Merge branch 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
[pandora-kernel.git] / tools / perf / builtin-top.c
index 7a87171..c9cdedb 100644 (file)
@@ -89,6 +89,7 @@ static bool                   vmlinux_warned;
 static bool                    inherit                         =  false;
 static int                     realtime_prio                   =      0;
 static bool                    group                           =  false;
+static bool                    sample_id_all_avail             =   true;
 static unsigned int            mmap_pages                      =    128;
 
 static bool                    dump_symtab                     =  false;
@@ -199,7 +200,8 @@ static void record_precise_ip(struct hist_entry *he, int counter, u64 ip)
        struct symbol *sym;
 
        if (he == NULL || he->ms.sym == NULL ||
-           (he != top.sym_filter_entry && use_browser != 1))
+           ((top.sym_filter_entry == NULL ||
+             top.sym_filter_entry->ms.sym != he->ms.sym) && use_browser != 1))
                return;
 
        sym = he->ms.sym;
@@ -289,11 +291,13 @@ static void print_sym_table(void)
 
        printf("%-*.*s\n", win_width, win_width, graph_dotted_line);
 
-       if (top.total_lost_warned != top.session->hists.stats.total_lost) {
-               top.total_lost_warned = top.session->hists.stats.total_lost;
-               color_fprintf(stdout, PERF_COLOR_RED, "WARNING:");
-               printf(" LOST %" PRIu64 " events, Check IO/CPU overload\n",
-                      top.total_lost_warned);
+       if (top.sym_evsel->hists.stats.nr_lost_warned !=
+           top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]) {
+               top.sym_evsel->hists.stats.nr_lost_warned =
+                       top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST];
+               color_fprintf(stdout, PERF_COLOR_RED,
+                             "WARNING: LOST %d chunks, Check IO/CPU overload",
+                             top.sym_evsel->hists.stats.nr_lost_warned);
                ++printed;
        }
 
@@ -561,7 +565,6 @@ static void perf_top__sort_new_samples(void *arg)
        hists__decay_entries_threaded(&t->sym_evsel->hists,
                                      top.hide_user_symbols,
                                      top.hide_kernel_symbols);
-       hists__output_recalc_col_len(&t->sym_evsel->hists, winsize.ws_row - 3);
 }
 
 static void *display_thread_tui(void *arg __used)
@@ -671,6 +674,7 @@ static int symbol_filter(struct map *map __used, struct symbol *sym)
 }
 
 static void perf_event__process_sample(const union perf_event *event,
+                                      struct perf_evsel *evsel,
                                       struct perf_sample *sample,
                                       struct perf_session *session)
 {
@@ -770,12 +774,8 @@ static void perf_event__process_sample(const union perf_event *event,
        }
 
        if (al.sym == NULL || !al.sym->ignore) {
-               struct perf_evsel *evsel;
                struct hist_entry *he;
 
-               evsel = perf_evlist__id2evsel(top.evlist, sample->id);
-               assert(evsel != NULL);
-
                if ((sort__has_parent || symbol_conf.use_callchain) &&
                    sample->callchain) {
                        err = perf_session__resolve_callchain(session, al.thread,
@@ -807,6 +807,7 @@ static void perf_event__process_sample(const union perf_event *event,
 static void perf_session__mmap_read_idx(struct perf_session *self, int idx)
 {
        struct perf_sample sample;
+       struct perf_evsel *evsel;
        union perf_event *event;
        int ret;
 
@@ -817,10 +818,16 @@ static void perf_session__mmap_read_idx(struct perf_session *self, int idx)
                        continue;
                }
 
+               evsel = perf_evlist__id2evsel(self->evlist, sample.id);
+               assert(evsel != NULL);
+
                if (event->header.type == PERF_RECORD_SAMPLE)
-                       perf_event__process_sample(event, &sample, self);
-               else
+                       perf_event__process_sample(event, evsel, &sample, self);
+               else if (event->header.type < PERF_RECORD_MAX) {
+                       hists__inc_nr_events(&evsel->hists, event->header.type);
                        perf_event__process(event, &sample, self);
+               } else
+                       ++self->hists.stats.nr_unknown_events;
        }
 }
 
@@ -834,10 +841,16 @@ static void perf_session__mmap_read(struct perf_session *self)
 
 static void start_counters(struct perf_evlist *evlist)
 {
-       struct perf_evsel *counter;
+       struct perf_evsel *counter, *first;
+
+       first = list_entry(evlist->entries.next, struct perf_evsel, node);
 
        list_for_each_entry(counter, &evlist->entries, node) {
                struct perf_event_attr *attr = &counter->attr;
+               struct xyarray *group_fd = NULL;
+
+               if (group && counter != first)
+                       group_fd = first->fd;
 
                attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID;
 
@@ -858,14 +871,23 @@ static void start_counters(struct perf_evlist *evlist)
                attr->mmap = 1;
                attr->comm = 1;
                attr->inherit = inherit;
+retry_sample_id:
+               attr->sample_id_all = sample_id_all_avail ? 1 : 0;
 try_again:
                if (perf_evsel__open(counter, top.evlist->cpus,
-                                    top.evlist->threads, group) < 0) {
+                                    top.evlist->threads, group,
+                                    group_fd) < 0) {
                        int err = errno;
 
                        if (err == EPERM || err == EACCES) {
-                               ui__warning_paranoid();
+                               ui__error_paranoid();
                                goto out_err;
+                       } else if (err == EINVAL && sample_id_all_avail) {
+                               /*
+                                * Old kernel, no attr->sample_id_type_all field
+                                */
+                               sample_id_all_avail = false;
+                               goto retry_sample_id;
                        }
                        /*
                         * If it's cycles then fall back to hrtimer