perf evsel: Introduce mmap support
Out of the code in 'perf top'. Record is next in line.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 2abf949..6d41292 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -60,3 +60,11 @@
evlist->pollfd = malloc(sizeof(struct pollfd) * nfds);
return evlist->pollfd != NULL ? 0 : -ENOMEM;
}
+
+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++;
+}
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index a7d7e12..16bbfcb 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -21,5 +21,6 @@
int perf_evlist__add_default(struct perf_evlist *evlist);
int perf_evlist__alloc_pollfd(struct perf_evlist *evlist, int ncpus, int nthreads);
+void perf_evlist__add_pollfd(struct perf_evlist *evlist, int fd);
#endif /* __PERF_EVLIST_H */
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 82a0053..f500695 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -1,9 +1,13 @@
#include "evsel.h"
+#include "evlist.h"
#include "../perf.h"
#include "util.h"
#include "cpumap.h"
#include "thread.h"
+#include <unistd.h>
+#include <sys/mman.h>
+
#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
@@ -49,10 +53,32 @@
}
}
+void perf_evsel__munmap(struct perf_evsel *evsel, int ncpus, int nthreads)
+{
+ struct perf_mmap *mm;
+ int cpu, thread;
+
+ for (cpu = 0; cpu < ncpus; cpu++)
+ for (thread = 0; thread < nthreads; ++thread) {
+ mm = xyarray__entry(evsel->mmap, cpu, thread);
+ if (mm->base != NULL) {
+ munmap(mm->base, evsel->mmap_len);
+ mm->base = NULL;
+ }
+ }
+}
+
+int perf_evsel__alloc_mmap(struct perf_evsel *evsel, int ncpus, int nthreads)
+{
+ evsel->mmap = xyarray__new(ncpus, nthreads, sizeof(struct perf_mmap));
+ return evsel->mmap != NULL ? 0 : -ENOMEM;
+}
+
void perf_evsel__delete(struct perf_evsel *evsel)
{
assert(list_empty(&evsel->node));
xyarray__delete(evsel->fd);
+ xyarray__delete(evsel->mmap);
free(evsel);
}
@@ -208,3 +234,48 @@
{
return __perf_evsel__open(evsel, &empty_cpu_map.map, threads, group, inherit);
}
+
+int perf_evsel__mmap(struct perf_evsel *evsel, struct cpu_map *cpus,
+ struct thread_map *threads, int pages,
+ struct perf_evlist *evlist)
+{
+ unsigned int page_size = sysconf(_SC_PAGE_SIZE);
+ int mask = pages * page_size - 1, cpu;
+ struct perf_mmap *mm;
+ int thread;
+
+ if (evsel->mmap == NULL &&
+ perf_evsel__alloc_mmap(evsel, cpus->nr, threads->nr) < 0)
+ return -ENOMEM;
+
+ evsel->mmap_len = (pages + 1) * page_size;
+
+ for (cpu = 0; cpu < cpus->nr; cpu++) {
+ for (thread = 0; thread < threads->nr; thread++) {
+ mm = xyarray__entry(evsel->mmap, cpu, thread);
+ mm->prev = 0;
+ mm->mask = mask;
+ mm->base = mmap(NULL, evsel->mmap_len, PROT_READ,
+ MAP_SHARED, FD(evsel, cpu, thread), 0);
+ if (mm->base == MAP_FAILED)
+ goto out_unmap;
+
+ if (evlist != NULL)
+ perf_evlist__add_pollfd(evlist, FD(evsel, cpu, thread));
+ }
+ }
+
+ return 0;
+
+out_unmap:
+ do {
+ while (--thread >= 0) {
+ mm = xyarray__entry(evsel->mmap, cpu, thread);
+ munmap(mm->base, evsel->mmap_len);
+ mm->base = NULL;
+ }
+ thread = threads->nr;
+ } while (--cpu >= 0);
+
+ return -1;
+}
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 1594696..c8fbef2 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -29,19 +29,23 @@
struct perf_event_attr attr;
char *filter;
struct xyarray *fd;
+ struct xyarray *mmap;
struct perf_counts *counts;
+ size_t mmap_len;
int idx;
void *priv;
};
struct cpu_map;
struct thread_map;
+struct perf_evlist;
struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx);
void perf_evsel__delete(struct perf_evsel *evsel);
int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus);
+int perf_evsel__alloc_mmap(struct perf_evsel *evsel, int ncpus, int nthreads);
void perf_evsel__free_fd(struct perf_evsel *evsel);
void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
@@ -51,6 +55,10 @@
struct thread_map *threads, bool group, bool inherit);
int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
struct thread_map *threads, bool group, bool inherit);
+int perf_evsel__mmap(struct perf_evsel *evsel, struct cpu_map *cpus,
+ struct thread_map *threads, int pages,
+ struct perf_evlist *evlist);
+void perf_evsel__munmap(struct perf_evsel *evsel, int ncpus, int nthreads);
#define perf_evsel__match(evsel, t, c) \
(evsel->attr.type == PERF_TYPE_##t && \