diff options
Diffstat (limited to 'tools')
29 files changed, 677 insertions, 338 deletions
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index e032716..5a520f8 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt @@ -137,6 +137,17 @@ Do not update the builid cache. This saves some overhead in situations where the information in the perf.data file (which includes buildids) is sufficient. +-G name,...:: +--cgroup name,...:: +monitor only in the container (cgroup) called "name". This option is available only +in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to +container "name" are monitored when they run on the monitored CPUs. Multiple cgroups +can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup +to first event, second cgroup to second event and so on. It is possible to provide +an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have +corresponding events, i.e., they always refer to events defined earlier on the command +line. + SEE ALSO -------- linkperf:perf-stat[1], linkperf:perf-list[1] diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt index b6da7af..918cc38 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt @@ -83,6 +83,17 @@ This option is only valid in system-wide mode. print counts using a CSV-style output to make it easy to import directly into spreadsheets. Columns are separated by the string specified in SEP. +-G name:: +--cgroup name:: +monitor only in the container (cgroup) called "name". This option is available only +in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to +container "name" are monitored when they run on the monitored CPUs. Multiple cgroups +can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup +to first event, second cgroup to second event and so on. It is possible to provide +an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have +corresponding events, i.e., they always refer to events defined earlier on the command +line. + EXAMPLES -------- diff --git a/tools/perf/Makefile b/tools/perf/Makefile index be3eb1d..bc4d9bf 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -326,7 +326,7 @@ grep-libs = $(filter -l%,$(1)) strip-libs = $(filter-out -l%,$(1)) $(OUTPUT)python/perf.so: $(PYRF_OBJS) - @python util/setup.py --quiet build_ext --build-lib='$(OUTPUT)python' \ + $(QUIET_GEN)python util/setup.py --quiet build_ext --build-lib='$(OUTPUT)python' \ --build-temp='$(OUTPUT)python/temp' # # No Perl scripts right now: @@ -442,6 +442,7 @@ LIB_H += util/pstack.h LIB_H += util/cpumap.h LIB_H += util/top.h LIB_H += $(ARCH_INCLUDE) +LIB_H += util/cgroup.h LIB_OBJS += $(OUTPUT)util/abspath.o LIB_OBJS += $(OUTPUT)util/alias.o @@ -496,6 +497,7 @@ LIB_OBJS += $(OUTPUT)util/probe-event.o LIB_OBJS += $(OUTPUT)util/util.o LIB_OBJS += $(OUTPUT)util/xyarray.o LIB_OBJS += $(OUTPUT)util/cpumap.o +LIB_OBJS += $(OUTPUT)util/cgroup.o BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c index d9ab3ce..0c7454f 100644 --- a/tools/perf/bench/sched-pipe.c +++ b/tools/perf/bench/sched-pipe.c @@ -55,7 +55,7 @@ int bench_sched_pipe(int argc, const char **argv, * discarding returned value of read(), write() * causes error in building environment for perf */ - int ret, wait_stat; + int __used ret, wait_stat; pid_t pid, retpid; argc = parse_options(argc, argv, options, diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index ea6a116..4271829 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -62,7 +62,8 @@ static int hists__add_entry(struct hists *self, struct addr_location *al) * All aggregated on the first sym_hist. */ struct annotation *notes = symbol__annotation(he->ms.sym); - if (notes->histograms == NULL && symbol__alloc_hist(he->ms.sym, 1) < 0) + if (notes->src == NULL && + symbol__alloc_hist(he->ms.sym, 1) < 0) return -ENOMEM; return hist_entry__inc_addr_samples(he, 0, al->addr); @@ -77,7 +78,8 @@ static int process_sample_event(union perf_event *event, { struct addr_location al; - if (perf_event__preprocess_sample(event, session, &al, sample, NULL) < 0) { + if (perf_event__preprocess_sample(event, session, &al, sample, + symbol__annotate_init) < 0) { pr_warning("problem processing %d event, skipping it.\n", event->header.type); return -1; @@ -111,7 +113,7 @@ static void hists__find_annotations(struct hists *self) goto find_next; notes = symbol__annotation(he->ms.sym); - if (notes->histograms == NULL) { + if (notes->src == NULL) { find_next: if (key == KEY_LEFT) nd = rb_prev(nd); @@ -142,11 +144,11 @@ find_next: nd = rb_next(nd); /* * Since we have a hist_entry per IP for the same - * symbol, free he->ms.sym->histogram to signal we already + * symbol, free he->ms.sym->src to signal we already * processed this symbol. */ - free(notes->histograms); - notes->histograms = NULL; + free(notes->src); + notes->src = NULL; } } } diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 07f8d6d..a4aaadc 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -680,9 +680,9 @@ static int __cmd_record(int argc, const char **argv) perf_event__synthesize_guest_os); if (!system_wide) - perf_event__synthesize_thread(target_tid, - process_synthesized_event, - session); + perf_event__synthesize_thread_map(evsel_list->threads, + process_synthesized_event, + session); else perf_event__synthesize_threads(process_synthesized_event, session); @@ -807,6 +807,9 @@ const struct option record_options[] = { "do not update the buildid cache"), OPT_BOOLEAN('B', "no-buildid", &no_buildid, "do not collect buildids in perf.data"), + OPT_CALLBACK('G', "cgroup", &evsel_list, "name", + "monitor event in cgroup name only", + parse_cgroups), OPT_END() }; @@ -835,6 +838,12 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) write_mode = WRITE_FORCE; } + if (nr_cgroups && !system_wide) { + fprintf(stderr, "cgroup monitoring only available in" + " system-wide mode\n"); + usage_with_options(record_usage, record_options); + } + symbol__init(); if (no_buildid_cache || no_buildid) diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index de06bf5..f9a99a1 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -44,6 +44,7 @@ static const char default_pretty_printing_style[] = "normal"; static const char *pretty_printing_style = default_pretty_printing_style; static char callchain_default_opt[] = "fractal,0.5"; +static symbol_filter_t annotate_init; static struct hists *perf_session__hists_findnew(struct perf_session *self, u64 event_stream, u32 type, @@ -123,7 +124,7 @@ static int perf_session__add_hist_entry(struct perf_session *session, * All aggregated on the first sym_hist. */ struct annotation *notes = symbol__annotation(he->ms.sym); - if (notes->histograms == NULL && + if (notes->src == NULL && symbol__alloc_hist(he->ms.sym, 1) < 0) err = -ENOMEM; else @@ -166,7 +167,8 @@ static int process_sample_event(union perf_event *event, struct addr_location al; struct perf_event_attr *attr; - if (perf_event__preprocess_sample(event, session, &al, sample, NULL) < 0) { + if (perf_event__preprocess_sample(event, session, &al, sample, + annotate_init) < 0) { fprintf(stderr, "problem processing %d event, skipping it.\n", event->header.type); return -1; @@ -519,6 +521,7 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) */ if (use_browser > 0) { symbol_conf.priv_size = sizeof(struct annotation); + annotate_init = symbol__annotate_init; /* * For searching by name on the "Browse map details". * providing it only in verbose mode not to bloat too diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c index ae26211..a32f411 100644 --- a/tools/perf/builtin-sched.c +++ b/tools/perf/builtin-sched.c @@ -369,11 +369,6 @@ static void process_sched_event(struct task_desc *this_task __used, struct sched_atom *atom) { int ret = 0; - u64 now; - long long delta; - - now = get_nsecs(); - delta = start_time + atom->timestamp - now; switch (atom->type) { case SCHED_EVENT_RUN: @@ -562,7 +557,7 @@ static void wait_for_tasks(void) static void run_one_test(void) { - u64 T0, T1, delta, avg_delta, fluct, std_dev; + u64 T0, T1, delta, avg_delta, fluct; T0 = get_nsecs(); wait_for_tasks(); @@ -578,7 +573,6 @@ static void run_one_test(void) else fluct = delta - avg_delta; sum_fluct += fluct; - std_dev = sum_fluct / nr_runs / sqrt(nr_runs); if (!run_avg) run_avg = delta; run_avg = (run_avg*9 + delta)/10; @@ -799,7 +793,7 @@ replay_switch_event(struct trace_switch_event *switch_event, u64 timestamp, struct thread *thread __used) { - struct task_desc *prev, *next; + struct task_desc *prev, __used *next; u64 timestamp0; s64 delta; @@ -1404,7 +1398,7 @@ map_switch_event(struct trace_switch_event *switch_event, u64 timestamp, struct thread *thread __used) { - struct thread *sched_out, *sched_in; + struct thread *sched_out __used, *sched_in; int new_shortname; u64 timestamp0; s64 delta; diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c index 806a999..21c0252 100644 --- a/tools/perf/builtin-stat.c +++ b/tools/perf/builtin-stat.c @@ -390,6 +390,9 @@ static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg) fprintf(stderr, fmt, cpustr, msecs, csv_sep, event_name(evsel)); + if (evsel->cgrp) + fprintf(stderr, "%s%s", csv_sep, evsel->cgrp->name); + if (csv_output) return; @@ -420,6 +423,9 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg) fprintf(stderr, fmt, cpustr, avg, csv_sep, event_name(evsel)); + if (evsel->cgrp) + fprintf(stderr, "%s%s", csv_sep, evsel->cgrp->name); + if (csv_output) return; @@ -460,9 +466,17 @@ static void print_counter_aggr(struct perf_evsel *counter) int scaled = counter->counts->scaled; if (scaled == -1) { - fprintf(stderr, "%*s%s%-24s\n", + fprintf(stderr, "%*s%s%*s", csv_output ? 0 : 18, - "<not counted>", csv_sep, event_name(counter)); + "<not counted>", + csv_sep, + csv_output ? 0 : -24, + event_name(counter)); + + if (counter->cgrp) + fprintf(stderr, "%s%s", csv_sep, counter->cgrp->name); + + fputc('\n', stderr); return; } @@ -487,7 +501,6 @@ static void print_counter_aggr(struct perf_evsel *counter) fprintf(stderr, " (scaled from %.2f%%)", 100 * avg_running / avg_enabled); } - fprintf(stderr, "\n"); } @@ -505,14 +518,18 @@ static void print_counter(struct perf_evsel *counter) ena = counter->counts->cpu[cpu].ena; run = counter->counts->cpu[cpu].run; if (run == 0 || ena == 0) { - fprintf(stderr, "CPU%*d%s%*s%s%-24s", + fprintf(stderr, "CPU%*d%s%*s%s%*s", csv_output ? 0 : -4, evsel_list->cpus->map[cpu], csv_sep, csv_output ? 0 : 18, "<not counted>", csv_sep, + csv_output ? 0 : -24, event_name(counter)); - fprintf(stderr, "\n"); + if (counter->cgrp) + fprintf(stderr, "%s%s", csv_sep, counter->cgrp->name); + + fputc('\n', stderr); continue; } @@ -529,7 +546,7 @@ static void print_counter(struct perf_evsel *counter) 100.0 * run / ena); } } - fprintf(stderr, "\n"); + fputc('\n', stderr); } } @@ -642,6 +659,9 @@ static const struct option options[] = { "disable CPU count aggregation"), OPT_STRING('x', "field-separator", &csv_sep, "separator", "print counts with custom separator"), + OPT_CALLBACK('G', "cgroup", &evsel_list, "name", + "monitor event in cgroup name only", + parse_cgroups), OPT_END() }; @@ -682,9 +702,13 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used) if (run_count <= 0) usage_with_options(stat_usage, options); - /* no_aggr is for system-wide only */ - if (no_aggr && !system_wide) + /* no_aggr, cgroup are for system-wide only */ + if ((no_aggr || nr_cgroups) && !system_wide) { + fprintf(stderr, "both cgroup and no-aggregation " + "modes only available in system-wide mode\n"); + usage_with_options(stat_usage, options); + } /* Set attrs and nr_counters if no event is selected and !null_run */ if (!null_run && !evsel_list->nr_entries) { diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c index 716118a..c9fd66d 100644 --- a/tools/perf/builtin-top.c +++ b/tools/perf/builtin-top.c @@ -139,7 +139,7 @@ static void sig_winch_handler(int sig __used) static int parse_source(struct sym_entry *syme) { struct symbol *sym; - struct sym_entry_source *source; + struct annotation *notes; struct map *map; int err = -1; @@ -152,39 +152,35 @@ static int parse_source(struct sym_entry *syme) /* * We can't annotate with just /proc/kallsyms */ - if (map->dso->origin == DSO__ORIG_KERNEL) + if (map->dso->origin == DSO__ORIG_KERNEL) { + pr_err("Can't annotate %s: No vmlinux file was found in the " + "path\n", sym->name); + sleep(1); return -1; - - if (syme->src == NULL) { - syme->src = zalloc(sizeof(*source)); - if (syme->src == NULL) - return -1; - pthread_mutex_init(&syme->src->lock, NULL); - INIT_LIST_HEAD(&syme->src->head); } - source = syme->src; - - if (symbol__annotation(sym)->histograms != NULL) { - pthread_mutex_lock(&source->lock); + notes = symbol__annotation(sym); + if (notes->src != NULL) { + pthread_mutex_lock(¬es->lock); goto out_assign; } - pthread_mutex_lock(&source->lock); + pthread_mutex_lock(¬es->lock); if (symbol__alloc_hist(sym, top.evlist->nr_entries) < 0) { pr_err("Not enough memory for annotating '%s' symbol!\n", sym->name); + sleep(1); goto out_unlock; } - err = symbol__annotate(sym, syme->map, &source->head, 0); + err = symbol__annotate(sym, syme->map, 0); if (err == 0) { out_assign: sym_filter_entry = syme; } out_unlock: - pthread_mutex_unlock(&source->lock); + pthread_mutex_unlock(¬es->lock); return err; } @@ -196,20 +192,27 @@ static void __zero_source_counters(struct sym_entry *syme) static void record_precise_ip(struct sym_entry *syme, int counter, u64 ip) { + struct annotation *notes; + struct symbol *sym; + if (syme != sym_filter_entry) return; - if (pthread_mutex_trylock(&syme->src->lock)) + sym = sym_entry__symbol(syme); + notes = symbol__annotation(sym); + + if (pthread_mutex_trylock(¬es->lock)) return; ip = syme->map->map_ip(syme->map, ip); - symbol__inc_addr_samples(sym_entry__symbol(syme), syme->map, counter, ip); + symbol__inc_addr_samples(sym, syme->map, counter, ip); - pthread_mutex_unlock(&syme->src->lock); + pthread_mutex_unlock(¬es->lock); } static void show_details(struct sym_entry *syme) { + struct annotation *notes; struct symbol *symbol; int more; @@ -217,24 +220,26 @@ static void show_details(struct sym_entry *syme) return; symbol = sym_entry__symbol(syme); - if (!syme->src || symbol__annotation(symbol)->histograms == NULL) - return; + notes = symbol__annotation(symbol); + + pthread_mutex_lock(¬es->lock); + + if (notes->src == NULL) + goto out_unlock; printf("Showing %s for %s\n", event_name(top.sym_evsel), symbol->name); printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter); - pthread_mutex_lock(&syme->src->lock); - more = symbol__annotate_printf(symbol, syme->map, &syme->src->head, - top.sym_evsel->idx, 0, sym_pcnt_filter, - top.print_entries); + more = symbol__annotate_printf(symbol, syme->map, top.sym_evsel->idx, + 0, sym_pcnt_filter, top.print_entries, 4); if (top.zero) symbol__annotate_zero_histogram(symbol, top.sym_evsel->idx); else - symbol__annotate_decay_histogram(symbol, &syme->src->head, - top.sym_evsel->idx); - pthread_mutex_unlock(&syme->src->lock); + symbol__annotate_decay_histogram(symbol, top.sym_evsel->idx); if (more != 0) printf("%d lines not displayed, maybe increase display entries [e]\n", more); +out_unlock: + pthread_mutex_unlock(¬es->lock); } static const char CONSOLE_CLEAR[] = "[H[2J"; @@ -372,10 +377,8 @@ static void prompt_symbol(struct sym_entry **target, const char *msg) /* zero counters of active symbol */ if (syme) { - pthread_mutex_lock(&syme->src->lock); __zero_source_counters(syme); *target = NULL; - pthread_mutex_unlock(&syme->src->lock); } fprintf(stdout, "\n%s: ", msg); @@ -554,10 +557,8 @@ static void handle_keypress(struct perf_session *session, int c) else { struct sym_entry *syme = sym_filter_entry; - pthread_mutex_lock(&syme->src->lock); sym_filter_entry = NULL; __zero_source_counters(syme); - pthread_mutex_unlock(&syme->src->lock); } break; case 'U': @@ -653,7 +654,7 @@ static int symbol_filter(struct map *map, struct symbol *sym) syme = symbol__priv(sym); syme->map = map; - syme->src = NULL; + symbol__annotate_init(map, sym); if (!sym_filter_entry && sym_filter && !strcmp(name, sym_filter)) { /* schedule initial sym_filter_entry setup */ @@ -865,7 +866,7 @@ static int __cmd_top(void) { pthread_t thread; struct perf_evsel *first; - int ret; + int ret __used; /* * FIXME: perf_session__new should allow passing a O_MMAP, so that all this * mmap reading, etc is encapsulated in it. Use O_WRONLY for now. @@ -875,8 +876,8 @@ static int __cmd_top(void) return -ENOMEM; if (top.target_tid != -1) - perf_event__synthesize_thread(top.target_tid, perf_event__process, - session); + perf_event__synthesize_thread_map(top.evlist->threads, + perf_event__process, session); else perf_event__synthesize_threads(perf_event__process, session); diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 2973376..02976b8 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -14,25 +14,39 @@ #include "symbol.h" #include "debug.h" #include "annotate.h" +#include <pthread.h> -int symbol__alloc_hist(struct symbol *sym, int nevents) +int symbol__annotate_init(struct map *map __used, struct symbol *sym) { struct annotation *notes = symbol__annotation(sym); + pthread_mutex_init(¬es->lock, NULL); + return 0; +} - notes->sizeof_sym_hist = (sizeof(*notes->histograms) + +int symbol__alloc_hist(struct symbol *sym, int nevents) +{ + struct annotation *notes = symbol__annotation(sym); + size_t sizeof_sym_hist = (sizeof(struct sym_hist) + (sym->end - sym->start) * sizeof(u64)); - notes->histograms = calloc(nevents, notes->sizeof_sym_hist); - notes->nr_histograms = nevents; - return notes->histograms == NULL ? -1 : 0; + + notes->src = zalloc(sizeof(*notes->src) + nevents * sizeof_sym_hist); + if (notes->src == NULL) + return -1; + notes->src->sizeof_sym_hist = sizeof_sym_hist; + notes->src->nr_histograms = nevents; + INIT_LIST_HEAD(¬es->src->source); + return 0; } void symbol__annotate_zero_histograms(struct symbol *sym) { struct annotation *notes = symbol__annotation(sym); - if (notes->histograms != NULL) - memset(notes->histograms, 0, - notes->nr_histograms * notes->sizeof_sym_hist); + pthread_mutex_lock(¬es->lock); + if (notes->src != NULL) + memset(notes->src->histograms, 0, + notes->src->nr_histograms * notes->src->sizeof_sym_hist); + pthread_mutex_unlock(¬es->lock); } int symbol__inc_addr_samples(struct symbol *sym, struct map *map, @@ -43,7 +57,7 @@ int symbol__inc_addr_samples(struct symbol *sym, struct map *map, struct sym_hist *h; notes = symbol__annotation(sym); - if (notes->histograms == NULL) + if (notes->src == NULL) return -ENOMEM; pr_debug3("%s: addr=%#" PRIx64 "\n", __func__, map->unmap_ip(map, addr)); @@ -95,10 +109,10 @@ struct objdump_line *objdump__get_next_ip_line(struct list_head *head, return NULL; } -static int objdump_line__print(struct objdump_line *oline, - struct list_head *head, struct symbol *sym, +static int objdump_line__print(struct objdump_line *oline, struct symbol *sym, int evidx, u64 len, int min_pcnt, - int printed, int max_lines) + int printed, int max_lines, + struct objdump_line *queue) { static const char *prev_line; static const char *prev_color; @@ -109,10 +123,12 @@ static int objdump_line__print(struct objdump_line *oline, double percent = 0.0; const char *color; struct annotation *notes = symbol__annotation(sym); - struct source_line *src_line = notes->src_line; + struct source_line *src_line = notes->src->lines; struct sym_hist *h = annotation__histogram(notes, evidx); s64 offset = oline->offset; - struct objdump_line *next = objdump__get_next_ip_line(head, oline); + struct objdump_line *next; + + next = objdump__get_next_ip_line(¬es->src->source, oline); while (offset < (s64)len && (next == NULL || offset < next->offset)) { @@ -132,9 +148,18 @@ static int objdump_line__print(struct objdump_line *oline, if (percent < min_pcnt) return -1; - if (printed >= max_lines) + if (max_lines && printed >= max_lines) return 1; + if (queue != NULL) { + list_for_each_entry_from(queue, ¬es->src->source, node) { + if (queue == oline) + break; + objdump_line__print(queue, sym, evidx, len, + 0, 0, 1, NULL); + } + } + color = get_percent_color(percent); /* @@ -154,9 +179,12 @@ static int objdump_line__print(struct objdump_line *oline, color_fprintf(stdout, color, " %7.2f", percent); printf(" : "); color_fprintf(stdout, PERF_COLOR_BLUE, "%s\n", oline->line); - } else if (printed >= max_lines) + } else if (max_lines && printed >= max_lines) return 1; else { + if (queue) + return -1; + if (!*oline->line) printf(" :\n"); else @@ -166,9 +194,10 @@ static int objdump_line__print(struct objdump_line *oline, return 0; } -static int symbol__parse_objdump_line(struct symbol *sym, struct map *map, FILE *file, - struct list_head *head, size_t privsize) +static int symbol__parse_objdump_line(struct symbol *sym, struct map *map, + FILE *file, size_t privsize) { + struct annotation *notes = symbol__annotation(sym); struct objdump_line *objdump_line; char *line = NULL, *tmp, *tmp2, *c; size_t line_len; @@ -222,13 +251,12 @@ static int symbol__parse_objdump_line(struct symbol *sym, struct map *map, FILE free(line); return -1; } - objdump__add_line(head, objdump_line); + objdump__add_line(¬es->src->source, objdump_line); return 0; } -int symbol__annotate(struct symbol *sym, struct map *map, - struct list_head *head, size_t privsize) +int symbol__annotate(struct symbol *sym, struct map *map, size_t privsize) { struct dso *dso = map->dso; char *filename = dso__build_id_filename(dso, NULL, 0); @@ -236,7 +264,6 @@ int symbol__annotate(struct symbol *sym, struct map *map, char command[PATH_MAX * 2]; FILE *file; int err = 0; - u64 len; char symfs_filename[PATH_MAX]; if (filename) { @@ -281,8 +308,6 @@ fallback: filename, sym->name, map->unmap_ip(map, sym->start), map->unmap_ip(map, sym->end)); - len = sym->end - sym->start; - pr_debug("annotating [%p] %30s : [%p] %30s\n", dso, dso->long_name, sym, sym->name); @@ -300,7 +325,7 @@ fallback: goto out_free_filename; while (!feof(file)) - if (symbol__parse_objdump_line(sym, map, file, head, privsize) < 0) + if (symbol__parse_objdump_line(sym, map, file, privsize) < 0) break; pclose(file); @@ -333,14 +358,14 @@ static void insert_source_line(struct rb_root *root, struct source_line *src_lin static void symbol__free_source_line(struct symbol *sym, int len) { struct annotation *notes = symbol__annotation(sym); - struct source_line *src_line = notes->src_line; + struct source_line *src_line = notes->src->lines; int i; for (i = 0; i < len; i++) free(src_line[i].path); free(src_line); - notes->src_line = NULL; + notes->src->lines = NULL; } /* Get the filename:line for the colored entries */ @@ -358,8 +383,8 @@ static int symbol__get_source_line(struct symbol *sym, struct map *map, if (!h->sum) return 0; - src_line = notes->src_line = calloc(len, sizeof(struct source_line)); - if (!notes->src_line) + src_line = notes->src->lines = calloc(len, sizeof(struct source_line)); + if (!notes->src->lines) return -1; start = map->unmap_ip(map, sym->start); @@ -439,14 +464,15 @@ static void symbol__annotate_hits(struct symbol *sym, int evidx) printf("%*s: %" PRIu64 "\n", BITS_PER_LONG / 2, "h->sum", h->sum); } -int symbol__annotate_printf(struct symbol *sym, struct map *map, - struct list_head *head, int evidx, bool full_paths, - int min_pcnt, int max_lines) +int symbol__annotate_printf(struct symbol *sym, struct map *map, int evidx, + bool full_paths, int min_pcnt, int max_lines, + int context) { struct dso *dso = map->dso; const char *filename = dso->long_name, *d_filename; - struct objdump_line *pos; - int printed = 2; + struct annotation *notes = symbol__annotation(sym); + struct objdump_line *pos, *queue = NULL; + int printed = 2, queue_len = 0; int more = 0; u64 len; @@ -463,11 +489,21 @@ int symbol__annotate_printf(struct symbol *sym, struct map *map, if (verbose) symbol__annotate_hits(sym, evidx); - list_for_each_entry(pos, head, node) { - switch (objdump_line__print(pos, head, sym, evidx, len, min_pcnt, - printed, max_lines)) { + list_for_each_entry(pos, ¬es->src->source, node) { + if (context && queue == NULL) { + queue = pos; + queue_len = 0; + } + + switch (objdump_line__print(pos, sym, evidx, len, min_pcnt, + printed, max_lines, queue)) { case 0: ++printed; + if (context) { + printed += queue_len; + queue = NULL; + queue_len = 0; + } break; case 1: /* filtered by max_lines */ @@ -475,7 +511,16 @@ int symbol__annotate_printf(struct symbol *sym, struct map *map, break; case -1: default: - /* filtered by min_pcnt */ + /* + * Filtered by min_pcnt or non IP lines when + * context != 0 + */ + if (!context) + break; + if (queue_len == context) + queue = list_entry(queue->node.next, typeof(*queue), node); + else + ++queue_len; break; } } @@ -488,11 +533,10 @@ void symbol__annotate_zero_histogram(struct symbol *sym, int evidx) struct annotation *notes = symbol__annotation(sym); struct sym_hist *h = annotation__histogram(notes, evidx); - memset(h, 0, notes->sizeof_sym_hist); + memset(h, 0, notes->src->sizeof_sym_hist); } -void symbol__annotate_decay_histogram(struct symbol *sym, - struct list_head *head, int evidx) +void symbol__annotate_decay_histogram(struct symbol *sym, int evidx) { struct annotation *notes = symbol__annotation(sym); struct sym_hist *h = annotation__histogram(notes, evidx); @@ -500,7 +544,7 @@ void symbol__annotate_decay_histogram(struct symbol *sym, h->sum = 0; - list_for_each_entry(pos, head, node) { + list_for_each_entry(pos, ¬es->src->source, node) { if (pos->offset != -1) { h->addr[pos->offset] = h->addr[pos->offset] * 7 / 8; h->sum += h->addr[pos->offset]; @@ -525,10 +569,9 @@ int symbol__tty_annotate(struct symbol *sym, struct map *map, int evidx, struct dso *dso = map->dso; const char *filename = dso->long_name; struct rb_root source_line = RB_ROOT; - LIST_HEAD(head); u64 len; - if (symbol__annotate(sym, map, &head, 0) < 0) + if (symbol__annotate(sym, map, 0) < 0) return -1; len = sym->end - sym->start; @@ -539,12 +582,12 @@ int symbol__tty_annotate(struct symbol *sym, struct map *map, int evidx, print_summary(&source_line, filename); } - symbol__annotate_printf(sym, map, &head, evidx, full_paths, - min_pcnt, max_lines); + symbol__annotate_printf(sym, map, evidx, full_paths, + min_pcnt, max_lines, 0); if (print_lines) symbol__free_source_line(sym, len); - objdump_line_list__purge(&head); + objdump_line_list__purge(&symbol__annotation(sym)->src->source); return 0; } diff --git a/tools/perf/util/annotate.h b/tools/perf/util/annotate.h index b1253aa..e848803 100644 --- a/tools/perf/util/annotate.h +++ b/tools/perf/util/annotate.h @@ -28,22 +28,29 @@ struct source_line { char *path; }; -/** struct annotation - symbols with hits have this attached as in sannotation +/** struct annotated_source - symbols with hits have this attached as in sannotation * * @histogram: Array of addr hit histograms per event being monitored - * @src_line: If 'print_lines' is specified, per source code line percentages + * @lines: If 'print_lines' is specified, per source code line percentages + * @source: source parsed from objdump -dS * - * src_line is allocated, percentages calculated and all sorted by percentage + * lines is allocated, percentages calculated and all sorted by percentage * when the annotation is about to be presented, so the percentages are for * one of the entries in the histogram array, i.e. for the event/counter being * presented. It is deallocated right after symbol__{tui,tty,etc}_annotate * returns. */ -struct annotation { - struct source_line *src_line; - struct sym_hist *histograms; +struct annotated_source { + struct list_head source; + struct source_line *lines; int nr_histograms; int sizeof_sym_hist; + struct sym_hist histograms[0]; +}; + +struct annotation { + pthread_mutex_t lock; + struct annotated_source *src; }; struct sannotation { @@ -53,7 +60,8 @@ struct sannotation { static inline struct sym_hist *annotation__histogram(struct annotation *notes, int idx) { - return ((void *)notes->histograms) + (notes->sizeof_sym_hist * idx); + return (((void *)¬es->src->histograms) + + (notes->src->sizeof_sym_hist * idx)); } static inline struct annotation *symbol__annotation(struct symbol *sym) @@ -67,14 +75,13 @@ int symbol__inc_addr_samples(struct symbol *sym, struct map *map, int symbol__alloc_hist(struct symbol *sym, int nevents); void symbol__annotate_zero_histograms(struct symbol *sym); -int symbol__annotate(struct symbol *sym, struct map *map, - struct list_head *head, size_t privsize); -int symbol__annotate_printf(struct symbol *sym, struct map *map, - struct list_head *head, int evidx, bool full_paths, - int min_pcnt, int max_lines); +int symbol__annotate(struct symbol *sym, struct map *map, size_t privsize); +int symbol__annotate_init(struct map *map __used, struct symbol *sym); +int symbol__annotate_printf(struct symbol *sym, struct map *map, int evidx, + bool full_paths, int min_pcnt, int max_lines, + int context); void symbol__annotate_zero_histogram(struct symbol *sym, int evidx); -void symbol__annotate_decay_histogram(struct symbol *sym, - struct list_head *head, int evidx); +void symbol__annotate_decay_histogram(struct symbol *sym, int evidx); void objdump_line_list__purge(struct list_head *head); int symbol__tty_annotate(struct symbol *sym, struct map *map, int evidx, @@ -82,7 +89,7 @@ int symbol__tty_annotate(struct symbol *sym, struct map *map, int evidx, int max_lines); #ifdef NO_NEWT_SUPPORT -static inline int symbol__tui_annotate(symbol *sym __used, +static inline int symbol__tui_annotate(struct symbol *sym __used, struct map *map __used, int evidx __used) { return 0; diff --git a/tools/perf/util/cgroup.c b/tools/perf/util/cgroup.c new file mode 100644 index 0000000..9fea755 --- /dev/null +++ b/tools/perf/util/cgroup.c @@ -0,0 +1,178 @@ +#include "util.h" +#include "../perf.h" +#include "parse-options.h" +#include "evsel.h" +#include "cgroup.h" +#include "debugfs.h" /* MAX_PATH, STR() */ +#include "evlist.h" + +int nr_cgroups; + +static int +cgroupfs_find_mountpoint(char *buf, size_t maxlen) +{ + FILE *fp; + char mountpoint[MAX_PATH+1], tokens[MAX_PATH+1], type[MAX_PATH+1]; + char *token, *saved_ptr; + int found = 0; + + fp = fopen("/proc/mounts", "r"); + if (!fp) + return -1; + + /* + * in order to handle split hierarchy, we need to scan /proc/mounts + * and inspect every cgroupfs mount point to find one that has + * perf_event subsystem + */ + while (fscanf(fp, "%*s %"STR(MAX_PATH)"s %"STR(MAX_PATH)"s %" + STR(MAX_PATH)"s %*d %*d\n", + mountpoint, type, tokens) == 3) { + + if (!strcmp(type, "cgroup")) { + + token = strtok_r(tokens, ",", &saved_ptr); + + while (token != NULL) { + if (!strcmp(token, "perf_event")) { + found = 1; + break; + } + token = strtok_r(NULL, ",", &saved_ptr); + } + } + if (found) + break; + } + fclose(fp); + if (!found) + return -1; + + if (strlen(mountpoint) < maxlen) { + strcpy(buf, mountpoint); + return 0; + } + return -1; +} + +static int open_cgroup(char *name) +{ + char path[MAX_PATH+1]; + char mnt[MAX_PATH+1]; + int fd; + + + if (cgroupfs_find_mountpoint(mnt, MAX_PATH+1)) + return -1; + + snprintf(path, MAX_PATH, "%s/%s", mnt, name); + + fd = open(path, O_RDONLY); + if (fd == -1) + fprintf(stderr, "no access to cgroup %s\n", path); + + return fd; +} + +static int add_cgroup(struct perf_evlist *evlist, char *str) +{ + struct perf_evsel *counter; + struct cgroup_sel *cgrp = NULL; + int n; + /* + * check if cgrp is already defined, if so we reuse it + */ + list_for_each_entry(counter, &evlist->entries, node) { + cgrp = counter->cgrp; + if (!cgrp) + continue; + if (!strcmp(cgrp->name, str)) + break; + + cgrp = NULL; + } + + if (!cgrp) { + cgrp = zalloc(sizeof(*cgrp)); + if (!cgrp) + return -1; + + cgrp->name = str; + + cgrp->fd = open_cgroup(str); + if (cgrp->fd == -1) { + free(cgrp); + return -1; + } + } + + /* + * find corresponding event + * if add cgroup N, then need to find event N + */ + n = 0; + list_for_each_entry(counter, &evlist->entries, node) { + if (n == nr_cgroups) + goto found; + n++; + } + if (cgrp->refcnt == 0) + free(cgrp); + + return -1; +found: + cgrp->refcnt++; + counter->cgrp = cgrp; + return 0; +} + +void close_cgroup(struct cgroup_sel *cgrp) +{ + if (!cgrp) + return; + + /* XXX: not reentrant */ + if (--cgrp->refcnt == 0) { + close(cgrp->fd); + free(cgrp->name); + free(cgrp); + } +} + +int parse_cgroups(const struct option *opt __used, const char *str, + int unset __used) +{ + struct perf_evlist *evlist = *(struct perf_evlist **)opt->value; + const char *p, *e, *eos = str + strlen(str); + char *s; + int ret; + + if (list_empty(&evlist->entries)) { + fprintf(stderr, "must define events before cgroups\n"); + return -1; + } + + for (;;) { + p = strchr(str, ','); + e = p ? p : eos; + + /* allow empty cgroups, i.e., skip */ + if (e - str) { + /* termination added */ + s = strndup(str, e - str); + if (!s) + return -1; + ret = add_cgroup(evlist, s); + if (ret) { + free(s); + return -1; + } + } + /* nr_cgroups is increased een for empty cgroups */ + nr_cgroups++; + if (!p) + break; + str = p+1; + } + return 0; +} diff --git a/tools/perf/util/cgroup.h b/tools/perf/util/cgroup.h new file mode 100644 index 0000000..89acd6d --- /dev/null +++ b/tools/perf/util/cgroup.h @@ -0,0 +1,17 @@ +#ifndef __CGROUP_H__ +#define __CGROUP_H__ + +struct option; + +struct cgroup_sel { + char *name; + int fd; + int refcnt; +}; + + +extern int nr_cgroups; /* number of explicit cgroups defined */ +extern void close_cgroup(struct cgroup_sel *cgrp); +extern int parse_cgroups(const struct option *opt, const char *str, int unset); + +#endif /* __CGROUP_H__ */ diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 731265f..fbf5754 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -6,6 +6,7 @@ #include "string.h" #include "strlist.h" #include "thread.h" +#include "thread_map.h" static const char *perf_event__names[] = { [0] = "TOTAL", @@ -265,11 +266,12 @@ static int __event__synthesize_thread(union perf_event *comm_event, process, session); } -int perf_event__synthesize_thread(pid_t pid, perf_event__handler_t process, - struct perf_session *session) +int perf_event__synthesize_thread_map(struct thread_map *threads, + perf_event__handler_t process, + struct perf_session *session) { union perf_event *comm_event, *mmap_event; - int err = -1; + int err = -1, thread; comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size); if (comm_event == NULL) @@ -279,8 +281,15 @@ int perf_event__synthesize_thread(pid_t pid, perf_event__handler_t process, if (mmap_event == NULL) goto out_free_comm; - err = __event__synthesize_thread(comm_event, mmap_event, pid, - process, session); + err = 0; + for (thread = 0; thread < threads->nr; ++thread) { + if (__event__synthesize_thread(comm_event, mmap_event, + threads->map[thread], + process, session)) { + err = -1; + break; + } + } free(mmap_event); out_free_comm: free(comm_event); diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h index eecb422..9c35170 100644 --- a/tools/perf/util/event.h +++ b/tools/perf/util/event.h @@ -135,6 +135,7 @@ union perf_event { void perf_event__print_totals(void); struct perf_session; +struct thread_map; typedef int (*perf_event__handler_synth_t)(union perf_event *event, struct perf_session *session); @@ -142,8 +143,9 @@ typedef int (*perf_event__handler_t)(union perf_event *event, struct perf_sample *sample, struct perf_session *session); -int perf_event__synthesize_thread(pid_t pid, perf_event__handler_t process, - struct perf_session *session); +int perf_event__synthesize_thread_map(struct thread_map *threads, + perf_event__handler_t process, + struct perf_session *session); int perf_event__synthesize_threads(perf_event__handler_t process, struct perf_session *session); int perf_event__synthesize_kernel_mmap(perf_event__handler_t process, diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 211063e..c974e08 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -85,6 +85,7 @@ void perf_evsel__exit(struct perf_evsel *evsel) void perf_evsel__delete(struct perf_evsel *evsel) { perf_evsel__exit(evsel); + close_cgroup(evsel->cgrp); free(evsel); } @@ -163,21 +164,32 @@ static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus, struct thread_map *threads, bool group, bool inherit) { int cpu, thread; + unsigned long flags = 0; + int pid = -1; if (evsel->fd == NULL && perf_evsel__alloc_fd(evsel, cpus->nr, threads->nr) < 0) return -1; + if (evsel->cgrp) { + flags = PERF_FLAG_PID_CGROUP; + pid = evsel->cgrp->fd; + } + for (cpu = 0; cpu < cpus->nr; cpu++) { int group_fd = -1; evsel->attr.inherit = (cpus->map[cpu] < 0) && inherit; for (thread = 0; thread < threads->nr; thread++) { + + if (!evsel->cgrp) + pid = threads->map[thread]; + FD(evsel, cpu, thread) = sys_perf_event_open(&evsel->attr, - threads->map[thread], + pid, cpus->map[cpu], - group_fd, 0); + group_fd, flags); if (FD(evsel, cpu, thread) < 0) goto out_close; diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index eecdc3a..1d3d5a3 100644 --- a/tools/perf/util/evsel.h +++ b/tools/perf/util/evsel.h @@ -6,6 +6,7 @@ #include "../../../include/linux/perf_event.h" #include "types.h" #include "xyarray.h" +#include "cgroup.h" struct perf_counts_values { union { @@ -45,6 +46,7 @@ struct perf_evsel { struct perf_counts *counts; int idx; void *priv; + struct cgroup_sel *cgrp; }; struct cpu_map; diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c index c0de5ec..72c124d 100644 --- a/tools/perf/util/header.c +++ b/tools/perf/util/header.c @@ -1145,7 +1145,7 @@ int perf_event__synthesize_tracing_data(int fd, struct perf_evlist *evlist, { union perf_event ev; ssize_t size = 0, aligned_size = 0, padding; - int err = 0; + int err __used = 0; memset(&ev, 0, sizeof(ev)); diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c index bac5ab6..3f43723 100644 --- a/tools/perf/util/hist.c +++ b/tools/perf/util/hist.c @@ -955,10 +955,9 @@ int hist_entry__inc_addr_samples(struct hist_entry *he, int evidx, u64 ip) return symbol__inc_addr_samples(he->ms.sym, he->ms.map, evidx, ip); } -int hist_entry__annotate(struct hist_entry *he, struct list_head *head, - size_t privsize) +int hist_entry__annotate(struct hist_entry *he, size_t privsize) { - return symbol__annotate(he->ms.sym, he->ms.map, head, privsize); + return symbol__annotate(he->ms.sym, he->ms.map, privsize); } void hists__inc_nr_events(struct hists *self, u32 type) diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h index 2c6cdae..37c7908 100644 --- a/tools/perf/util/hist.h +++ b/tools/perf/util/hist.h @@ -78,8 +78,7 @@ size_t hists__fprintf(struct hists *self, struct hists *pair, bool show_displacement, FILE *fp); int hist_entry__inc_addr_samples(struct hist_entry *self, int evidx, u64 addr); -int hist_entry__annotate(struct hist_entry *self, struct list_head *head, - size_t privsize); +int hist_entry__annotate(struct hist_entry *self, size_t privsize); void hists__filter_by_dso(struct hists *self, const struct dso *dso); void hists__filter_by_thread(struct hists *self, const struct thread *thread); diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c index 69215bf..fe461f6 100644 --- a/tools/perf/util/probe-finder.c +++ b/tools/perf/util/probe-finder.c @@ -33,6 +33,7 @@ #include <ctype.h> #include <dwarf-regs.h> +#include <linux/bitops.h> #include "event.h" #include "debug.h" #include "util.h" @@ -333,13 +334,23 @@ static Dwarf_Die *die_get_real_type(Dwarf_Die *vr_die, Dwarf_Die *die_mem) return vr_die; } -static bool die_is_signed_type(Dwarf_Die *tp_die) +static int die_get_attr_udata(Dwarf_Die *tp_die, unsigned int attr_name, + Dwarf_Word *result) { Dwarf_Attribute attr; + + if (dwarf_attr(tp_die, attr_name, &attr) == NULL || + dwarf_formudata(&attr, result) != 0) + return -ENOENT; + + return 0; +} + +static bool die_is_signed_type(Dwarf_Die *tp_die) +{ Dwarf_Word ret; - if (dwarf_attr(tp_die, DW_AT_encoding, &attr) == NULL || - dwarf_formudata(&attr, &ret) != 0) + if (die_get_attr_udata(tp_die, DW_AT_encoding, &ret)) return false; return (ret == DW_ATE_signed_char || ret == DW_ATE_signed || @@ -348,11 +359,29 @@ static bool die_is_signed_type(Dwarf_Die *tp_die) static int die_get_byte_size(Dwarf_Die *tp_die) { - Dwarf_Attribute attr; Dwarf_Word ret; - if (dwarf_attr(tp_die, DW_AT_byte_size, &attr) == NULL || - dwarf_formudata(&attr, &ret) != 0) + if (die_get_attr_udata(tp_die, DW_AT_byte_size, &ret)) + return 0; + + return (int)ret; +} + +static int die_get_bit_size(Dwarf_Die *tp_die) +{ + Dwarf_Word ret; + + if (die_get_attr_udata(tp_die, DW_AT_bit_size, &ret)) + return 0; + + return (int)ret; +} + +static int die_get_bit_offset(Dwarf_Die *tp_die) +{ + Dwarf_Word ret; + + if (die_get_attr_udata(tp_die, DW_AT_bit_offset, &ret)) return 0; return (int)ret; @@ -827,6 +856,8 @@ static_var: return 0; } +#define BYTES_TO_BITS(nb) ((nb) * BITS_PER_LONG / sizeof(long)) + static int convert_variable_type(Dwarf_Die *vr_die, struct probe_trace_arg *tvar, const char *cast) @@ -843,6 +874,14 @@ static int convert_variable_type(Dwarf_Die *vr_die, return (tvar->type == NULL) ? -ENOMEM : 0; } + if (die_get_bit_size(vr_die) != 0) { + /* This is a bitfield */ + ret = snprintf(buf, 16, "b%d@%d/%zd", die_get_bit_size(vr_die), + die_get_bit_offset(vr_die), + BYTES_TO_BITS(die_get_byte_size(vr_die))); + goto formatted; + } + if (die_get_real_type(vr_die, &type) == NULL) { pr_warning("Failed to get a type information of %s.\n", dwarf_diename(vr_die)); @@ -887,29 +926,31 @@ static int convert_variable_type(Dwarf_Die *vr_die, return (tvar->type == NULL) ? -ENOMEM : 0; } - ret = die_get_byte_size(&type) * 8; - if (ret) { - /* Check the bitwidth */ - if (ret > MAX_BASIC_TYPE_BITS) { - pr_info("%s exceeds max-bitwidth." - " Cut down to %d bits.\n", - dwarf_diename(&type), MAX_BASIC_TYPE_BITS); - ret = MAX_BASIC_TYPE_BITS; - } + ret = BYTES_TO_BITS(die_get_byte_size(&type)); + if (!ret) + /* No size ... try to use default type */ + return 0; - ret = snprintf(buf, 16, "%c%d", - die_is_signed_type(&type) ? 's' : 'u', ret); - if (ret < 0 || ret >= 16) { - if (ret >= 16) - ret = -E2BIG; - pr_warning("Failed to convert variable type: %s\n", - strerror(-ret)); - return ret; - } - tvar->type = strdup(buf); - if (tvar->type == NULL) - return -ENOMEM; + /* Check the bitwidth */ + if (ret > MAX_BASIC_TYPE_BITS) { + pr_info("%s exceeds max-bitwidth. Cut down to %d bits.\n", + dwarf_diename(&type), MAX_BASIC_TYPE_BITS); + ret = MAX_BASIC_TYPE_BITS; + } + ret = snprintf(buf, 16, "%c%d", + die_is_signed_type(&type) ? 's' : 'u', ret); + +formatted: + if (ret < 0 || ret >= 16) { + if (ret >= 16) + ret = -E2BIG; + pr_warning("Failed to convert variable type: %s\n", + strerror(-ret)); + return ret; } + tvar->type = strdup(buf); + if (tvar->type == NULL) + return -ENOMEM; return 0; } @@ -1234,51 +1275,38 @@ static int find_probe_point_by_line(struct probe_finder *pf) static int find_lazy_match_lines(struct list_head *head, const char *fname, const char *pat) { - char *fbuf, *p1, *p2; - int fd, line, nlines = -1; - struct stat st; + FILE *fp; + char *line = NULL; + size_t line_len; + ssize_t len; + int count = 0, linenum = 1; - fd = open(fname, O_RDONLY); - if (fd < 0) { - pr_warning("Failed to open %s: %s\n", fname, strerror(-fd)); + fp = fopen(fname, "r"); + if (!fp) { + pr_warning("Failed to open %s: %s\n", fname, strerror(errno)); return -errno; } - if (fstat(fd, &st) < 0) { - pr_warning("Failed to get the size of %s: %s\n", - fname, strerror(errno)); - nlines = -errno; - goto out_close; - } - - nlines = -ENOMEM; - fbuf = malloc(st.st_size + 2); - if (fbuf == NULL) - goto out_close; - if (read(fd, fbuf, st.st_size) < 0) { - pr_warning("Failed to read %s: %s\n", fname, strerror(errno)); - nlines = -errno; - goto out_free_fbuf; - } - fbuf[st.st_size] = '\n'; /* Dummy line */ - fbuf[st.st_size + 1] = '\0'; - p1 = fbuf; - line = 1; - nlines = 0; - while ((p2 = strchr(p1, '\n')) != NULL) { - *p2 = '\0'; - if (strlazymatch(p1, pat)) { - line_list__add_line(head, line); - nlines++; + while ((len = getline(&line, &line_len, fp)) > 0) { + + if (line[len - 1] == '\n') + line[len - 1] = '\0'; + + if (strlazymatch(line, pat)) { + line_list__add_line(head, linenum); + count++; } - line++; - p1 = p2 + 1; + linenum++; } -out_free_fbuf: - free(fbuf); -out_close: - close(fd); - return nlines; + + if (ferror(fp)) + count = -errno; + free(line); + fclose(fp); + + if (count == 0) + pr_debug("No matched lines found in %s.\n", fname); + return count; } static int probe_point_lazy_walker(const char *fname, int lineno, @@ -1312,10 +1340,7 @@ static int find_probe_point_lazy(Dwarf_Die *sp_die, struct probe_finder *pf) /* Matching lazy line pattern */ ret = find_lazy_match_lines(&pf->lcache, pf->fname, pf->pev->point.lazy_line); - if (ret == 0) { - pr_debug("No matched lines found in %s.\n", pf->fname); - return 0; - } else if (ret < 0) + if (ret <= 0) return ret; } diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c index c6d9933..2040b85 100644 --- a/tools/perf/util/scripting-engines/trace-event-python.c +++ b/tools/perf/util/scripting-engines/trace-event-python.c @@ -248,8 +248,7 @@ static void python_process_event(int cpu, void *data, context = PyCObject_FromVoidPtr(scripting_context, NULL); PyTuple_SetItem(t, n++, PyString_FromString(handler_name)); - PyTuple_SetItem(t, n++, - PyCObject_FromVoidPtr(scripting_context, NULL)); + PyTuple_SetItem(t, n++, context); if (handler) { PyTuple_SetItem(t, n++, PyInt_FromLong(cpu)); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 7821d0e..3e193f8 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1525,8 +1525,8 @@ int dso__load(struct dso *self, struct map *map, symbol_filter_t filter) symbol_conf.symfs, self->long_name); break; case DSO__ORIG_GUEST_KMODULE: - if (map->groups && map->groups->machine) - root_dir = map->groups->machine->root_dir; + if (map->groups && machine) + root_dir = machine->root_dir; else root_dir = ""; snprintf(name, size, "%s%s%s", symbol_conf.symfs, diff --git a/tools/perf/util/top.h b/tools/perf/util/top.h index fe44afb..4f769f4 100644 --- a/tools/perf/util/top.h +++ b/tools/perf/util/top.h @@ -11,11 +11,6 @@ struct perf_evlist; struct perf_evsel; -struct sym_entry_source { - struct list_head head; - pthread_mutex_t lock; -}; - struct sym_entry { struct rb_node rb_node; struct list_head node; @@ -24,7 +19,6 @@ struct sym_entry { int skip; u8 origin; struct map *map; - struct sym_entry_source *src; unsigned long count[0]; }; @@ -46,8 +40,8 @@ struct perf_top { u64 exact_samples; u64 guest_us_samples, guest_kernel_samples; int print_entries, count_filter, delay_secs; - int display_weighted, freq, rb_entries; - int sym_counter, target_pid, target_tid; + int display_weighted, freq, rb_entries, sym_counter; + pid_t target_pid, target_tid; bool hide_kernel_symbols, hide_user_symbols, zero; const char *cpu_list; struct perf_evsel *sym_evsel; diff --git a/tools/perf/util/trace-event-parse.c b/tools/perf/util/trace-event-parse.c index 73a0222..d8e622d 100644 --- a/tools/perf/util/trace-event-parse.c +++ b/tools/perf/util/trace-event-parse.c @@ -153,7 +153,7 @@ void parse_proc_kallsyms(char *file, unsigned int size __unused) char *next = NULL; char *addr_str; char ch; - int ret; + int ret __used; int i; line = strtok_r(file, "\n", &next); diff --git a/tools/perf/util/ui/browsers/annotate.c b/tools/perf/util/ui/browsers/annotate.c index 8d8a168..1aa3965 100644 --- a/tools/perf/util/ui/browsers/annotate.c +++ b/tools/perf/util/ui/browsers/annotate.c @@ -60,7 +60,6 @@ static void annotate_browser__write(struct ui_browser *self, void *entry, int ro } static double objdump_line__calc_percent(struct objdump_line *self, - struct list_head *head, struct symbol *sym, int evidx) { double percent = 0.0; @@ -69,11 +68,12 @@ static double objdump_line__calc_percent(struct objdump_line *self, int len = sym->end - sym->start; unsigned int hits = 0; struct annotation *notes = symbol__annotation(sym); - struct source_line *src_line = notes->src_line; + struct source_line *src_line = notes->src->lines; struct sym_hist *h = annotation__histogram(notes, evidx); s64 offset = self->offset; - struct objdump_line *next = objdump__get_next_ip_line(head, self); + struct objdump_line *next; + next = objdump__get_next_ip_line(¬es->src->source, self); while (offset < (s64)len && (next == NULL || offset < next->offset)) { if (src_line) { @@ -192,10 +192,10 @@ int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx) { struct objdump_line *pos, *n; struct objdump_line_rb_node *rbpos; - LIST_HEAD(head); + struct annotation *notes = symbol__annotation(sym); struct annotate_browser browser = { .b = { - .entries = &head, + .entries = ¬es->src->source, .refresh = ui_browser__list_head_refresh, .seek = ui_browser__list_head_seek, .write = annotate_browser__write, @@ -210,20 +210,20 @@ int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx) if (map->dso->annotate_warned) return -1; - if (symbol__annotate(sym, map, &head, sizeof(*rbpos)) < 0) { + if (symbol__annotate(sym, map, sizeof(*rbpos)) < 0) { ui__error_window(ui_helpline__last_msg); return -1; } ui_helpline__push("Press <- or ESC to exit"); - list_for_each_entry(pos, &head, node) { + list_for_each_entry(pos, ¬es->src->source, node) { size_t line_len = strlen(pos->line); if (browser.b.width < line_len) browser.b.width = line_len; rbpos = objdump_line__rb(pos); rbpos->idx = browser.b.nr_entries++; - rbpos->percent = objdump_line__calc_percent(pos, &head, sym, evidx); + rbpos->percent = objdump_line__calc_percent(pos, sym, evidx); if (rbpos->percent < 0.01) continue; objdump__insert_line(&browser.entries, rbpos); @@ -238,7 +238,7 @@ int symbol__tui_annotate(struct symbol *sym, struct map *map, int evidx) browser.b.width += 18; /* Percentage */ ret = annotate_browser__run(&browser); - list_for_each_entry_safe(pos, n, &head, node) { + list_for_each_entry_safe(pos, n, ¬es->src->source, node) { list_del(&pos->node); objdump_line__free(pos); } diff --git a/tools/perf/util/ui/browsers/map.c b/tools/perf/util/ui/browsers/map.c index e515836..8462bff 100644 --- a/tools/perf/util/ui/browsers/map.c +++ b/tools/perf/util/ui/browsers/map.c @@ -41,7 +41,7 @@ static int ui_entry__read(const char *title, char *bf, size_t size, int width) out_free_form: newtPopWindow(); newtFormDestroy(form); - return 0; + return err; } struct map_browser { diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c index 4c6983d..362a0cb 100644 --- a/tools/power/x86/turbostat/turbostat.c +++ b/tools/power/x86/turbostat/turbostat.c @@ -72,7 +72,7 @@ int need_reinitialize; int num_cpus; -typedef struct per_cpu_counters { +struct counters { unsigned long long tsc; /* per thread */ unsigned long long aperf; /* per thread */ unsigned long long mperf; /* per thread */ @@ -88,13 +88,13 @@ typedef struct per_cpu_counters { int pkg; int core; int cpu; - struct per_cpu_counters *next; -} PCC; + struct counters *next; +}; -PCC *pcc_even; -PCC *pcc_odd; -PCC *pcc_delta; -PCC *pcc_average; +struct counters *cnt_even; +struct counters *cnt_odd; +struct counters *cnt_delta; +struct counters *cnt_average; struct timeval tv_even; struct timeval tv_odd; struct timeval tv_delta; @@ -125,7 +125,7 @@ unsigned long long get_msr(int cpu, off_t offset) return msr; } -void print_header() +void print_header(void) { if (show_pkg) fprintf(stderr, "pkg "); @@ -160,39 +160,39 @@ void print_header() putc('\n', stderr); } -void dump_pcc(PCC *pcc) +void dump_cnt(struct counters *cnt) { - fprintf(stderr, "package: %d ", pcc->pkg); - fprintf(stderr, "core:: %d ", pcc->core); - fprintf(stderr, "CPU: %d ", pcc->cpu); - fprintf(stderr, "TSC: %016llX\n", pcc->tsc); - fprintf(stderr, "c3: %016llX\n", pcc->c3); - fprintf(stderr, "c6: %016llX\n", pcc->c6); - fprintf(stderr, "c7: %016llX\n", pcc->c7); - fprintf(stderr, "aperf: %016llX\n", pcc->aperf); - fprintf(stderr, "pc2: %016llX\n", pcc->pc2); - fprintf(stderr, "pc3: %016llX\n", pcc->pc3); - fprintf(stderr, "pc6: %016llX\n", pcc->pc6); - fprintf(stderr, "pc7: %016llX\n", pcc->pc7); - fprintf(stderr, "msr0x%x: %016llX\n", extra_msr_offset, pcc->extra_msr); + fprintf(stderr, "package: %d ", cnt->pkg); + fprintf(stderr, "core:: %d ", cnt->core); + fprintf(stderr, "CPU: %d ", cnt->cpu); + fprintf(stderr, "TSC: %016llX\n", cnt->tsc); + fprintf(stderr, "c3: %016llX\n", cnt->c3); + fprintf(stderr, "c6: %016llX\n", cnt->c6); + fprintf(stderr, "c7: %016llX\n", cnt->c7); + fprintf(stderr, "aperf: %016llX\n", cnt->aperf); + fprintf(stderr, "pc2: %016llX\n", cnt->pc2); + fprintf(stderr, "pc3: %016llX\n", cnt->pc3); + fprintf(stderr, "pc6: %016llX\n", cnt->pc6); + fprintf(stderr, "pc7: %016llX\n", cnt->pc7); + fprintf(stderr, "msr0x%x: %016llX\n", extra_msr_offset, cnt->extra_msr); } -void dump_list(PCC *pcc) +void dump_list(struct counters *cnt) { - printf("dump_list 0x%p\n", pcc); + printf("dump_list 0x%p\n", cnt); - for (; pcc; pcc = pcc->next) - dump_pcc(pcc); + for (; cnt; cnt = cnt->next) + dump_cnt(cnt); } -void print_pcc(PCC *p) +void print_cnt(struct counters *p) { double interval_float; interval_float = tv_delta.tv_sec + tv_delta.tv_usec/1000000.0; /* topology columns, print blanks on 1st (average) line */ - if (p == pcc_average) { + if (p == cnt_average) { if (show_pkg) fprintf(stderr, " "); if (show_core) @@ -262,24 +262,24 @@ void print_pcc(PCC *p) putc('\n', stderr); } -void print_counters(PCC *cnt) +void print_counters(struct counters *counters) { - PCC *pcc; + struct counters *cnt; print_header(); if (num_cpus > 1) - print_pcc(pcc_average); + print_cnt(cnt_average); - for (pcc = cnt; pcc != NULL; pcc = pcc->next) - print_pcc(pcc); + for (cnt = counters; cnt != NULL; cnt = cnt->next) + print_cnt(cnt); } #define SUBTRACT_COUNTER(after, before, delta) (delta = (after - before), (before > after)) - -int compute_delta(PCC *after, PCC *before, PCC *delta) +int compute_delta(struct counters *after, + struct counters *before, struct counters *delta) { int errors = 0; int perf_err = 0; @@ -391,20 +391,20 @@ int compute_delta(PCC *after, PCC *before, PCC *delta) delta->extra_msr = after->extra_msr; if (errors) { fprintf(stderr, "ERROR cpu%d before:\n", before->cpu); - dump_pcc(before); + dump_cnt(before); fprintf(stderr, "ERROR cpu%d after:\n", before->cpu); - dump_pcc(after); + dump_cnt(after); errors = 0; } } return 0; } -void compute_average(PCC *delta, PCC *avg) +void compute_average(struct counters *delta, struct counters *avg) { - PCC *sum; + struct counters *sum; - sum = calloc(1, sizeof(PCC)); + sum = calloc(1, sizeof(struct counters)); if (sum == NULL) { perror("calloc sum"); exit(1); @@ -438,35 +438,34 @@ void compute_average(PCC *delta, PCC *avg) free(sum); } -void get_counters(PCC *pcc) +void get_counters(struct counters *cnt) { - for ( ; pcc; pcc = pcc->next) { - pcc->tsc = get_msr(pcc->cpu, MSR_TSC); + for ( ; cnt; cnt = cnt->next) { + cnt->tsc = get_msr(cnt->cpu, MSR_TSC); if (do_nhm_cstates) - pcc->c3 = get_msr(pcc->cpu, MSR_CORE_C3_RESIDENCY); + cnt->c3 = get_msr(cnt->cpu, MSR_CORE_C3_RESIDENCY); if (do_nhm_cstates) - pcc->c6 = get_msr(pcc->cpu, MSR_CORE_C6_RESIDENCY); + cnt->c6 = get_msr(cnt->cpu, MSR_CORE_C6_RESIDENCY); if (do_snb_cstates) - pcc->c7 = get_msr(pcc->cpu, MSR_CORE_C7_RESIDENCY); + cnt->c7 = get_msr(cnt->cpu, MSR_CORE_C7_RESIDENCY); if (has_aperf) - pcc->aperf = get_msr(pcc->cpu, MSR_APERF); + cnt->aperf = get_msr(cnt->cpu, MSR_APERF); if (has_aperf) - pcc->mperf = get_msr(pcc->cpu, MSR_MPERF); + cnt->mperf = get_msr(cnt->cpu, MSR_MPERF); if (do_snb_cstates) - pcc->pc2 = get_msr(pcc->cpu, MSR_PKG_C2_RESIDENCY); + cnt->pc2 = get_msr(cnt->cpu, MSR_PKG_C2_RESIDENCY); if (do_nhm_cstates) - pcc->pc3 = get_msr(pcc->cpu, MSR_PKG_C3_RESIDENCY); + cnt->pc3 = get_msr(cnt->cpu, MSR_PKG_C3_RESIDENCY); if (do_nhm_cstates) - pcc->pc6 = get_msr(pcc->cpu, MSR_PKG_C6_RESIDENCY); + cnt->pc6 = get_msr(cnt->cpu, MSR_PKG_C6_RESIDENCY); if (do_snb_cstates) - pcc->pc7 = get_msr(pcc->cpu, MSR_PKG_C7_RESIDENCY); + cnt->pc7 = get_msr(cnt->cpu, MSR_PKG_C7_RESIDENCY); if (extra_msr_offset) - pcc->extra_msr = get_msr(pcc->cpu, extra_msr_offset); + cnt->extra_msr = get_msr(cnt->cpu, extra_msr_offset); } } - -void print_nehalem_info() +void print_nehalem_info(void) { unsigned long long msr; unsigned int ratio; @@ -514,38 +513,38 @@ void print_nehalem_info() } -void free_counter_list(PCC *list) +void free_counter_list(struct counters *list) { - PCC *p; + struct counters *p; for (p = list; p; ) { - PCC *free_me; + struct counters *free_me; free_me = p; p = p->next; free(free_me); } - return; } void free_all_counters(void) { - free_counter_list(pcc_even); - pcc_even = NULL; + free_counter_list(cnt_even); + cnt_even = NULL; - free_counter_list(pcc_odd); - pcc_odd = NULL; + free_counter_list(cnt_odd); + cnt_odd = NULL; - free_counter_list(pcc_delta); - pcc_delta = NULL; + free_counter_list(cnt_delta); + cnt_delta = NULL; - free_counter_list(pcc_average); - pcc_average = NULL; + free_counter_list(cnt_average); + cnt_average = NULL; } -void insert_cpu_counters(PCC **list, PCC *new) +void insert_counters(struct counters **list, + struct counters *new) { - PCC *prev; + struct counters *prev; /* * list was empty @@ -594,18 +593,16 @@ void insert_cpu_counters(PCC **list, PCC *new) */ new->next = prev->next; prev->next = new; - - return; } -void alloc_new_cpu_counters(int pkg, int core, int cpu) +void alloc_new_counters(int pkg, int core, int cpu) { - PCC *new; + struct counters *new; if (verbose > 1) printf("pkg%d core%d, cpu%d\n", pkg, core, cpu); - new = (PCC *)calloc(1, sizeof(PCC)); + new = (struct counters *)calloc(1, sizeof(struct counters)); if (new == NULL) { perror("calloc"); exit(1); @@ -613,9 +610,10 @@ void alloc_new_cpu_counters(int pkg, int core, int cpu) new->pkg = pkg; new->core = core; new->cpu = cpu; - insert_cpu_counters(&pcc_odd, new); + insert_counters(&cnt_odd, new); - new = (PCC *)calloc(1, sizeof(PCC)); + new = (struct counters *)calloc(1, + sizeof(struct counters)); if (new == NULL) { perror("calloc"); exit(1); @@ -623,9 +621,9 @@ void alloc_new_cpu_counters(int pkg, int core, int cpu) new->pkg = pkg; new->core = core; new->cpu = cpu; - insert_cpu_counters(&pcc_even, new); + insert_counters(&cnt_even, new); - new = (PCC *)calloc(1, sizeof(PCC)); + new = (struct counters *)calloc(1, sizeof(struct counters)); if (new == NULL) { perror("calloc"); exit(1); @@ -633,9 +631,9 @@ void alloc_new_cpu_counters(int pkg, int core, int cpu) new->pkg = pkg; new->core = core; new->cpu = cpu; - insert_cpu_counters(&pcc_delta, new); + insert_counters(&cnt_delta, new); - new = (PCC *)calloc(1, sizeof(PCC)); + new = (struct counters *)calloc(1, sizeof(struct counters)); if (new == NULL) { perror("calloc"); exit(1); @@ -643,7 +641,7 @@ void alloc_new_cpu_counters(int pkg, int core, int cpu) new->pkg = pkg; new->core = core; new->cpu = cpu; - pcc_average = new; + cnt_average = new; } int get_physical_package_id(int cpu) @@ -719,7 +717,7 @@ void re_initialize(void) { printf("turbostat: topology changed, re-initializing.\n"); free_all_counters(); - num_cpus = for_all_cpus(alloc_new_cpu_counters); + num_cpus = for_all_cpus(alloc_new_counters); need_reinitialize = 0; printf("num_cpus is now %d\n", num_cpus); } @@ -728,7 +726,7 @@ void dummy(int pkg, int core, int cpu) { return; } /* * check to see if a cpu came on-line */ -void verify_num_cpus() +void verify_num_cpus(void) { int new_num_cpus; @@ -740,14 +738,12 @@ void verify_num_cpus() num_cpus, new_num_cpus); need_reinitialize = 1; } - - return; } void turbostat_loop() { restart: - get_counters(pcc_even); + get_counters(cnt_even); gettimeofday(&tv_even, (struct timezone *)NULL); while (1) { @@ -757,24 +753,24 @@ restart: goto restart; } sleep(interval_sec); - get_counters(pcc_odd); + get_counters(cnt_odd); gettimeofday(&tv_odd, (struct timezone *)NULL); - compute_delta(pcc_odd, pcc_even, pcc_delta); + compute_delta(cnt_odd, cnt_even, cnt_delta); timersub(&tv_odd, &tv_even, &tv_delta); - compute_average(pcc_delta, pcc_average); - print_counters(pcc_delta); + compute_average(cnt_delta, cnt_average); + print_counters(cnt_delta); if (need_reinitialize) { re_initialize(); goto restart; } sleep(interval_sec); - get_counters(pcc_even); + get_counters(cnt_even); gettimeofday(&tv_even, (struct timezone *)NULL); - compute_delta(pcc_even, pcc_odd, pcc_delta); + compute_delta(cnt_even, cnt_odd, cnt_delta); timersub(&tv_even, &tv_odd, &tv_delta); - compute_average(pcc_delta, pcc_average); - print_counters(pcc_delta); + compute_average(cnt_delta, cnt_average); + print_counters(cnt_delta); } } @@ -892,7 +888,7 @@ void check_cpuid() * this check is valid for both Intel and AMD */ asm("cpuid" : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx) : "a" (0x80000007)); - has_invariant_tsc = edx && (1 << 8); + has_invariant_tsc = edx & (1 << 8); if (!has_invariant_tsc) { fprintf(stderr, "No invariant TSC\n"); @@ -905,7 +901,7 @@ void check_cpuid() */ asm("cpuid" : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx) : "a" (0x6)); - has_aperf = ecx && (1 << 0); + has_aperf = ecx & (1 << 0); if (!has_aperf) { fprintf(stderr, "No APERF MSR\n"); exit(1); @@ -952,7 +948,7 @@ void turbostat_init() check_dev_msr(); check_super_user(); - num_cpus = for_all_cpus(alloc_new_cpu_counters); + num_cpus = for_all_cpus(alloc_new_counters); if (verbose) print_nehalem_info(); @@ -962,7 +958,7 @@ int fork_it(char **argv) { int retval; pid_t child_pid; - get_counters(pcc_even); + get_counters(cnt_even); gettimeofday(&tv_even, (struct timezone *)NULL); child_pid = fork(); @@ -985,14 +981,14 @@ int fork_it(char **argv) exit(1); } } - get_counters(pcc_odd); + get_counters(cnt_odd); gettimeofday(&tv_odd, (struct timezone *)NULL); - retval = compute_delta(pcc_odd, pcc_even, pcc_delta); + retval = compute_delta(cnt_odd, cnt_even, cnt_delta); timersub(&tv_odd, &tv_even, &tv_delta); - compute_average(pcc_delta, pcc_average); + compute_average(cnt_delta, cnt_average); if (!retval) - print_counters(pcc_delta); + print_counters(cnt_delta); fprintf(stderr, "%.6f sec\n", tv_delta.tv_sec + tv_delta.tv_usec/1000000.0);; |
