event.c 15.1 KB
Newer Older
1 2 3
#include <linux/types.h>
#include "event.h"
#include "debug.h"
4
#include "session.h"
5
#include "sort.h"
6
#include "string.h"
7
#include "strlist.h"
8
#include "thread.h"
9 10

static pid_t event__synthesize_comm(pid_t pid, int full,
11
				    event__handler_t process,
12
				    struct perf_session *session)
13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 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
{
	event_t ev;
	char filename[PATH_MAX];
	char bf[BUFSIZ];
	FILE *fp;
	size_t size = 0;
	DIR *tasks;
	struct dirent dirent, *next;
	pid_t tgid = 0;

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

	fp = fopen(filename, "r");
	if (fp == NULL) {
out_race:
		/*
		 * We raced with a task exiting - just return:
		 */
		pr_debug("couldn't open %s\n", filename);
		return 0;
	}

	memset(&ev.comm, 0, sizeof(ev.comm));
	while (!ev.comm.comm[0] || !ev.comm.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(ev.comm.comm, name, size++);
		} else if (memcmp(bf, "Tgid:", 5) == 0) {
			char *tgids = bf + 5;
			while (*tgids && isspace(*tgids))
				++tgids;
			tgid = ev.comm.pid = atoi(tgids);
		}
	}

	ev.comm.header.type = PERF_RECORD_COMM;
	size = ALIGN(size, sizeof(u64));
	ev.comm.header.size = sizeof(ev.comm) - (sizeof(ev.comm.comm) - size);

	if (!full) {
		ev.comm.tid = pid;

61
		process(&ev, session);
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78
		goto out_fclose;
	}

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

	tasks = opendir(filename);
	if (tasks == NULL)
		goto out_race;

	while (!readdir_r(tasks, &dirent, &next) && next) {
		char *end;
		pid = strtol(dirent.d_name, &end, 10);
		if (*end)
			continue;

		ev.comm.tid = pid;

79
		process(&ev, session);
80 81 82 83 84 85 86 87 88 89 90 91 92
	}
	closedir(tasks);

out_fclose:
	fclose(fp);
	return tgid;

out_failure:
	pr_warning("couldn't get COMM and pgid, malformed %s\n", filename);
	return -1;
}

static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
93
					 event__handler_t process,
94
					 struct perf_session *session)
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
{
	char filename[PATH_MAX];
	FILE *fp;

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

	fp = fopen(filename, "r");
	if (fp == NULL) {
		/*
		 * We raced with a task exiting - just return:
		 */
		pr_debug("couldn't open %s\n", filename);
		return -1;
	}

	while (1) {
		char bf[BUFSIZ], *pbf = bf;
		event_t ev = {
113 114 115 116
			.header = {
				.type = PERF_RECORD_MMAP,
				.misc = 0, /* Just like the kernel, see kernel/perf_event.c __perf_event_mmap */
			 },
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132
		};
		int n;
		size_t size;
		if (fgets(bf, sizeof(bf), fp) == NULL)
			break;

		/* 00400000-0040c000 r-xp 00000000 fd:01 41038  /bin/cat */
		n = hex2u64(pbf, &ev.mmap.start);
		if (n < 0)
			continue;
		pbf += n + 1;
		n = hex2u64(pbf, &ev.mmap.len);
		if (n < 0)
			continue;
		pbf += n + 3;
		if (*pbf == 'x') { /* vm_exec */
133
			u64 vm_pgoff;
134 135 136 137 138 139 140 141 142
			char *execname = strchr(bf, '/');

			/* Catch VDSO */
			if (execname == NULL)
				execname = strstr(bf, "[vdso]");

			if (execname == NULL)
				continue;

143 144 145 146 147 148 149 150
			pbf += 3;
			n = hex2u64(pbf, &vm_pgoff);
			/* pgoff is in bytes, not pages */
			if (n >= 0)
				ev.mmap.pgoff = vm_pgoff << getpagesize();
			else
				ev.mmap.pgoff = 0;

151 152 153 154 155 156 157 158 159 160
			size = strlen(execname);
			execname[size - 1] = '\0'; /* Remove \n */
			memcpy(ev.mmap.filename, execname, size);
			size = ALIGN(size, sizeof(u64));
			ev.mmap.len -= ev.mmap.start;
			ev.mmap.header.size = (sizeof(ev.mmap) -
					       (sizeof(ev.mmap.filename) - size));
			ev.mmap.pid = tgid;
			ev.mmap.tid = pid;

161
			process(&ev, session);
162 163 164 165 166 167 168
		}
	}

	fclose(fp);
	return 0;
}

169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184
int event__synthesize_modules(event__handler_t process,
			      struct perf_session *session)
{
	struct rb_node *nd;

	for (nd = rb_first(&session->kmaps.maps[MAP__FUNCTION]);
	     nd; nd = rb_next(nd)) {
		event_t ev;
		size_t size;
		struct map *pos = rb_entry(nd, struct map, rb_node);

		if (pos->dso->kernel)
			continue;

		size = ALIGN(pos->dso->long_name_len + 1, sizeof(u64));
		memset(&ev, 0, sizeof(ev));
185
		ev.mmap.header.misc = 1; /* kernel uses 0 for user space maps, see kernel/perf_event.c __perf_event_mmap */
186 187 188 189 190 191 192 193 194 195 196 197 198 199
		ev.mmap.header.type = PERF_RECORD_MMAP;
		ev.mmap.header.size = (sizeof(ev.mmap) -
				        (sizeof(ev.mmap.filename) - size));
		ev.mmap.start = pos->start;
		ev.mmap.len   = pos->end - pos->start;

		memcpy(ev.mmap.filename, pos->dso->long_name,
		       pos->dso->long_name_len + 1);
		process(&ev, session);
	}

	return 0;
}

200
int event__synthesize_thread(pid_t pid, event__handler_t process,
201
			     struct perf_session *session)
202
{
203
	pid_t tgid = event__synthesize_comm(pid, 1, process, session);
204 205
	if (tgid == -1)
		return -1;
206
	return event__synthesize_mmap_events(pid, tgid, process, session);
207 208
}

209
void event__synthesize_threads(event__handler_t process,
210
			       struct perf_session *session)
211 212 213 214 215 216 217 218 219 220 221 222 223
{
	DIR *proc;
	struct dirent dirent, *next;

	proc = opendir("/proc");

	while (!readdir_r(proc, &dirent, &next) && next) {
		char *end;
		pid_t pid = strtol(dirent.d_name, &end, 10);

		if (*end) /* only interested in proper numerical dirents */
			continue;

224
		event__synthesize_thread(pid, process, session);
225 226 227 228
	}

	closedir(proc);
}
229

230 231 232 233 234 235 236 237 238
struct process_symbol_args {
	const char *name;
	u64	   start;
};

static int find_symbol_cb(void *arg, const char *name, char type, u64 start)
{
	struct process_symbol_args *args = arg;

239 240 241 242 243 244
	/*
	 * Must be a function or at least an alias, as in PARISC64, where "_text" is
	 * an 'A' to the same address as "_stext".
	 */
	if (!(symbol_type__is_a(type, MAP__FUNCTION) ||
	      type == 'A') || strcmp(name, args->name))
245 246 247 248 249 250
		return 0;

	args->start = start;
	return 1;
}

251
int event__synthesize_kernel_mmap(event__handler_t process,
252 253 254 255 256
				  struct perf_session *session,
				  const char *symbol_name)
{
	size_t size;
	event_t ev = {
257 258 259 260
		.header = {
			.type = PERF_RECORD_MMAP,
			.misc = 1, /* kernel uses 0 for user space maps, see kernel/perf_event.c __perf_event_mmap */
		},
261 262 263 264 265 266 267 268
	};
	/*
	 * We should get this from /sys/kernel/sections/.text, but till that is
	 * available use this, and after it is use this as a fallback for older
	 * kernels.
	 */
	struct process_symbol_args args = { .name = symbol_name, };

269
	if (kallsyms__parse("/proc/kallsyms", &args, find_symbol_cb) <= 0)
270 271 272 273 274 275
		return -ENOENT;

	size = snprintf(ev.mmap.filename, sizeof(ev.mmap.filename),
			"[kernel.kallsyms.%s]", symbol_name) + 1;
	size = ALIGN(size, sizeof(u64));
	ev.mmap.header.size = (sizeof(ev.mmap) - (sizeof(ev.mmap.filename) - size));
276 277 278
	ev.mmap.pgoff = args.start;
	ev.mmap.start = session->vmlinux_maps[MAP__FUNCTION]->start;
	ev.mmap.len   = session->vmlinux_maps[MAP__FUNCTION]->end - ev.mmap.start ;
279 280 281 282

	return process(&ev, session);
}

283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
static void thread__comm_adjust(struct thread *self)
{
	char *comm = self->comm;

	if (!symbol_conf.col_width_list_str && !symbol_conf.field_sep &&
	    (!symbol_conf.comm_list ||
	     strlist__has_entry(symbol_conf.comm_list, comm))) {
		unsigned int slen = strlen(comm);

		if (slen > comms__col_width) {
			comms__col_width = slen;
			threads__col_width = slen + 6;
		}
	}
}

static int thread__set_comm_adjust(struct thread *self, const char *comm)
{
	int ret = thread__set_comm(self, comm);

	if (ret)
		return ret;

	thread__comm_adjust(self);

	return 0;
}

311
int event__process_comm(event_t *self, struct perf_session *session)
312
{
313
	struct thread *thread = perf_session__findnew(session, self->comm.pid);
314

L
Li Zefan 已提交
315
	dump_printf(": %s:%d\n", self->comm.comm, self->comm.pid);
316

317
	if (thread == NULL || thread__set_comm_adjust(thread, self->comm.comm)) {
318 319 320 321 322 323 324
		dump_printf("problem processing PERF_RECORD_COMM, skipping event.\n");
		return -1;
	}

	return 0;
}

325
int event__process_lost(event_t *self, struct perf_session *session)
326 327
{
	dump_printf(": id:%Ld: lost:%Ld\n", self->lost.id, self->lost.lost);
328
	session->events_stats.lost += self->lost.lost;
329 330 331
	return 0;
}

332
int event__process_mmap(event_t *self, struct perf_session *session)
333
{
334 335
	struct thread *thread;
	struct map *map;
336

337 338 339
	dump_printf(" %d/%d: [%#Lx(%#Lx) @ %#Lx]: %s\n",
		    self->mmap.pid, self->mmap.tid, self->mmap.start,
		    self->mmap.len, self->mmap.pgoff, self->mmap.filename);
340

341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361
	if (self->mmap.pid == 0) {
		static const char kmmap_prefix[] = "[kernel.kallsyms.";

		if (self->mmap.filename[0] == '/') {
			char short_module_name[1024];
			char *name = strrchr(self->mmap.filename, '/'), *dot;

			if (name == NULL)
				goto out_problem;

			++name; /* skip / */
			dot = strrchr(name, '.');
			if (dot == NULL)
				goto out_problem;

			snprintf(short_module_name, sizeof(short_module_name),
				 "[%.*s]", (int)(dot - name), name);
			strxfrchar(short_module_name, '-', '_');

			map = perf_session__new_module_map(session,
							   self->mmap.start,
362
							   self->mmap.filename);
363 364 365
			if (map == NULL)
				goto out_problem;

366
			name = strdup(short_module_name);
367 368 369
			if (name == NULL)
				goto out_problem;

370
			map->dso->short_name = name;
371 372 373 374 375 376 377 378 379 380 381 382 383 384
			map->end = map->start + self->mmap.len;
		} else if (memcmp(self->mmap.filename, kmmap_prefix,
				sizeof(kmmap_prefix) - 1) == 0) {
			const char *symbol_name = (self->mmap.filename +
						   sizeof(kmmap_prefix) - 1);
			/*
			 * Should be there already, from the build-id table in
			 * the header.
			 */
			struct dso *kernel = __dsos__findnew(&dsos__kernel,
							     "[kernel.kallsyms]");
			if (kernel == NULL)
				goto out_problem;

385
			kernel->kernel = 1;
386
			if (__perf_session__create_kernel_maps(session, kernel) < 0)
387 388 389 390
				goto out_problem;

			session->vmlinux_maps[MAP__FUNCTION]->start = self->mmap.start;
			session->vmlinux_maps[MAP__FUNCTION]->end   = self->mmap.start + self->mmap.len;
391 392 393 394 395 396
			/*
			 * Be a bit paranoid here, some perf.data file came with
			 * a zero sized synthesized MMAP event for the kernel.
			 */
			if (session->vmlinux_maps[MAP__FUNCTION]->end == 0)
				session->vmlinux_maps[MAP__FUNCTION]->end = ~0UL;
397 398 399 400

			perf_session__set_kallsyms_ref_reloc_sym(session, symbol_name,
								 self->mmap.pgoff);
		}
401 402 403 404
		return 0;
	}

	thread = perf_session__findnew(session, self->mmap.pid);
405 406
	map = map__new(self->mmap.start, self->mmap.len, self->mmap.pgoff,
		       self->mmap.pid, self->mmap.filename, MAP__FUNCTION,
407 408
		       session->cwd, session->cwdlen);

409
	if (thread == NULL || map == NULL)
410 411 412 413
		goto out_problem;

	thread__insert_map(thread, map);
	return 0;
414

415 416
out_problem:
	dump_printf("problem processing PERF_RECORD_MMAP, skipping event.\n");
417 418 419
	return 0;
}

420
int event__process_task(event_t *self, struct perf_session *session)
421
{
422 423
	struct thread *thread = perf_session__findnew(session, self->fork.pid);
	struct thread *parent = perf_session__findnew(session, self->fork.ppid);
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443

	dump_printf("(%d:%d):(%d:%d)\n", self->fork.pid, self->fork.tid,
		    self->fork.ppid, self->fork.ptid);
	/*
	 * A thread clone will have the same PID for both parent and child.
	 */
	if (thread == parent)
		return 0;

	if (self->header.type == PERF_RECORD_EXIT)
		return 0;

	if (thread == NULL || parent == NULL ||
	    thread__fork(thread, parent) < 0) {
		dump_printf("problem processing PERF_RECORD_FORK, skipping event.\n");
		return -1;
	}

	return 0;
}
444

445 446 447 448
void thread__find_addr_map(struct thread *self,
			   struct perf_session *session, u8 cpumode,
			   enum map_type type, u64 addr,
			   struct addr_location *al)
449
{
450
	struct map_groups *mg = &self->mg;
451

452
	al->thread = self;
453 454
	al->addr = addr;

455
	if (cpumode == PERF_RECORD_MISC_KERNEL) {
456
		al->level = 'k';
457
		mg = &session->kmaps;
458
	} else if (cpumode == PERF_RECORD_MISC_USER)
459 460 461 462 463 464 465
		al->level = '.';
	else {
		al->level = 'H';
		al->map = NULL;
		return;
	}
try_again:
466
	al->map = map_groups__find(mg, type, al->addr);
467 468 469 470 471 472 473 474 475 476
	if (al->map == NULL) {
		/*
		 * If this is outside of all known maps, and is a negative
		 * address, try to look it up in the kernel dso, as it might be
		 * a vsyscall or vdso (which executes in user-mode).
		 *
		 * XXX This is nasty, we should have a symbol list in the
		 * "[vdso]" dso, but for now lets use the old trick of looking
		 * in the whole kernel symbol list.
		 */
477 478
		if ((long long)al->addr < 0 && mg != &session->kmaps) {
			mg = &session->kmaps;
479 480
			goto try_again;
		}
481
	} else
482
		al->addr = al->map->map_ip(al->map, al->addr);
483 484 485 486 487 488 489 490 491 492
}

void thread__find_addr_location(struct thread *self,
				struct perf_session *session, u8 cpumode,
				enum map_type type, u64 addr,
				struct addr_location *al,
				symbol_filter_t filter)
{
	thread__find_addr_map(self, session, cpumode, type, addr, al);
	if (al->map != NULL)
493
		al->sym = map__find_symbol(al->map, al->addr, filter);
494 495
	else
		al->sym = NULL;
496 497
}

498 499 500 501 502 503 504 505 506 507 508 509 510
static void dso__calc_col_width(struct dso *self)
{
	if (!symbol_conf.col_width_list_str && !symbol_conf.field_sep &&
	    (!symbol_conf.dso_list ||
	     strlist__has_entry(symbol_conf.dso_list, self->name))) {
		unsigned int slen = strlen(self->name);
		if (slen > dsos__col_width)
			dsos__col_width = slen;
	}

	self->slen_calculated = 1;
}

511 512
int event__preprocess_sample(const event_t *self, struct perf_session *session,
			     struct addr_location *al, symbol_filter_t filter)
513 514
{
	u8 cpumode = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
515
	struct thread *thread = perf_session__findnew(session, self->ip.pid);
516 517 518 519

	if (thread == NULL)
		return -1;

520 521 522 523
	if (symbol_conf.comm_list &&
	    !strlist__has_entry(symbol_conf.comm_list, thread->comm))
		goto out_filtered;

524 525
	dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);

526 527
	thread__find_addr_map(thread, session, cpumode, MAP__FUNCTION,
			      self->ip.ip, al);
528 529 530
	dump_printf(" ...... dso: %s\n",
		    al->map ? al->map->dso->long_name :
			al->level == 'H' ? "[hypervisor]" : "<not found>");
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551
	al->sym = NULL;

	if (al->map) {
		if (symbol_conf.dso_list &&
		    (!al->map || !al->map->dso ||
		     !(strlist__has_entry(symbol_conf.dso_list,
					  al->map->dso->short_name) ||
		       (al->map->dso->short_name != al->map->dso->long_name &&
			strlist__has_entry(symbol_conf.dso_list,
					   al->map->dso->long_name)))))
			goto out_filtered;
		/*
		 * We have to do this here as we may have a dso with no symbol
		 * hit that has a name longer than the ones with symbols
		 * sampled.
		 */
		if (!sort_dso.elide && !al->map->dso->slen_calculated)
			dso__calc_col_width(al->map->dso);

		al->sym = map__find_symbol(al->map, al->addr, filter);
	}
552 553 554 555 556 557 558 559 560 561

	if (symbol_conf.sym_list && al->sym &&
	    !strlist__has_entry(symbol_conf.sym_list, al->sym->name))
		goto out_filtered;

	al->filtered = false;
	return 0;

out_filtered:
	al->filtered = true;
562 563
	return 0;
}
564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630

int event__parse_sample(event_t *event, u64 type, struct sample_data *data)
{
	u64 *array = event->sample.array;

	if (type & PERF_SAMPLE_IP) {
		data->ip = event->ip.ip;
		array++;
	}

	if (type & PERF_SAMPLE_TID) {
		u32 *p = (u32 *)array;
		data->pid = p[0];
		data->tid = p[1];
		array++;
	}

	if (type & PERF_SAMPLE_TIME) {
		data->time = *array;
		array++;
	}

	if (type & PERF_SAMPLE_ADDR) {
		data->addr = *array;
		array++;
	}

	if (type & PERF_SAMPLE_ID) {
		data->id = *array;
		array++;
	}

	if (type & PERF_SAMPLE_STREAM_ID) {
		data->stream_id = *array;
		array++;
	}

	if (type & PERF_SAMPLE_CPU) {
		u32 *p = (u32 *)array;
		data->cpu = *p;
		array++;
	}

	if (type & PERF_SAMPLE_PERIOD) {
		data->period = *array;
		array++;
	}

	if (type & PERF_SAMPLE_READ) {
		pr_debug("PERF_SAMPLE_READ is unsuported for now\n");
		return -1;
	}

	if (type & PERF_SAMPLE_CALLCHAIN) {
		data->callchain = (struct ip_callchain *)array;
		array += 1 + data->callchain->nr;
	}

	if (type & PERF_SAMPLE_RAW) {
		u32 *p = (u32 *)array;
		data->raw_size = *p;
		p++;
		data->raw_data = p;
	}

	return 0;
}