From bca647aac5067fec8dfcbf8ddb79a4c0d5afdfdd Mon Sep 17 00:00:00 2001 From: Tom Zanussi Date: Wed, 10 Nov 2010 08:11:30 -0600 Subject: perf record: make the record options available outside perf record Other perf commands that invoke perf record, such as perf trace, may want to reuse the options used by perf record. This makes them non-static and renames them to avoid clashes with other 'options' variables. Signed-off-by: Tom Zanussi Acked-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4e75583ddd6d..93bd2ff001fb 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -790,7 +790,7 @@ static const char * const record_usage[] = { static bool force, append_file; -static const struct option options[] = { +const struct option record_options[] = { OPT_CALLBACK('e', "event", NULL, "event", "event selector. use 'perf list' to list available events", parse_events), @@ -839,16 +839,16 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) { int i, j, err = -ENOMEM; - argc = parse_options(argc, argv, options, record_usage, + argc = parse_options(argc, argv, record_options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); if (!argc && target_pid == -1 && target_tid == -1 && !system_wide && !cpu_list) - usage_with_options(record_usage, options); + usage_with_options(record_usage, record_options); if (force && append_file) { fprintf(stderr, "Can't overwrite and append at the same time." " You need to choose between -f and -A"); - usage_with_options(record_usage, options); + usage_with_options(record_usage, record_options); } else if (append_file) { write_mode = WRITE_APPEND; } else { @@ -871,7 +871,7 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) if (thread_num <= 0) { fprintf(stderr, "Can't find all threads of pid %d\n", target_pid); - usage_with_options(record_usage, options); + usage_with_options(record_usage, record_options); } } else { all_tids=malloc(sizeof(pid_t)); -- cgit From d9cf837ef9629ab34167bd6fc0141383ddb8813a Mon Sep 17 00:00:00 2001 From: Corey Ashford Date: Fri, 19 Nov 2010 17:37:24 -0800 Subject: perf stat: Change and clean up sys_perf_event_open error handling This patch makes several changes to "perf stat": - "perf stat" will no longer go ahead and run the application when one or more of the specified events could not be opened. - Use error() and die() instead of pr_err() so that the output is more consistent with "perf top" and "perf record". - Handle permission errors in a more robust way, and in a similar way to "perf record" and "perf top". In addition, the sys_perf_event_open() error handling of "perf top" and "perf record" is made more consistent and adds the following phrase when an event doesn't open (with something ther than an access or permission error): "/bin/dmesg may provide additional information." This is added because kernel code doesn't have a good way of expressing detailed errors to user space, so its only avenue is to use printk's. However, many users may not think of looking at dmesg to find out why an event is being rejected. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Ian Munsie Cc: Michael Ellerman LKML-Reference: <1290217044-26293-1-git-send-email-cjashfor@linux.vnet.ibm.com> Signed-off-by: Corey Ashford Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 93bd2ff001fb..d9dd47885218 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -326,7 +326,7 @@ try_again: goto try_again; } printf("\n"); - error("perfcounter syscall returned with %d (%s)\n", + error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n", fd[nr_cpu][counter][thread_index], strerror(err)); #if defined(__i386__) || defined(__x86_64__) -- cgit From c1a3a4b90a5a47adcca0e587f5d7e9ea61329b26 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 22 Nov 2010 14:01:55 -0200 Subject: perf record: Handle restrictive permissions in /proc/{kallsyms,modules} The 59365d1 commit, even being reverted by 33e0d57, showed a non robust behavior in 'perf record': it really should just warn the user that some functionality will not be available. The new behavior then becomes: [acme@felicio linux]$ ls -la /proc/{kallsyms,modules} -r-------- 1 root root 0 Nov 22 12:19 /proc/kallsyms -r-------- 1 root root 0 Nov 22 12:19 /proc/modules [acme@felicio linux]$ perf record ls -R > /dev/null Couldn't record kernel reference relocation symbol Symbol resolution may be skewed if relocation was used (e.g. kexec). Check /proc/kallsyms permission or run as root. [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.004 MB perf.data (~161 samples) ] [acme@felicio linux]$ perf report --stdio [kernel.kallsyms] with build id 77b05e00e64e4de1c9347d83879779b540d69f00 not found, continuing without symbols # Events: 98 cycles # # Overhead Command Shared Object Symbol # ........ ....... ............... .................... # 48.26% ls [kernel] [k] ffffffff8102b92b 22.49% ls libc-2.12.90.so [.] __strlen_sse2 8.35% ls libc-2.12.90.so [.] __GI___strcoll_l 8.17% ls ls [.] 11580 3.35% ls libc-2.12.90.so [.] _IO_new_file_xsputn 3.33% ls libc-2.12.90.so [.] _int_malloc 1.88% ls libc-2.12.90.so [.] _int_free 0.84% ls libc-2.12.90.so [.] malloc_consolidate 0.84% ls libc-2.12.90.so [.] __readdir64 0.83% ls ls [.] strlen@plt 0.83% ls libc-2.12.90.so [.] __GI_fwrite_unlocked 0.83% ls libc-2.12.90.so [.] __memcpy_sse2 # # (For a higher level overview, try: perf report --sort comm,dso) # [acme@felicio linux]$ It still has the build-ids for DSOs in the maps with hits: [acme@felicio linux]$ perf buildid-list 77b05e00e64e4de1c9347d83879779b540d69f00 [kernel.kallsyms] 09c4a431a4a8b648fcfc2c2bdda70f56050ddff1 /bin/ls af75ea9ad951d25e0f038901a11b3846dccb29a4 /lib64/libc-2.12.90.so [acme@felicio linux]$ That can be used in another machine to resolve kernel symbols. Cc: Eugene Teo Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Jesper Juhl Cc: Marcus Meissner Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Sarah Sharp Cc: Stephane Eranian Cc: Tejun Heo Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 93bd2ff001fb..e2c2de201eec 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -697,17 +697,18 @@ static int __cmd_record(int argc, const char **argv) if (err < 0) err = event__synthesize_kernel_mmap(process_synthesized_event, session, machine, "_stext"); - if (err < 0) { - pr_err("Couldn't record kernel reference relocation symbol.\n"); - return err; - } + if (err < 0) + pr_err("Couldn't record kernel reference relocation symbol\n" + "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" + "Check /proc/kallsyms permission or run as root.\n"); err = event__synthesize_modules(process_synthesized_event, session, machine); - if (err < 0) { - pr_err("Couldn't record kernel reference relocation symbol.\n"); - return err; - } + if (err < 0) + pr_err("Couldn't record kernel module information.\n" + "Symbol resolution may be skewed if relocation was used (e.g. kexec).\n" + "Check /proc/modules permission or run as root.\n"); + if (perf_guest) perf_session__process_machines(session, event__synthesize_guest_os); -- cgit From baa2f6cedbfae962f04281a31f08ec29667d31a0 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 26 Nov 2010 19:39:15 -0200 Subject: perf record: Add option to disable collecting build-ids Collecting build-ids for long running sessions may take a long time because it needs to traverse the whole just collected perf.data stream of events, marking the DSOs that had hits and then looking for the .note.gnu.build-id ELF section. For things like the 'trace' tool that records and right away consumes the data on systems where its unlikely that the DSOs being monitored will change while 'trace' runs, it is desirable to remove build id collection, so add a -B/--no-buildid option to perf record to allow such use case. Longer term we'll avoid all this if we, at DSO load time, in the kernel, take advantage of this slow code path to collect the build-id and stash it somewhere, so that we can insert it in the PERF_RECORD_MMAP event. Reported-by: Thomas Gleixner Cc: Frederic Weisbecker Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 3d2cb4899807..024e1441d76b 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -61,6 +61,7 @@ static bool inherit_stat = false; static bool no_samples = false; static bool sample_address = false; static bool no_buildid = false; +static bool no_buildid_cache = false; static long samples = 0; static u64 bytes_written = 0; @@ -437,7 +438,8 @@ static void atexit_header(void) if (!pipe_output) { session->header.data_size += bytes_written; - process_buildids(); + if (!no_buildid) + process_buildids(); perf_header__write(&session->header, output, true); perf_session__delete(session); symbol__exit(); @@ -557,6 +559,9 @@ static int __cmd_record(int argc, const char **argv) return -1; } + if (!no_buildid) + perf_header__set_feat(&session->header, HEADER_BUILD_ID); + if (!file_new) { err = perf_header__read(session, output); if (err < 0) @@ -831,8 +836,10 @@ const struct option record_options[] = { "Sample addresses"), OPT_BOOLEAN('n', "no-samples", &no_samples, "don't sample"), - OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid, + OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache, "do not update the buildid cache"), + OPT_BOOLEAN('B', "no-buildid", &no_buildid, + "do not collect buildids in perf.data"), OPT_END() }; @@ -857,7 +864,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) } symbol__init(); - if (no_buildid) + + if (no_buildid_cache || no_buildid) disable_buildid_cache(); if (!nr_counters) { -- cgit From 640c03ce837fe8d4b56342aba376ea0da3960459 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 2 Dec 2010 14:10:21 -0200 Subject: perf session: Parse sample earlier MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit At perf_session__process_event, so that we reduce the number of lines in eache tool sample processing routine that now receives a sample_data pointer already parsed. This will also be useful in the next patch, where we'll allow sample the identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu, timestamp) just after before every event. Also validate callchains in perf_session__process_event, i.e. as early as possible, and keep a counter of the number of events discarded due to invalid callchains, warning the user about it if it happens. There is an assumption that was kept that all events have the same sample_type, that will be dealt with in the future, when this preexisting limitation will be removed. Tested-by: Thomas Gleixner Reviewed-by: Thomas Gleixner Acked-by: Ian Munsie Acked-by: Thomas Gleixner Cc: Frédéric Weisbecker Cc: Ian Munsie Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Stephane Eranian LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org> Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 024e1441d76b..b34de9291c27 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -36,6 +36,7 @@ static int *fd[MAX_NR_CPUS][MAX_COUNTERS]; static u64 user_interval = ULLONG_MAX; static u64 default_interval = 0; +static u64 sample_type; static int nr_cpus = 0; static unsigned int page_size; @@ -129,6 +130,7 @@ static void write_output(void *buf, size_t size) } static int process_synthesized_event(event_t *event, + struct sample_data *sample __used, struct perf_session *self __used) { write_output(event, event->header.size); @@ -287,6 +289,9 @@ static void create_counter(int counter, int cpu) attr->sample_type |= PERF_SAMPLE_CPU; } + if (!sample_type) + sample_type = attr->sample_type; + attr->mmap = track; attr->comm = track; attr->inherit = !no_inherit; @@ -642,6 +647,8 @@ static int __cmd_record(int argc, const char **argv) open_counters(cpumap[i]); } + perf_session__set_sample_type(session, sample_type); + if (pipe_output) { err = perf_header__write_pipe(output); if (err < 0) -- cgit From 9c90a61c7e4286aa5a38b314a2d8f5a1e70b5135 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Thu, 2 Dec 2010 10:25:28 -0200 Subject: perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit So that we can use -T == --timestamp, asking for PERF_SAMPLE_TIME: $ perf record -aT $ perf report -D | grep PERF_RECORD_ 3 5951915425 0x47530 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff8138c1a2 period: 215979 cpu:3 3 5952026879 0x47588 [0x90]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff810cb480 period: 215979 cpu:3 3 5952059959 0x47618 [0x38]: PERF_RECORD_FORK(6853:6853):(16811:16811) 3 5952138878 0x47650 [0x78]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff811bac35 period: 431478 cpu:3 3 5952375068 0x476c8 [0x30]: PERF_RECORD_COMM: find:6853 3 5952395923 0x476f8 [0x50]: PERF_RECORD_MMAP 6853/6853: [0x400000(0x25000) @ 0]: /usr/bin/find 3 5952413756 0x47748 [0xa0]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff810d080f period: 859332 cpu:3 3 5952419837 0x477e8 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44600000(0x21d000) @ 0]: /lib64/ld-2.5.so 3 5952437929 0x47840 [0x48]: PERF_RECORD_MMAP 6853/6853: [0x7fff7e1c9000(0x1000) @ 0x7fff7e1c9000]: [vdso] 3 5952570127 0x47888 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f46200000(0x218000) @ 0]: /lib64/libselinux.so.1 3 5952623637 0x478e0 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44a00000(0x356000) @ 0]: /lib64/libc-2.5.so 3 5952675720 0x47938 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44e00000(0x204000) @ 0]: /lib64/libdl-2.5.so 3 5952710080 0x47990 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f45a00000(0x246000) @ 0]: /lib64/libsepol.so.1 3 5952847802 0x479e8 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff813897f0 period: 1142536 cpu:3 First column is the cpu and the second the timestamp. That way we can investigate problems in the event stream. If the new perf binary is run on an older kernel, it will disable this feature automatically. Tested-by: Thomas Gleixner Reviewed-by: Thomas Gleixner Acked-by: Ian Munsie Acked-by: Thomas Gleixner Cc: Frédéric Weisbecker Cc: Ian Munsie Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Paul Mackerras Cc: Stephane Eranian LKML-Reference: <1291318772-30880-5-git-send-email-acme@infradead.org> Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index b34de9291c27..699dd2149c4b 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -49,6 +49,7 @@ static const char *output_name = "perf.data"; static int group = 0; static int realtime_prio = 0; static bool raw_samples = false; +static bool sample_id_all_avail = true; static bool system_wide = false; static pid_t target_pid = -1; static pid_t target_tid = -1; @@ -61,6 +62,7 @@ static bool call_graph = false; static bool inherit_stat = false; static bool no_samples = false; static bool sample_address = false; +static bool sample_time = false; static bool no_buildid = false; static bool no_buildid_cache = false; @@ -283,6 +285,9 @@ static void create_counter(int counter, int cpu) if (system_wide) attr->sample_type |= PERF_SAMPLE_CPU; + if (sample_time) + attr->sample_type |= PERF_SAMPLE_TIME; + if (raw_samples) { attr->sample_type |= PERF_SAMPLE_TIME; attr->sample_type |= PERF_SAMPLE_RAW; @@ -299,6 +304,8 @@ static void create_counter(int counter, int cpu) attr->disabled = 1; attr->enable_on_exec = 1; } +retry_sample_id: + attr->sample_id_all = sample_id_all_avail ? 1 : 0; for (thread_index = 0; thread_index < thread_num; thread_index++) { try_again: @@ -315,6 +322,12 @@ try_again: else if (err == ENODEV && cpu_list) { die("No such device - did you specify" " an out-of-range profile CPU?\n"); + } else if (err == EINVAL && sample_id_all_avail) { + /* + * Old kernel, no attr->sample_id_type_all field + */ + sample_id_all_avail = false; + goto retry_sample_id; } /* @@ -661,6 +674,8 @@ static int __cmd_record(int argc, const char **argv) post_processing_offset = lseek(output, 0, SEEK_CUR); + perf_session__set_sample_id_all(session, sample_id_all_avail); + if (pipe_output) { err = event__synthesize_attrs(&session->header, process_synthesized_event, @@ -841,6 +856,7 @@ const struct option record_options[] = { "per thread counts"), OPT_BOOLEAN('d', "data", &sample_address, "Sample addresses"), + OPT_BOOLEAN('T', "timestamp", &sample_time, "Sample timestamps"), OPT_BOOLEAN('n', "no-samples", &no_samples, "don't sample"), OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache, -- cgit From 18483b81ee7e70ee68d4b18be618be5cfcc0b290 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 6 Dec 2010 15:13:38 -0200 Subject: perf record: Fix eternal wait for stillborn child When execvp fails to find the specified command on the path we won't get SIGCHLD, so send a SIGUSR1 and exit right away. Current situation would require a SIGINT performed by the user and would produce meaningless summary. Now: [acme@emilia linux]$ ./foo -bash: ./foo: No such file or directory [acme@emilia linux]$ perf record ./foo ./foo: No such file or directory [acme@emilia linux]$ Acked-by: Thomas Gleixner Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi Cc: Thomas Gleixner LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e2c2de201eec..564491fa18b2 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -197,7 +197,7 @@ static void sig_atexit(void) if (child_pid > 0) kill(child_pid, SIGTERM); - if (signr == -1) + if (signr == -1 || signr == SIGUSR1) return; signal(signr, SIG_DFL); @@ -515,6 +515,7 @@ static int __cmd_record(int argc, const char **argv) atexit(sig_atexit); signal(SIGCHLD, sig_handler); signal(SIGINT, sig_handler); + signal(SIGUSR1, sig_handler); if (forks && (pipe(child_ready_pipe) < 0 || pipe(go_pipe) < 0)) { perror("failed to create pipes"); @@ -606,6 +607,7 @@ static int __cmd_record(int argc, const char **argv) execvp(argv[0], (char **)argv); perror(argv[0]); + kill(getppid(), SIGUSR1); exit(-1); } @@ -762,7 +764,7 @@ static int __cmd_record(int argc, const char **argv) } } - if (quiet) + if (quiet || signr == SIGUSR1) return 0; fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking); -- cgit From 21ef97f05a7da5bc23b26cb34d6746f83ca9bf20 Mon Sep 17 00:00:00 2001 From: Ian Munsie Date: Fri, 10 Dec 2010 14:09:16 +1100 Subject: perf session: Fallback to unordered processing if no sample_id_all If we are running the new perf on an old kernel without support for sample_id_all, we should fall back to the old unordered processing of events. If we didn't than we would *always* process events without timestamps out of order, whether or not we hit a reordering race. In other words, instead of there being a chance of not attributing samples correctly, we would guarantee that samples would not be attributed. While processing all events without timestamps before events with timestamps may seem like an intuitive solution, it falls down as PERF_RECORD_EXIT events would also be processed before any samples. Even with a workaround for that case, samples before/after an exec would not be attributed correctly. This patch allows commands to indicate whether they need to fall back to unordered processing, so that commands that do not care about timestamps on every event will not be affected. If we do fallback, this will print out a warning if report -D was invoked. This patch adds the test in perf_session__new so that we only need to test once per session. Commands that do not use an event_ops (such as record and top) can simply pass NULL in it's place. Acked-by: Thomas Gleixner Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Thomas Gleixner LKML-Reference: <1291951882-sup-6069@au1.ibm.com> Signed-off-by: Ian Munsie Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e9be6ae87a27..efd1b3c3d4a0 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -572,7 +572,7 @@ static int __cmd_record(int argc, const char **argv) } session = perf_session__new(output_name, O_WRONLY, - write_mode == WRITE_FORCE, false); + write_mode == WRITE_FORCE, false, NULL); if (session == NULL) { pr_err("Not enough memory for reading perf file header\n"); return -1; -- cgit From eac23d1c384b55e4bbb89ea9e5a6bb77fb4d1140 Mon Sep 17 00:00:00 2001 From: Ian Munsie Date: Thu, 9 Dec 2010 16:33:53 +1100 Subject: perf record,report,annotate,diff: Process events in order This patch changes perf report to ask for the ID info on all events be default if recording from multiple CPUs. Perf report, annotate and diff will now process the events in order if the kernel is able to provide timestamps on all events. This ensures that events such as COMM and MMAP which are necessary to correctly interpret samples are processed prior to those samples so that they are attributed correctly. Before: # perf record ./cachetest # perf report # Events: 6K cycles # # Overhead Command Shared Object Symbol # ........ ....... ................. ............................... # 74.11% :3259 [unknown] [k] 0x4a6c 1.50% cachetest ld-2.11.2.so [.] 0x1777c 1.46% :3259 [kernel.kallsyms] [k] .perf_event_mmap_ctx 1.25% :3259 [kernel.kallsyms] [k] restore 0.74% :3259 [kernel.kallsyms] [k] ._raw_spin_lock 0.71% :3259 [kernel.kallsyms] [k] .filemap_fault 0.66% :3259 [kernel.kallsyms] [k] .memset 0.54% cachetest [kernel.kallsyms] [k] .sha_transform 0.54% :3259 [kernel.kallsyms] [k] .copy_4K_page 0.54% :3259 [kernel.kallsyms] [k] .find_get_page 0.52% :3259 [kernel.kallsyms] [k] .trace_hardirqs_off 0.50% :3259 [kernel.kallsyms] [k] .__do_fault After: # perf report # Events: 6K cycles # # Overhead Command Shared Object Symbol # ........ ....... ................. ............................... # 44.28% cachetest cachetest [.] sumArrayNaive 22.53% cachetest cachetest [.] sumArrayOptimal 6.59% cachetest ld-2.11.2.so [.] 0x1777c 2.13% cachetest [unknown] [k] 0x340 1.46% cachetest [kernel.kallsyms] [k] .perf_event_mmap_ctx 1.25% cachetest [kernel.kallsyms] [k] restore 0.74% cachetest [kernel.kallsyms] [k] ._raw_spin_lock 0.71% cachetest [kernel.kallsyms] [k] .filemap_fault 0.66% cachetest [kernel.kallsyms] [k] .memset 0.54% cachetest [kernel.kallsyms] [k] .copy_4K_page 0.54% cachetest [kernel.kallsyms] [k] .find_get_page 0.54% cachetest [kernel.kallsyms] [k] .sha_transform 0.52% cachetest [kernel.kallsyms] [k] .trace_hardirqs_off 0.50% cachetest [kernel.kallsyms] [k] .__do_fault Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Thomas Gleixner LKML-Reference: <1291872833-839-1-git-send-email-imunsie@au1.ibm.com> Signed-off-by: Ian Munsie Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index efd1b3c3d4a0..5149e3deb7bc 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -285,7 +285,7 @@ static void create_counter(int counter, int cpu) if (system_wide) attr->sample_type |= PERF_SAMPLE_CPU; - if (sample_time) + if (sample_time || system_wide || !no_inherit || cpu_list) attr->sample_type |= PERF_SAMPLE_TIME; if (raw_samples) { @@ -327,6 +327,9 @@ try_again: * Old kernel, no attr->sample_id_type_all field */ sample_id_all_avail = false; + if (!sample_time && !raw_samples) + attr->sample_type &= ~PERF_SAMPLE_TIME; + goto retry_sample_id; } -- cgit From a43d3f08c64125edbdfdc3d3aa512d3e37321f37 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Sat, 25 Dec 2010 12:12:25 -0200 Subject: perf record: Fix use of sample_id_all userspace with !sample_id_all kernels Check if parse_single_tracepoint_event has already asked for PERF_SAMPLE_TIME. This is kludgy but short term fix for problems introduced by eac23d1c that broke 'perf script' by having different sample_types when using multiple tracepoint events when we use a perf binary that tries to use sample_id_all on an older kernel. We need to move counter creation to perf_session, support different sample_types, etc. Ongoing work on the perf test infrastructure needs this so that we can create counters to monitor threads generating specific events, etc. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi Cc: Torok Edwin Cc: Ian Munsie LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 24 +++++++++++++++++++----- 1 file changed, 19 insertions(+), 5 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 5149e3deb7bc..50efbd509b8f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -243,6 +243,19 @@ static void create_counter(int counter, int cpu) u64 time_running; u64 id; } read_data; + /* + * Check if parse_single_tracepoint_event has already asked for + * PERF_SAMPLE_TIME. + * + * XXX this is kludgy but short term fix for problems introduced by + * eac23d1c that broke 'perf script' by having different sample_types + * when using multiple tracepoint events when we use a perf binary + * that tries to use sample_id_all on an older kernel. + * + * We need to move counter creation to perf_session, support + * different sample_types, etc. + */ + bool time_needed = attr->sample_type & PERF_SAMPLE_TIME; attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | PERF_FORMAT_TOTAL_TIME_RUNNING | @@ -285,7 +298,8 @@ static void create_counter(int counter, int cpu) if (system_wide) attr->sample_type |= PERF_SAMPLE_CPU; - if (sample_time || system_wide || !no_inherit || cpu_list) + if (sample_id_all_avail && + (sample_time || system_wide || !no_inherit || cpu_list)) attr->sample_type |= PERF_SAMPLE_TIME; if (raw_samples) { @@ -294,9 +308,6 @@ static void create_counter(int counter, int cpu) attr->sample_type |= PERF_SAMPLE_CPU; } - if (!sample_type) - sample_type = attr->sample_type; - attr->mmap = track; attr->comm = track; attr->inherit = !no_inherit; @@ -327,7 +338,7 @@ try_again: * Old kernel, no attr->sample_id_type_all field */ sample_id_all_avail = false; - if (!sample_time && !raw_samples) + if (!sample_time && !raw_samples && !time_needed) attr->sample_type &= ~PERF_SAMPLE_TIME; goto retry_sample_id; @@ -428,6 +439,9 @@ try_again: } } } + + if (!sample_type) + sample_type = attr->sample_type; } static void open_counters(int cpu) -- cgit From 69aad6f1ee69546dea8535ab8f3da9f445d57328 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 3 Jan 2011 16:39:04 -0200 Subject: perf tools: Introduce event selectors Out of ad-hoc code and global arrays with hard coded sizes. This is the first step on having a library that will be first used on regression tests in the 'perf test' tool. [acme@felicio linux]$ size /tmp/perf.before text data bss dec hex filename 1273776 97384 5104416 6475576 62cf38 /tmp/perf.before [acme@felicio linux]$ size /tmp/perf.new text data bss dec hex filename 1275422 97416 1392416 2765254 2a31c6 /tmp/perf.new Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 113 +++++++++++++++++++++----------------------- 1 file changed, 54 insertions(+), 59 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 50efbd509b8f..e68aee33bc19 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -18,6 +18,7 @@ #include "util/header.h" #include "util/event.h" +#include "util/evsel.h" #include "util/debug.h" #include "util/session.h" #include "util/symbol.h" @@ -27,13 +28,13 @@ #include #include +#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y)) + enum write_mode_t { WRITE_FORCE, WRITE_APPEND }; -static int *fd[MAX_NR_CPUS][MAX_COUNTERS]; - static u64 user_interval = ULLONG_MAX; static u64 default_interval = 0; static u64 sample_type; @@ -81,7 +82,6 @@ static struct perf_session *session; static const char *cpu_list; struct mmap_data { - int counter; void *base; unsigned int mask; unsigned int prev; @@ -229,12 +229,12 @@ static struct perf_header_attr *get_header_attr(struct perf_event_attr *a, int n return h_attr; } -static void create_counter(int counter, int cpu) +static void create_counter(struct perf_evsel *evsel, int cpu) { - char *filter = filters[counter]; - struct perf_event_attr *attr = attrs + counter; + char *filter = evsel->filter; + struct perf_event_attr *attr = &evsel->attr; struct perf_header_attr *h_attr; - int track = !counter; /* only the first counter needs these */ + int track = !evsel->idx; /* only the first counter needs these */ int thread_index; int ret; struct { @@ -320,10 +320,9 @@ retry_sample_id: for (thread_index = 0; thread_index < thread_num; thread_index++) { try_again: - fd[nr_cpu][counter][thread_index] = sys_perf_event_open(attr, - all_tids[thread_index], cpu, group_fd, 0); + FD(evsel, nr_cpu, thread_index) = sys_perf_event_open(attr, all_tids[thread_index], cpu, group_fd, 0); - if (fd[nr_cpu][counter][thread_index] < 0) { + if (FD(evsel, nr_cpu, thread_index) < 0) { int err = errno; if (err == EPERM || err == EACCES) @@ -360,7 +359,7 @@ try_again: } printf("\n"); error("sys_perf_event_open() syscall returned with %d (%s). /bin/dmesg may provide additional information.\n", - fd[nr_cpu][counter][thread_index], strerror(err)); + FD(evsel, nr_cpu, thread_index), strerror(err)); #if defined(__i386__) || defined(__x86_64__) if (attr->type == PERF_TYPE_HARDWARE && err == EOPNOTSUPP) @@ -374,7 +373,7 @@ try_again: exit(-1); } - h_attr = get_header_attr(attr, counter); + h_attr = get_header_attr(attr, evsel->idx); if (h_attr == NULL) die("nomem\n"); @@ -385,7 +384,7 @@ try_again: } } - if (read(fd[nr_cpu][counter][thread_index], &read_data, sizeof(read_data)) == -1) { + if (read(FD(evsel, nr_cpu, thread_index), &read_data, sizeof(read_data)) == -1) { perror("Unable to read perf file descriptor"); exit(-1); } @@ -395,43 +394,44 @@ try_again: exit(-1); } - assert(fd[nr_cpu][counter][thread_index] >= 0); - fcntl(fd[nr_cpu][counter][thread_index], F_SETFL, O_NONBLOCK); + assert(FD(evsel, nr_cpu, thread_index) >= 0); + fcntl(FD(evsel, nr_cpu, thread_index), F_SETFL, O_NONBLOCK); /* * First counter acts as the group leader: */ if (group && group_fd == -1) - group_fd = fd[nr_cpu][counter][thread_index]; - - if (counter || thread_index) { - ret = ioctl(fd[nr_cpu][counter][thread_index], - PERF_EVENT_IOC_SET_OUTPUT, - fd[nr_cpu][0][0]); + group_fd = FD(evsel, nr_cpu, thread_index); + + if (evsel->idx || thread_index) { + struct perf_evsel *first; + first = list_entry(evsel_list.next, struct perf_evsel, node); + ret = ioctl(FD(evsel, nr_cpu, thread_index), + PERF_EVENT_IOC_SET_OUTPUT, + FD(first, nr_cpu, 0)); if (ret) { error("failed to set output: %d (%s)\n", errno, strerror(errno)); exit(-1); } } else { - mmap_array[nr_cpu].counter = counter; mmap_array[nr_cpu].prev = 0; mmap_array[nr_cpu].mask = mmap_pages*page_size - 1; mmap_array[nr_cpu].base = mmap(NULL, (mmap_pages+1)*page_size, - PROT_READ|PROT_WRITE, MAP_SHARED, fd[nr_cpu][counter][thread_index], 0); + PROT_READ | PROT_WRITE, MAP_SHARED, FD(evsel, nr_cpu, thread_index), 0); if (mmap_array[nr_cpu].base == MAP_FAILED) { error("failed to mmap with %d (%s)\n", errno, strerror(errno)); exit(-1); } - event_array[nr_poll].fd = fd[nr_cpu][counter][thread_index]; + event_array[nr_poll].fd = FD(evsel, nr_cpu, thread_index); event_array[nr_poll].events = POLLIN; nr_poll++; } if (filter != NULL) { - ret = ioctl(fd[nr_cpu][counter][thread_index], - PERF_EVENT_IOC_SET_FILTER, filter); + ret = ioctl(FD(evsel, nr_cpu, thread_index), + PERF_EVENT_IOC_SET_FILTER, filter); if (ret) { error("failed to set filter with %d (%s)\n", errno, strerror(errno)); @@ -446,11 +446,12 @@ try_again: static void open_counters(int cpu) { - int counter; + struct perf_evsel *pos; group_fd = -1; - for (counter = 0; counter < nr_counters; counter++) - create_counter(counter, cpu); + + list_for_each_entry(pos, &evsel_list, node) + create_counter(pos, cpu); nr_cpu++; } @@ -537,7 +538,7 @@ static void mmap_read_all(void) static int __cmd_record(int argc, const char **argv) { - int i, counter; + int i; struct stat st; int flags; int err; @@ -604,7 +605,7 @@ static int __cmd_record(int argc, const char **argv) goto out_delete_session; } - if (have_tracepoints(attrs, nr_counters)) + if (have_tracepoints(&evsel_list)) perf_header__set_feat(&session->header, HEADER_TRACE_INFO); /* @@ -666,12 +667,6 @@ static int __cmd_record(int argc, const char **argv) close(child_ready_pipe[0]); } - nr_cpus = read_cpu_map(cpu_list); - if (nr_cpus < 1) { - perror("failed to collect number of CPUs"); - return -1; - } - if (!system_wide && no_inherit && !cpu_list) { open_counters(-1); } else { @@ -711,7 +706,7 @@ static int __cmd_record(int argc, const char **argv) return err; } - if (have_tracepoints(attrs, nr_counters)) { + if (have_tracepoints(&evsel_list)) { /* * FIXME err <= 0 here actually means that * there were no tracepoints so its not really @@ -720,8 +715,7 @@ static int __cmd_record(int argc, const char **argv) * return this more properly and also * propagate errors that now are calling die() */ - err = event__synthesize_tracing_data(output, attrs, - nr_counters, + err = event__synthesize_tracing_data(output, &evsel_list, process_synthesized_event, session); if (err <= 0) { @@ -795,13 +789,13 @@ static int __cmd_record(int argc, const char **argv) if (done) { for (i = 0; i < nr_cpu; i++) { - for (counter = 0; - counter < nr_counters; - counter++) { + struct perf_evsel *pos; + + list_for_each_entry(pos, &evsel_list, node) { for (thread = 0; thread < thread_num; thread++) - ioctl(fd[i][counter][thread], + ioctl(FD(pos, i, thread), PERF_EVENT_IOC_DISABLE); } } @@ -887,7 +881,8 @@ const struct option record_options[] = { int cmd_record(int argc, const char **argv, const char *prefix __used) { - int i, j, err = -ENOMEM; + int err = -ENOMEM; + struct perf_evsel *pos; argc = parse_options(argc, argv, record_options, record_usage, PARSE_OPT_STOP_AT_NON_OPTION); @@ -910,10 +905,9 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) if (no_buildid_cache || no_buildid) disable_buildid_cache(); - if (!nr_counters) { - nr_counters = 1; - attrs[0].type = PERF_TYPE_HARDWARE; - attrs[0].config = PERF_COUNT_HW_CPU_CYCLES; + if (list_empty(&evsel_list) && perf_evsel_list__create_default() < 0) { + pr_err("Not enough memory for event selector list\n"); + goto out_symbol_exit; } if (target_pid != -1) { @@ -933,12 +927,15 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) thread_num = 1; } - for (i = 0; i < MAX_NR_CPUS; i++) { - for (j = 0; j < MAX_COUNTERS; j++) { - fd[i][j] = malloc(sizeof(int)*thread_num); - if (!fd[i][j]) - goto out_free_fd; - } + nr_cpus = read_cpu_map(cpu_list); + if (nr_cpus < 1) { + perror("failed to collect number of CPUs"); + return -1; + } + + list_for_each_entry(pos, &evsel_list, node) { + if (perf_evsel__alloc_fd(pos, nr_cpus, thread_num) < 0) + goto out_free_fd; } event_array = malloc( sizeof(struct pollfd)*MAX_NR_CPUS*MAX_COUNTERS*thread_num); @@ -968,10 +965,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) out_free_event_array: free(event_array); out_free_fd: - for (i = 0; i < MAX_NR_CPUS; i++) { - for (j = 0; j < MAX_COUNTERS; j++) - free(fd[i][j]); - } + list_for_each_entry(pos, &evsel_list, node) + perf_evsel__free_fd(pos); free(all_tids); all_tids = NULL; out_symbol_exit: -- cgit From 70d544d0576775a2b3923a7e68cb49b0313d80c9 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 3 Jan 2011 16:51:39 -0200 Subject: perf evsel: Delete the event selectors at exit Freeing all the possibly allocated resources, reducing complexity on each tool exit path. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e68aee33bc19..052de1780f76 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -965,8 +965,6 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) out_free_event_array: free(event_array); out_free_fd: - list_for_each_entry(pos, &evsel_list, node) - perf_evsel__free_fd(pos); free(all_tids); all_tids = NULL; out_symbol_exit: -- cgit From 60d567e2d9187379d642f6aba7c8a52b3fd5d261 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 3 Jan 2011 17:49:48 -0200 Subject: perf tools: Refactor cpumap to hold nr and the map So that later, we can pass the cpu_map instance instead of (nr_cpus, cpu_map) for things like perf_evsel__open and friends. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 052de1780f76..220e6e7f0b97 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -39,7 +39,7 @@ static u64 user_interval = ULLONG_MAX; static u64 default_interval = 0; static u64 sample_type; -static int nr_cpus = 0; +static struct cpu_map *cpus; static unsigned int page_size; static unsigned int mmap_pages = 128; static unsigned int user_freq = UINT_MAX; @@ -670,8 +670,8 @@ static int __cmd_record(int argc, const char **argv) if (!system_wide && no_inherit && !cpu_list) { open_counters(-1); } else { - for (i = 0; i < nr_cpus; i++) - open_counters(cpumap[i]); + for (i = 0; i < cpus->nr; i++) + open_counters(cpus->map[i]); } perf_session__set_sample_type(session, sample_type); @@ -927,14 +927,14 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) thread_num = 1; } - nr_cpus = read_cpu_map(cpu_list); - if (nr_cpus < 1) { - perror("failed to collect number of CPUs"); + cpus = cpu_map__new(cpu_list); + if (cpus == NULL) { + perror("failed to parse CPUs map"); return -1; } list_for_each_entry(pos, &evsel_list, node) { - if (perf_evsel__alloc_fd(pos, nr_cpus, thread_num) < 0) + if (perf_evsel__alloc_fd(pos, cpus->nr, thread_num) < 0) goto out_free_fd; } event_array = malloc( -- cgit From 5c98d466e49267a9221f30958d45cd06f794269a Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Mon, 3 Jan 2011 17:53:33 -0200 Subject: perf tools: Refactor all_tids to hold nr and the map So that later, we can pass the thread_map instance instead of (thread_num, thread_map) for things like perf_evsel__open and friends, just like was done with cpu_map. Cc: Frederic Weisbecker Cc: Ingo Molnar Cc: Mike Galbraith Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Tom Zanussi LKML-Reference: Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/builtin-record.c | 39 +++++++++++++++------------------------ 1 file changed, 15 insertions(+), 24 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 220e6e7f0b97..7bc049035484 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -54,8 +54,7 @@ static bool sample_id_all_avail = true; static bool system_wide = false; static pid_t target_pid = -1; static pid_t target_tid = -1; -static pid_t *all_tids = NULL; -static int thread_num = 0; +static struct thread_map *threads; static pid_t child_pid = -1; static bool no_inherit = false; static enum write_mode_t write_mode = WRITE_FORCE; @@ -318,9 +317,9 @@ static void create_counter(struct perf_evsel *evsel, int cpu) retry_sample_id: attr->sample_id_all = sample_id_all_avail ? 1 : 0; - for (thread_index = 0; thread_index < thread_num; thread_index++) { + for (thread_index = 0; thread_index < threads->nr; thread_index++) { try_again: - FD(evsel, nr_cpu, thread_index) = sys_perf_event_open(attr, all_tids[thread_index], cpu, group_fd, 0); + FD(evsel, nr_cpu, thread_index) = sys_perf_event_open(attr, threads->map[thread_index], cpu, group_fd, 0); if (FD(evsel, nr_cpu, thread_index) < 0) { int err = errno; @@ -653,7 +652,7 @@ static int __cmd_record(int argc, const char **argv) } if (!system_wide && target_tid == -1 && target_pid == -1) - all_tids[0] = child_pid; + threads->map[0] = child_pid; close(child_ready_pipe[1]); close(go_pipe[0]); @@ -793,7 +792,7 @@ static int __cmd_record(int argc, const char **argv) list_for_each_entry(pos, &evsel_list, node) { for (thread = 0; - thread < thread_num; + thread < threads->nr; thread++) ioctl(FD(pos, i, thread), PERF_EVENT_IOC_DISABLE); @@ -910,21 +909,13 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) goto out_symbol_exit; } - if (target_pid != -1) { + if (target_pid != -1) target_tid = target_pid; - thread_num = find_all_tid(target_pid, &all_tids); - if (thread_num <= 0) { - fprintf(stderr, "Can't find all threads of pid %d\n", - target_pid); - usage_with_options(record_usage, record_options); - } - } else { - all_tids=malloc(sizeof(pid_t)); - if (!all_tids) - goto out_symbol_exit; - all_tids[0] = target_tid; - thread_num = 1; + threads = thread_map__new(target_pid, target_tid); + if (threads == NULL) { + pr_err("Problems finding threads of monitor\n"); + usage_with_options(record_usage, record_options); } cpus = cpu_map__new(cpu_list); @@ -934,11 +925,11 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) } list_for_each_entry(pos, &evsel_list, node) { - if (perf_evsel__alloc_fd(pos, cpus->nr, thread_num) < 0) + if (perf_evsel__alloc_fd(pos, cpus->nr, threads->nr) < 0) goto out_free_fd; } - event_array = malloc( - sizeof(struct pollfd)*MAX_NR_CPUS*MAX_COUNTERS*thread_num); + event_array = malloc((sizeof(struct pollfd) * MAX_NR_CPUS * + MAX_COUNTERS * threads->nr)); if (!event_array) goto out_free_fd; @@ -965,8 +956,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) out_free_event_array: free(event_array); out_free_fd: - free(all_tids); - all_tids = NULL; + thread_map__delete(threads); + threads = NULL; out_symbol_exit: symbol__exit(); return err; -- cgit