-
+/*
+ * builtin-record.c
+ *
+ * Builtin record command: Record the profile of a workload
+ * (or a CPU, or a PID) into the perf.data output file - for
+ * later analysis via perf report.
+ */
+#include "builtin.h"
#include "perf.h"
-#include "builtin.h"
+
#include "util/util.h"
#include "util/parse-options.h"
#include "util/parse-events.h"
+#include "util/string.h"
+#include <unistd.h>
#include <sched.h>
#define ALIGN(x, a) __ALIGN_MASK(x, (typeof(x))(a)-1)
#define __ALIGN_MASK(x, mask) (((x)+(mask))&~(mask))
-static int default_interval = 100000;
-static int event_count[MAX_COUNTERS];
+static long default_interval = 100000;
+static long event_count[MAX_COUNTERS];
static int fd[MAX_NR_CPUS][MAX_COUNTERS];
-static int nr_cpus = 0;
+static int nr_cpus = 0;
static unsigned int page_size;
-static unsigned int mmap_pages = 16;
+static unsigned int mmap_pages = 128;
+static int freq = 0;
static int output;
static const char *output_name = "perf.data";
static int group = 0;
static int system_wide = 0;
static pid_t target_pid = -1;
static int inherit = 1;
-static int nmi = 1;
+static int force = 0;
+static int append_file = 0;
const unsigned int default_count[] = {
1000000,
return head;
}
-static long events;
+static long samples;
static struct timeval last_read, this_read;
+static __u64 bytes_written;
+
static void mmap_read(struct mmap_data *md)
{
unsigned int head = mmap_read_head(md);
/*
* If we're further behind than half the buffer, there's a chance
- * the writer will bite our tail and screw up the events under us.
+ * the writer will bite our tail and mess up the samples under us.
*
* If we somehow ended up ahead of the head, we got messed up.
*
last_read = this_read;
if (old != head)
- events++;
+ samples++;
size = head - old;
buf = &data[old & md->mask];
size = md->mask + 1 - (old & md->mask);
old += size;
+
while (size) {
int ret = write(output, buf, size);
- if (ret < 0) {
- perror("failed to write");
- exit(-1);
- }
+
+ if (ret < 0)
+ die("failed to write");
+
size -= ret;
buf += ret;
+
+ bytes_written += ret;
}
}
buf = &data[old & md->mask];
size = head - old;
old += size;
+
while (size) {
int ret = write(output, buf, size);
- if (ret < 0) {
- perror("failed to write");
- exit(-1);
- }
+
+ if (ret < 0)
+ die("failed to write");
+
size -= ret;
buf += ret;
+
+ bytes_written += ret;
}
md->prev = old;
char comm[16];
};
-static pid_t pid_synthesize_comm_event(pid_t pid)
+static void pid_synthesize_comm_event(pid_t pid, int full)
{
struct comm_event comm_ev;
char filename[PATH_MAX];
- pid_t spid, ppid;
char bf[BUFSIZ];
- int fd, nr, ret;
- char comm[18];
+ int fd, ret;
size_t size;
- char state;
+ char *field, *sep;
+ DIR *tasks;
+ struct dirent dirent, *next;
snprintf(filename, sizeof(filename), "/proc/%d/stat", pid);
}
close(fd);
+ /* 9027 (cat) R 6747 9027 6747 34816 9027 ... */
memset(&comm_ev, 0, sizeof(comm_ev));
- nr = sscanf(bf, "%d %s %c %d %d ",
- &spid, comm, &state, &ppid, &comm_ev.pid);
- if (nr != 5) {
- fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n",
- filename);
- exit(EXIT_FAILURE);
- }
+ 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;
comm_ev.header.type = PERF_EVENT_COMM;
- comm_ev.tid = pid;
- size = strlen(comm);
- comm[--size] = '\0'; /* Remove the ')' at the end */
- --size; /* Remove the '(' at the begin */
- memcpy(comm_ev.comm, comm + 1, size);
size = ALIGN(size, sizeof(uint64_t));
comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size);
- ret = write(output, &comm_ev, comm_ev.header.size);
- if (ret < 0) {
- perror("failed to write");
- exit(-1);
+ if (!full) {
+ comm_ev.tid = pid;
+
+ ret = write(output, &comm_ev, comm_ev.header.size);
+ if (ret < 0) {
+ perror("failed to write");
+ exit(-1);
+ }
+ return;
}
- return comm_ev.pid;
+
+ snprintf(filename, sizeof(filename), "/proc/%d/task", pid);
+
+ tasks = opendir(filename);
+ while (!readdir_r(tasks, &dirent, &next) && next) {
+ char *end;
+ pid = strtol(dirent.d_name, &end, 10);
+ if (*end)
+ continue;
+
+ comm_ev.tid = pid;
+
+ ret = write(output, &comm_ev, comm_ev.header.size);
+ if (ret < 0) {
+ perror("failed to write");
+ exit(-1);
+ }
+ }
+ closedir(tasks);
+ return;
+
+out_failure:
+ fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n",
+ filename);
+ exit(EXIT_FAILURE);
}
-static void pid_synthesize_mmap_events(pid_t pid, pid_t pgid)
+static void pid_synthesize_mmap_samples(pid_t pid)
{
char filename[PATH_MAX];
FILE *fp;
exit(EXIT_FAILURE);
}
while (1) {
- char bf[BUFSIZ];
- unsigned char vm_read, vm_write, vm_exec, vm_mayshare;
+ char bf[BUFSIZ], *pbf = bf;
struct mmap_event mmap_ev = {
.header.type = PERF_EVENT_MMAP,
};
- unsigned long ino;
- int major, minor;
+ int n;
size_t size;
if (fgets(bf, sizeof(bf), fp) == NULL)
break;
/* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */
- sscanf(bf, "%llx-%llx %c%c%c%c %llx %x:%x %lu",
- &mmap_ev.start, &mmap_ev.len,
- &vm_read, &vm_write, &vm_exec, &vm_mayshare,
- &mmap_ev.pgoff, &major, &minor, &ino);
- if (vm_exec == 'x') {
+ n = hex2u64(pbf, &mmap_ev.start);
+ if (n < 0)
+ continue;
+ pbf += n + 1;
+ n = hex2u64(pbf, &mmap_ev.len);
+ if (n < 0)
+ continue;
+ pbf += n + 3;
+ if (*pbf == 'x') { /* vm_exec */
char *execname = strrchr(bf, ' ');
if (execname == NULL || execname[1] != '/')
mmap_ev.len -= mmap_ev.start;
mmap_ev.header.size = (sizeof(mmap_ev) -
(sizeof(mmap_ev.filename) - size));
- mmap_ev.pid = pgid;
+ mmap_ev.pid = pid;
mmap_ev.tid = pid;
if (write(output, &mmap_ev, mmap_ev.header.size) < 0) {
fclose(fp);
}
-static void open_counters(int cpu, pid_t pid)
+static void synthesize_samples(void)
{
- struct perf_counter_hw_event hw_event;
- int counter, group_fd;
+ DIR *proc;
+ struct dirent dirent, *next;
+
+ proc = opendir("/proc");
+
+ while (!readdir_r(proc, &dirent, &next) && next) {
+ char *end;
+ pid_t pid;
+
+ 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);
+ }
+
+ closedir(proc);
+}
+
+static int group_fd;
+
+static void create_counter(int counter, int cpu, pid_t pid)
+{
+ struct perf_counter_attr attr;
int track = 1;
- if (pid > 0) {
- pid_t pgid = pid_synthesize_comm_event(pid);
- pid_synthesize_mmap_events(pid, pgid);
+ memset(&attr, 0, sizeof(attr));
+ attr.config = event_id[counter];
+ attr.sample_period = event_count[counter];
+ attr.sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID | PERF_SAMPLE_PERIOD;
+ attr.freq = freq;
+ attr.mmap = track;
+ attr.comm = track;
+ attr.inherit = (cpu < 0) && inherit;
+
+ track = 0; /* only the first counter needs these */
+
+ fd[nr_cpu][counter] = sys_perf_counter_open(&attr, pid, cpu, group_fd, 0);
+
+ if (fd[nr_cpu][counter] < 0) {
+ int err = errno;
+
+ error("syscall returned with %d (%s)\n",
+ fd[nr_cpu][counter], strerror(err));
+ if (err == EPERM)
+ printf("Are you root?\n");
+ exit(-1);
}
+ assert(fd[nr_cpu][counter] >= 0);
+ fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK);
- group_fd = -1;
- for (counter = 0; counter < nr_counters; counter++) {
+ /*
+ * First counter acts as the group leader:
+ */
+ if (group && group_fd == -1)
+ group_fd = fd[nr_cpu][counter];
+
+ event_array[nr_poll].fd = fd[nr_cpu][counter];
+ event_array[nr_poll].events = POLLIN;
+ nr_poll++;
+
+ mmap_array[nr_cpu][counter].counter = counter;
+ 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);
+ if (mmap_array[nr_cpu][counter].base == MAP_FAILED) {
+ error("failed to mmap with %d (%s)\n", errno, strerror(errno));
+ exit(-1);
+ }
+}
- memset(&hw_event, 0, sizeof(hw_event));
- hw_event.config = event_id[counter];
- hw_event.irq_period = event_count[counter];
- hw_event.record_type = PERF_RECORD_IP | PERF_RECORD_TID;
- hw_event.nmi = nmi;
- hw_event.mmap = track;
- hw_event.comm = track;
- hw_event.inherit = (cpu < 0) && inherit;
-
- track = 0; // only the first counter needs these
-
- fd[nr_cpu][counter] =
- sys_perf_counter_open(&hw_event, pid, cpu, group_fd, 0);
-
- if (fd[nr_cpu][counter] < 0) {
- int err = errno;
- printf("kerneltop error: syscall returned with %d (%s)\n",
- fd[nr_cpu][counter], strerror(err));
- if (err == EPERM)
- printf("Are you root?\n");
- exit(-1);
- }
- assert(fd[nr_cpu][counter] >= 0);
- fcntl(fd[nr_cpu][counter], F_SETFL, O_NONBLOCK);
+static void open_counters(int cpu, pid_t pid)
+{
+ int counter;
- /*
- * First counter acts as the group leader:
- */
- if (group && group_fd == -1)
- group_fd = fd[nr_cpu][counter];
-
- event_array[nr_poll].fd = fd[nr_cpu][counter];
- event_array[nr_poll].events = POLLIN;
- nr_poll++;
-
- mmap_array[nr_cpu][counter].counter = counter;
- 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);
- if (mmap_array[nr_cpu][counter].base == MAP_FAILED) {
- printf("kerneltop error: failed to mmap with %d (%s)\n",
- errno, strerror(errno));
- exit(-1);
- }
+ 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);
+
nr_cpu++;
}
static int __cmd_record(int argc, const char **argv)
{
int i, counter;
+ struct stat st;
pid_t pid;
+ int flags;
int ret;
page_size = sysconf(_SC_PAGE_SIZE);
assert(nr_cpus <= MAX_NR_CPUS);
assert(nr_cpus >= 0);
- output = open(output_name, O_CREAT|O_EXCL|O_RDWR, S_IRWXU);
+ 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);
+ }
+
+ flags = O_CREAT|O_RDWR;
+ if (append_file)
+ flags |= O_APPEND;
+ else
+ flags |= O_TRUNC;
+
+ output = open(output_name, flags, S_IRUSR|S_IWUSR);
if (output < 0) {
perror("failed to create output file");
exit(-1);
}
if (!system_wide) {
- open_counters(-1, target_pid != -1 ? target_pid : 0);
+ open_counters(-1, target_pid != -1 ? target_pid : getpid());
} else for (i = 0; i < nr_cpus; i++)
open_counters(i, target_pid);
}
}
- /*
- * TODO: store the current /proc/$/maps information somewhere
- */
+ if (system_wide)
+ synthesize_samples();
while (!done) {
- int hits = events;
+ int hits = samples;
for (i = 0; i < nr_cpu; i++) {
for (counter = 0; counter < nr_counters; counter++)
mmap_read(&mmap_array[i][counter]);
}
- if (hits == events)
+ if (hits == samples)
ret = poll(event_array, nr_poll, 100);
}
+ /*
+ * Approximate RIP event size: 24 bytes.
+ */
+ fprintf(stderr,
+ "[ perf record: Captured and wrote %.3f MB %s (~%lld samples) ]\n",
+ (double)bytes_written / 1024.0 / 1024.0,
+ output_name,
+ bytes_written / 24);
+
return 0;
}
static const struct option options[] = {
OPT_CALLBACK('e', "event", NULL, "event",
events_help_msg, parse_events),
- OPT_INTEGER('c', "count", &default_interval,
- "event period to sample"),
- OPT_INTEGER('m', "mmap-pages", &mmap_pages,
- "number of mmap data pages"),
- OPT_STRING('o', "output", &output_name, "file",
- "output file name"),
- OPT_BOOLEAN('i', "inherit", &inherit,
- "child tasks inherit counters"),
OPT_INTEGER('p', "pid", &target_pid,
"record events on existing pid"),
OPT_INTEGER('r', "realtime", &realtime_prio,
"collect data with this RT SCHED_FIFO priority"),
OPT_BOOLEAN('a', "all-cpus", &system_wide,
"system-wide collection from all CPUs"),
+ OPT_BOOLEAN('A', "append", &append_file,
+ "append to the output file to do incremental profiling"),
+ OPT_BOOLEAN('f', "force", &force,
+ "overwrite existing data file"),
+ OPT_LONG('c', "count", &default_interval,
+ "event period to sample"),
+ OPT_STRING('o', "output", &output_name, "file",
+ "output file name"),
+ OPT_BOOLEAN('i', "inherit", &inherit,
+ "child tasks inherit counters"),
+ OPT_INTEGER('F', "freq", &freq,
+ "profile at this frequency"),
+ OPT_INTEGER('m', "mmap-pages", &mmap_pages,
+ "number of mmap data pages"),
OPT_END()
};
event_id[0] = 0;
}
+ if (freq) {
+ default_interval = freq;
+ freq = 1;
+ }
for (counter = 0; counter < nr_counters; counter++) {
if (event_count[counter])
continue;