
|
|
if (evlist->overwrite) {
|
2011-01-15 05:40:59 -07:00
|
|
|
/*
|
2011-01-29 04:08:13 -07:00
|
|
|
* If we're further behind than half the buffer, there's a chance
|
|
|
|
* the writer will bite our tail and mess up the samples under us.
|
|
|
|
*
|
|
|
|
* If we somehow ended up ahead of the head, we got messed up.
|
|
|
|
*
|
|
|
|
* In either case, truncate and restart at head.
|
2011-01-15 05:40:59 -07:00
|
|
|
*/
|
2011-01-29 04:08:13 -07:00
|
|
|
int diff = head - old;
|
|
|
|
if (diff > md->mask / 2 || diff < 0) {
|
|
|
|
fprintf(stderr, "WARNING: failed to keep up with mmap data.\n");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* head points to a known good entry, start there.
|
|
|
|
*/
|
|
|
|
old = head;
|
|
|
|
}
|
2011-01-15 05:40:59 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
if (old != head) {
|
|
|
|
size_t size;
|
|
|
|
|
2011-01-29 09:01:45 -07:00
|
|
|
event = (union perf_event *)&data[old & md->mask];
|
2011-01-15 05:40:59 -07:00
|
|
|
size = event->header.size;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Event straddles the mmap boundary -- header should always
|
|
|
|
* be inside due to u64 alignment of output.
|
|
|
|
*/
|
|
|
|
if ((old & md->mask) + size != ((old + size) & md->mask)) {
|
|
|
|
unsigned int offset = old;
|
|
|
|
unsigned int len = min(sizeof(*event), size), cpy;
|
|
|
|
void *dst = &evlist->event_copy;
|
|
|
|
|
|
|
|
do {
|
|
|
|
cpy = min(md->mask + 1 - (offset & md->mask), len);
|
|
|
|
memcpy(dst, &data[offset & md->mask], cpy);
|
|
|
|
offset += cpy;
|
|
|
|
dst += cpy;
|
|
|
|
len -= cpy;
|
|
|
|
} while (len);
|
|
|
|
|
|
|
|
event = &evlist->event_copy;
|
|
|
|
}
|
|
|
|
|
|
|
|
old += size;
|
|
|
|
}
|
|
|
|
|
|
|
|
md->prev = old;
|
2011-01-29 04:08:13 -07:00
|
|
|
|
|
|
|
if (!evlist->overwrite)
|
|
|
|
perf_mmap__write_tail(md, old);
|
|
|
|
|
2011-01-15 05:40:59 -07:00
|
|
|
return event;
|
|
|
|
}
|
2011-01-30 05:46:46 -07:00
|
|
|
|
|
|
|
void perf_evlist__munmap(struct perf_evlist *evlist, int ncpus)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for (cpu = 0; cpu < ncpus; cpu++) {
|
|
|
|
if (evlist->mmap[cpu].base != NULL) {
|
|
|
|
munmap(evlist->mmap[cpu].base, evlist->mmap_len);
|
|
|
|
evlist->mmap[cpu].base = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
int perf_evlist__alloc_mmap(struct perf_evlist *evlist, int ncpus)
|
|
|
|
{
|
|
|
|
evlist->mmap = zalloc(ncpus * sizeof(struct perf_mmap));
|
|
|
|
return evlist->mmap != NULL ? 0 : -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __perf_evlist__mmap(struct perf_evlist *evlist, int cpu, int prot,
|
|
|
|
int mask, int fd)
|
|
|
|
{
|
|
|
|
evlist->mmap[cpu].prev = 0;
|
|
|
|
evlist->mmap[cpu].mask = mask;
|
|
|
|
evlist->mmap[cpu].base = mmap(NULL, evlist->mmap_len, prot,
|
|
|
|
MAP_SHARED, fd, 0);
|
|
|
|
if (evlist->mmap[cpu].base == MAP_FAILED)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
perf_evlist__add_pollfd(evlist, fd);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/** perf_evlist__mmap - Create per cpu maps to receive events
|
|
|
|
*
|
|
|
|
* @evlist - list of events
|
|
|
|
* @cpus - cpu map being monitored
|
|
|
|
* @threads - threads map being monitored
|
|
|
|
* @pages - map length in pages
|
|
|
|
* @overwrite - overwrite older events?
|
|
|
|
*
|
|
|
|
* If overwrite is false the user needs to signal event consuption using:
|
|
|
|
*
|
|
|
|
* struct perf_mmap *m = &evlist->mmap[cpu];
|
|
|
|
* unsigned int head = perf_mmap__read_head(m);
|
|
|
|
*
|
|
|
|
* perf_mmap__write_tail(m, head)
|
|
|
|
*/
|
|
|
|
int perf_evlist__mmap(struct perf_evlist *evlist, struct cpu_map *cpus,
|
|
|
|
struct thread_map *threads, int pages, bool overwrite)
|
|
|
|
{
|
|
|
|
unsigned int page_size = sysconf(_SC_PAGE_SIZE);
|
|
|
|
int mask = pages * page_size - 1, cpu;
|
|
|
|
struct perf_evsel *first_evsel, *evsel;
|
|
|
|
int thread, prot = PROT_READ | (overwrite ? 0 : PROT_WRITE);
|
|
|
|
|
|
|
|
if (evlist->mmap == NULL &&
|
|
|
|
perf_evlist__alloc_mmap(evlist, cpus->nr) < 0)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
if (evlist->pollfd == NULL &&
|
|
|
|
perf_evlist__alloc_pollfd(evlist, cpus->nr, threads->nr) < 0)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
evlist->overwrite = overwrite;
|
|
|
|
evlist->mmap_len = (pages + 1) * page_size;
|
|
|
|
first_evsel = list_entry(evlist->entries.next, struct perf_evsel, node);
|
|
|
|
|
|
|
|
list_for_each_entry(evsel, &evlist->entries, node) {
|
|
|
|
if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
|
|
|
|
evsel->id == NULL &&
|
|
|
|
perf_evsel__alloc_id(evsel, cpus->nr, threads->nr) < 0)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for (cpu = 0; cpu < cpus->nr; cpu++) {
|
|
|
|
for (thread = 0; thread < threads->nr; thread++) {
|
|
|
|
int fd = FD(evsel, cpu, thread);
|
|
|
|
|
|
|
|
if (evsel->idx || thread) {
|
|
|
|
if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT,
|
|
|
|
FD(first_evsel, cpu, 0)) != 0)
|
|
|
|
goto out_unmap;
|
|
|
|
} else if (__perf_evlist__mmap(evlist, cpu, prot, mask, fd) < 0)
|
|
|
|
goto out_unmap;
|
|
|
|
|
|
|
|
if ((evsel->attr.read_format & PERF_FORMAT_ID) &&
|
|
|
|
perf_evlist__id_hash(evlist, evsel, cpu, thread, fd) < 0)
|
|
|
|
goto out_unmap;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_unmap:
|
|
|
|
for (cpu = 0; cpu < cpus->nr; cpu++) {
|
|
|
|
if (evlist->mmap[cpu].base != NULL) {
|
|
|
|
munmap(evlist->mmap[cpu].base, evlist->mmap_len);
|
|
|
|
evlist->mmap[cpu].base = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return -1;
|
|
|
|
}
|