perf evsel: Introduce per cpu and per thread open helpers

Abstracting away the loops needed to create the various event fd handlers.

The users have to pass a confiruged perf->evsel.attr field, which is already
usable after perf_evsel__new (constructor) time, using defaults.

Comes out of the ad-hoc routines in builtin-stat, that now uses it.

Fixed a small silly bug where we were die()ing before killing our
children, dysfunctional family this one 8-)

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/builtin-stat.c b/tools/perf/builtin-stat.c
index a8b00b4..065e79e 100644 (file)
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -53,8 +53,6 @@
 #include <math.h>
 #include <locale.h>
 
-#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
-
 #define DEFAULT_SEPARATOR      " "
 
 static struct perf_event_attr default_attrs[] = {
@@ -160,56 +158,24 @@ struct stats                      runtime_cycles_stats[MAX_NR_CPUS];
 struct stats                   runtime_branches_stats[MAX_NR_CPUS];
 struct stats                   walltime_nsecs_stats;
 
-#define ERR_PERF_OPEN \
-"counter %d, sys_perf_event_open() syscall returned with %d (%s).  /bin/dmesg may provide additional information."
-
-static int create_perf_stat_counter(struct perf_evsel *evsel, bool *perm_err)
+static int create_perf_stat_counter(struct perf_evsel *evsel)
 {
        struct perf_event_attr *attr = &evsel->attr;
-       int thread;
-       int ncreated = 0;
 
        if (scale)
                attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
                                    PERF_FORMAT_TOTAL_TIME_RUNNING;
 
-       if (system_wide) {
-               int cpu;
-
-               for (cpu = 0; cpu < nr_cpus; cpu++) {
-                       FD(evsel, cpu, 0) = sys_perf_event_open(attr,
-                                       -1, cpumap[cpu], -1, 0);
-                       if (FD(evsel, cpu, 0) < 0) {
-                               if (errno == EPERM || errno == EACCES)
-                                       *perm_err = true;
-                               error(ERR_PERF_OPEN, evsel->idx,
-                                       FD(evsel, cpu, 0), strerror(errno));
-                       } else {
-                               ++ncreated;
-                       }
-               }
-       } else {
-               attr->inherit = !no_inherit;
-               if (target_pid == -1 && target_tid == -1) {
-                       attr->disabled = 1;
-                       attr->enable_on_exec = 1;
-               }
-               for (thread = 0; thread < thread_num; thread++) {
-                       FD(evsel, 0, thread) = sys_perf_event_open(attr,
-                               all_tids[thread], -1, -1, 0);
-                       if (FD(evsel, 0, thread) < 0) {
-                               if (errno == EPERM || errno == EACCES)
-                                       *perm_err = true;
-                               error(ERR_PERF_OPEN, evsel->idx,
-                                       FD(evsel, 0, thread),
-                                        strerror(errno));
-                       } else {
-                               ++ncreated;
-                       }
-               }
+       if (system_wide)
+               return perf_evsel__open_per_cpu(evsel, nr_cpus, cpumap);
+
+       attr->inherit = !no_inherit;
+       if (target_pid == -1 && target_tid == -1) {
+               attr->disabled = 1;
+               attr->enable_on_exec = 1;
        }
 
-       return ncreated;
+       return perf_evsel__open_per_thread(evsel, thread_num, all_tids);
 }
 
 /*
@@ -289,9 +255,7 @@ static int run_perf_stat(int argc __used, const char **argv)
        unsigned long long t0, t1;
        struct perf_evsel *counter;
        int status = 0;
-       int ncreated = 0;
        int child_ready_pipe[2], go_pipe[2];
-       bool perm_err = false;
        const bool forks = (argc > 0);
        char buf;
 
@@ -349,19 +313,23 @@ static int run_perf_stat(int argc __used, const char **argv)
                close(child_ready_pipe[0]);
        }
 
-       list_for_each_entry(counter, &evsel_list, node)
-               ncreated += create_perf_stat_counter(counter, &perm_err);
-
-       if (ncreated < nr_counters) {
-               if (perm_err)
-                       error("You may not have permission to collect %sstats.\n"
-                             "\t Consider tweaking"
-                             " /proc/sys/kernel/perf_event_paranoid or running as root.",
-                             system_wide ? "system-wide " : "");
-               die("Not all events could be opened.\n");
-               if (child_pid != -1)
-                       kill(child_pid, SIGTERM);
-               return -1;
+       list_for_each_entry(counter, &evsel_list, node) {
+               if (create_perf_stat_counter(counter) < 0) {
+                       if (errno == -EPERM || errno == -EACCES) {
+                               error("You may not have permission to collect %sstats.\n"
+                                     "\t Consider tweaking"
+                                     " /proc/sys/kernel/perf_event_paranoid or running as root.",
+                                     system_wide ? "system-wide " : "");
+                       } else {
+                               error("open_counter returned with %d (%s). "
+                                     "/bin/dmesg may provide additional information.\n",
+                                      errno, strerror(errno));
+                       }
+                       if (child_pid != -1)
+                               kill(child_pid, SIGTERM);
+                       die("Not all events could be opened.\n");
+                       return -1;
+               }
        }
 
        /*

diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 3f5de51..e62cc5e 100644 (file)
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -1,4 +1,5 @@
 #include "evsel.h"
+#include "../perf.h"
 #include "util.h"
 
 #define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
@@ -121,3 +122,54 @@ int __perf_evsel__read(struct perf_evsel *evsel,
 
        return 0;
 }
+
+int perf_evsel__open_per_cpu(struct perf_evsel *evsel, int ncpus, int *cpu_map)
+{
+       int cpu;
+
+       for (cpu = 0; cpu < ncpus; cpu++) {
+               FD(evsel, cpu, 0) = sys_perf_event_open(&evsel->attr, -1,
+                                                       cpu_map[cpu], -1, 0);
+               if (FD(evsel, cpu, 0) < 0)
+                       goto out_close;
+       }
+
+       return 0;
+
+out_close:
+       while (--cpu >= 0) {
+               close(FD(evsel, cpu, 0));
+               FD(evsel, cpu, 0) = -1;
+       }
+       return -1;
+}
+
+int perf_evsel__open_per_thread(struct perf_evsel *evsel, int nthreads, int *thread_map)
+{
+       int thread;
+
+       for (thread = 0; thread < nthreads; thread++) {
+               FD(evsel, 0, thread) = sys_perf_event_open(&evsel->attr,
+                                                          thread_map[thread], -1, -1, 0);
+               if (FD(evsel, 0, thread) < 0)
+                       goto out_close;
+       }
+
+       return 0;
+
+out_close:
+       while (--thread >= 0) {
+               close(FD(evsel, 0, thread));
+               FD(evsel, 0, thread) = -1;
+       }
+       return -1;
+}
+
+int perf_evsel__open(struct perf_evsel *evsel, int ncpus, int nthreads,
+                    int *cpu_map, int *thread_map)
+{
+       if (nthreads < 0)
+               return perf_evsel__open_per_cpu(evsel, ncpus, cpu_map);
+
+       return perf_evsel__open_per_thread(evsel, nthreads, thread_map);
+}

diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 8b48ef1..a62fb55 100644 (file)
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -42,6 +42,11 @@ int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus);
 void perf_evsel__free_fd(struct perf_evsel *evsel);
 void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads);
 
+int perf_evsel__open_per_cpu(struct perf_evsel *evsel, int ncpus, int *cpu_map);
+int perf_evsel__open_per_thread(struct perf_evsel *evsel, int nthreads, int *thread_map);
+int perf_evsel__open(struct perf_evsel *evsel, int ncpus, int nthreads,
+                    int *cpu_map, int *thread_map);
+
 #define perf_evsel__match(evsel, t, c)         \
        (evsel->attr.type == PERF_TYPE_##t &&   \
         evsel->attr.config == PERF_COUNT_##c)