builtin-record.c 11.4 KB
Newer Older
I
Ingo Molnar 已提交
1 2 3 4
/*
 * perf record: Record the profile of a workload (or a CPU, or a PID) into
 * the perf.data output file - for later analysis via perf report.
 */
5
#include "perf.h"
6
#include "builtin.h"
7
#include "util/util.h"
8
#include "util/parse-options.h"
9
#include "util/parse-events.h"
10
#include "util/string.h"
11

12
#include <unistd.h>
13 14
#include <sched.h>

15 16
#define ALIGN(x, a)		__ALIGN_MASK(x, (typeof(x))(a)-1)
#define __ALIGN_MASK(x, mask)	(((x)+(mask))&~(mask))
17

18 19
static int			default_interval = 100000;
static int			event_count[MAX_COUNTERS];
20

21
static int			fd[MAX_NR_CPUS][MAX_COUNTERS];
22
static int			nr_cpus				= 0;
23
static unsigned int		page_size;
24
static unsigned int		mmap_pages			= 128;
25
static int			output;
26
static const char		*output_name			= "perf.data";
27
static int			group				= 0;
28 29
static unsigned int		realtime_prio			= 0;
static int			system_wide			= 0;
30
static pid_t			target_pid			= -1;
31
static int			inherit				= 1;
32
static int			force				= 0;
I
Ingo Molnar 已提交
33
static int			append_file			= 0;
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

const unsigned int default_count[] = {
	1000000,
	1000000,
	  10000,
	  10000,
	1000000,
	  10000,
};

struct mmap_data {
	int counter;
	void *base;
	unsigned int mask;
	unsigned int prev;
};

static unsigned int mmap_read_head(struct mmap_data *md)
{
	struct perf_counter_mmap_page *pc = md->base;
	int head;

	head = pc->data_head;
	rmb();

	return head;
}

static long events;
static struct timeval last_read, this_read;

static void mmap_read(struct mmap_data *md)
{
	unsigned int head = mmap_read_head(md);
	unsigned int old = md->prev;
	unsigned char *data = md->base + page_size;
	unsigned long size;
	void *buf;
	int diff;

	gettimeofday(&this_read, NULL);

	/*
	 * 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.
	 *
	 * If we somehow ended up ahead of the head, we got messed up.
	 *
	 * In either case, truncate and restart at head.
	 */
	diff = head - old;
	if (diff > md->mask / 2 || diff < 0) {
		struct timeval iv;
		unsigned long msecs;

		timersub(&this_read, &last_read, &iv);
		msecs = iv.tv_sec*1000 + iv.tv_usec/1000;

		fprintf(stderr, "WARNING: failed to keep up with mmap data."
				"  Last read %lu msecs ago.\n", msecs);

		/*
		 * head points to a known good entry, start there.
		 */
		old = head;
	}

	last_read = this_read;

	if (old != head)
		events++;

	size = head - old;

	if ((old & md->mask) + size != (head & md->mask)) {
		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);
			}
			size -= ret;
			buf += 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);
		}
		size -= ret;
		buf += ret;
	}

	md->prev = old;
}

static volatile int done = 0;

141
static void sig_handler(int sig)
142
{
143
	done = 1;
144 145
}

146 147 148 149 150 151
static struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
static struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];

static int nr_poll;
static int nr_cpu;

152
struct mmap_event {
153 154 155 156 157 158 159
	struct perf_event_header	header;
	__u32				pid;
	__u32				tid;
	__u64				start;
	__u64				len;
	__u64				pgoff;
	char				filename[PATH_MAX];
160
};
161

162
struct comm_event {
163 164 165 166
	struct perf_event_header	header;
	__u32				pid;
	__u32				tid;
	char				comm[16];
167 168
};

169
static void pid_synthesize_comm_event(pid_t pid, int full)
170
{
171
	struct comm_event comm_ev;
172 173
	char filename[PATH_MAX];
	char bf[BUFSIZ];
174
	int fd, ret;
175
	size_t size;
176
	char *field, *sep;
177 178
	DIR *tasks;
	struct dirent dirent, *next;
179 180 181 182 183 184 185 186 187 188 189 190 191 192

	snprintf(filename, sizeof(filename), "/proc/%d/stat", pid);

	fd = open(filename, O_RDONLY);
	if (fd < 0) {
		fprintf(stderr, "couldn't open %s\n", filename);
		exit(EXIT_FAILURE);
	}
	if (read(fd, bf, sizeof(bf)) < 0) {
		fprintf(stderr, "couldn't read %s\n", filename);
		exit(EXIT_FAILURE);
	}
	close(fd);

193
	/* 9027 (cat) R 6747 9027 6747 34816 9027 ... */
194
	memset(&comm_ev, 0, sizeof(comm_ev));
195 196 197 198 199 200 201 202
	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++);
203 204

	comm_ev.pid = pid;
205 206 207
	comm_ev.header.type = PERF_EVENT_COMM;
	size = ALIGN(size, sizeof(uint64_t));
	comm_ev.header.size = sizeof(comm_ev) - (sizeof(comm_ev.comm) - size);
208

209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
	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;
	}

	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);
		}
236
	}
237 238 239
	closedir(tasks);
	return;

240 241 242 243
out_failure:
	fprintf(stderr, "couldn't get COMM and pgid, malformed %s\n",
		filename);
	exit(EXIT_FAILURE);
244 245
}

246
static void pid_synthesize_mmap_events(pid_t pid)
247 248 249 250 251 252 253 254 255 256 257 258
{
	char filename[PATH_MAX];
	FILE *fp;

	snprintf(filename, sizeof(filename), "/proc/%d/maps", pid);

	fp = fopen(filename, "r");
	if (fp == NULL) {
		fprintf(stderr, "couldn't open %s\n", filename);
		exit(EXIT_FAILURE);
	}
	while (1) {
259
		char bf[BUFSIZ], *pbf = bf;
260 261 262
		struct mmap_event mmap_ev = {
			.header.type = PERF_EVENT_MMAP,
		};
263
		int n;
264 265 266 267 268
		size_t size;
		if (fgets(bf, sizeof(bf), fp) == NULL)
			break;

		/* 00400000-0040c000 r-xp 00000000 fd:01 41038  /bin/cat */
269 270 271 272 273 274 275 276 277
		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 */
278 279 280 281 282 283 284 285 286 287 288 289 290
			char *execname = strrchr(bf, ' ');

			if (execname == NULL || execname[1] != '/')
				continue;

			execname += 1;
			size = strlen(execname);
			execname[size - 1] = '\0'; /* Remove \n */
			memcpy(mmap_ev.filename, execname, size);
			size = ALIGN(size, sizeof(uint64_t));
			mmap_ev.len -= mmap_ev.start;
			mmap_ev.header.size = (sizeof(mmap_ev) -
					       (sizeof(mmap_ev.filename) - size));
291
			mmap_ev.pid = pid;
292 293 294 295 296 297 298 299 300 301 302 303
			mmap_ev.tid = pid;

			if (write(output, &mmap_ev, mmap_ev.header.size) < 0) {
				perror("failed to write");
				exit(-1);
			}
		}
	}

	fclose(fp);
}

304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
static void synthesize_events(void)
{
	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_events(pid);
	}

	closedir(proc);
}

326
static void open_counters(int cpu, pid_t pid)
327
{
328
	struct perf_counter_attr attr;
329 330 331
	int counter, group_fd;
	int track = 1;

332
	if (pid > 0) {
333 334
		pid_synthesize_comm_event(pid, 0);
		pid_synthesize_mmap_events(pid);
335
	}
336 337 338 339

	group_fd = -1;
	for (counter = 0; counter < nr_counters; counter++) {

340 341 342 343 344 345 346
		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;
		attr.mmap		= track;
		attr.comm		= track;
		attr.inherit	= (cpu < 0) && inherit;
347 348 349 350

		track = 0; // only the first counter needs these

		fd[nr_cpu][counter] =
351
			sys_perf_counter_open(&attr, pid, cpu, group_fd, 0);
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387

		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);

		/*
		 * 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);
		}
	}
	nr_cpu++;
}

388
static int __cmd_record(int argc, const char **argv)
389 390
{
	int i, counter;
I
Ingo Molnar 已提交
391
	struct stat st;
392
	pid_t pid;
I
Ingo Molnar 已提交
393
	int flags;
394 395 396 397 398 399 400
	int ret;

	page_size = sysconf(_SC_PAGE_SIZE);
	nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
	assert(nr_cpus <= MAX_NR_CPUS);
	assert(nr_cpus >= 0);

I
Ingo Molnar 已提交
401 402
	if (!stat(output_name, &st) && !force && !append_file) {
		fprintf(stderr, "Error, output file %s exists, use -A to append or -f to overwrite.\n",
403 404 405 406
				output_name);
		exit(-1);
	}

I
Ingo Molnar 已提交
407 408 409 410 411 412 413
	flags = O_CREAT|O_RDWR;
	if (append_file)
		flags |= O_APPEND;
	else
		flags |= O_TRUNC;

	output = open(output_name, flags, S_IRUSR|S_IWUSR);
414 415 416 417 418
	if (output < 0) {
		perror("failed to create output file");
		exit(-1);
	}

419 420 421 422
	if (!system_wide) {
		open_counters(-1, target_pid != -1 ? target_pid : 0);
	} else for (i = 0; i < nr_cpus; i++)
		open_counters(i, target_pid);
423

424 425
	signal(SIGCHLD, sig_handler);
	signal(SIGINT, sig_handler);
426

427
	if (target_pid == -1 && argc) {
428 429 430
		pid = fork();
		if (pid < 0)
			perror("failed to fork");
431

432
		if (!pid) {
433
			if (execvp(argv[0], (char **)argv)) {
434 435 436
				perror(argv[0]);
				exit(-1);
			}
437 438 439 440 441 442 443 444 445 446 447 448 449
		}
	}

	if (realtime_prio) {
		struct sched_param param;

		param.sched_priority = realtime_prio;
		if (sched_setscheduler(0, SCHED_FIFO, &param)) {
			printf("Could not set realtime priority.\n");
			exit(-1);
		}
	}

450 451
	if (system_wide)
		synthesize_events();
452 453 454 455

	while (!done) {
		int hits = events;

456
		for (i = 0; i < nr_cpu; i++) {
457 458 459 460 461 462 463 464 465 466
			for (counter = 0; counter < nr_counters; counter++)
				mmap_read(&mmap_array[i][counter]);
		}

		if (hits == events)
			ret = poll(event_array, nr_poll, 100);
	}

	return 0;
}
467 468

static const char * const record_usage[] = {
469 470
	"perf record [<options>] [<command>]",
	"perf record [<options>] -- <command> [<options>]",
471 472 473
	NULL
};

474 475
static char events_help_msg[EVENTS_HELP_MAX];

476
static const struct option options[] = {
477
	OPT_CALLBACK('e', "event", NULL, "event",
478
		     events_help_msg, parse_events),
479 480 481 482 483 484
	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"),
I
Ingo Molnar 已提交
485 486
	OPT_BOOLEAN('A', "append", &append_file,
			    "append to the output file to do incremental profiling"),
487 488
	OPT_BOOLEAN('f', "force", &force,
			"overwrite existing data file"),
I
Ingo Molnar 已提交
489 490 491 492 493 494 495 496
	OPT_INTEGER('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('m', "mmap-pages", &mmap_pages,
		    "number of mmap data pages"),
497 498 499 500 501 502 503
	OPT_END()
};

int cmd_record(int argc, const char **argv, const char *prefix)
{
	int counter;

504
	create_events_help(events_help_msg);
505 506

	argc = parse_options(argc, argv, options, record_usage, 0);
507
	if (!argc && target_pid == -1 && !system_wide)
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523
		usage_with_options(record_usage, options);

	if (!nr_counters) {
		nr_counters = 1;
		event_id[0] = 0;
	}

	for (counter = 0; counter < nr_counters; counter++) {
		if (event_count[counter])
			continue;

		event_count[counter] = default_interval;
	}

	return __cmd_record(argc, argv);
}