diff options
Diffstat (limited to 'tools/perf')
54 files changed, 677 insertions, 455 deletions
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt index c407897f043..82bffac036e 100644 --- a/tools/perf/Documentation/perf-record.txt +++ b/tools/perf/Documentation/perf-record.txt @@ -209,6 +209,10 @@ overrides that and uses per-thread mmaps. A side-effect of that is that inheritance is automatically disabled. --per-thread is ignored with a warning if combined with -a or -C options. +--initial-delay msecs:: +After starting the program, wait msecs before measuring. This is useful to +filter out the startup phase of the program, which is often very different. + 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 80c7da6732f..29ee857c09c 100644 --- a/tools/perf/Documentation/perf-stat.txt +++ b/tools/perf/Documentation/perf-stat.txt @@ -133,7 +133,7 @@ use --per-core in addition to -a. (system-wide). The output includes the core number and the number of online logical processors on that physical processor. -D msecs:: ---initial-delay msecs:: +--delay msecs:: After starting the program, wait msecs before measuring. This is useful to filter out the startup phase of the program, which is often very different. diff --git a/tools/perf/MANIFEST b/tools/perf/MANIFEST index 3170a7ff578..f41572d0dd7 100644 --- a/tools/perf/MANIFEST +++ b/tools/perf/MANIFEST @@ -1,9 +1,11 @@ tools/perf tools/scripts tools/lib/traceevent -tools/lib/lk +tools/lib/api tools/lib/symbol/kallsyms.c tools/lib/symbol/kallsyms.h +tools/include/asm/bug.h +tools/include/linux/compiler.h include/linux/const.h include/linux/perf_event.h include/linux/rbtree.h diff --git a/tools/perf/Makefile.perf b/tools/perf/Makefile.perf index 3638b0bd20d..87d7726cee2 100644 --- a/tools/perf/Makefile.perf +++ b/tools/perf/Makefile.perf @@ -211,7 +211,7 @@ LIB_H += ../../include/linux/hash.h LIB_H += ../../include/linux/stringify.h LIB_H += util/include/linux/bitmap.h LIB_H += util/include/linux/bitops.h -LIB_H += util/include/linux/compiler.h +LIB_H += ../include/linux/compiler.h LIB_H += util/include/linux/const.h LIB_H += util/include/linux/ctype.h LIB_H += util/include/linux/kernel.h @@ -226,7 +226,7 @@ LIB_H += util/include/linux/string.h LIB_H += util/include/linux/types.h LIB_H += util/include/linux/linkage.h LIB_H += util/include/asm/asm-offsets.h -LIB_H += util/include/asm/bug.h +LIB_H += ../include/asm/bug.h LIB_H += util/include/asm/byteorder.h LIB_H += util/include/asm/hweight.h LIB_H += util/include/asm/swab.h diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c index ab65057a031..0da603b79b6 100644 --- a/tools/perf/builtin-annotate.c +++ b/tools/perf/builtin-annotate.c @@ -232,7 +232,7 @@ static int __cmd_annotate(struct perf_annotate *ann) perf_session__fprintf_dsos(session, stdout); total_nr_samples = 0; - list_for_each_entry(pos, &session->evlist->entries, node) { + evlist__for_each(session->evlist, pos) { struct hists *hists = &pos->hists; u32 nr_samples = hists->stats.nr_events[PERF_RECORD_SAMPLE]; diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c index e6a0844bc2f..a77e31246c0 100644 --- a/tools/perf/builtin-diff.c +++ b/tools/perf/builtin-diff.c @@ -356,9 +356,10 @@ static struct perf_evsel *evsel_match(struct perf_evsel *evsel, { struct perf_evsel *e; - list_for_each_entry(e, &evlist->entries, node) + evlist__for_each(evlist, e) { if (perf_evsel__match2(evsel, e)) return e; + } return NULL; } @@ -367,7 +368,7 @@ static void perf_evlist__collapse_resort(struct perf_evlist *evlist) { struct perf_evsel *evsel; - list_for_each_entry(evsel, &evlist->entries, node) { + evlist__for_each(evlist, evsel) { struct hists *hists = &evsel->hists; hists__collapse_resort(hists, NULL); @@ -614,7 +615,7 @@ static void data_process(void) struct perf_evsel *evsel_base; bool first = true; - list_for_each_entry(evsel_base, &evlist_base->entries, node) { + evlist__for_each(evlist_base, evsel_base) { struct data__file *d; int i; @@ -769,6 +770,81 @@ static int hpp__entry_baseline(struct hist_entry *he, char *buf, size_t size) return ret; } +static int __hpp__color_compare(struct perf_hpp_fmt *fmt, + struct perf_hpp *hpp, struct hist_entry *he, + int comparison_method) +{ + struct diff_hpp_fmt *dfmt = + container_of(fmt, struct diff_hpp_fmt, fmt); + struct hist_entry *pair = get_pair_fmt(he, dfmt); + double diff; + s64 wdiff; + char pfmt[20] = " "; + + if (!pair) + goto dummy_print; + + switch (comparison_method) { + case COMPUTE_DELTA: + if (pair->diff.computed) + diff = pair->diff.period_ratio_delta; + else + diff = compute_delta(he, pair); + + if (fabs(diff) < 0.01) + goto dummy_print; + scnprintf(pfmt, 20, "%%%+d.2f%%%%", dfmt->header_width - 1); + return percent_color_snprintf(hpp->buf, hpp->size, + pfmt, diff); + case COMPUTE_RATIO: + if (he->dummy) + goto dummy_print; + if (pair->diff.computed) + diff = pair->diff.period_ratio; + else + diff = compute_ratio(he, pair); + + scnprintf(pfmt, 20, "%%%d.6f", dfmt->header_width); + return value_color_snprintf(hpp->buf, hpp->size, + pfmt, diff); + case COMPUTE_WEIGHTED_DIFF: + if (he->dummy) + goto dummy_print; + if (pair->diff.computed) + wdiff = pair->diff.wdiff; + else + wdiff = compute_wdiff(he, pair); + + scnprintf(pfmt, 20, "%%14ld", dfmt->header_width); + return color_snprintf(hpp->buf, hpp->size, + get_percent_color(wdiff), + pfmt, wdiff); + default: + BUG_ON(1); + } +dummy_print: + return scnprintf(hpp->buf, hpp->size, "%*s", + dfmt->header_width, pfmt); +} + +static int hpp__color_delta(struct perf_hpp_fmt *fmt, + struct perf_hpp *hpp, struct hist_entry *he) +{ + return __hpp__color_compare(fmt, hpp, he, COMPUTE_DELTA); +} + +static int hpp__color_ratio(struct perf_hpp_fmt *fmt, + struct perf_hpp *hpp, struct hist_entry *he) +{ + return __hpp__color_compare(fmt, hpp, he, COMPUTE_RATIO); +} + +static int hpp__color_wdiff(struct perf_hpp_fmt *fmt, + struct perf_hpp *hpp, struct hist_entry *he) +{ + return __hpp__color_compare(fmt, hpp, he, COMPUTE_WEIGHTED_DIFF); +} + static void hpp__entry_unpair(struct hist_entry *he, int idx, char *buf, size_t size) { @@ -940,8 +1016,22 @@ static void data__hpp_register(struct data__file *d, int idx) fmt->entry = hpp__entry_global; /* TODO more colors */ - if (idx == PERF_HPP_DIFF__BASELINE) + switch (idx) { + case PERF_HPP_DIFF__BASELINE: fmt->color = hpp__color_baseline; + break; + case PERF_HPP_DIFF__DELTA: + fmt->color = hpp__color_delta; + break; + case PERF_HPP_DIFF__RATIO: + fmt->color = hpp__color_ratio; + break; + case PERF_HPP_DIFF__WEIGHTED_DIFF: + fmt->color = hpp__color_wdiff; + break; + default: + break; + } init_header(d, dfmt); perf_hpp__column_register(fmt); diff --git a/tools/perf/builtin-evlist.c b/tools/perf/builtin-evlist.c index 20b0f12763b..c99e0de7e54 100644 --- a/tools/perf/builtin-evlist.c +++ b/tools/perf/builtin-evlist.c @@ -29,7 +29,7 @@ static int __cmd_evlist(const char *file_name, struct perf_attr_details *details if (session == NULL) return -ENOMEM; - list_for_each_entry(pos, &session->evlist->entries, node) + evlist__for_each(session->evlist, pos) perf_evsel__fprintf(pos, details, stdout); perf_session__delete(session); diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c index c9f6d74e1fd..b3466018bbd 100644 --- a/tools/perf/builtin-inject.c +++ b/tools/perf/builtin-inject.c @@ -369,7 +369,7 @@ static int __cmd_inject(struct perf_inject *inject) inject->tool.ordered_samples = true; - list_for_each_entry(evsel, &session->evlist->entries, node) { + evlist__for_each(session->evlist, evsel) { const char *name = perf_evsel__name(evsel); if (!strcmp(name, "sched:sched_switch")) { diff --git a/tools/perf/builtin-kvm.c b/tools/perf/builtin-kvm.c index a6ec1052c29..a7350519c63 100644 --- a/tools/perf/builtin-kvm.c +++ b/tools/perf/builtin-kvm.c @@ -1174,7 +1174,7 @@ static int kvm_live_open_events(struct perf_kvm_stat *kvm) * Note: exclude_{guest,host} do not apply here. * This command processes KVM tracepoints from host only */ - list_for_each_entry(pos, &evlist->entries, node) { + evlist__for_each(evlist, pos) { struct perf_event_attr *attr = &pos->attr; /* make sure these *are* set */ @@ -1556,10 +1556,8 @@ out: if (kvm->session) perf_session__delete(kvm->session); kvm->session = NULL; - if (kvm->evlist) { - perf_evlist__delete_maps(kvm->evlist); + if (kvm->evlist) perf_evlist__delete(kvm->evlist); - } return err; } diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 6ec0cbc2a5d..07d4cf8d3fd 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -183,7 +183,7 @@ static int record__open(struct record *rec) perf_evlist__config(evlist, opts); - list_for_each_entry(pos, &evlist->entries, node) { + evlist__for_each(evlist, pos) { try_again: if (perf_evsel__open(pos, evlist->cpus, evlist->threads) < 0) { if (perf_evsel__fallback(pos, errno, msg, sizeof(msg))) { @@ -324,7 +324,6 @@ out: static void record__init_features(struct record *rec) { - struct perf_evlist *evsel_list = rec->evlist; struct perf_session *session = rec->session; int feat; @@ -334,13 +333,29 @@ static void record__init_features(struct record *rec) if (rec->no_buildid) perf_header__clear_feat(&session->header, HEADER_BUILD_ID); - if (!have_tracepoints(&evsel_list->entries)) + if (!have_tracepoints(&rec->evlist->entries)) perf_header__clear_feat(&session->header, HEADER_TRACING_DATA); if (!rec->opts.branch_stack) perf_header__clear_feat(&session->header, HEADER_BRANCH_STACK); } +static volatile int workload_exec_errno; + +/* + * perf_evlist__prepare_workload will send a SIGUSR1 + * if the fork fails, since we asked by setting its + * want_signal to true. + */ +static void workload_exec_failed_signal(int signo, siginfo_t *info, + void *ucontext __maybe_unused) +{ + workload_exec_errno = info->si_value.sival_int; + done = 1; + signr = signo; + child_finished = 1; +} + static int __cmd_record(struct record *rec, int argc, const char **argv) { int err; @@ -349,7 +364,6 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) struct machine *machine; struct perf_tool *tool = &rec->tool; struct record_opts *opts = &rec->opts; - struct perf_evlist *evsel_list = rec->evlist; struct perf_data_file *file = &rec->file; struct perf_session *session; bool disabled = false; @@ -359,7 +373,6 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) on_exit(record__sig_exit, rec); signal(SIGCHLD, sig_handler); signal(SIGINT, sig_handler); - signal(SIGUSR1, sig_handler); signal(SIGTERM, sig_handler); session = perf_session__new(file, false, NULL); @@ -373,9 +386,9 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) record__init_features(rec); if (forks) { - err = perf_evlist__prepare_workload(evsel_list, &opts->target, + err = perf_evlist__prepare_workload(rec->evlist, &opts->target, argv, file->is_pipe, - true); + workload_exec_failed_signal); if (err < 0) { pr_err("Couldn't run the workload!\n"); goto out_delete_session; @@ -387,7 +400,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) goto out_delete_session; } - if (!evsel_list->nr_groups) + if (!rec->evlist->nr_groups) perf_header__clear_feat(&session->header, HEADER_GROUP_DESC); /* @@ -400,7 +413,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) if (err < 0) goto out_delete_session; } else { - err = perf_session__write_header(session, evsel_list, + err = perf_session__write_header(session, rec->evlist, file->fd, false); if (err < 0) goto out_delete_session; @@ -424,7 +437,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) goto out_delete_session; } - if (have_tracepoints(&evsel_list->entries)) { + if (have_tracepoints(&rec->evlist->entries)) { /* * FIXME err <= 0 here actually means that * there were no tracepoints so its not really @@ -433,7 +446,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) * return this more properly and also * propagate errors that now are calling die() */ - err = perf_event__synthesize_tracing_data(tool, file->fd, evsel_list, + err = perf_event__synthesize_tracing_data(tool, file->fd, rec->evlist, process_synthesized_event); if (err <= 0) { pr_err("Couldn't record tracing data.\n"); @@ -465,7 +478,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) perf_event__synthesize_guest_os, tool); } - err = __machine__synthesize_threads(machine, tool, &opts->target, evsel_list->threads, + err = __machine__synthesize_threads(machine, tool, &opts->target, rec->evlist->threads, process_synthesized_event, opts->sample_address); if (err != 0) goto out_delete_session; @@ -486,14 +499,19 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) * (apart from group members) have enable_on_exec=1 set, * so don't spoil it by prematurely enabling them. */ - if (!target__none(&opts->target)) - perf_evlist__enable(evsel_list); + if (!target__none(&opts->target) && !opts->initial_delay) + perf_evlist__enable(rec->evlist); /* * Let the child rip */ if (forks) - perf_evlist__start_workload(evsel_list); + perf_evlist__start_workload(rec->evlist); + + if (opts->initial_delay) { + usleep(opts->initial_delay * 1000); + perf_evlist__enable(rec->evlist); + } for (;;) { int hits = rec->samples; @@ -506,7 +524,7 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) if (hits == rec->samples) { if (done) break; - err = poll(evsel_list->pollfd, evsel_list->nr_fds, -1); + err = poll(rec->evlist->pollfd, rec->evlist->nr_fds, -1); waking++; } @@ -516,11 +534,19 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) * disable events in this case. */ if (done && !disabled && !target__none(&opts->target)) { - perf_evlist__disable(evsel_list); + perf_evlist__disable(rec->evlist); disabled = true; } } + if (forks && workload_exec_errno) { + char msg[512]; + const char *emsg = strerror_r(workload_exec_errno, msg, sizeof(msg)); + pr_err("Workload failed: %s\n", emsg); + err = -1; + goto out_delete_session; + } + if (quiet || signr == SIGUSR1) return 0; @@ -856,6 +882,8 @@ const struct option record_options[] = { OPT_CALLBACK('G', "cgroup", &record.evlist, "name", "monitor event in cgroup name only", parse_cgroups), + OPT_UINTEGER(0, "initial-delay", &record.opts.initial_delay, + "ms to wait before starting measurement after program start"), OPT_STRING('u', "uid", &record.opts.target.uid_str, "user", "user to profile"), @@ -878,16 +906,13 @@ const struct option record_options[] = { int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused) { int err = -ENOMEM; - struct perf_evlist *evsel_list; struct record *rec = &record; char errbuf[BUFSIZ]; - evsel_list = perf_evlist__new(); - if (evsel_list == NULL) + rec->evlist = perf_evlist__new(); + if (rec->evlist == NULL) return -ENOMEM; - rec->evlist = evsel_list; - argc = parse_options(argc, argv, record_options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); if (!argc && target__none(&rec->opts.target)) @@ -914,8 +939,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused) if (rec->no_buildid_cache || rec->no_buildid) disable_buildid_cache(); - if (evsel_list->nr_entries == 0 && - perf_evlist__add_default(evsel_list) < 0) { + if (rec->evlist->nr_entries == 0 && + perf_evlist__add_default(rec->evlist) < 0) { pr_err("Not enough memory for event selector list\n"); goto out_symbol_exit; } @@ -941,20 +966,15 @@ int cmd_record(int argc, const char **argv, const char *prefix __maybe_unused) } err = -ENOMEM; - if (perf_evlist__create_maps(evsel_list, &rec->opts.target) < 0) + if (perf_evlist__create_maps(rec->evlist, &rec->opts.target) < 0) usage_with_options(record_usage, record_options); if (record_opts__config(&rec->opts)) { err = -EINVAL; - goto out_free_fd; + goto out_symbol_exit; } err = __cmd_record(&record, argc, argv); - - perf_evlist__munmap(evsel_list); - perf_evlist__close(evsel_list); -out_free_fd: - perf_evlist__delete_maps(evsel_list); out_symbol_exit: symbol__exit(); return err; diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index bf8dd2e893e..46864dd7eb8 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -384,7 +384,7 @@ static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist, { struct perf_evsel *pos; - list_for_each_entry(pos, &evlist->entries, node) { + evlist__for_each(evlist, pos) { struct hists *hists = &pos->hists; const char *evname = perf_evsel__name(pos); @@ -412,40 +412,11 @@ static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist, return 0; } -static int __cmd_report(struct report *rep) +static void report__warn_kptr_restrict(const struct report *rep) { - int ret = -EINVAL; - u64 nr_samples; - struct perf_session *session = rep->session; - struct perf_evsel *pos; - struct map *kernel_map; - struct kmap *kernel_kmap; - const char *help = "For a higher level overview, try: perf report --sort comm,dso"; - struct ui_progress prog; - struct perf_data_file *file = session->file; + struct map *kernel_map = rep->session->machines.host.vmlinux_maps[MAP__FUNCTION]; + struct kmap *kernel_kmap = map__kmap(kernel_map); - signal(SIGINT, sig_handler); - - if (rep->cpu_list) { - ret = perf_session__cpu_bitmap(session, rep->cpu_list, - rep->cpu_bitmap); - if (ret) - return ret; - } - - if (rep->show_threads) - perf_read_values_init(&rep->show_threads_values); - - ret = report__setup_sample_type(rep); - if (ret) - return ret; - - ret = perf_session__process_events(session, &rep->tool); - if (ret) - return ret; - - kernel_map = session->machines.host.vmlinux_maps[MAP__FUNCTION]; - kernel_kmap = map__kmap(kernel_map); if (kernel_map == NULL || (kernel_map->dso->hit && (kernel_kmap->ref_reloc_sym == NULL || @@ -468,28 +439,73 @@ static int __cmd_report(struct report *rep) "Samples in kernel modules can't be resolved as well.\n\n", desc); } +} - if (use_browser == 0) { - if (verbose > 3) - perf_session__fprintf(session, stdout); +static int report__gtk_browse_hists(struct report *rep, const char *help) +{ + int (*hist_browser)(struct perf_evlist *evlist, const char *help, + struct hist_browser_timer *timer, float min_pcnt); - if (verbose > 2) - perf_session__fprintf_dsos(session, stdout); + hist_browser = dlsym(perf_gtk_handle, "perf_evlist__gtk_browse_hists"); - if (dump_trace) { - perf_session__fprintf_nr_events(session, stdout); - return 0; - } + if (hist_browser == NULL) { + ui__error("GTK browser not found!\n"); + return -1; } - nr_samples = 0; - list_for_each_entry(pos, &session->evlist->entries, node) + return hist_browser(rep->session->evlist, help, NULL, rep->min_percent); +} + +static int report__browse_hists(struct report *rep) +{ + int ret; + struct perf_session *session = rep->session; + struct perf_evlist *evlist = session->evlist; + const char *help = "For a higher level overview, try: perf report --sort comm,dso"; + + switch (use_browser) { + case 1: + ret = perf_evlist__tui_browse_hists(evlist, help, NULL, + rep->min_percent, + &session->header.env); + /* + * Usually "ret" is the last pressed key, and we only + * care if the key notifies us to switch data file. + */ + if (ret != K_SWITCH_INPUT_DATA) + ret = 0; + break; + case 2: + ret = report__gtk_browse_hists(rep, help); + break; + default: + ret = perf_evlist__tty_browse_hists(evlist, rep, help); + break; + } + + return ret; +} + +static u64 report__collapse_hists(struct report *rep) +{ + struct ui_progress prog; + struct perf_evsel *pos; + u64 nr_samples = 0; + /* + * Count number of histogram entries to use when showing progress, + * reusing nr_samples variable. + */ + evlist__for_each(rep->session->evlist, pos) nr_samples += pos->hists.nr_entries; ui_progress__init(&prog, nr_samples, "Merging related events..."); < |