From 3efa1cc99ec51bc7a7ae0011a16619fd20dbe6ea Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 14 Jun 2009 15:04:15 +0200 Subject: perf record/report: Add call graph / call chain profiling Add the first steps of call-graph profiling: - add the -c (--call-graph) option to perf record - parse the call-graph record and printout out under -D (--dump-trace) The call-graph data is not put into the histogram yet, but it can be seen that it's being processed correctly: 0x3ce0 [0x38]: event: 35 . . ... raw event: size 56 bytes . 0000: 23 00 00 00 05 00 38 00 d4 df 0e 81 ff ff ff ff #.....8........ . 0010: 60 0b 00 00 60 0b 00 00 03 00 00 00 01 00 02 00 `...`.......... . 0020: d4 df 0e 81 ff ff ff ff a0 61 ed 41 36 00 00 00 .........a.A6.. . 0030: 04 92 e6 41 36 00 00 00 .a.A6.. . 0x3ce0 [0x38]: PERF_EVENT (IP, 5): 2912: 0xffffffff810edfd4 period: 1 ... chain: u:2, k:1, nr:3 ..... 0: 0xffffffff810edfd4 ..... 1: 0x3641ed61a0 ..... 2: 0x3641e69204 ... thread: perf:2912 ...... dso: [kernel] This shows a 3-entry call-graph: with 1 kernel-space and two user-space entries Cc: Frederic Weisbecker Cc: Pekka Enberg Cc: Arjan van de Ven Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 0f5771f615da..a177a591b52c 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -37,6 +37,7 @@ static pid_t target_pid = -1; static int inherit = 1; static int force = 0; static int append_file = 0; +static int call_graph = 0; static int verbose = 0; static long samples; @@ -351,11 +352,16 @@ static void create_counter(int counter, int cpu, pid_t pid) int track = 1; attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; + if (freq) { attr->sample_type |= PERF_SAMPLE_PERIOD; attr->freq = 1; attr->sample_freq = freq; } + + if (call_graph) + attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + attr->mmap = track; attr->comm = track; attr->inherit = (cpu < 0) && inherit; @@ -555,6 +561,8 @@ static const struct option options[] = { "profile at this frequency"), OPT_INTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"), + OPT_BOOLEAN('g', "call-graph", &call_graph, + "do call-graph (stack chain/backtrace) recording"), OPT_BOOLEAN('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), OPT_END() -- cgit v1.2.3 From 613d8602292165f86ba1969784fea01a06d55900 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 15 Jun 2009 08:17:12 +0200 Subject: perf record: Fix fast task-exit race Recording with -a (or with -p) can race with tasks going away: couldn't open /proc/8440/maps Causing an early exit() and no recording done. Do not abort the recording session - instead just skip that task. Also, only print the warnings under -v. Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index a177a591b52c..e1dfef24887f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -202,8 +202,12 @@ static void pid_synthesize_comm_event(pid_t pid, int full) fd = open(filename, O_RDONLY); if (fd < 0) { - fprintf(stderr, "couldn't open %s\n", filename); - exit(EXIT_FAILURE); + /* + * We raced with a task exiting - just return: + */ + if (verbose) + fprintf(stderr, "couldn't open %s\n", filename); + return; } if (read(fd, bf, sizeof(bf)) < 0) { fprintf(stderr, "couldn't read %s\n", filename); @@ -273,8 +277,12 @@ static void pid_synthesize_mmap_samples(pid_t pid) fp = fopen(filename, "r"); if (fp == NULL) { - fprintf(stderr, "couldn't open %s\n", filename); - exit(EXIT_FAILURE); + /* + * We raced with a task exiting - just return: + */ + if (verbose) + fprintf(stderr, "couldn't open %s\n", filename); + return; } while (1) { char bf[BUFSIZ], *pbf = bf; -- cgit v1.2.3 From 9d91a6f7a489eb914c16b82d927f9d81d629c259 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 18 Jun 2009 11:40:28 +0200 Subject: perf_counter tools: Handle lost events Make use of the new ->data_tail mechanism to tell kernel-space about user-space draining the data stream. Emit lost events (and display them) if they happen. Signed-off-by: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 20 ++++++++++++++++---- 1 file changed, 16 insertions(+), 4 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index e1dfef24887f..06fdfb8b4828 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -78,10 +78,10 @@ struct mmap_data { static struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS]; -static unsigned int mmap_read_head(struct mmap_data *md) +static unsigned long mmap_read_head(struct mmap_data *md) { struct perf_counter_mmap_page *pc = md->base; - int head; + long head; head = pc->data_head; rmb(); @@ -89,6 +89,17 @@ static unsigned int mmap_read_head(struct mmap_data *md) return head; } +static void mmap_write_tail(struct mmap_data *md, unsigned long tail) +{ + struct perf_counter_mmap_page *pc = md->base; + + /* + * ensure all reads are done before we write the tail out. + */ + /* mb(); */ + pc->data_tail = tail; +} + static void mmap_read(struct mmap_data *md) { unsigned int head = mmap_read_head(md); @@ -109,7 +120,7 @@ static void mmap_read(struct mmap_data *md) * In either case, truncate and restart at head. */ diff = head - old; - if (diff > md->mask / 2 || diff < 0) { + if (diff < 0) { struct timeval iv; unsigned long msecs; @@ -167,6 +178,7 @@ static void mmap_read(struct mmap_data *md) } md->prev = old; + mmap_write_tail(md, old); } static volatile int done = 0; @@ -424,7 +436,7 @@ try_again: mmap_array[nr_cpu][counter].prev = 0; mmap_array[nr_cpu][counter].mask = mmap_pages*page_size - 1; mmap_array[nr_cpu][counter].base = mmap(NULL, (mmap_pages+1)*page_size, - PROT_READ, MAP_SHARED, fd[nr_cpu][counter], 0); + PROT_READ|PROT_WRITE, MAP_SHARED, fd[nr_cpu][counter], 0); if (mmap_array[nr_cpu][counter].base == MAP_FAILED) { error("failed to mmap with %d (%s)\n", errno, strerror(errno)); exit(-1); -- cgit v1.2.3 From f5970550d5ccf90453cbd7d260370ea99d1f6513 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 18 Jun 2009 23:22:55 +0200 Subject: perf_counter tools: Add a data file header Add a data file header so we can transfer data between record and report. LKML-Reference: Signed-off-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 94 +++++++++++++++++++++++++-------------------- 1 file changed, 52 insertions(+), 42 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 06fdfb8b4828..28304677c73e 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -51,6 +51,9 @@ static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS]; static int nr_poll; static int nr_cpu; +static int file_new = 1; +static struct perf_file_header file_header; + struct mmap_event { struct perf_event_header header; __u32 pid; @@ -100,6 +103,21 @@ static void mmap_write_tail(struct mmap_data *md, unsigned long tail) pc->data_tail = tail; } +static void write_output(void *buf, size_t size) +{ + while (size) { + int ret = write(output, buf, size); + + if (ret < 0) + die("failed to write"); + + size -= ret; + buf += ret; + + bytes_written += ret; + } +} + static void mmap_read(struct mmap_data *md) { unsigned int head = mmap_read_head(md); @@ -148,34 +166,14 @@ static void mmap_read(struct mmap_data *md) size = md->mask + 1 - (old & md->mask); old += size; - while (size) { - int ret = write(output, buf, size); - - if (ret < 0) - die("failed to write"); - - size -= ret; - buf += ret; - - bytes_written += ret; - } + write_output(buf, size); } buf = &data[old & md->mask]; size = head - old; old += size; - while (size) { - int ret = write(output, buf, size); - - if (ret < 0) - die("failed to write"); - - size -= ret; - buf += ret; - - bytes_written += ret; - } + write_output(buf, size); md->prev = old; mmap_write_tail(md, old); @@ -204,7 +202,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) struct comm_event comm_ev; char filename[PATH_MAX]; char bf[BUFSIZ]; - int fd, ret; + int fd; size_t size; char *field, *sep; DIR *tasks; @@ -246,11 +244,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) if (!full) { comm_ev.tid = pid; - ret = write(output, &comm_ev, comm_ev.header.size); - if (ret < 0) { - perror("failed to write"); - exit(-1); - } + write_output(&comm_ev, comm_ev.header.size); return; } @@ -265,11 +259,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) comm_ev.tid = pid; - ret = write(output, &comm_ev, comm_ev.header.size); - if (ret < 0) { - perror("failed to write"); - exit(-1); - } + write_output(&comm_ev, comm_ev.header.size); } closedir(tasks); return; @@ -332,10 +322,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) mmap_ev.pid = pid; mmap_ev.tid = pid; - if (write(output, &mmap_ev, mmap_ev.header.size) < 0) { - perror("failed to write"); - exit(-1); - } + write_output(&mmap_ev, mmap_ev.header.size); } } @@ -382,6 +369,15 @@ static void create_counter(int counter, int cpu, pid_t pid) if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + if (file_new) { + file_header.sample_type = attr->sample_type; + } else { + if (file_header.sample_type != attr->sample_type) { + fprintf(stderr, "incompatible append\n"); + exit(-1); + } + } + attr->mmap = track; attr->comm = track; attr->inherit = (cpu < 0) && inherit; @@ -461,6 +457,13 @@ static void open_counters(int cpu, pid_t pid) nr_cpu++; } +static void atexit_header(void) +{ + file_header.data_size += bytes_written; + + pwrite(output, &file_header, sizeof(file_header), 0); +} + static int __cmd_record(int argc, const char **argv) { int i, counter; @@ -474,6 +477,10 @@ static int __cmd_record(int argc, const char **argv) assert(nr_cpus <= MAX_NR_CPUS); assert(nr_cpus >= 0); + atexit(sig_atexit); + signal(SIGCHLD, sig_handler); + signal(SIGINT, sig_handler); + if (!stat(output_name, &st) && !force && !append_file) { fprintf(stderr, "Error, output file %s exists, use -A to append or -f to overwrite.\n", output_name); @@ -482,7 +489,7 @@ static int __cmd_record(int argc, const char **argv) flags = O_CREAT|O_RDWR; if (append_file) - flags |= O_APPEND; + file_new = 0; else flags |= O_TRUNC; @@ -492,15 +499,18 @@ static int __cmd_record(int argc, const char **argv) exit(-1); } + if (!file_new) { + read(output, &file_header, sizeof(file_header)); + lseek(output, file_header.data_size, SEEK_CUR); + } + + atexit(atexit_header); + if (!system_wide) { open_counters(-1, target_pid != -1 ? target_pid : getpid()); } else for (i = 0; i < nr_cpus; i++) open_counters(i, target_pid); - atexit(sig_atexit); - signal(SIGCHLD, sig_handler); - signal(SIGINT, sig_handler); - if (target_pid == -1 && argc) { pid = fork(); if (pid < 0) -- cgit v1.2.3 From 9cffa8d53335d891cc0ecb3824a67118b3ee4b2f Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Fri, 19 Jun 2009 22:21:42 +1000 Subject: perf_counter tools: Define and use our own u64, s64 etc. definitions On 64-bit powerpc, __u64 is defined to be unsigned long rather than unsigned long long. This causes compiler warnings every time we print a __u64 value with %Lx. Rather than changing __u64, we define our own u64 to be unsigned long long on all architectures, and similarly s64 as signed long long. For consistency we also define u32, s32, u16, s16, u8 and s8. These definitions are put in a new header, types.h, because these definitions are needed in util/string.h and util/symbol.h. The main change here is the mechanical change of __[us]{64,32,16,8} to remove the "__". The other changes are: * Create types.h * Include types.h in perf.h, util/string.h and util/symbol.h * Add types.h to the LIB_H definition in Makefile * Added (u64) casts in process_overflow_event() and print_sym_table() to kill two remaining warnings. Signed-off-by: Paul Mackerras Acked-by: Peter Zijlstra Cc: benh@kernel.crashing.org LKML-Reference: <19003.33494.495844.956580@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 28304677c73e..e2cebc053bd7 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -44,7 +44,7 @@ static long samples; static struct timeval last_read; static struct timeval this_read; -static __u64 bytes_written; +static u64 bytes_written; static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS]; @@ -56,18 +56,18 @@ static struct perf_file_header file_header; struct mmap_event { struct perf_event_header header; - __u32 pid; - __u32 tid; - __u64 start; - __u64 len; - __u64 pgoff; + u32 pid; + u32 tid; + u64 start; + u64 len; + u64 pgoff; char filename[PATH_MAX]; }; struct comm_event { struct perf_event_header header; - __u32 pid; - __u32 tid; + u32 pid; + u32 tid; char comm[16]; }; @@ -238,7 +238,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) comm_ev.pid = pid; comm_ev.header.type = PERF_EVENT_COMM; - size = ALIGN(size, sizeof(__u64)); + size = ALIGN(size, sizeof(u64)); comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size); if (!full) { @@ -315,7 +315,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) size = strlen(execname); execname[size - 1] = '\0'; /* Remove \n */ memcpy(mmap_ev.filename, execname, size); - size = ALIGN(size, sizeof(__u64)); + size = ALIGN(size, sizeof(u64)); mmap_ev.len -= mmap_ev.start; mmap_ev.header.size = (sizeof(mmap_ev) - (sizeof(mmap_ev.filename) - size)); -- cgit v1.2.3 From eadc84cc01e04f9f74ec2de0c9355be035c7b396 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 20 Jun 2009 02:01:40 +0200 Subject: perfcounter: Handle some IO return values MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Building perfcounter tools raises the following warnings: builtin-record.c: In function ‘atexit_header’: builtin-record.c:464: erreur: ignoring return value of ‘pwrite’, declared with attribute warn_unused_result builtin-record.c: In function ‘__cmd_record’: builtin-record.c:503: erreur: ignoring return value of ‘read’, declared with attribute warn_unused_result builtin-report.c: In function ‘__cmd_report’: builtin-report.c:1403: erreur: ignoring return value of ‘read’, declared with attribute warn_unused_result This patch handles these IO return values. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1245456100-5477-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 9 +++++++-- 1 file changed, 7 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 e2cebc053bd7..d7ebbd757543 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -461,7 +461,8 @@ static void atexit_header(void) { file_header.data_size += bytes_written; - pwrite(output, &file_header, sizeof(file_header), 0); + if (pwrite(output, &file_header, sizeof(file_header), 0) == -1) + perror("failed to write on file headers"); } static int __cmd_record(int argc, const char **argv) @@ -500,7 +501,11 @@ static int __cmd_record(int argc, const char **argv) } if (!file_new) { - read(output, &file_header, sizeof(file_header)); + if (read(output, &file_header, sizeof(file_header)) == -1) { + perror("failed to read file headers"); + exit(-1); + } + lseek(output, file_header.data_size, SEEK_CUR); } -- cgit v1.2.3 From 76c64c5e4c47b6d28deb3cae8dfa07a93c2229dc Mon Sep 17 00:00:00 2001 From: Johannes Weiner Date: Wed, 24 Jun 2009 21:08:36 +0200 Subject: perf record: Fix filemap pathname parsing in /proc/pid/maps Looking backward for the first space from the end of a line in /proc/pid/maps does not find the start of the pathname of the mapped file if it contains a space. Since the only slashes we have in this file occur in the (absolute!) pathname column of file mappings, looking for the first slash in a line is a safe method to find the name. Signed-off-by: Johannes Weiner Cc: Stefani Seibold Cc: Andrew Morton Cc: "Eric W. Biederman" Cc: Alexey Dobriyan Cc: Peter Zijlstra LKML-Reference: <20090624190835.GA25548@cmpxchg.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 ++--- 1 file changed, 2 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 d7ebbd757543..9b899ba1b410 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -306,12 +306,11 @@ static void pid_synthesize_mmap_samples(pid_t pid) continue; pbf += n + 3; if (*pbf == 'x') { /* vm_exec */ - char *execname = strrchr(bf, ' '); + char *execname = strchr(bf, '/'); - if (execname == NULL || execname[1] != '/') + if (execname == NULL) continue; - execname += 1; size = strlen(execname); execname[size - 1] = '\0'; /* Remove \n */ memcpy(mmap_ev.filename, execname, size); -- cgit v1.2.3 From 7c6a1c65bbd3be688e581511f45818663efc1877 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 25 Jun 2009 17:05:54 +0200 Subject: perf_counter tools: Rework the file format Create a structured file format that includes the full perf_counter_attr and all its relevant counter IDs so that the reporting program has full information. Signed-off-by: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 100 ++++++++++++++++++++++++++++---------------- 1 file changed, 65 insertions(+), 35 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 9b899ba1b410..f4f0240d2302 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -14,6 +14,8 @@ #include "util/parse-events.h" #include "util/string.h" +#include "util/header.h" + #include #include @@ -52,7 +54,8 @@ static int nr_poll; static int nr_cpu; static int file_new = 1; -static struct perf_file_header file_header; + +struct perf_header *header; struct mmap_event { struct perf_event_header header; @@ -328,7 +331,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) fclose(fp); } -static void synthesize_samples(void) +static void synthesize_all(void) { DIR *proc; struct dirent dirent, *next; @@ -352,10 +355,35 @@ static void synthesize_samples(void) static int group_fd; +static struct perf_header_attr *get_header_attr(struct perf_counter_attr *a, int nr) +{ + struct perf_header_attr *h_attr; + + if (nr < header->attrs) { + h_attr = header->attr[nr]; + } else { + h_attr = perf_header_attr__new(a); + perf_header__add_attr(header, h_attr); + } + + return h_attr; +} + static void create_counter(int counter, int cpu, pid_t pid) { struct perf_counter_attr *attr = attrs + counter; - int track = 1; + struct perf_header_attr *h_attr; + int track = !counter; /* only the first counter needs these */ + struct { + u64 count; + u64 time_enabled; + u64 time_running; + u64 id; + } read_data; + + attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED | + PERF_FORMAT_TOTAL_TIME_RUNNING | + PERF_FORMAT_ID; attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; @@ -368,22 +396,11 @@ static void create_counter(int counter, int cpu, pid_t pid) if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; - if (file_new) { - file_header.sample_type = attr->sample_type; - } else { - if (file_header.sample_type != attr->sample_type) { - fprintf(stderr, "incompatible append\n"); - exit(-1); - } - } - attr->mmap = track; attr->comm = track; attr->inherit = (cpu < 0) && inherit; attr->disabled = 1; - track = 0; /* only the first counter needs these */ - try_again: fd[nr_cpu][counter] = sys_perf_counter_open(attr, pid, cpu, group_fd, 0); @@ -414,6 +431,19 @@ try_again: exit(-1); } + h_attr = get_header_attr(attr, counter); + + if (!file_new) { + if (memcmp(&h_attr->attr, attr, sizeof(*attr))) { + fprintf(stderr, "incompatible append\n"); + exit(-1); + } + } + + read(fd[nr_cpu][counter], &read_data, sizeof(read_data)); + + perf_header_attr__add_id(h_attr, read_data.id); + assert(fd[nr_cpu][counter] >= 0); fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK); @@ -444,11 +474,6 @@ static void open_counters(int cpu, pid_t pid) { int counter; - if (pid > 0) { - pid_synthesize_comm_event(pid, 0); - pid_synthesize_mmap_samples(pid); - } - group_fd = -1; for (counter = 0; counter < nr_counters; counter++) create_counter(counter, cpu, pid); @@ -458,17 +483,16 @@ static void open_counters(int cpu, pid_t pid) static void atexit_header(void) { - file_header.data_size += bytes_written; + header->data_size += bytes_written; - if (pwrite(output, &file_header, sizeof(file_header), 0) == -1) - perror("failed to write on file headers"); + perf_header__write(header, output); } static int __cmd_record(int argc, const char **argv) { int i, counter; struct stat st; - pid_t pid; + pid_t pid = 0; int flags; int ret; @@ -499,22 +523,31 @@ static int __cmd_record(int argc, const char **argv) exit(-1); } - if (!file_new) { - if (read(output, &file_header, sizeof(file_header)) == -1) { - perror("failed to read file headers"); - exit(-1); - } - - lseek(output, file_header.data_size, SEEK_CUR); - } + if (!file_new) + header = perf_header__read(output); + else + header = perf_header__new(); atexit(atexit_header); if (!system_wide) { - open_counters(-1, target_pid != -1 ? target_pid : getpid()); + pid = target_pid; + if (pid == -1) + pid = getpid(); + + open_counters(-1, pid); } else for (i = 0; i < nr_cpus; i++) open_counters(i, target_pid); + if (file_new) + perf_header__write(header, output); + + if (!system_wide) { + pid_synthesize_comm_event(pid, 0); + pid_synthesize_mmap_samples(pid); + } else + synthesize_all(); + if (target_pid == -1 && argc) { pid = fork(); if (pid < 0) @@ -538,9 +571,6 @@ static int __cmd_record(int argc, const char **argv) } } - if (system_wide) - synthesize_samples(); - while (!done) { int hits = samples; -- cgit v1.2.3 From 649c48a9e7fafcc72bfcc99471d9dea98d789d59 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Wed, 24 Jun 2009 21:12:48 +0200 Subject: perf-report: Add modes for inherited stats and no-samples Now that we can collect per task statistics, add modes that make use of that facility. Signed-off-by: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 19 +++++++++++++++++-- 1 file changed, 17 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 f4f0240d2302..798a56d890e5 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -41,6 +41,8 @@ static int force = 0; static int append_file = 0; static int call_graph = 0; static int verbose = 0; +static int inherit_stat = 0; +static int no_samples = 0; static long samples; static struct timeval last_read; @@ -393,6 +395,12 @@ static void create_counter(int counter, int cpu, pid_t pid) attr->sample_freq = freq; } + if (no_samples) + attr->sample_freq = 0; + + if (inherit_stat) + attr->inherit_stat = 1; + if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; @@ -571,7 +579,7 @@ static int __cmd_record(int argc, const char **argv) } } - while (!done) { + for (;;) { int hits = samples; for (i = 0; i < nr_cpu; i++) { @@ -579,8 +587,11 @@ static int __cmd_record(int argc, const char **argv) mmap_read(&mmap_array[i][counter]); } - if (hits == samples) + if (hits == samples) { + if (done) + break; ret = poll(event_array, nr_poll, 100); + } } /* @@ -629,6 +640,10 @@ static const struct option options[] = { "do call-graph (stack chain/backtrace) recording"), OPT_BOOLEAN('v', "verbose", &verbose, "be more verbose (show counter open errors, etc)"), + OPT_BOOLEAN('s', "stat", &inherit_stat, + "per thread counts"), + OPT_BOOLEAN('n', "no-samples", &no_samples, + "don't sample"), OPT_END() }; -- cgit v1.2.3 From 3928ddbe994cce1da1b6365b0db04d5765f254f4 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 25 Jun 2009 22:21:27 +0200 Subject: perf record: Fix unhandled io return value MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Building latest perfcounter fails on the following error: builtin-record.c: In function ‘create_counter’: builtin-record.c:451: erreur: ignoring return value of ‘read’, declared with attribute warn_unused_result make: *** [builtin-record.o] Erreur 1 Just check if we successfully read the perf file descriptor. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Frederic Weisbecker LKML-Reference: <1245961287-5327-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- 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 798a56d890e5..d18546f37d7c 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -448,7 +448,10 @@ try_again: } } - read(fd[nr_cpu][counter], &read_data, sizeof(read_data)); + if (read(fd[nr_cpu][counter], &read_data, sizeof(read_data)) == -1) { + perror("Unable to read perf file descriptor\n"); + exit(-1); + } perf_header_attr__add_id(h_attr, read_data.id); -- cgit v1.2.3 From f37a291c527c954df4da568de718ebb36b8261c0 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Wed, 1 Jul 2009 12:37:06 +0200 Subject: perf_counter tools: Add more warnings and fix/annotate them Enable -Wextra. This found a few real bugs plus a number of signed/unsigned type mismatches/uncleanlinesses. It also required a few annotations All things considered it was still worth it so lets try with this enabled for now. Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 4 ++-- 1 file changed, 2 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 d18546f37d7c..4ef78a5e6f32 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -294,7 +294,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) while (1) { char bf[BUFSIZ], *pbf = bf; struct mmap_event mmap_ev = { - .header.type = PERF_EVENT_MMAP, + .header = { .type = PERF_EVENT_MMAP }, }; int n; size_t size; @@ -650,7 +650,7 @@ static const struct option options[] = { OPT_END() }; -int cmd_record(int argc, const char **argv, const char *prefix) +int cmd_record(int argc, const char **argv, const char *prefix __used) { int counter; -- cgit v1.2.3 From 11b5f81e1b0ea0bc84fe32f0a27054e052b2bf84 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Thu, 16 Jul 2009 15:44:29 +0200 Subject: perf_counter: Synthesize VDSO mmap event perf record synthesizes mmap events for the running process. Right now it just catches file mappings, but we can check for the vdso symbol and add that too. Signed-off-by: Anton Blanchard Signed-off-by: Peter Zijlstra LKML-Reference: <20090716104817.517264409@samba.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 4ef78a5e6f32..072aaf0369f8 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -313,6 +313,10 @@ static void pid_synthesize_mmap_samples(pid_t pid) if (*pbf == 'x') { /* vm_exec */ char *execname = strchr(bf, '/'); + /* Catch VDSO */ + if (execname == NULL) + execname = strstr(bf, "[vdso]"); + if (execname == NULL) continue; -- cgit v1.2.3 From 4bba828dd9bb950ad1fe340ef148a5436a10f131 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Thu, 16 Jul 2009 15:44:29 +0200 Subject: perf_counter: Add perf record option to log addresses Add the -d or --data option to log event addresses (eg page faults). Signed-off-by: Anton Blanchard Signed-off-by: Peter Zijlstra LKML-Reference: <20090716104817.697698033@samba.org> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 072aaf0369f8..68a9be0d1513 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -43,6 +43,7 @@ static int call_graph = 0; static int verbose = 0; static int inherit_stat = 0; static int no_samples = 0; +static int sample_address = 0; static long samples; static struct timeval last_read; @@ -405,6 +406,9 @@ static void create_counter(int counter, int cpu, pid_t pid) if (inherit_stat) attr->inherit_stat = 1; + if (sample_address) + attr->sample_type |= PERF_SAMPLE_ADDR; + if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; @@ -649,6 +653,8 @@ static const struct option options[] = { "be more verbose (show counter open errors, etc)"), OPT_BOOLEAN('s', "stat", &inherit_stat, "per thread counts"), + OPT_BOOLEAN('d', "data", &sample_address, + "Sample addresses"), OPT_BOOLEAN('n', "no-samples", &no_samples, "don't sample"), OPT_END() -- cgit v1.2.3 From a0541234f89c93f313961ce7b28676e11488a5f0 Mon Sep 17 00:00:00 2001 From: Anton Blanchard Date: Wed, 22 Jul 2009 23:04:12 +1000 Subject: perf_counter: Improve perf stat and perf record option parsing perf stat and perf record currently look for all options on the command line. This can lead to some confusion: # perf stat ls -l Error: unknown switch `l' While we can work around this by adding '--' before the command, the git option parsing code can stop at the first non option: # perf stat ls -l Performance counter stats for 'ls -l': .... Signed-off-by: Anton Blanchard Signed-off-by: Peter Zijlstra LKML-Reference: <20090722130412.GD9029@kryten> --- tools/perf/builtin-record.c | 3 ++- 1 file changed, 2 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 68a9be0d1513..6da09928130f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -664,7 +664,8 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) { int counter; - argc = parse_options(argc, argv, options, record_usage, 0); + argc = parse_options(argc, argv, options, record_usage, + PARSE_OPT_STOP_AT_NON_OPTION); if (!argc && target_pid == -1 && !system_wide) usage_with_options(record_usage, options); -- cgit v1.2.3 From f413cdb80ce00ec1a4d0ab949b5d96c81cae7f75 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Fri, 7 Aug 2009 01:25:54 +0200 Subject: perf_counter: Fix/complete ftrace event records sampling This patch implements the kernel side support for ftrace event record sampling. A new counter sampling attribute is added: PERF_SAMPLE_TP_RECORD which requests ftrace events record sampling. In this case if a PERF_TYPE_TRACEPOINT counter is active and a tracepoint fires, we emit the tracepoint binary record to the perfcounter event buffer, as a sample. Result, after setting PERF_SAMPLE_TP_RECORD attribute from perf record: perf record -f -F 1 -a -e workqueue:workqueue_execution perf report -D 0x21e18 [0x48]: event: 9 . . ... raw event: size 72 bytes . 0000: 09 00 00 00 01 00 48 00 d0 c7 00 81 ff ff ff ff ......H........ . 0010: 0a 00 00 00 0a 00 00 00 21 00 00 00 00 00 00 00 ........!...... . 0020: 2b 00 01 02 0a 00 00 00 0a 00 00 00 65 76 65 6e +...........eve . 0030: 74 73 2f 31 00 00 00 00 00 00 00 00 0a 00 00 00 ts/1........... . 0040: e0 b1 31 81 ff ff ff ff ....... . 0x21e18 [0x48]: PERF_EVENT_SAMPLE (IP, 1): 10: 0xffffffff8100c7d0 period: 33 The raw ftrace binary record starts at offset 0020. Translation: struct trace_entry { type = 0x2b = 43; flags = 1; preempt_count = 2; pid = 0xa = 10; tgid = 0xa = 10; } thread_comm = "events/1" thread_pid = 0xa = 10; func = 0xffffffff8131b1e0 = flush_to_ldisc() What will come next? - Userspace support ('perf trace'), 'flight data recorder' mode for perf trace, etc. - The unconditional copy from the profiling callback brings some costs however if someone wants no such sampling to occur, and needs to be fixed in the future. For that we need to have an instant access to the perf counter attribute. This is a matter of a flag to add in the struct ftrace_event. - Take care of the events recursivity! Don't ever try to record a lock event for example, it seems some locking is used in the profiling fast path and lead to a tracing recursivity. That will be fixed using raw spinlock or recursivity protection. - [...] - Profit! :-) Signed-off-by: Frederic Weisbecker Cc: Li Zefan Cc: Tom Zanussi Cc: Arnaldo Carvalho de Melo Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Steven Rostedt Cc: Paul Mackerras Cc: Pekka Enberg Cc: Gabriel Munteanu Cc: Lai Jiangshan Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 6da09928130f..90c98082af10 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -412,6 +412,7 @@ static void create_counter(int counter, int cpu, pid_t pid) if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + attr->mmap = track; attr->comm = track; attr->inherit = (cpu < 0) && inherit; -- cgit v1.2.3 From 266e0e219888420a1a7cafc82e82891cf7b5a979 Mon Sep 17 00:00:00 2001 From: Pierre Habouzit Date: Fri, 7 Aug 2009 14:16:01 +0200 Subject: perf record: Fix the -A UI for empty or non-existent perf.data 1. Ignore the -A argument if there is no perf.data file 2. Treat an empty file like a non existent file. Else, perf will try to read the perf.data header, and fail with an error. Treating an empty file like a non-existent file makes sense, since an interupted (as in SIGKILLed) perf could leave such files around, and you don't want to annoy the user with errors for files with no data in it. Signed-off-by: Pierre Habouzit Acked-by: Peter Zijlstra Cc: Paul Mackerras Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 90c98082af10..0345aad8eba5 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -525,10 +525,14 @@ static int __cmd_record(int argc, const char **argv) signal(SIGCHLD, sig_handler); signal(SIGINT, sig_handler); - if (!stat(output_name, &st) && !force && !append_file) { - fprintf(stderr, "Error, output file %s exists, use -A to append or -f to overwrite.\n", - output_name); - exit(-1); + if (!stat(output_name, &st) && st.st_size) { + if (!force && !append_file) { + fprintf(stderr, "Error, output file %s exists, use -A to append or -f to overwrite.\n", + output_name); + exit(-1); + } + } else { + append_file = 0; } flags = O_CREAT|O_RDWR; -- cgit v1.2.3 From 2a8083f063472f27c253545dd64e1a7bbbb1ab61 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 11 Aug 2009 16:22:00 -0300 Subject: perf record: Fix .tid and .pid fill-in when synthesizing events MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Noticed when trying to record events for a firefox thread. We were synthesizing both .tid and .pid with the pid passed via --pid. Fix it by reading /proc/PID/status and getting the tgid to use in .pid, .tid gets the specified "pid". Signed-off-by: Arnaldo Carvalho de Melo Cc: "H. Peter Anvin" Cc: Frédéric Weisbecker Cc: Peter Zijlstra Cc: Mike Galbraith LKML-Reference: <20090811192200.GF18061@ghostprotocols.net> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 71 ++++++++++++++++++++++++--------------------- 1 file changed, 38 insertions(+), 33 deletions(-) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 0345aad8eba5..30b83def03d4 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -203,46 +203,48 @@ static void sig_atexit(void) kill(getpid(), signr); } -static void pid_synthesize_comm_event(pid_t pid, int full) +static pid_t pid_synthesize_comm_event(pid_t pid, int full) { struct comm_event comm_ev; char filename[PATH_MAX]; char bf[BUFSIZ]; - int fd; - size_t size; - char *field, *sep; + FILE *fp; + size_t size = 0; DIR *tasks; struct dirent dirent, *next; + pid_t tgid = 0; - snprintf(filename, sizeof(filename), "/proc/%d/stat", pid); + snprintf(filename, sizeof(filename), "/proc/%d/status", pid); - fd = open(filename, O_RDONLY); - if (fd < 0) { + fp = fopen(filename, "r"); + if (fd == NULL) { /* * We raced with a task exiting - just return: */ if (verbose) fprintf(stderr, "couldn't open %s\n", filename); - return; + return 0; } - if (read(fd, bf, sizeof(bf)) < 0) { - fprintf(stderr, "couldn't read %s\n", filename); - exit(EXIT_FAILURE); - } - close(fd); - /* 9027 (cat) R 6747 9027 6747 34816 9027 ... */ memset(&comm_ev, 0, sizeof(comm_ev)); - field = strchr(bf, '('); - if (field == NULL) - goto out_failure; - sep = strchr(++field, ')'); - if (sep == NULL) - goto out_failure; - size = sep - field; - memcpy(comm_ev.comm, field, size++); - - comm_ev.pid = pid; + while (!comm_ev.comm[0] || !comm_ev.pid) { + if (fgets(bf, sizeof(bf), fp) == NULL) + goto out_failure; + + if (memcmp(bf, "Name:", 5) == 0) { + char *name = bf + 5; + while (*name && isspace(*name)) + ++name; + size = strlen(name) - 1; + memcpy(comm_ev.comm, name, size++); + } else if (memcmp(bf, "Tgid:", 5) == 0) { + char *tgids = bf + 5; + while (*tgids && isspace(*tgids)) + ++tgids; + tgid = comm_ev.pid = atoi(tgids); + } + } + comm_ev.header.type = PERF_EVENT_COMM; size = ALIGN(size, sizeof(u64)); comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size); @@ -251,7 +253,7 @@ static void pid_synthesize_comm_event(pid_t pid, int full) comm_ev.tid = pid; write_output(&comm_ev, comm_ev.header.size); - return; + goto out_fclose; } snprintf(filename, sizeof(filename), "/proc/%d/task", pid); @@ -268,7 +270,10 @@ static void pid_synthesize_comm_event(pid_t pid, int full) write_output(&comm_ev, comm_ev.header.size); } closedir(tasks); - return; + +out_fclose: + fclose(fp); + return tgid; out_failure: fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n", @@ -276,7 +281,7 @@ out_failure: exit(EXIT_FAILURE); } -static void pid_synthesize_mmap_samples(pid_t pid) +static void pid_synthesize_mmap_samples(pid_t pid, pid_t tgid) { char filename[PATH_MAX]; FILE *fp; @@ -328,7 +333,7 @@ static void pid_synthesize_mmap_samples(pid_t pid) mmap_ev.len -= mmap_ev.start; mmap_ev.header.size = (sizeof(mmap_ev) - (sizeof(mmap_ev.filename) - size)); - mmap_ev.pid = pid; + mmap_ev.pid = tgid; mmap_ev.tid = pid; write_output(&mmap_ev, mmap_ev.header.size); @@ -347,14 +352,14 @@ static void synthesize_all(void) while (!readdir_r(proc, &dirent, &next) && next) { char *end; - pid_t pid; + pid_t pid, tgid; pid = strtol(dirent.d_name, &end, 10); if (*end) /* only interested in proper numerical dirents */ continue; - pid_synthesize_comm_event(pid, 1); - pid_synthesize_mmap_samples(pid); + tgid = pid_synthesize_comm_event(pid, 1); + pid_synthesize_mmap_samples(pid, tgid); } closedir(proc); @@ -567,8 +572,8 @@ static int __cmd_record(int argc, const char **argv) perf_header__write(header, output); if (!system_wide) { - pid_synthesize_comm_event(pid, 0); - pid_synthesize_mmap_samples(pid); + pid_t tgid = pid_synthesize_comm_event(pid, 0); + pid_synthesize_mmap_samples(pid, tgid); } else synthesize_all(); -- cgit v1.2.3 From 0a5ac84650fb7a7f226814103d95724e34b012ae Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 12 Aug 2009 11:18:01 +0200 Subject: perf record: Add missing -C option support for specifying profile cpu perf top supports a -C for setting the profile CPU, but perf record does not. This adds the same option for perf record, allowing the user to specify a specific target profile CPU. Signed-off-by: Jens Axboe Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <20090812091801.GC12579@kernel.dk> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 17 ++++++++++++++--- 1 file changed, 14 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 30b83def03d4..de76008fbaa2 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -35,6 +35,7 @@ static const char *output_name = "perf.data"; static int group = 0; static unsigned int realtime_prio = 0; static int system_wide = 0; +static int profile_cpu = -1; static pid_t target_pid = -1; static int inherit = 1; static int force = 0; @@ -431,6 +432,8 @@ try_again: if (err == EPERM) die("Permission error - are you root?\n"); + else if (err == ENODEV && profile_cpu != -1) + die("No such device - did you specify an out-of-range profile CPU?\n"); /* * If it's cycles then fall back to hrtimer @@ -564,9 +567,15 @@ static int __cmd_record(int argc, const char **argv) if (pid == -1) pid = getpid(); - open_counters(-1, pid); - } else for (i = 0; i < nr_cpus; i++) - open_counters(i, target_pid); + open_counters(profile_cpu, pid); + } else { + if (profile_cpu != -1) { + open_counters(profile_cpu, target_pid); + } else { + for (i = 0; i < nr_cpus; i++) + open_counters(i, target_pid); + } + } if (file_new) perf_header__write(header, output); @@ -645,6 +654,8 @@ static const struct option options[] = { "system-wide collection from all CPUs"), OPT_BOOLEAN('A', "append", &append_file, "append to the output file to do incremental profiling"), + OPT_INTEGER('C', "profile_cpu", &profile_cpu, + "CPU to profile on"), OPT_BOOLEAN('f', "force", &force, "overwrite existing data file"), OPT_LONG('c', "count", &default_interval, -- cgit v1.2.3 From 3a9f131fb00b8ac5950a11ad1599e45edfb5ae44 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 Aug 2009 10:27:18 +0200 Subject: perf tools: Add a per tracepoint counter attribute to get raw sample Add a new flag field while opening a tracepoint perf counter: -e tracepoint_subsystem:tracepoint_name:flags This is intended to be generic although for now it only supports the r[e[c[o[r[d]]]]] flag: ./perf record -e workqueue:workqueue_insertion:record ./perf record -e workqueue:workqueue_insertion:r will have the same effect: enabling the raw samples record for the given tracepoint counter. In the future, we may want to support further flags, separated by commas. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith LKML-Reference: <1250152039-7284-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- 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 de76008fbaa2..78adc47da869 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -398,7 +398,7 @@ static void create_counter(int counter, int cpu, pid_t pid) PERF_FORMAT_TOTAL_TIME_RUNNING | PERF_FORMAT_ID; - attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID; + attr->sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID; if (freq) { attr->sample_type |= PERF_SAMPLE_PERIOD; -- cgit v1.2.3 From daac07b2e6b77f1bd44104aa2f0593e5505f27d4 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 Aug 2009 10:27:19 +0200 Subject: perf tools: Add a general option to enable raw sample records While we can enable the perf sample records per tracepoint counter, we may also want to enable this option for every tracepoint counters to open, so that we don't need to add a :record flag for all of them. Add the -R, --raw-samples options for this purpose. Signed-off-by: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Mike Galbraith LKML-Reference: <1250152039-7284-2-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- tools/perf/builtin-record.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'tools/perf/builtin-record.c') diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c index 78adc47da869..3d051b9cf25f 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -34,6 +34,7 @@ static int output; static const char *output_name = "perf.data"; static int group = 0; static unsigned int realtime_prio = 0; +static int raw_samples = 0; static int system_wide = 0; static int profile_cpu = -1; static pid_t target_pid = -1; @@ -418,6 +419,8 @@ static void create_counter(int counter, int cpu, pid_t pid) if (call_graph) attr->sample_type |= PERF_SAMPLE_CALLCHAIN; + if (raw_samples) + attr->sample_type |= PERF_SAMPLE_RAW; attr->mmap = track; attr->comm = track; @@ -650,6 +653,8 @@ static const struct option options[] = { "record events on existing pid"), OPT_INTEGER('r', "realtime", &realtime_prio, "collect data with this RT SCHED_FIFO priority"), + OPT_BOOLEAN('R', "raw-samples", &raw_samples, + "collect raw sample records from all opened counters"), OPT_BOOLEAN('a', "all-cpus", &system_wide, "system-wide collection from all CPUs"), OPT_BOOLEAN('A', "append", &append_file, -- cgit v1.2.3 From 39e6dd73502f64e2ae3236b304e160ae30de9384 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Fri, 14 Aug 2009 15:26:32 -0300 Subject: perf record: Fix typo in pid_synthesize_comm_event We were using 'fd' locally, but there was a global 'fd' too, so when converting from open to fopen the test made against fd should be made against 'fp', but since we have that global it didnt get discovered ... Reported-by: Ulrich Drepper Signed-off-by: Arnaldo Carvalho de Melo Cc: Peter Zijlstra LKML-Reference: <20090814182632.GF3490@ghostprotocols.net> Signed-off-by: Ingo Molnar --- 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 3d051b9cf25f..89a5ddcd1ded 100644 --- a/tools/perf/builtin-record.c +++ b/tools/perf/builtin-record.c @@ -219,7 +219,7 @@ static pid_t pid_synthesize_comm_event(pid_t pid, int full) snprintf(filename, sizeof(filename), "/proc/%d/status", pid); fp = fopen(filename, "r"); - if (fd == NULL) { + if (fp == NULL) { /* * We raced with a task exiting - just return: */ -- cgit v1.2.3