From 3e76ac78b08479e84a3eca3fb1b3066fb8230461 Mon Sep 17 00:00:00 2001 From: Andrew Vagin Date: Tue, 20 Dec 2011 17:32:45 +0300 Subject: [PATCH 01/10] perf record: Add ability to record event period The problem is that when SAMPLE_PERIOD is not set, the kernel generates a number of samples in proportion to an event's period. Number of these samples may be too big and the kernel throttles all samples above a defined limit. E.g.: I want to trace when a process sleeps. I created a process which sleeps for 1ms and for 4ms. perf got 100 events in both cases. swapper 0 [000] 1141.371830: sched_stat_sleep: comm=foo pid=1801 delay=1386750 [ns] swapper 0 [000] 1141.369444: sched_stat_sleep: comm=foo pid=1801 delay=4499585 [ns] In the first case a kernel want to send 4499585 events and in the second case it wants to send 1386750 events. perf-reports shows that process sleeps in both places equal time. Instead of this we can get only one sample with an attribute period. As result we have less data transferring between kernel and user-space and we avoid throttling of samples. The patch "events: Don't divide events if it has field period" added a kernel part of this functionality. Acked-by: Arun Sharma Cc: Arun Sharma Cc: David Ahern Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: devel@openvz.org Link: http://lkml.kernel.org/r/1324391565-1369947-1-git-send-email-avagin@openvz.org Signed-off-by: Andrew Vagin Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 1 + tools/perf/perf.h | 1 + tools/perf/util/evsel.c | 3 +++ 3 files changed, 5 insertions(+) diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 766fa0a91a32..f8fd14fb62ec 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -700,6 +700,7 @@ const struct option record_options[] = { OPT_BOOLEAN('d', "data", &record.opts.sample_address, "Sample addresses"), OPT_BOOLEAN('T', "timestamp", &record.opts.sample_time, "Sample timestamps"), + OPT_BOOLEAN('P', "period", &record.opts.period, "Sample period"), OPT_BOOLEAN('n', "no-samples", &record.opts.no_samples, "don't sample"), OPT_BOOLEAN('N', "no-buildid-cache", &record.no_buildid_cache, diff --git a/tools/perf/perf.h b/tools/perf/perf.h index ea804f5a8cc2..64f8bee31ced 100644 --- a/tools/perf/perf.h +++ b/tools/perf/perf.h @@ -200,6 +200,7 @@ struct perf_record_opts { bool sample_time; bool sample_id_all_avail; bool system_wide; + bool period; unsigned int freq; unsigned int mmap_pages; unsigned int user_freq; diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 4a8c8b02e9cc..60ad0286759e 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -108,6 +108,9 @@ void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts) if (opts->system_wide) attr->sample_type |= PERF_SAMPLE_CPU; + if (opts->period) + attr->sample_type |= PERF_SAMPLE_PERIOD; + if (opts->sample_id_all_avail && (opts->sample_time || opts->system_wide || !opts->no_inherit || opts->cpu_list)) From f3bda2c9a689b38c059f7cb2d761ff58a2996370 Mon Sep 17 00:00:00 2001 From: Robert Richter Date: Thu, 15 Dec 2011 17:32:39 +0100 Subject: [PATCH 02/10] perf evsel: Fix uninitialized memory access to struct perf_sample Memory in struct perf_sample is not fully initialized during parsing. Depending on sampling data some parts may left unchanged. Zero out struct perf_sample first to avoid access to uninitialized memory. Cc: Ingo Molnar Link: http://lkml.kernel.org/r/1323966762-8574-2-git-send-email-robert.richter@amd.com Signed-off-by: Robert Richter Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evsel.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 60ad0286759e..667f3b78bb2c 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -460,7 +460,7 @@ int perf_event__parse_sample(const union perf_event *event, u64 type, u32 val32[2]; } u; - + memset(data, 0, sizeof(*data)); data->cpu = data->pid = data->tid = -1; data->stream_id = data->id = data->time = -1ULL; From 6581f6e35f7d0338f699fce660adb48e863f2b59 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:50 +0900 Subject: [PATCH 03/10] perf report: Document '--call-graph' for optional print_limit argument The '--call-graph' command line option can receive undocumented optional print_limit argument. Besides, use strtoul() to parse the option since its type is u32. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-2-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/perf-report.txt | 5 +++-- tools/perf/builtin-report.c | 6 +++--- 2 files changed, 6 insertions(+), 5 deletions(-) diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index dc85392a5ac7..35af0dc8ccb4 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -80,9 +80,10 @@ OPTIONS --dump-raw-trace:: Dump raw trace in ASCII. --g [type,min,order]:: +-g [type,min[,limit],order]:: --call-graph:: - Display call chains using type, min percent threshold and order. + Display call chains using type, min percent threshold, optional print + limit and order. type can be either: - flat: single column, linear exposure of call chains. - graph: use a graph tree, displaying absolute overhead rates. diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c index ece7c5d3f504..b2654c9fb5c6 100644 --- a/tools/perf/builtin-report.c +++ b/tools/perf/builtin-report.c @@ -407,7 +407,7 @@ parse_callchain_opt(const struct option *opt, const char *arg, int unset) goto setup; if (tok2[0] != 'c') { - callchain_param.print_limit = strtod(tok2, &endptr); + callchain_param.print_limit = strtoul(tok2, &endptr, 0); tok2 = strtok(NULL, ","); if (!tok2) goto setup; @@ -485,8 +485,8 @@ int cmd_report(int argc, const char **argv, const char *prefix __used) "regex filter to identify parent, see: '--sort parent'"), OPT_BOOLEAN('x', "exclude-other", &symbol_conf.exclude_other, "Only display entries with parent-match"), - OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent, call_order", - "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold and callchain order. " + OPT_CALLBACK_DEFAULT('g', "call-graph", &report, "output_type,min_percent[,print_limit],call_order", + "Display callchains using output_type (graph, flat, fractal, or none) , min percent threshold, optional print limit and callchain order. " "Default: fractal,0.5,callee", &parse_callchain_opt, callchain_default_opt), OPT_BOOLEAN('G', "inverted", &report.inverted_callchain, "alias for inverted call graph"), From 301b195db179241da8be25f345f3c4e64960f1d5 Mon Sep 17 00:00:00 2001 From: Nelson Elhage Date: Mon, 19 Dec 2011 08:39:30 -0500 Subject: [PATCH 04/10] perf evlist: Fix errno value reporting on failed mmap On failure, perf_evlist__mmap_per_{cpu,thread} will try to munmap() every map that doesn't have a NULL base. This will fail with EINVAL if one of them has base == MAP_FAILED, clobbering errno, so that perf_evlist__map will return EINVAL on any failure regardless of the root cause. Fix this by resetting failed maps to a NULL base. Acked-by: Namhyung Kim Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1324301972-22740-2-git-send-email-nelhage@nelhage.com Signed-off-by: Nelson Elhage Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 8b19e7a1e881..963d63dde457 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -447,8 +447,10 @@ static int __perf_evlist__mmap(struct perf_evlist *evlist, evlist->mmap[idx].mask = mask; evlist->mmap[idx].base = mmap(NULL, evlist->mmap_len, prot, MAP_SHARED, fd, 0); - if (evlist->mmap[idx].base == MAP_FAILED) + if (evlist->mmap[idx].base == MAP_FAILED) { + evlist->mmap[idx].base = NULL; return -1; + } perf_evlist__add_pollfd(evlist, fd); return 0; From 2b600f9578852d12af59420011e3dadfaa58b043 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:51 +0900 Subject: [PATCH 05/10] perf symbols: Get rid of duplicated snprintf() The 'path' variable is set on a upper line, don't need to do it again. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-3-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 632b50c7bc26..e54b13d4c357 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1757,7 +1757,7 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, struct stat st; /*sshfs might return bad dent->d_type, so we have to stat*/ - sprintf(path, "%s/%s", dir_name, dent->d_name); + snprintf(path, sizeof(path), "%s/%s", dir_name, dent->d_name); if (stat(path, &st)) continue; @@ -1766,8 +1766,6 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, !strcmp(dent->d_name, "..")) continue; - snprintf(path, sizeof(path), "%s/%s", - dir_name, dent->d_name); ret = map_groups__set_modules_path_dir(mg, path); if (ret < 0) goto out; @@ -1788,9 +1786,6 @@ static int map_groups__set_modules_path_dir(struct map_groups *mg, if (map == NULL) continue; - snprintf(path, sizeof(path), "%s/%s", - dir_name, dent->d_name); - long_name = strdup(path); if (long_name == NULL) { ret = -1; From d74c896b7e3250a07f7d0315eecdd2ae1a7bc3c3 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:52 +0900 Subject: [PATCH 06/10] perf symbols: Fix error path on symbol__init() The order of freeing comm_list and dso_list should be reversed. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-4-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/symbol.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index e54b13d4c357..215d50f2042e 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -2604,10 +2604,10 @@ int symbol__init(void) symbol_conf.initialized = true; return 0; -out_free_dso_list: - strlist__delete(symbol_conf.dso_list); out_free_comm_list: strlist__delete(symbol_conf.comm_list); +out_free_dso_list: + strlist__delete(symbol_conf.dso_list); return -1; } From 0161d82e9b740caa90f508138d1ae1b9d981b6d3 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:53 +0900 Subject: [PATCH 07/10] perf tools: Fix a memory leak on perf_read_values_destroy After freeing each elements of the @values->value, we should free itself too. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-5-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/values.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tools/perf/util/values.c b/tools/perf/util/values.c index bdd33470b235..697c8b4e59cc 100644 --- a/tools/perf/util/values.c +++ b/tools/perf/util/values.c @@ -32,6 +32,7 @@ void perf_read_values_destroy(struct perf_read_values *values) for (i = 0; i < values->threads; i++) free(values->value[i]); + free(values->value); free(values->pid); free(values->tid); free(values->counterrawid); From 5f9273d64a5ccbd3c2b4446cc8b71123ed5d6366 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 22:52:03 +0900 Subject: [PATCH 08/10] perf tools: Remove stale git headlines from top comment These files are part of PERF not GIT although they're come from there :) Cc: Ingo Molnar Cc: Johannes Schindelin Cc: Linus Torvalds Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323784323-2150-1-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/config.c | 5 ++++- tools/perf/util/usage.c | 5 ++++- 2 files changed, 8 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/config.c b/tools/perf/util/config.c index 80d9598db31a..0deac6a14b65 100644 --- a/tools/perf/util/config.c +++ b/tools/perf/util/config.c @@ -1,5 +1,8 @@ /* - * GIT - The information manager from hell + * config.c + * + * Helper functions for parsing config items. + * Originally copied from GIT source. * * Copyright (C) Linus Torvalds, 2005 * Copyright (C) Johannes Schindelin, 2005 diff --git a/tools/perf/util/usage.c b/tools/perf/util/usage.c index e16bf9a707e8..d76d1c0ff98f 100644 --- a/tools/perf/util/usage.c +++ b/tools/perf/util/usage.c @@ -1,5 +1,8 @@ /* - * GIT - The information manager from hell + * usage.c + * + * Various reporting routines. + * Originally copied from GIT source. * * Copyright (C) Linus Torvalds, 2005 */ From cb8f4e9aa37c469ddd80dda51469f327606c0118 Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 13 Dec 2011 00:16:55 +0900 Subject: [PATCH 09/10] perf events: Tidy up perf_event__preprocess_sample Use local variable 'dso' to reduce typing a bit and rearrange the if condition. Also NULL check of al->map in the condition is not necessary. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1323703017-6060-7-git-send-email-namhyung@gmail.com Signed-off-by: Namhyung Kim Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/event.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c index 97c479bcb0dc..b7c7f39a8f6d 100644 --- a/tools/perf/util/event.c +++ b/tools/perf/util/event.c @@ -814,13 +814,14 @@ int perf_event__preprocess_sample(const union perf_event *event, al->cpu = sample->cpu; if (al->map) { + struct dso *dso = al->map->dso; + if (symbol_conf.dso_list && - (!al->map || !al->map->dso || - !(strlist__has_entry(symbol_conf.dso_list, - al->map->dso->short_name) || - (al->map->dso->short_name != al->map->dso->long_name && - strlist__has_entry(symbol_conf.dso_list, - al->map->dso->long_name))))) + (!dso || !(strlist__has_entry(symbol_conf.dso_list, + dso->short_name) || + (dso->short_name != dso->long_name && + strlist__has_entry(symbol_conf.dso_list, + dso->long_name))))) goto out_filtered; al->sym = map__find_symbol(al->map, al->addr, filter); From 65c1e0452a3389f9b7b8c1b23305ed2922fafb2d Mon Sep 17 00:00:00 2001 From: Jiri Olsa Date: Thu, 15 Dec 2011 16:30:39 +0100 Subject: [PATCH 10/10] perf test: Add more automated tests for event parsing Adding automated tests for event parsing to include testing for modifier and ',' operator. Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: cjashfor@linux.vnet.ibm.com Link: http://lkml.kernel.org/r/1323963039-7602-4-git-send-email-jolsa@redhat.com Signed-off-by: Jiri Olsa [ committer note: Remove some tests that need group_leader & bison patchkits ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-test.c | 127 +++++++++++++++++++++++++++++++++++++- 1 file changed, 126 insertions(+), 1 deletion(-) diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c index 6173f780dce0..2b9a7f497a20 100644 --- a/tools/perf/builtin-test.c +++ b/tools/perf/builtin-test.c @@ -603,7 +603,7 @@ out_free_threads: #define TEST_ASSERT_VAL(text, cond) \ do { \ - if (!cond) { \ + if (!(cond)) { \ pr_debug("FAILED %s:%d %s\n", __FILE__, __LINE__, text); \ return -1; \ } \ @@ -759,6 +759,103 @@ static int test__checkevent_breakpoint_w(struct perf_evlist *evlist) return 0; } +static int test__checkevent_tracepoint_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_tracepoint(evlist); +} + +static int +test__checkevent_tracepoint_multi_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel; + + TEST_ASSERT_VAL("wrong number of entries", evlist->nr_entries > 1); + + list_for_each_entry(evsel, &evlist->entries, node) { + TEST_ASSERT_VAL("wrong exclude_user", + !evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", + evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + } + + return test__checkevent_tracepoint_multi(evlist); +} + +static int test__checkevent_raw_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_raw(evlist); +} + +static int test__checkevent_numeric_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", !evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_numeric(evlist); +} + +static int test__checkevent_symbolic_name_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", !evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_symbolic_name(evlist); +} + +static int test__checkevent_symbolic_alias_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", !evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", !evsel->attr.precise_ip); + + return test__checkevent_symbolic_alias(evlist); +} + +static int test__checkevent_genhw_modifier(struct perf_evlist *evlist) +{ + struct perf_evsel *evsel = list_entry(evlist->entries.next, + struct perf_evsel, node); + + TEST_ASSERT_VAL("wrong exclude_user", evsel->attr.exclude_user); + TEST_ASSERT_VAL("wrong exclude_kernel", !evsel->attr.exclude_kernel); + TEST_ASSERT_VAL("wrong exclude_hv", evsel->attr.exclude_hv); + TEST_ASSERT_VAL("wrong precise_ip", evsel->attr.precise_ip); + + return test__checkevent_genhw(evlist); +} + static struct test__event_st { const char *name; __u32 type; @@ -808,6 +905,34 @@ static struct test__event_st { .name = "mem:0:w", .check = test__checkevent_breakpoint_w, }, + { + .name = "syscalls:sys_enter_open:k", + .check = test__checkevent_tracepoint_modifier, + }, + { + .name = "syscalls:*:u", + .check = test__checkevent_tracepoint_multi_modifier, + }, + { + .name = "r1:kp", + .check = test__checkevent_raw_modifier, + }, + { + .name = "1:1:hp", + .check = test__checkevent_numeric_modifier, + }, + { + .name = "instructions:h", + .check = test__checkevent_symbolic_name_modifier, + }, + { + .name = "faults:u", + .check = test__checkevent_symbolic_alias_modifier, + }, + { + .name = "L1-dcache-load-miss:kp", + .check = test__checkevent_genhw_modifier, + }, }; #define TEST__EVENTS_CNT (sizeof(test__events) / sizeof(struct test__event_st))