code-reading.c 13.2 KB
Newer Older
B
Borislav Petkov 已提交
1
#include <linux/types.h>
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
#include <stdlib.h>
#include <unistd.h>
#include <stdio.h>
#include <ctype.h>
#include <string.h>

#include "parse-events.h"
#include "evlist.h"
#include "evsel.h"
#include "thread_map.h"
#include "cpumap.h"
#include "machine.h"
#include "event.h"
#include "thread.h"

#include "tests.h"

#define BUFSZ	1024
#define READLEN	128

22 23 24 25 26
struct state {
	u64 done[1024];
	size_t done_cnt;
};

27 28 29 30 31 32 33 34 35
static unsigned int hex(char c)
{
	if (c >= '0' && c <= '9')
		return c - '0';
	if (c >= 'a' && c <= 'f')
		return c - 'a' + 10;
	return c - 'A' + 10;
}

36 37
static size_t read_objdump_line(const char *line, size_t line_len, void *buf,
			      size_t len)
38 39
{
	const char *p;
40
	size_t i, j = 0;
41 42 43 44

	/* Skip to a colon */
	p = strchr(line, ':');
	if (!p)
45
		return 0;
46 47 48
	i = p + 1 - line;

	/* Read bytes */
49
	while (j < len) {
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67
		char c1, c2;

		/* Skip spaces */
		for (; i < line_len; i++) {
			if (!isspace(line[i]))
				break;
		}
		/* Get 2 hex digits */
		if (i >= line_len || !isxdigit(line[i]))
			break;
		c1 = line[i++];
		if (i >= line_len || !isxdigit(line[i]))
			break;
		c2 = line[i++];
		/* Followed by a space */
		if (i < line_len && line[i] && !isspace(line[i]))
			break;
		/* Store byte */
68 69 70
		*(unsigned char *)buf = (hex(c1) << 4) | hex(c2);
		buf += 1;
		j++;
71
	}
72 73
	/* return number of successfully read bytes */
	return j;
74 75
}

76
static int read_objdump_output(FILE *f, void *buf, size_t *len, u64 start_addr)
77 78
{
	char *line = NULL;
79
	size_t line_len, off_last = 0;
80 81
	ssize_t ret;
	int err = 0;
82
	u64 addr, last_addr = start_addr;
83 84 85 86

	while (off_last < *len) {
		size_t off, read_bytes, written_bytes;
		unsigned char tmp[BUFSZ];
87 88 89 90 91 92 93 94 95

		ret = getline(&line, &line_len, f);
		if (feof(f))
			break;
		if (ret < 0) {
			pr_debug("getline failed\n");
			err = -1;
			break;
		}
96 97 98 99 100 101 102 103

		/* read objdump data into temporary buffer */
		read_bytes = read_objdump_line(line, ret, tmp, sizeof(tmp));
		if (!read_bytes)
			continue;

		if (sscanf(line, "%"PRIx64, &addr) != 1)
			continue;
104 105 106 107 108
		if (addr < last_addr) {
			pr_debug("addr going backwards, read beyond section?\n");
			break;
		}
		last_addr = addr;
109 110 111 112 113 114 115 116 117

		/* copy it from temporary buffer to 'buf' according
		 * to address on current objdump line */
		off = addr - start_addr;
		if (off >= *len)
			break;
		written_bytes = MIN(read_bytes, *len - off);
		memcpy(buf + off, tmp, written_bytes);
		off_last = off + written_bytes;
118 119
	}

120 121 122
	/* len returns number of bytes that could not be read */
	*len -= off_last;

123 124 125 126 127 128 129 130 131 132 133 134 135
	free(line);

	return err;
}

static int read_via_objdump(const char *filename, u64 addr, void *buf,
			    size_t len)
{
	char cmd[PATH_MAX * 2];
	const char *fmt;
	FILE *f;
	int ret;

136
	fmt = "%s -z -d --start-address=0x%"PRIx64" --stop-address=0x%"PRIx64" %s";
137 138 139 140 141 142 143
	ret = snprintf(cmd, sizeof(cmd), fmt, "objdump", addr, addr + len,
		       filename);
	if (ret <= 0 || (size_t)ret >= sizeof(cmd))
		return -1;

	pr_debug("Objdump command is: %s\n", cmd);

144 145 146
	/* Ignore objdump errors */
	strcat(cmd, " 2>/dev/null");

147 148 149 150 151 152
	f = popen(cmd, "r");
	if (!f) {
		pr_debug("popen failed\n");
		return -1;
	}

153
	ret = read_objdump_output(f, buf, &len, addr);
154 155 156 157 158 159 160 161 162 163 164
	if (len) {
		pr_debug("objdump read too few bytes\n");
		if (!ret)
			ret = len;
	}

	pclose(f);

	return ret;
}

165 166 167 168 169 170 171 172 173 174 175 176
static void dump_buf(unsigned char *buf, size_t len)
{
	size_t i;

	for (i = 0; i < len; i++) {
		pr_debug("0x%02x ", buf[i]);
		if (i % 16 == 15)
			pr_debug("\n");
	}
	pr_debug("\n");
}

177
static int read_object_code(u64 addr, size_t len, u8 cpumode,
178
			    struct thread *thread, struct state *state)
179 180 181 182 183 184 185 186 187 188
{
	struct addr_location al;
	unsigned char buf1[BUFSZ];
	unsigned char buf2[BUFSZ];
	size_t ret_len;
	u64 objdump_addr;
	int ret;

	pr_debug("Reading object code for memory address: %#"PRIx64"\n", addr);

189
	thread__find_addr_map(thread, cpumode, MAP__FUNCTION, addr, &al);
190 191 192 193 194 195 196
	if (!al.map || !al.map->dso) {
		pr_debug("thread__find_addr_map failed\n");
		return -1;
	}

	pr_debug("File is: %s\n", al.map->dso->long_name);

197 198
	if (al.map->dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS &&
	    !dso__is_kcore(al.map->dso)) {
199 200 201 202 203 204 205 206 207 208 209 210 211 212
		pr_debug("Unexpected kernel address - skipping\n");
		return 0;
	}

	pr_debug("On file address is: %#"PRIx64"\n", al.addr);

	if (len > BUFSZ)
		len = BUFSZ;

	/* Do not go off the map */
	if (addr + len > al.map->end)
		len = al.map->end - addr;

	/* Read the object code using perf */
213 214
	ret_len = dso__data_read_offset(al.map->dso, thread->mg->machine,
					al.addr, buf1, len);
215 216 217 218 219 220 221 222 223 224 225 226
	if (ret_len != len) {
		pr_debug("dso__data_read_offset failed\n");
		return -1;
	}

	/*
	 * Converting addresses for use by objdump requires more information.
	 * map__load() does that.  See map__rip_2objdump() for details.
	 */
	if (map__load(al.map, NULL))
		return -1;

227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
	/* objdump struggles with kcore - try each map only once */
	if (dso__is_kcore(al.map->dso)) {
		size_t d;

		for (d = 0; d < state->done_cnt; d++) {
			if (state->done[d] == al.map->start) {
				pr_debug("kcore map tested already");
				pr_debug(" - skipping\n");
				return 0;
			}
		}
		if (state->done_cnt >= ARRAY_SIZE(state->done)) {
			pr_debug("Too many kcore maps - skipping\n");
			return 0;
		}
		state->done[state->done_cnt++] = al.map->start;
	}

245 246 247 248 249 250 251 252 253 254 255
	/* Read the object code using objdump */
	objdump_addr = map__rip_2objdump(al.map, al.addr);
	ret = read_via_objdump(al.map->dso->long_name, objdump_addr, buf2, len);
	if (ret > 0) {
		/*
		 * The kernel maps are inaccurate - assume objdump is right in
		 * that case.
		 */
		if (cpumode == PERF_RECORD_MISC_KERNEL ||
		    cpumode == PERF_RECORD_MISC_GUEST_KERNEL) {
			len -= ret;
256
			if (len) {
257
				pr_debug("Reducing len to %zu\n", len);
258 259 260 261 262 263 264 265 266
			} else if (dso__is_kcore(al.map->dso)) {
				/*
				 * objdump cannot handle very large segments
				 * that may be found in kcore.
				 */
				pr_debug("objdump failed for kcore");
				pr_debug(" - skipping\n");
				return 0;
			} else {
267
				return -1;
268
			}
269 270 271 272 273 274 275 276 277 278
		}
	}
	if (ret < 0) {
		pr_debug("read_via_objdump failed\n");
		return -1;
	}

	/* The results should be identical */
	if (memcmp(buf1, buf2, len)) {
		pr_debug("Bytes read differ from those read by objdump\n");
279 280 281 282
		pr_debug("buf1 (dso):\n");
		dump_buf(buf1, len);
		pr_debug("buf2 (objdump):\n");
		dump_buf(buf2, len);
283 284 285 286 287 288 289 290 291
		return -1;
	}
	pr_debug("Bytes read match those read by objdump\n");

	return 0;
}

static int process_sample_event(struct machine *machine,
				struct perf_evlist *evlist,
292
				union perf_event *event, struct state *state)
293 294 295 296
{
	struct perf_sample sample;
	struct thread *thread;
	u8 cpumode;
297
	int ret;
298 299 300 301 302 303

	if (perf_evlist__parse_sample(evlist, event, &sample)) {
		pr_debug("perf_evlist__parse_sample failed\n");
		return -1;
	}

304
	thread = machine__findnew_thread(machine, sample.pid, sample.tid);
305 306 307 308 309 310 311
	if (!thread) {
		pr_debug("machine__findnew_thread failed\n");
		return -1;
	}

	cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;

312 313 314
	ret = read_object_code(sample.ip, READLEN, cpumode, thread, state);
	thread__put(thread);
	return ret;
315 316 317
}

static int process_event(struct machine *machine, struct perf_evlist *evlist,
318
			 union perf_event *event, struct state *state)
319 320
{
	if (event->header.type == PERF_RECORD_SAMPLE)
321
		return process_sample_event(machine, evlist, event, state);
322

323 324 325 326 327 328 329 330 331 332 333 334 335
	if (event->header.type == PERF_RECORD_THROTTLE ||
	    event->header.type == PERF_RECORD_UNTHROTTLE)
		return 0;

	if (event->header.type < PERF_RECORD_MAX) {
		int ret;

		ret = machine__process_event(machine, event, NULL);
		if (ret < 0)
			pr_debug("machine__process_event failed, event type %u\n",
				 event->header.type);
		return ret;
	}
336 337 338 339

	return 0;
}

340 341
static int process_events(struct machine *machine, struct perf_evlist *evlist,
			  struct state *state)
342 343 344 345 346 347
{
	union perf_event *event;
	int i, ret;

	for (i = 0; i < evlist->nr_mmaps; i++) {
		while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
348
			ret = process_event(machine, evlist, event, state);
349
			perf_evlist__mmap_consume(evlist, i);
350 351 352 353 354 355 356 357 358 359 360 361 362 363
			if (ret < 0)
				return ret;
		}
	}
	return 0;
}

static int comp(const void *a, const void *b)
{
	return *(int *)a - *(int *)b;
}

static void do_sort_something(void)
{
364
	int buf[40960], i;
365

366 367
	for (i = 0; i < (int)ARRAY_SIZE(buf); i++)
		buf[i] = ARRAY_SIZE(buf) - i - 1;
368

369
	qsort(buf, ARRAY_SIZE(buf), sizeof(int), comp);
370

371
	for (i = 0; i < (int)ARRAY_SIZE(buf); i++) {
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
		if (buf[i] != i) {
			pr_debug("qsort failed\n");
			break;
		}
	}
}

static void sort_something(void)
{
	int i;

	for (i = 0; i < 10; i++)
		do_sort_something();
}

static void syscall_something(void)
{
	int pipefd[2];
	int i;

	for (i = 0; i < 1000; i++) {
		if (pipe(pipefd) < 0) {
			pr_debug("pipe failed\n");
			break;
		}
		close(pipefd[1]);
		close(pipefd[0]);
	}
}

static void fs_something(void)
{
	const char *test_file_name = "temp-perf-code-reading-test-file--";
	FILE *f;
	int i;

	for (i = 0; i < 1000; i++) {
		f = fopen(test_file_name, "w+");
		if (f) {
			fclose(f);
			unlink(test_file_name);
		}
	}
}

static void do_something(void)
{
	fs_something();

	sort_something();

	syscall_something();
}

enum {
	TEST_CODE_READING_OK,
	TEST_CODE_READING_NO_VMLINUX,
429
	TEST_CODE_READING_NO_KCORE,
430
	TEST_CODE_READING_NO_ACCESS,
431
	TEST_CODE_READING_NO_KERNEL_OBJ,
432 433
};

434
static int do_test_code_reading(bool try_kcore)
435 436 437
{
	struct machine *machine;
	struct thread *thread;
438
	struct record_opts opts = {
439 440 441 442 443 444 445 446
		.mmap_pages	     = UINT_MAX,
		.user_freq	     = UINT_MAX,
		.user_interval	     = ULLONG_MAX,
		.freq		     = 4000,
		.target		     = {
			.uses_mmap   = true,
		},
	};
447 448 449
	struct state state = {
		.done_cnt = 0,
	};
450 451 452 453 454 455 456
	struct thread_map *threads = NULL;
	struct cpu_map *cpus = NULL;
	struct perf_evlist *evlist = NULL;
	struct perf_evsel *evsel = NULL;
	int err = -1, ret;
	pid_t pid;
	struct map *map;
457
	bool have_vmlinux, have_kcore, excl_kernel = false;
458 459 460

	pid = getpid();

461
	machine = machine__new_host();
462 463 464 465 466 467 468

	ret = machine__create_kernel_maps(machine);
	if (ret < 0) {
		pr_debug("machine__create_kernel_maps failed\n");
		goto out_err;
	}

469 470 471 472
	/* Force the use of kallsyms instead of vmlinux to try kcore */
	if (try_kcore)
		symbol_conf.kallsyms_name = "/proc/kallsyms";

473
	/* Load kernel map */
474
	map = machine__kernel_map(machine);
475 476 477 478 479
	ret = map__load(map, NULL);
	if (ret < 0) {
		pr_debug("map__load failed\n");
		goto out_err;
	}
480 481 482 483 484 485 486 487 488
	have_vmlinux = dso__is_vmlinux(map->dso);
	have_kcore = dso__is_kcore(map->dso);

	/* 2nd time through we just try kcore */
	if (try_kcore && !have_kcore)
		return TEST_CODE_READING_NO_KCORE;

	/* No point getting kernel events if there is no kernel object */
	if (!have_vmlinux && !have_kcore)
489 490 491 492 493 494 495 496 497
		excl_kernel = true;

	threads = thread_map__new_by_tid(pid);
	if (!threads) {
		pr_debug("thread_map__new_by_tid failed\n");
		goto out_err;
	}

	ret = perf_event__synthesize_thread_map(NULL, threads,
498
						perf_event__process, machine, false, 500);
499 500 501 502 503
	if (ret < 0) {
		pr_debug("perf_event__synthesize_thread_map failed\n");
		goto out_err;
	}

504
	thread = machine__findnew_thread(machine, pid, pid);
505 506
	if (!thread) {
		pr_debug("machine__findnew_thread failed\n");
507
		goto out_put;
508 509 510 511 512
	}

	cpus = cpu_map__new(NULL);
	if (!cpus) {
		pr_debug("cpu_map__new failed\n");
513
		goto out_put;
514 515 516 517 518 519 520 521
	}

	while (1) {
		const char *str;

		evlist = perf_evlist__new();
		if (!evlist) {
			pr_debug("perf_evlist__new failed\n");
522
			goto out_put;
523 524 525 526 527 528 529 530 531
		}

		perf_evlist__set_maps(evlist, cpus, threads);

		if (excl_kernel)
			str = "cycles:u";
		else
			str = "cycles";
		pr_debug("Parsing event '%s'\n", str);
532
		ret = parse_events(evlist, str, NULL);
533 534
		if (ret < 0) {
			pr_debug("parse_events failed\n");
535
			goto out_put;
536 537 538 539 540 541 542 543 544 545 546 547 548 549
		}

		perf_evlist__config(evlist, &opts);

		evsel = perf_evlist__first(evlist);

		evsel->attr.comm = 1;
		evsel->attr.disabled = 1;
		evsel->attr.enable_on_exec = 0;

		ret = perf_evlist__open(evlist);
		if (ret < 0) {
			if (!excl_kernel) {
				excl_kernel = true;
550 551 552 553 554 555 556
				/*
				 * Both cpus and threads are now owned by evlist
				 * and will be freed by following perf_evlist__set_maps
				 * call. Getting refference to keep them alive.
				 */
				cpu_map__get(cpus);
				thread_map__get(threads);
557
				perf_evlist__set_maps(evlist, NULL, NULL);
558 559 560 561 562
				perf_evlist__delete(evlist);
				evlist = NULL;
				continue;
			}
			pr_debug("perf_evlist__open failed\n");
563
			goto out_put;
564 565 566 567 568 569 570
		}
		break;
	}

	ret = perf_evlist__mmap(evlist, UINT_MAX, false);
	if (ret < 0) {
		pr_debug("perf_evlist__mmap failed\n");
571
		goto out_put;
572 573 574 575 576 577 578 579
	}

	perf_evlist__enable(evlist);

	do_something();

	perf_evlist__disable(evlist);

580
	ret = process_events(machine, evlist, &state);
581
	if (ret < 0)
582
		goto out_put;
583

584 585 586
	if (!have_vmlinux && !have_kcore && !try_kcore)
		err = TEST_CODE_READING_NO_KERNEL_OBJ;
	else if (!have_vmlinux && !try_kcore)
587 588 589 590 591
		err = TEST_CODE_READING_NO_VMLINUX;
	else if (excl_kernel)
		err = TEST_CODE_READING_NO_ACCESS;
	else
		err = TEST_CODE_READING_OK;
592 593
out_put:
	thread__put(thread);
594
out_err:
595

596 597
	if (evlist) {
		perf_evlist__delete(evlist);
598
	} else {
599
		cpu_map__put(cpus);
600
		thread_map__put(threads);
601
	}
602
	machine__delete_threads(machine);
603
	machine__delete(machine);
604 605 606 607

	return err;
}

608
int test__code_reading(int subtest __maybe_unused)
609 610 611
{
	int ret;

612 613 614
	ret = do_test_code_reading(false);
	if (!ret)
		ret = do_test_code_reading(true);
615 616 617 618 619

	switch (ret) {
	case TEST_CODE_READING_OK:
		return 0;
	case TEST_CODE_READING_NO_VMLINUX:
620
		pr_debug("no vmlinux\n");
621
		return 0;
622
	case TEST_CODE_READING_NO_KCORE:
623
		pr_debug("no kcore\n");
624
		return 0;
625
	case TEST_CODE_READING_NO_ACCESS:
626
		pr_debug("no access\n");
627
		return 0;
628
	case TEST_CODE_READING_NO_KERNEL_OBJ:
629
		pr_debug("no kernel obj\n");
630
		return 0;
631 632 633 634
	default:
		return -1;
	};
}