From 6e7eeb51106d2e9ef7975214747e76d23c5d01af Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 2 Sep 2013 10:39:21 -0300 Subject: perf trace: Allow syscall arg formatters to mask args The futex syscall ignores some arguments according to the 'operation' arg, so allow arg formatters to mask those. Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-abqrg3oldgfsdnltfrvso9f7@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 69a065e..c29692a 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -15,14 +15,16 @@ #include #include -static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, unsigned long arg) +static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, + unsigned long arg, u8 *arg_mask __maybe_unused) { return scnprintf(bf, size, "%#lx", arg); } #define SCA_HEX syscall_arg__scnprintf_hex -static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, unsigned long arg) +static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, + unsigned long arg, u8 *arg_mask __maybe_unused) { int printed = 0, prot = arg; @@ -52,7 +54,8 @@ static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, unsigned l #define SCA_MMAP_PROT syscall_arg__scnprintf_mmap_prot -static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size, unsigned long arg) +static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size, + unsigned long arg, u8 *arg_mask __maybe_unused) { int printed = 0, flags = arg; @@ -92,7 +95,8 @@ static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size, unsigned #define SCA_MMAP_FLAGS syscall_arg__scnprintf_mmap_flags -static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, unsigned long arg) +static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, + unsigned long arg, u8 *arg_mask __maybe_unused) { int behavior = arg; @@ -136,7 +140,7 @@ static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, uns static struct syscall_fmt { const char *name; const char *alias; - size_t (*arg_scnprintf[6])(char *bf, size_t size, unsigned long arg); + size_t (*arg_scnprintf[6])(char *bf, size_t size, unsigned long arg, u8 *arg_mask); bool errmsg; bool timeout; bool hexret; @@ -198,7 +202,8 @@ struct syscall { const char *name; bool filtered; struct syscall_fmt *fmt; - size_t (**arg_scnprintf)(char *bf, size_t size, unsigned long arg); + size_t (**arg_scnprintf)(char *bf, size_t size, + unsigned long arg, u8 *args_mask); }; static size_t fprintf_duration(unsigned long t, FILE *fp) @@ -443,17 +448,23 @@ static size_t syscall__scnprintf_args(struct syscall *sc, char *bf, size_t size, if (sc->tp_format != NULL) { struct format_field *field; + u8 mask = 0, bit = 1; + + for (field = sc->tp_format->format.fields->next; field; + field = field->next, ++i, bit <<= 1) { + if (mask & bit) + continue; - for (field = sc->tp_format->format.fields->next; field; field = field->next) { printed += scnprintf(bf + printed, size - printed, "%s%s: ", printed ? ", " : "", field->name); - if (sc->arg_scnprintf && sc->arg_scnprintf[i]) - printed += sc->arg_scnprintf[i](bf + printed, size - printed, args[i]); - else + if (sc->arg_scnprintf && sc->arg_scnprintf[i]) { + printed += sc->arg_scnprintf[i](bf + printed, size - printed, + args[i], &mask); + } else { printed += scnprintf(bf + printed, size - printed, "%ld", args[i]); - ++i; + } } } else { while (i < 6) { -- cgit v0.10.2 From f9da0b0c74af25a68b9ac43a2c81f2eea970de5a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 2 Sep 2013 13:46:44 -0300 Subject: perf trace: Add beautifier for futex 'operation' parm That uses the arg mask mechanism just introduced to suppress ignored arguments according to the futex operation. Based on an initial patch from David Ahern that showed the need for some way to allow args to tell how many further args should be shown. Initial-patch-by: David Ahern Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-0k30it46r4hv5eanefbdmj5t@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index c29692a..1a6cb74 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -14,6 +14,7 @@ #include #include #include +#include static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, unsigned long arg, u8 *arg_mask __maybe_unused) @@ -137,6 +138,49 @@ static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, #define SCA_MADV_BHV syscall_arg__scnprintf_madvise_behavior +static size_t syscall_arg__scnprintf_futex_op(char *bf, size_t size, unsigned long arg, u8 *arg_mask) +{ + enum syscall_futex_args { + SCF_UADDR = (1 << 0), + SCF_OP = (1 << 1), + SCF_VAL = (1 << 2), + SCF_TIMEOUT = (1 << 3), + SCF_UADDR2 = (1 << 4), + SCF_VAL3 = (1 << 5), + }; + int op = arg; + int cmd = op & FUTEX_CMD_MASK; + size_t printed = 0; + + switch (cmd) { +#define P_FUTEX_OP(n) case FUTEX_##n: printed = scnprintf(bf, size, #n); + P_FUTEX_OP(WAIT); *arg_mask |= SCF_VAL3|SCF_UADDR2; break; + P_FUTEX_OP(WAKE); *arg_mask |= SCF_VAL3|SCF_UADDR2|SCF_TIMEOUT; break; + P_FUTEX_OP(FD); *arg_mask |= SCF_VAL3|SCF_UADDR2|SCF_TIMEOUT; break; + P_FUTEX_OP(REQUEUE); *arg_mask |= SCF_VAL3|SCF_TIMEOUT; break; + P_FUTEX_OP(CMP_REQUEUE); *arg_mask |= SCF_TIMEOUT; break; + P_FUTEX_OP(CMP_REQUEUE_PI); *arg_mask |= SCF_TIMEOUT; break; + P_FUTEX_OP(WAKE_OP); break; + P_FUTEX_OP(LOCK_PI); *arg_mask |= SCF_VAL3|SCF_UADDR2|SCF_TIMEOUT; break; + P_FUTEX_OP(UNLOCK_PI); *arg_mask |= SCF_VAL3|SCF_UADDR2|SCF_TIMEOUT; break; + P_FUTEX_OP(TRYLOCK_PI); *arg_mask |= SCF_VAL3|SCF_UADDR2; break; + P_FUTEX_OP(WAIT_BITSET); *arg_mask |= SCF_UADDR2; break; + P_FUTEX_OP(WAKE_BITSET); *arg_mask |= SCF_UADDR2; break; + P_FUTEX_OP(WAIT_REQUEUE_PI); break; + default: printed = scnprintf(bf, size, "%#x", cmd); break; + } + + if (op & FUTEX_PRIVATE_FLAG) + printed += scnprintf(bf + printed, size - printed, "|PRIV"); + + if (op & FUTEX_CLOCK_REALTIME) + printed += scnprintf(bf + printed, size - printed, "|CLKRT"); + + return printed; +} + +#define SCA_FUTEX_OP syscall_arg__scnprintf_futex_op + static struct syscall_fmt { const char *name; const char *alias; @@ -153,7 +197,8 @@ static struct syscall_fmt { { .name = "connect", .errmsg = true, }, { .name = "fstat", .errmsg = true, .alias = "newfstat", }, { .name = "fstatat", .errmsg = true, .alias = "newfstatat", }, - { .name = "futex", .errmsg = true, }, + { .name = "futex", .errmsg = true, + .arg_scnprintf = { [1] = SCA_FUTEX_OP, /* op */ }, }, { .name = "ioctl", .errmsg = true, .arg_scnprintf = { [2] = SCA_HEX, /* arg */ }, }, { .name = "lstat", .errmsg = true, .alias = "newlstat", }, -- cgit v0.10.2 From fa0097ee690693006ab1aea6c01ad3c851b65c77 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Sat, 31 Aug 2013 21:50:51 +0300 Subject: perf: Add a dummy software event to keep tracking When an event is disabled the "tracking" events selected by the 'mmap', 'comm' and 'task' bits of struct perf_event_attr, are also disabled. However, the information those events provide is necessary to resolve symbols for when the main event is re-enabled. The "tracking" events can be kept enabled by putting them on another event, but that requires an event that otherwise does nothing. A new software event PERF_COUNT_SW_DUMMY is added for that purpose. Signed-off-by: Adrian Hunter Acked-by: Peter Zijlstra Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1377975053-3811-2-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h index 408b8c7..ca1d90b 100644 --- a/include/uapi/linux/perf_event.h +++ b/include/uapi/linux/perf_event.h @@ -109,6 +109,7 @@ enum perf_sw_ids { PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6, PERF_COUNT_SW_ALIGNMENT_FAULTS = 7, PERF_COUNT_SW_EMULATION_FAULTS = 8, + PERF_COUNT_SW_DUMMY = 9, PERF_COUNT_SW_MAX, /* non-ABI */ }; -- cgit v0.10.2 From d22d1a2a2c224b3b378d873589ced27add7ebde4 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Sat, 31 Aug 2013 21:50:52 +0300 Subject: perf tools: Add support for PERF_COUNT_SW_DUMMY Add support for the new dummy software event PERF_COUNT_SW_DUMMY. Signed-off-by: Adrian Hunter Acked-by: Jiri Olsa Tested-by: Jiri Olsa Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1377975053-3811-3-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index e8745fb..3612183 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -323,6 +323,7 @@ const char *perf_evsel__sw_names[PERF_COUNT_SW_MAX] = { "major-faults", "alignment-faults", "emulation-faults", + "dummy", }; static const char *__perf_evsel__sw_name(u64 config) diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index 9cba923..b2dd077 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -108,6 +108,10 @@ static struct event_symbol event_symbols_sw[PERF_COUNT_SW_MAX] = { .symbol = "emulation-faults", .alias = "", }, + [PERF_COUNT_SW_DUMMY] = { + .symbol = "dummy", + .alias = "", + }, }; #define __PERF_EVENT_FIELD(config, name) \ diff --git a/tools/perf/util/parse-events.l b/tools/perf/util/parse-events.l index 0790452..91346b7 100644 --- a/tools/perf/util/parse-events.l +++ b/tools/perf/util/parse-events.l @@ -145,6 +145,7 @@ context-switches|cs { return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW cpu-migrations|migrations { return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_CPU_MIGRATIONS); } alignment-faults { return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_ALIGNMENT_FAULTS); } emulation-faults { return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_EMULATION_FAULTS); } +dummy { return sym(yyscanner, PERF_TYPE_SOFTWARE, PERF_COUNT_SW_DUMMY); } L1-dcache|l1-d|l1d|L1-data | L1-icache|l1-i|l1i|L1-instruction | diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c index 381f4fd..71b5412 100644 --- a/tools/perf/util/python.c +++ b/tools/perf/util/python.c @@ -987,6 +987,7 @@ static struct { { "COUNT_SW_PAGE_FAULTS_MAJ", PERF_COUNT_SW_PAGE_FAULTS_MAJ }, { "COUNT_SW_ALIGNMENT_FAULTS", PERF_COUNT_SW_ALIGNMENT_FAULTS }, { "COUNT_SW_EMULATION_FAULTS", PERF_COUNT_SW_EMULATION_FAULTS }, + { "COUNT_SW_DUMMY", PERF_COUNT_SW_DUMMY }, { "SAMPLE_IP", PERF_SAMPLE_IP }, { "SAMPLE_TID", PERF_SAMPLE_TID }, -- cgit v0.10.2 From 395c307089c9f5f0d82c63c11c79227b57bb7ac5 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Sat, 31 Aug 2013 21:50:53 +0300 Subject: perf tests: Add 'keep tracking' test Add a test for the newly added PERF_COUNT_SW_DUMMY event. The test checks that tracking events continue when an event is disabled but a dummy software event is not disabled. Signed-off-by: Adrian Hunter Acked-by: Jiri Olsa Tested-by: Jiri Olsa Cc: David Ahern Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jiri Olsa Cc: Mike Galbraith Cc: Namhyung Kim Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1377975053-3811-4-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/Makefile b/tools/perf/Makefile index ecebfd0..c5dc1ad 100644 --- a/tools/perf/Makefile +++ b/tools/perf/Makefile @@ -465,6 +465,7 @@ endif # NO_LIBELF ifndef NO_LIBUNWIND LIB_OBJS += $(OUTPUT)util/unwind.o endif +LIB_OBJS += $(OUTPUT)tests/keep-tracking.o ifndef NO_LIBAUDIT BUILTIN_OBJS += $(OUTPUT)builtin-trace.o diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c index 8ad9415..8bbeba3 100644 --- a/tools/perf/tests/builtin-test.c +++ b/tools/perf/tests/builtin-test.c @@ -108,6 +108,10 @@ static struct test { .func = test__sample_parsing, }, { + .desc = "Test using a dummy software event to keep tracking", + .func = test__keep_tracking, + }, + { .func = NULL, }, }; diff --git a/tools/perf/tests/keep-tracking.c b/tools/perf/tests/keep-tracking.c new file mode 100644 index 0000000..d444ea2 --- /dev/null +++ b/tools/perf/tests/keep-tracking.c @@ -0,0 +1,154 @@ +#include +#include +#include + +#include "parse-events.h" +#include "evlist.h" +#include "evsel.h" +#include "thread_map.h" +#include "cpumap.h" +#include "tests.h" + +#define CHECK__(x) { \ + while ((x) < 0) { \ + pr_debug(#x " failed!\n"); \ + goto out_err; \ + } \ +} + +#define CHECK_NOT_NULL__(x) { \ + while ((x) == NULL) { \ + pr_debug(#x " failed!\n"); \ + goto out_err; \ + } \ +} + +static int find_comm(struct perf_evlist *evlist, const char *comm) +{ + union perf_event *event; + int i, found; + + found = 0; + for (i = 0; i < evlist->nr_mmaps; i++) { + while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) { + if (event->header.type == PERF_RECORD_COMM && + (pid_t)event->comm.pid == getpid() && + (pid_t)event->comm.tid == getpid() && + strcmp(event->comm.comm, comm) == 0) + found += 1; + } + } + return found; +} + +/** + * test__keep_tracking - test using a dummy software event to keep tracking. + * + * This function implements a test that checks that tracking events continue + * when an event is disabled but a dummy software event is not disabled. If the + * test passes %0 is returned, otherwise %-1 is returned. + */ +int test__keep_tracking(void) +{ + struct perf_record_opts opts = { + .mmap_pages = UINT_MAX, + .user_freq = UINT_MAX, + .user_interval = ULLONG_MAX, + .freq = 4000, + .target = { + .uses_mmap = true, + }, + }; + struct thread_map *threads = NULL; + struct cpu_map *cpus = NULL; + struct perf_evlist *evlist = NULL; + struct perf_evsel *evsel = NULL; + int found, err = -1; + const char *comm; + + threads = thread_map__new(-1, getpid(), UINT_MAX); + CHECK_NOT_NULL__(threads); + + cpus = cpu_map__new(NULL); + CHECK_NOT_NULL__(cpus); + + evlist = perf_evlist__new(); + CHECK_NOT_NULL__(evlist); + + perf_evlist__set_maps(evlist, cpus, threads); + + CHECK__(parse_events(evlist, "dummy:u")); + CHECK__(parse_events(evlist, "cycles:u")); + + perf_evlist__config(evlist, &opts); + + evsel = perf_evlist__first(evlist); + + evsel->attr.comm = 1; + evsel->attr.disabled = 1; + evsel->attr.enable_on_exec = 0; + + if (perf_evlist__open(evlist) < 0) { + fprintf(stderr, " (not supported)"); + err = 0; + goto out_err; + } + + CHECK__(perf_evlist__mmap(evlist, UINT_MAX, false)); + + /* + * First, test that a 'comm' event can be found when the event is + * enabled. + */ + + perf_evlist__enable(evlist); + + comm = "Test COMM 1"; + CHECK__(prctl(PR_SET_NAME, (unsigned long)comm, 0, 0, 0)); + + perf_evlist__disable(evlist); + + found = find_comm(evlist, comm); + if (found != 1) { + pr_debug("First time, failed to find tracking event.\n"); + goto out_err; + } + + /* + * Secondly, test that a 'comm' event can be found when the event is + * disabled with the dummy event still enabled. + */ + + perf_evlist__enable(evlist); + + evsel = perf_evlist__last(evlist); + + CHECK__(perf_evlist__disable_event(evlist, evsel)); + + comm = "Test COMM 2"; + CHECK__(prctl(PR_SET_NAME, (unsigned long)comm, 0, 0, 0)); + + perf_evlist__disable(evlist); + + found = find_comm(evlist, comm); + if (found != 1) { + pr_debug("Seconf time, failed to find tracking event.\n"); + goto out_err; + } + + err = 0; + +out_err: + if (evlist) { + perf_evlist__disable(evlist); + perf_evlist__munmap(evlist); + perf_evlist__close(evlist); + perf_evlist__delete(evlist); + } + if (cpus) + cpu_map__delete(cpus); + if (threads) + thread_map__delete(threads); + + return err; +} diff --git a/tools/perf/tests/tests.h b/tools/perf/tests/tests.h index 83d5b71..c048b58 100644 --- a/tools/perf/tests/tests.h +++ b/tools/perf/tests/tests.h @@ -38,5 +38,6 @@ int test__sw_clock_freq(void); int test__perf_time_to_tsc(void); int test__code_reading(void); int test__sample_parsing(void); +int test__keep_tracking(void); #endif /* TESTS_H */ diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 5df4ca9..b8727ae 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -246,7 +246,7 @@ void perf_evlist__disable(struct perf_evlist *evlist) for (cpu = 0; cpu < nr_cpus; cpu++) { list_for_each_entry(pos, &evlist->entries, node) { - if (!perf_evsel__is_group_leader(pos)) + if (!perf_evsel__is_group_leader(pos) || !pos->fd) continue; for (thread = 0; thread < nr_threads; thread++) ioctl(FD(pos, cpu, thread), @@ -264,7 +264,7 @@ void perf_evlist__enable(struct perf_evlist *evlist) for (cpu = 0; cpu < nr_cpus; cpu++) { list_for_each_entry(pos, &evlist->entries, node) { - if (!perf_evsel__is_group_leader(pos)) + if (!perf_evsel__is_group_leader(pos) || !pos->fd) continue; for (thread = 0; thread < nr_threads; thread++) ioctl(FD(pos, cpu, thread), @@ -273,6 +273,44 @@ void perf_evlist__enable(struct perf_evlist *evlist) } } +int perf_evlist__disable_event(struct perf_evlist *evlist, + struct perf_evsel *evsel) +{ + int cpu, thread, err; + + if (!evsel->fd) + return 0; + + for (cpu = 0; cpu < evlist->cpus->nr; cpu++) { + for (thread = 0; thread < evlist->threads->nr; thread++) { + err = ioctl(FD(evsel, cpu, thread), + PERF_EVENT_IOC_DISABLE, 0); + if (err) + return err; + } + } + return 0; +} + +int perf_evlist__enable_event(struct perf_evlist *evlist, + struct perf_evsel *evsel) +{ + int cpu, thread, err; + + if (!evsel->fd) + return -EINVAL; + + for (cpu = 0; cpu < evlist->cpus->nr; cpu++) { + for (thread = 0; thread < evlist->threads->nr; thread++) { + err = ioctl(FD(evsel, cpu, thread), + PERF_EVENT_IOC_ENABLE, 0); + if (err) + return err; + } + } + return 0; +} + static int perf_evlist__alloc_pollfd(struct perf_evlist *evlist) { int nr_cpus = cpu_map__nr(evlist->cpus); diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h index 841a394..880d713 100644 --- a/tools/perf/util/evlist.h +++ b/tools/perf/util/evlist.h @@ -110,6 +110,11 @@ void perf_evlist__munmap(struct perf_evlist *evlist); void perf_evlist__disable(struct perf_evlist *evlist); void perf_evlist__enable(struct perf_evlist *evlist); +int perf_evlist__disable_event(struct perf_evlist *evlist, + struct perf_evsel *evsel); +int perf_evlist__enable_event(struct perf_evlist *evlist, + struct perf_evsel *evsel); + void perf_evlist__set_selected(struct perf_evlist *evlist, struct perf_evsel *evsel); -- cgit v0.10.2 From b41f1cec91c37eeea6fdb15effbfa24ea0a5536b Mon Sep 17 00:00:00 2001 From: Namhyung Kim Date: Tue, 27 Aug 2013 11:41:53 +0900 Subject: perf list: Skip unsupported events Some hardware events might not be supported on a system. Listing those events seems meaningless and confusing to users. Let's skip them. Before: $ perf list cache | wc -l 33 After: $ perf list cache | wc -l 27 Signed-off-by: Namhyung Kim Cc: David Ahern Cc: Ingo Molnar Cc: Jiri Olsa Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/r/1377571313-14722-1-git-send-email-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c index b2dd077..9812531 100644 --- a/tools/perf/util/parse-events.c +++ b/tools/perf/util/parse-events.c @@ -15,6 +15,7 @@ #define YY_EXTRA_TYPE int #include "parse-events-flex.h" #include "pmu.h" +#include "thread_map.h" #define MAX_NAME_LEN 100 @@ -1076,6 +1077,33 @@ int is_valid_tracepoint(const char *event_string) return 0; } +static bool is_event_supported(u8 type, unsigned config) +{ + bool ret = true; + struct perf_evsel *evsel; + struct perf_event_attr attr = { + .type = type, + .config = config, + .disabled = 1, + .exclude_kernel = 1, + }; + struct { + struct thread_map map; + int threads[1]; + } tmap = { + .map.nr = 1, + .threads = { 0 }, + }; + + evsel = perf_evsel__new(&attr, 0); + if (evsel) { + ret = perf_evsel__open(evsel, NULL, &tmap.map) >= 0; + perf_evsel__delete(evsel); + } + + return ret; +} + static void __print_events_type(u8 type, struct event_symbol *syms, unsigned max) { @@ -1083,14 +1111,16 @@ static void __print_events_type(u8 type, struct event_symbol *syms, unsigned i; for (i = 0; i < max ; i++, syms++) { + if (!is_event_supported(type, i)) + continue; + if (strlen(syms->alias)) snprintf(name, sizeof(name), "%s OR %s", syms->symbol, syms->alias); else snprintf(name, sizeof(name), "%s", syms->symbol); - printf(" %-50s [%s]\n", name, - event_type_descriptors[type]); + printf(" %-50s [%s]\n", name, event_type_descriptors[type]); } } @@ -1119,6 +1149,10 @@ int print_hwcache_events(const char *event_glob, bool name_only) if (event_glob != NULL && !strglobmatch(name, event_glob)) continue; + if (!is_event_supported(PERF_TYPE_HW_CACHE, + type | (op << 8) | (i << 16))) + continue; + if (name_only) printf("%s ", name); else @@ -1148,6 +1182,9 @@ static void print_symbol_events(const char *event_glob, unsigned type, (syms->alias && strglobmatch(syms->alias, event_glob)))) continue; + if (!is_event_supported(type, i)) + continue; + if (name_only) { printf("%s ", syms->symbol); continue; -- cgit v0.10.2 From 0b8c25d949e010274c6697c570b772797ebc7c27 Mon Sep 17 00:00:00 2001 From: David Ahern Date: Sun, 28 Jul 2013 09:48:32 -0600 Subject: perf tools: Fix symbol offset computation for some dsos For some dsos (e.g., libc, libpthread, kernel modules) the symbol offset is huge. e.g., qemu-kvm 17238/17242 [007] 762235.640311: ffffffff816288a1 __schedule+0x451 ([kernel.kallsyms]) ffffffff81629609 schedule+0x29 ([kernel.kallsyms]) ffffffffa00a6ded kvm_vcpu_block+0xffffffffa00a106d (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffffa00bae6b kvm_arch_vcpu_ioctl_run+0xffffffffa00a118b (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffffa00a4d7a kvm_vcpu_ioctl+0xffffffffa00a141a (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffff811a7bdb do_vfs_ioctl+0x8b ([kernel.kallsyms]) ffffffff811a80c1 sys_ioctl+0x91 ([kernel.kallsyms]) ffffffff81633182 system_call+0x72 ([kernel.kallsyms]) 7f882a97af27 __GI___ioctl+0x7f882a891007 (/lib64/libc-2.14.90.so) 100000002 [unknown] ([unknown]) It seems to be maps with a non-0 start. Taking that into account the offsets are correct: qemu-kvm 17238/17242 [007] 762235.640311: ffffffff816288a1 __schedule+0x451 ([kernel.kallsyms]) ffffffff81629609 schedule+0x29 ([kernel.kallsyms]) ffffffffa00a6ded kvm_vcpu_block+0x6d (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffffa00bae6b kvm_arch_vcpu_ioctl_run+0x18b (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffffa00a4d7a kvm_vcpu_ioctl+0x41a (/lib/modules/3.11.0-rc1+/kernel/arch/x86/kvm/kvm.ko) ffffffff811a7bdb do_vfs_ioctl+0x8b ([kernel.kallsyms]) ffffffff811a80c1 sys_ioctl+0x91 ([kernel.kallsyms]) ffffffff81633182 system_call+0x72 ([kernel.kallsyms]) 7f882a97af27 __GI___ioctl+0x7 (/lib64/libc-2.14.90.so) 100000002 [unknown] ([unknown]) Signed-off-by: David Ahern Link: http://lkml.kernel.org/r/1375026512-45826-1-git-send-email-dsahern@gmail.com Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 07642a7..1fc0c62 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -1513,6 +1513,7 @@ void perf_evsel__print_ip(struct perf_evsel *evsel, union perf_event *event, printf(" "); if (print_symoffset) { al.addr = node->ip; + al.map = node->map; symbol__fprintf_symname_offs(node->sym, &al, stdout); } else symbol__fprintf_symname(node->sym, stdout); diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 77f3b95..7eb0362 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -259,7 +259,10 @@ size_t symbol__fprintf_symname_offs(const struct symbol *sym, if (sym && sym->name) { length = fprintf(fp, "%s", sym->name); if (al) { - offset = al->addr - sym->start; + if (al->addr < sym->end) + offset = al->addr - sym->start; + else + offset = al->addr - al->map->start - sym->start; length += fprintf(fp, "+0x%lx", offset); } return length; -- cgit v0.10.2 From 579e7865b2d431bb7d380a1b4ea0aa8eb8a10fd4 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 2 Sep 2013 15:37:32 -0300 Subject: perf trace: Add beautifier for lseek's whence arg [root@zoo ~]# perf trace -a -e lseek | head -1 546.922 ( 0.004 ms): 1184 lseek(fd: 26, offset: 0, whence: CUR) = 2 [root@zoo ~]# Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-2eiuhwz9jbnhj80q6jaqeji4@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 1a6cb74..02aaea6 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -24,6 +24,31 @@ static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, #define SCA_HEX syscall_arg__scnprintf_hex +static size_t syscall_arg__scnprintf_whence(char *bf, size_t size, + unsigned long arg, u8 *arg_mask __maybe_unused) +{ + int whence = arg; + + switch (whence) { +#define P_WHENCE(n) case SEEK_##n: return scnprintf(bf, size, #n) + P_WHENCE(SET); + P_WHENCE(CUR); + P_WHENCE(END); +#ifdef SEEK_DATA + P_WHENCE(DATA); +#endif +#ifdef SEEK_HOLE + P_WHENCE(HOLE); +#endif +#undef P_WHENCE + default: break; + } + + return scnprintf(bf, size, "%#x", whence); +} + +#define SCA_WHENCE syscall_arg__scnprintf_whence + static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, unsigned long arg, u8 *arg_mask __maybe_unused) { @@ -201,6 +226,8 @@ static struct syscall_fmt { .arg_scnprintf = { [1] = SCA_FUTEX_OP, /* op */ }, }, { .name = "ioctl", .errmsg = true, .arg_scnprintf = { [2] = SCA_HEX, /* arg */ }, }, + { .name = "lseek", .errmsg = true, + .arg_scnprintf = { [2] = SCA_WHENCE, /* whence */ }, }, { .name = "lstat", .errmsg = true, .alias = "newlstat", }, { .name = "madvise", .errmsg = true, .arg_scnprintf = { [0] = SCA_HEX, /* start */ -- cgit v0.10.2 From be65a89a0b558cb5b6863be71861f29b36feb88e Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 2 Sep 2013 16:22:31 -0300 Subject: perf trace: Add beautifier for open's flags arg Suppressing the mode when O_CREAT not present, needs improvements on the arg masking mechanism to be reused in openat, open_by_handle_at, mq_open: [root@zoo ~]# perf trace -a -e open | grep -v 'flags: RDONLY' | head -5 147.541 ( 0.028 ms): 1188 open(filename: 0x33c17782fb, flags: CLOEXEC ) = 23 229.898 ( 0.020 ms): 2071 open(filename: 0x3d93c80, flags: NOATIME ) = -1 EPERM Operation not permitted [root@zoo ~]# perf trace -a -e open | grep CREAT 1406.697 ( 0.024 ms): 616 open(filename: 0x7fffc3a0f910, flags: CREAT|TRUNC|WRONLY, mode: 438 ) = -1 ENOENT No such file or directory 2032.770 ( 0.804 ms): 4354 open(filename: 0x7f33ac814368, flags: CREAT|EXCL|RDWR, mode: 384 ) = 115 ^C[root@zoo ~]# Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-c7vm6klaf995qw1vqdih5t7q@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 02aaea6..5b6b287 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -206,6 +206,62 @@ static size_t syscall_arg__scnprintf_futex_op(char *bf, size_t size, unsigned lo #define SCA_FUTEX_OP syscall_arg__scnprintf_futex_op +static size_t syscall_arg__scnprintf_open_flags(char *bf, size_t size, + unsigned long arg, u8 *arg_mask) +{ + int printed = 0, flags = arg; + + if (!(flags & O_CREAT)) + *arg_mask |= 1 << 2; /* Mask the mode parm */ + + if (flags == 0) + return scnprintf(bf, size, "RDONLY"); +#define P_FLAG(n) \ + if (flags & O_##n) { \ + printed += scnprintf(bf + printed, size - printed, "%s%s", printed ? "|" : "", #n); \ + flags &= ~O_##n; \ + } + + P_FLAG(APPEND); + P_FLAG(ASYNC); + P_FLAG(CLOEXEC); + P_FLAG(CREAT); + P_FLAG(DIRECT); + P_FLAG(DIRECTORY); + P_FLAG(EXCL); + P_FLAG(LARGEFILE); + P_FLAG(NOATIME); + P_FLAG(NOCTTY); +#ifdef O_NONBLOCK + P_FLAG(NONBLOCK); +#elif O_NDELAY + P_FLAG(NDELAY); +#endif +#ifdef O_PATH + P_FLAG(PATH); +#endif + P_FLAG(RDWR); +#ifdef O_DSYNC + if ((flags & O_SYNC) == O_SYNC) + printed += scnprintf(bf + printed, size - printed, "%s%s", printed ? "|" : "", "SYNC"); + else { + P_FLAG(DSYNC); + } +#else + P_FLAG(SYNC); +#endif + P_FLAG(TRUNC); + P_FLAG(WRONLY); +#undef P_FLAG + + if (flags) + printed += scnprintf(bf + printed, size - printed, "%s%#x", printed ? "|" : "", flags); + + return printed; +} + +#define SCA_OPEN_FLAGS syscall_arg__scnprintf_open_flags + static struct syscall_fmt { const char *name; const char *alias; @@ -244,7 +300,8 @@ static struct syscall_fmt { [4] = SCA_HEX, /* new_addr */ }, }, { .name = "munmap", .errmsg = true, .arg_scnprintf = { [0] = SCA_HEX, /* addr */ }, }, - { .name = "open", .errmsg = true, }, + { .name = "open", .errmsg = true, + .arg_scnprintf = { [1] = SCA_OPEN_FLAGS, /* flags */ }, }, { .name = "poll", .errmsg = true, .timeout = true, }, { .name = "ppoll", .errmsg = true, .timeout = true, }, { .name = "pread", .errmsg = true, .alias = "pread64", }, -- cgit v0.10.2 From 31cd3855c98119cae287b761d8d2e75018714c5d Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 2 Sep 2013 16:40:40 -0300 Subject: perf trace: Tell arg formatters the arg index ... so that it can mask args relative to its position, like the 'mode' arg that may or not be printed according to the 'flags' (O_CREAT) value. [root@zoo ~]# perf trace -a -e openat,open_by_handle_at | head -1 469.754 ( 0.034 ms): 1183 openat(dfd: -100, filename: 0x7fbde40014b0, flags: CLOEXEC|DIRECTORY|NONBLOCK) = 23 [root@zoo ~]# Cc: Adrian Hunter Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-bgokqpkufd4sio7ixxknf1ux@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c index 5b6b287..b6f0725 100644 --- a/tools/perf/builtin-trace.c +++ b/tools/perf/builtin-trace.c @@ -17,7 +17,9 @@ #include static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, - unsigned long arg, u8 *arg_mask __maybe_unused) + unsigned long arg, + u8 arg_idx __maybe_unused, + u8 *arg_mask __maybe_unused) { return scnprintf(bf, size, "%#lx", arg); } @@ -25,7 +27,9 @@ static size_t syscall_arg__scnprintf_hex(char *bf, size_t size, #define SCA_HEX syscall_arg__scnprintf_hex static size_t syscall_arg__scnprintf_whence(char *bf, size_t size, - unsigned long arg, u8 *arg_mask __maybe_unused) + unsigned long arg, + u8 arg_idx __maybe_unused, + u8 *arg_mask __maybe_unused) { int whence = arg; @@ -50,7 +54,9 @@ static size_t syscall_arg__scnprintf_whence(char *bf, size_t size, #define SCA_WHENCE syscall_arg__scnprintf_whence static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, - unsigned long arg, u8 *arg_mask __maybe_unused) + unsigned long arg, + u8 arg_idx __maybe_unused, + u8 *arg_mask __maybe_unused) { int printed = 0, prot = arg; @@ -81,7 +87,8 @@ static size_t syscall_arg__scnprintf_mmap_prot(char *bf, size_t size, #define SCA_MMAP_PROT syscall_arg__scnprintf_mmap_prot static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size, - unsigned long arg, u8 *arg_mask __maybe_unused) + unsigned long arg, u8 arg_idx __maybe_unused, + u8 *arg_mask __maybe_unused) { int printed = 0, flags = arg; @@ -122,7 +129,8 @@ static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size, #define SCA_MMAP_FLAGS syscall_arg__scnprintf_mmap_flags static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, - unsigned long arg, u8 *arg_mask __maybe_unused) + unsigned long arg, u8 arg_idx __maybe_unused, + u8 *arg_mask __maybe_unused) { int behavior = arg; @@ -163,7 +171,8 @@ static size_t syscall_arg__scnprintf_madvise_behavior(char *bf, size_t size, #define SCA_MADV_BHV syscall_arg__scnprintf_madvise_behavior -static size_t syscall_arg__scnprintf_futex_op(char *bf, size_t size, unsigned long arg, u8 *arg_mask) +static size_t syscall_arg__scnprintf_futex_op(char *bf, size_t size, unsigned long arg, + u8 arg_idx __maybe_unused, u8 *arg_mask) { enum syscall_futex_args { SCF_UADDR = (1 << 0), @@ -207,12 +216,13 @@ static size_t syscall_arg__scnprintf_futex_op(char *bf, size_t size, unsigned lo #define SCA_FUTEX_OP syscall_arg__scnprintf_futex_op static size_t syscall_arg__scnprintf_open_flags(char *bf, size_t size, - unsigned long arg, u8 *arg_mask) + unsigned long arg, + u8 arg_idx, u8 *arg_mask) { int printed = 0, flags = arg; if (!(flags & O_CREAT)) - *arg_mask |= 1 << 2; /* Mask the mode parm */ + *arg_mask |= 1 << (arg_idx + 1); /* Mask the mode parm */ if (flags == 0) return scnprintf(bf, size, "RDONLY"); @@ -265,7 +275,7 @@ static size_t syscall_arg__scnprintf_open_flags(char *bf, size_t size, static struct syscall_fmt { const char *name; const char *alias; - size_t (*arg_scnprintf[6])(char *bf, size_t size, unsigned long arg, u8 *arg_mask); + size_t (*arg_scnprintf[6])(char *bf, size_t size, unsigned long arg, u8 arg_idx, u8 *arg_mask); bool errmsg; bool timeout; bool hexret; @@ -302,6 +312,10 @@ static struct syscall_fmt { .arg_scnprintf = { [0] = SCA_HEX, /* addr */ }, }, { .name = "open", .errmsg = true, .arg_scnprintf = { [1] = SCA_OPEN_FLAGS, /* flags */ }, }, + { .name = "open_by_handle_at", .errmsg = true, + .arg_scnprintf = { [2] = SCA_OPEN_FLAGS, /* flags */ }, }, + { .name = "openat", .errmsg = true, + .arg_scnprintf = { [2] = SCA_OPEN_FLAGS, /* flags */ }, }, { .name = "poll", .errmsg = true, .timeout = true, }, { .name = "ppoll", .errmsg = true, .timeout = true, }, { .name = "pread", .errmsg = true, .alias = "pread64", }, @@ -332,7 +346,7 @@ struct syscall { bool filtered; struct syscall_fmt *fmt; size_t (**arg_scnprintf)(char *bf, size_t size, - unsigned long arg, u8 *args_mask); + unsigned long arg, u8 arg_idx, u8 *args_mask); }; static size_t fprintf_duration(unsigned long t, FILE *fp) @@ -589,7 +603,7 @@ static size_t syscall__scnprintf_args(struct syscall *sc, char *bf, size_t size, if (sc->arg_scnprintf && sc->arg_scnprintf[i]) { printed += sc->arg_scnprintf[i](bf + printed, size - printed, - args[i], &mask); + args[i], i, &mask); } else { printed += scnprintf(bf + printed, size - printed, "%ld", args[i]); -- cgit v0.10.2