1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * builtin-ftrace.c 4 * 5 * Copyright (c) 2013 LG Electronics, Namhyung Kim <namhyung@kernel.org> 6 * Copyright (c) 2020 Changbin Du <changbin.du@gmail.com>, significant enhancement. 7 */ 8 9 #include "builtin.h" 10 11 #include <errno.h> 12 #include <unistd.h> 13 #include <signal.h> 14 #include <stdlib.h> 15 #include <fcntl.h> 16 #include <inttypes.h> 17 #include <math.h> 18 #include <poll.h> 19 #include <ctype.h> 20 #include <linux/capability.h> 21 #include <linux/string.h> 22 #include <sys/stat.h> 23 24 #include "debug.h" 25 #include <subcmd/pager.h> 26 #include <subcmd/parse-options.h> 27 #include <api/io.h> 28 #include <api/fs/tracing_path.h> 29 #include "evlist.h" 30 #include "target.h" 31 #include "cpumap.h" 32 #include "hashmap.h" 33 #include "thread_map.h" 34 #include "strfilter.h" 35 #include "util/cap.h" 36 #include "util/config.h" 37 #include "util/ftrace.h" 38 #include "util/stat.h" 39 #include "util/units.h" 40 #include "util/parse-sublevel-options.h" 41 42 #define DEFAULT_TRACER "function_graph" 43 44 static volatile sig_atomic_t workload_exec_errno; 45 static volatile sig_atomic_t done; 46 47 static struct stats latency_stats; /* for tracepoints */ 48 49 static char tracing_instance[PATH_MAX]; /* Trace instance directory */ 50 51 static void sig_handler(int sig __maybe_unused) 52 { 53 done = true; 54 } 55 56 /* 57 * evlist__prepare_workload will send a SIGUSR1 if the fork fails, since 58 * we asked by setting its exec_error to the function below, 59 * ftrace__workload_exec_failed_signal. 60 * 61 * XXX We need to handle this more appropriately, emitting an error, etc. 62 */ 63 static void ftrace__workload_exec_failed_signal(int signo __maybe_unused, 64 siginfo_t *info __maybe_unused, 65 void *ucontext __maybe_unused) 66 { 67 workload_exec_errno = info->si_value.sival_int; 68 done = true; 69 } 70 71 static bool check_ftrace_capable(void) 72 { 73 bool used_root; 74 75 if (perf_cap__capable(CAP_PERFMON, &used_root)) 76 return true; 77 78 if (!used_root && perf_cap__capable(CAP_SYS_ADMIN, &used_root)) 79 return true; 80 81 pr_err("ftrace only works for %s!\n", 82 used_root ? "root" 83 : "users with the CAP_PERFMON or CAP_SYS_ADMIN capability" 84 ); 85 return false; 86 } 87 88 static bool is_ftrace_supported(void) 89 { 90 char *file; 91 bool supported = false; 92 93 file = get_tracing_file("set_ftrace_pid"); 94 if (!file) { 95 pr_debug("cannot get tracing file set_ftrace_pid\n"); 96 return false; 97 } 98 99 if (!access(file, F_OK)) 100 supported = true; 101 102 put_tracing_file(file); 103 return supported; 104 } 105 106 /* 107 * Wrapper to test if a file in directory .../tracing/instances/XXX 108 * exists. If so return the .../tracing/instances/XXX file for use. 109 * Otherwise the file exists only in directory .../tracing and 110 * is applicable to all instances, for example file available_filter_functions. 111 * Return that file name in this case. 112 * 113 * This functions works similar to get_tracing_file() and expects its caller 114 * to free the returned file name. 115 * 116 * The global variable tracing_instance is set in init_tracing_instance() 117 * called at the beginning to a process specific tracing subdirectory. 118 */ 119 static char *get_tracing_instance_file(const char *name) 120 { 121 char *file; 122 123 if (asprintf(&file, "%s/%s", tracing_instance, name) < 0) 124 return NULL; 125 126 if (!access(file, F_OK)) 127 return file; 128 129 free(file); 130 file = get_tracing_file(name); 131 return file; 132 } 133 134 static int __write_tracing_file(const char *name, const char *val, bool append) 135 { 136 char *file; 137 int fd, ret = -1; 138 ssize_t size = strlen(val); 139 int flags = O_WRONLY; 140 char errbuf[512]; 141 char *val_copy; 142 143 file = get_tracing_instance_file(name); 144 if (!file) { 145 pr_debug("cannot get tracing file: %s\n", name); 146 return -1; 147 } 148 149 if (append) 150 flags |= O_APPEND; 151 else 152 flags |= O_TRUNC; 153 154 fd = open(file, flags); 155 if (fd < 0) { 156 pr_debug("cannot open tracing file: %s: %s\n", 157 name, str_error_r(errno, errbuf, sizeof(errbuf))); 158 goto out; 159 } 160 161 /* 162 * Copy the original value and append a '\n'. Without this, 163 * the kernel can hide possible errors. 164 */ 165 val_copy = strdup(val); 166 if (!val_copy) 167 goto out_close; 168 val_copy[size] = '\n'; 169 170 if (write(fd, val_copy, size + 1) == size + 1) 171 ret = 0; 172 else 173 pr_debug("write '%s' to tracing/%s failed: %s\n", 174 val, name, str_error_r(errno, errbuf, sizeof(errbuf))); 175 176 free(val_copy); 177 out_close: 178 close(fd); 179 out: 180 put_tracing_file(file); 181 return ret; 182 } 183 184 static int write_tracing_file(const char *name, const char *val) 185 { 186 return __write_tracing_file(name, val, false); 187 } 188 189 static int append_tracing_file(const char *name, const char *val) 190 { 191 return __write_tracing_file(name, val, true); 192 } 193 194 static int read_tracing_file_to_stdout(const char *name) 195 { 196 char buf[4096]; 197 char *file; 198 int fd; 199 int ret = -1; 200 201 file = get_tracing_instance_file(name); 202 if (!file) { 203 pr_debug("cannot get tracing file: %s\n", name); 204 return -1; 205 } 206 207 fd = open(file, O_RDONLY); 208 if (fd < 0) { 209 pr_debug("cannot open tracing file: %s: %s\n", 210 name, str_error_r(errno, buf, sizeof(buf))); 211 goto out; 212 } 213 214 /* read contents to stdout */ 215 while (true) { 216 int n = read(fd, buf, sizeof(buf)); 217 if (n == 0) 218 break; 219 else if (n < 0) 220 goto out_close; 221 222 if (fwrite(buf, n, 1, stdout) != 1) 223 goto out_close; 224 } 225 ret = 0; 226 227 out_close: 228 close(fd); 229 out: 230 put_tracing_file(file); 231 return ret; 232 } 233 234 static int read_tracing_file_by_line(const char *name, 235 void (*cb)(char *str, void *arg), 236 void *cb_arg) 237 { 238 char *line = NULL; 239 size_t len = 0; 240 char *file; 241 FILE *fp; 242 243 file = get_tracing_instance_file(name); 244 if (!file) { 245 pr_debug("cannot get tracing file: %s\n", name); 246 return -1; 247 } 248 249 fp = fopen(file, "r"); 250 if (fp == NULL) { 251 pr_debug("cannot open tracing file: %s\n", name); 252 put_tracing_file(file); 253 return -1; 254 } 255 256 while (getline(&line, &len, fp) != -1) { 257 cb(line, cb_arg); 258 } 259 260 if (line) 261 free(line); 262 263 fclose(fp); 264 put_tracing_file(file); 265 return 0; 266 } 267 268 static int write_tracing_file_int(const char *name, int value) 269 { 270 char buf[16]; 271 272 snprintf(buf, sizeof(buf), "%d", value); 273 if (write_tracing_file(name, buf) < 0) 274 return -1; 275 276 return 0; 277 } 278 279 static int write_tracing_option_file(const char *name, const char *val) 280 { 281 char *file; 282 int ret; 283 284 if (asprintf(&file, "options/%s", name) < 0) 285 return -1; 286 287 ret = __write_tracing_file(file, val, false); 288 free(file); 289 return ret; 290 } 291 292 static int reset_tracing_cpu(void); 293 static void reset_tracing_filters(void); 294 295 static void reset_tracing_options(struct perf_ftrace *ftrace __maybe_unused) 296 { 297 write_tracing_option_file("function-fork", "0"); 298 write_tracing_option_file("func_stack_trace", "0"); 299 write_tracing_option_file("sleep-time", "1"); 300 write_tracing_option_file("funcgraph-irqs", "1"); 301 write_tracing_option_file("funcgraph-proc", "0"); 302 write_tracing_option_file("funcgraph-abstime", "0"); 303 write_tracing_option_file("funcgraph-tail", "0"); 304 write_tracing_option_file("funcgraph-args", "0"); 305 write_tracing_option_file("funcgraph-retval", "0"); 306 write_tracing_option_file("funcgraph-retval-hex", "0"); 307 write_tracing_option_file("funcgraph-retaddr", "0"); 308 write_tracing_option_file("latency-format", "0"); 309 write_tracing_option_file("irq-info", "0"); 310 } 311 312 static int reset_tracing_files(struct perf_ftrace *ftrace __maybe_unused) 313 { 314 if (write_tracing_file("tracing_on", "0") < 0) 315 return -1; 316 317 if (write_tracing_file("current_tracer", "nop") < 0) 318 return -1; 319 320 if (write_tracing_file("set_ftrace_pid", " ") < 0) 321 return -1; 322 323 if (reset_tracing_cpu() < 0) 324 return -1; 325 326 if (write_tracing_file("max_graph_depth", "0") < 0) 327 return -1; 328 329 if (write_tracing_file("tracing_thresh", "0") < 0) 330 return -1; 331 332 reset_tracing_filters(); 333 reset_tracing_options(ftrace); 334 return 0; 335 } 336 337 /* Remove .../tracing/instances/XXX subdirectory created with 338 * init_tracing_instance(). 339 */ 340 static void exit_tracing_instance(void) 341 { 342 if (rmdir(tracing_instance)) 343 pr_err("failed to delete tracing/instances directory\n"); 344 } 345 346 /* Create subdirectory within .../tracing/instances/XXX to have session 347 * or process specific setup. To delete this setup, simply remove the 348 * subdirectory. 349 */ 350 static int init_tracing_instance(void) 351 { 352 char dirname[] = "instances/perf-ftrace-XXXXXX"; 353 char *path; 354 355 path = get_tracing_file(dirname); 356 if (!path) 357 goto error; 358 strncpy(tracing_instance, path, sizeof(tracing_instance) - 1); 359 put_tracing_file(path); 360 path = mkdtemp(tracing_instance); 361 if (!path) 362 goto error; 363 return 0; 364 365 error: 366 pr_err("failed to create tracing/instances directory\n"); 367 return -1; 368 } 369 370 static int set_tracing_pid(struct perf_ftrace *ftrace) 371 { 372 int i; 373 char buf[16]; 374 375 if (target__has_cpu(&ftrace->target)) 376 return 0; 377 378 for (i = 0; i < perf_thread_map__nr(ftrace->evlist->core.threads); i++) { 379 scnprintf(buf, sizeof(buf), "%d", 380 perf_thread_map__pid(ftrace->evlist->core.threads, i)); 381 if (append_tracing_file("set_ftrace_pid", buf) < 0) 382 return -1; 383 } 384 return 0; 385 } 386 387 static int set_tracing_cpumask(struct perf_cpu_map *cpumap) 388 { 389 char *cpumask; 390 size_t mask_size; 391 int ret; 392 int last_cpu; 393 394 last_cpu = perf_cpu_map__cpu(cpumap, perf_cpu_map__nr(cpumap) - 1).cpu; 395 mask_size = last_cpu / 4 + 2; /* one more byte for EOS */ 396 mask_size += last_cpu / 32; /* ',' is needed for every 32th cpus */ 397 398 cpumask = malloc(mask_size); 399 if (cpumask == NULL) { 400 pr_debug("failed to allocate cpu mask\n"); 401 return -1; 402 } 403 404 cpu_map__snprint_mask(cpumap, cpumask, mask_size); 405 406 ret = write_tracing_file("tracing_cpumask", cpumask); 407 408 free(cpumask); 409 return ret; 410 } 411 412 static int set_tracing_cpu(struct perf_ftrace *ftrace) 413 { 414 struct perf_cpu_map *cpumap = ftrace->evlist->core.user_requested_cpus; 415 416 if (!target__has_cpu(&ftrace->target)) 417 return 0; 418 419 return set_tracing_cpumask(cpumap); 420 } 421 422 static int set_tracing_func_stack_trace(struct perf_ftrace *ftrace) 423 { 424 if (!ftrace->func_stack_trace) 425 return 0; 426 427 if (write_tracing_option_file("func_stack_trace", "1") < 0) 428 return -1; 429 430 return 0; 431 } 432 433 static int set_tracing_func_irqinfo(struct perf_ftrace *ftrace) 434 { 435 if (!ftrace->func_irq_info) 436 return 0; 437 438 if (write_tracing_option_file("irq-info", "1") < 0) 439 return -1; 440 441 return 0; 442 } 443 444 static int reset_tracing_cpu(void) 445 { 446 struct perf_cpu_map *cpumap = perf_cpu_map__new_online_cpus(); 447 int ret; 448 449 ret = set_tracing_cpumask(cpumap); 450 perf_cpu_map__put(cpumap); 451 return ret; 452 } 453 454 static int __set_tracing_filter(const char *filter_file, struct list_head *funcs) 455 { 456 struct filter_entry *pos; 457 458 list_for_each_entry(pos, funcs, list) { 459 if (append_tracing_file(filter_file, pos->name) < 0) 460 return -1; 461 } 462 463 return 0; 464 } 465 466 static int set_tracing_filters(struct perf_ftrace *ftrace) 467 { 468 int ret; 469 470 ret = __set_tracing_filter("set_ftrace_filter", &ftrace->filters); 471 if (ret < 0) 472 return ret; 473 474 ret = __set_tracing_filter("set_ftrace_notrace", &ftrace->notrace); 475 if (ret < 0) 476 return ret; 477 478 ret = __set_tracing_filter("set_graph_function", &ftrace->graph_funcs); 479 if (ret < 0) 480 return ret; 481 482 /* old kernels do not have this filter */ 483 __set_tracing_filter("set_graph_notrace", &ftrace->nograph_funcs); 484 485 return ret; 486 } 487 488 static void reset_tracing_filters(void) 489 { 490 write_tracing_file("set_ftrace_filter", " "); 491 write_tracing_file("set_ftrace_notrace", " "); 492 write_tracing_file("set_graph_function", " "); 493 write_tracing_file("set_graph_notrace", " "); 494 } 495 496 static int set_tracing_depth(struct perf_ftrace *ftrace) 497 { 498 if (ftrace->graph_depth == 0) 499 return 0; 500 501 if (ftrace->graph_depth < 0) { 502 pr_err("invalid graph depth: %d\n", ftrace->graph_depth); 503 return -1; 504 } 505 506 if (write_tracing_file_int("max_graph_depth", ftrace->graph_depth) < 0) 507 return -1; 508 509 return 0; 510 } 511 512 static int set_tracing_percpu_buffer_size(struct perf_ftrace *ftrace) 513 { 514 int ret; 515 516 if (ftrace->percpu_buffer_size == 0) 517 return 0; 518 519 ret = write_tracing_file_int("buffer_size_kb", 520 ftrace->percpu_buffer_size / 1024); 521 if (ret < 0) 522 return ret; 523 524 return 0; 525 } 526 527 static int set_tracing_trace_inherit(struct perf_ftrace *ftrace) 528 { 529 if (!ftrace->inherit) 530 return 0; 531 532 if (write_tracing_option_file("function-fork", "1") < 0) 533 return -1; 534 535 return 0; 536 } 537 538 static int set_tracing_sleep_time(struct perf_ftrace *ftrace) 539 { 540 if (!ftrace->graph_nosleep_time) 541 return 0; 542 543 if (write_tracing_option_file("sleep-time", "0") < 0) 544 return -1; 545 546 return 0; 547 } 548 549 static int set_tracing_funcgraph_args(struct perf_ftrace *ftrace) 550 { 551 if (ftrace->graph_args) { 552 if (write_tracing_option_file("funcgraph-args", "1") < 0) 553 return -1; 554 } 555 556 return 0; 557 } 558 559 static int set_tracing_funcgraph_retval(struct perf_ftrace *ftrace) 560 { 561 if (ftrace->graph_retval || ftrace->graph_retval_hex) { 562 if (write_tracing_option_file("funcgraph-retval", "1") < 0) 563 return -1; 564 } 565 566 if (ftrace->graph_retval_hex) { 567 if (write_tracing_option_file("funcgraph-retval-hex", "1") < 0) 568 return -1; 569 } 570 571 return 0; 572 } 573 574 static int set_tracing_funcgraph_retaddr(struct perf_ftrace *ftrace) 575 { 576 if (ftrace->graph_retaddr) { 577 if (write_tracing_option_file("funcgraph-retaddr", "1") < 0) 578 return -1; 579 } 580 581 return 0; 582 } 583 584 static int set_tracing_funcgraph_irqs(struct perf_ftrace *ftrace) 585 { 586 if (!ftrace->graph_noirqs) 587 return 0; 588 589 if (write_tracing_option_file("funcgraph-irqs", "0") < 0) 590 return -1; 591 592 return 0; 593 } 594 595 static int set_tracing_funcgraph_verbose(struct perf_ftrace *ftrace) 596 { 597 if (!ftrace->graph_verbose) 598 return 0; 599 600 if (write_tracing_option_file("funcgraph-proc", "1") < 0) 601 return -1; 602 603 if (write_tracing_option_file("funcgraph-abstime", "1") < 0) 604 return -1; 605 606 if (write_tracing_option_file("latency-format", "1") < 0) 607 return -1; 608 609 return 0; 610 } 611 612 static int set_tracing_funcgraph_tail(struct perf_ftrace *ftrace) 613 { 614 if (!ftrace->graph_tail) 615 return 0; 616 617 if (write_tracing_option_file("funcgraph-tail", "1") < 0) 618 return -1; 619 620 return 0; 621 } 622 623 static int set_tracing_thresh(struct perf_ftrace *ftrace) 624 { 625 int ret; 626 627 if (ftrace->graph_thresh == 0) 628 return 0; 629 630 ret = write_tracing_file_int("tracing_thresh", ftrace->graph_thresh); 631 if (ret < 0) 632 return ret; 633 634 return 0; 635 } 636 637 static int set_tracing_options(struct perf_ftrace *ftrace) 638 { 639 if (set_tracing_pid(ftrace) < 0) { 640 pr_err("failed to set ftrace pid\n"); 641 return -1; 642 } 643 644 if (set_tracing_cpu(ftrace) < 0) { 645 pr_err("failed to set tracing cpumask\n"); 646 return -1; 647 } 648 649 if (set_tracing_func_stack_trace(ftrace) < 0) { 650 pr_err("failed to set tracing option func_stack_trace\n"); 651 return -1; 652 } 653 654 if (set_tracing_func_irqinfo(ftrace) < 0) { 655 pr_err("failed to set tracing option irq-info\n"); 656 return -1; 657 } 658 659 if (set_tracing_filters(ftrace) < 0) { 660 pr_err("failed to set tracing filters\n"); 661 return -1; 662 } 663 664 if (set_tracing_depth(ftrace) < 0) { 665 pr_err("failed to set graph depth\n"); 666 return -1; 667 } 668 669 if (set_tracing_percpu_buffer_size(ftrace) < 0) { 670 pr_err("failed to set tracing per-cpu buffer size\n"); 671 return -1; 672 } 673 674 if (set_tracing_trace_inherit(ftrace) < 0) { 675 pr_err("failed to set tracing option function-fork\n"); 676 return -1; 677 } 678 679 if (set_tracing_sleep_time(ftrace) < 0) { 680 pr_err("failed to set tracing option sleep-time\n"); 681 return -1; 682 } 683 684 if (set_tracing_funcgraph_args(ftrace) < 0) { 685 pr_err("failed to set tracing option funcgraph-args\n"); 686 return -1; 687 } 688 689 if (set_tracing_funcgraph_retval(ftrace) < 0) { 690 pr_err("failed to set tracing option funcgraph-retval\n"); 691 return -1; 692 } 693 694 if (set_tracing_funcgraph_retaddr(ftrace) < 0) { 695 pr_err("failed to set tracing option funcgraph-retaddr\n"); 696 return -1; 697 } 698 699 if (set_tracing_funcgraph_irqs(ftrace) < 0) { 700 pr_err("failed to set tracing option funcgraph-irqs\n"); 701 return -1; 702 } 703 704 if (set_tracing_funcgraph_verbose(ftrace) < 0) { 705 pr_err("failed to set tracing option funcgraph-proc/funcgraph-abstime\n"); 706 return -1; 707 } 708 709 if (set_tracing_thresh(ftrace) < 0) { 710 pr_err("failed to set tracing thresh\n"); 711 return -1; 712 } 713 714 if (set_tracing_funcgraph_tail(ftrace) < 0) { 715 pr_err("failed to set tracing option funcgraph-tail\n"); 716 return -1; 717 } 718 719 return 0; 720 } 721 722 static void select_tracer(struct perf_ftrace *ftrace) 723 { 724 bool graph = !list_empty(&ftrace->graph_funcs) || 725 !list_empty(&ftrace->nograph_funcs); 726 bool func = !list_empty(&ftrace->filters) || 727 !list_empty(&ftrace->notrace); 728 729 /* The function_graph has priority over function tracer. */ 730 if (graph) 731 ftrace->tracer = "function_graph"; 732 else if (func) 733 ftrace->tracer = "function"; 734 /* Otherwise, the default tracer is used. */ 735 736 pr_debug("%s tracer is used\n", ftrace->tracer); 737 } 738 739 static int __cmd_ftrace(struct perf_ftrace *ftrace) 740 { 741 char *trace_file; 742 int trace_fd; 743 char buf[4096]; 744 struct pollfd pollfd = { 745 .events = POLLIN, 746 }; 747 748 select_tracer(ftrace); 749 750 if (init_tracing_instance() < 0) 751 goto out; 752 753 if (reset_tracing_files(ftrace) < 0) { 754 pr_err("failed to reset ftrace\n"); 755 goto out_reset; 756 } 757 758 /* reset ftrace buffer */ 759 if (write_tracing_file("trace", "0") < 0) 760 goto out_reset; 761 762 if (set_tracing_options(ftrace) < 0) 763 goto out_reset; 764 765 if (write_tracing_file("current_tracer", ftrace->tracer) < 0) { 766 pr_err("failed to set current_tracer to %s\n", ftrace->tracer); 767 goto out_reset; 768 } 769 770 setup_pager(); 771 772 trace_file = get_tracing_instance_file("trace_pipe"); 773 if (!trace_file) { 774 pr_err("failed to open trace_pipe\n"); 775 goto out_reset; 776 } 777 778 trace_fd = open(trace_file, O_RDONLY); 779 780 put_tracing_file(trace_file); 781 782 if (trace_fd < 0) { 783 pr_err("failed to open trace_pipe\n"); 784 goto out_reset; 785 } 786 787 fcntl(trace_fd, F_SETFL, O_NONBLOCK); 788 pollfd.fd = trace_fd; 789 790 /* display column headers */ 791 read_tracing_file_to_stdout("trace"); 792 793 if (!ftrace->target.initial_delay) { 794 if (write_tracing_file("tracing_on", "1") < 0) { 795 pr_err("can't enable tracing\n"); 796 goto out_close_fd; 797 } 798 } 799 800 evlist__start_workload(ftrace->evlist); 801 802 if (ftrace->target.initial_delay > 0) { 803 usleep(ftrace->target.initial_delay * 1000); 804 if (write_tracing_file("tracing_on", "1") < 0) { 805 pr_err("can't enable tracing\n"); 806 goto out_close_fd; 807 } 808 } 809 810 while (!done) { 811 if (poll(&pollfd, 1, -1) < 0) 812 break; 813 814 if (pollfd.revents & POLLIN) { 815 int n = read(trace_fd, buf, sizeof(buf)); 816 if (n < 0) 817 break; 818 if (fwrite(buf, n, 1, stdout) != 1) 819 break; 820 /* flush output since stdout is in full buffering mode due to pager */ 821 fflush(stdout); 822 } 823 } 824 825 write_tracing_file("tracing_on", "0"); 826 827 if (workload_exec_errno) { 828 const char *emsg = str_error_r(workload_exec_errno, buf, sizeof(buf)); 829 /* flush stdout first so below error msg appears at the end. */ 830 fflush(stdout); 831 pr_err("workload failed: %s\n", emsg); 832 goto out_close_fd; 833 } 834 835 /* read remaining buffer contents */ 836 while (true) { 837 int n = read(trace_fd, buf, sizeof(buf)); 838 if (n <= 0) 839 break; 840 if (fwrite(buf, n, 1, stdout) != 1) 841 break; 842 } 843 844 out_close_fd: 845 close(trace_fd); 846 out_reset: 847 exit_tracing_instance(); 848 out: 849 return (done && !workload_exec_errno) ? 0 : -1; 850 } 851 852 static void make_histogram(struct perf_ftrace *ftrace, int buckets[], 853 char *buf, size_t len, char *linebuf) 854 { 855 int min_latency = ftrace->min_latency; 856 int max_latency = ftrace->max_latency; 857 unsigned int bucket_num = ftrace->bucket_num; 858 char *p, *q; 859 char *unit; 860 double num; 861 int i; 862 863 /* ensure NUL termination */ 864 buf[len] = '\0'; 865 866 /* handle data line by line */ 867 for (p = buf; (q = strchr(p, '\n')) != NULL; p = q + 1) { 868 *q = '\0'; 869 /* move it to the line buffer */ 870 strcat(linebuf, p); 871 872 /* 873 * parse trace output to get function duration like in 874 * 875 * # tracer: function_graph 876 * # 877 * # CPU DURATION FUNCTION CALLS 878 * # | | | | | | | 879 * 1) + 10.291 us | do_filp_open(); 880 * 1) 4.889 us | do_filp_open(); 881 * 1) 6.086 us | do_filp_open(); 882 * 883 */ 884 if (linebuf[0] == '#') 885 goto next; 886 887 /* ignore CPU */ 888 p = strchr(linebuf, ')'); 889 if (p == NULL) 890 p = linebuf; 891 892 while (*p && !isdigit(*p) && (*p != '|')) 893 p++; 894 895 /* no duration */ 896 if (*p == '\0' || *p == '|') 897 goto next; 898 899 num = strtod(p, &unit); 900 if (!unit || strncmp(unit, " us", 3)) 901 goto next; 902 903 if (ftrace->use_nsec) 904 num *= 1000; 905 906 i = 0; 907 if (num < min_latency) 908 goto do_inc; 909 910 num -= min_latency; 911 912 if (!ftrace->bucket_range) { 913 i = log2(num); 914 if (i < 0) 915 i = 0; 916 } else { 917 // Less than 1 unit (ms or ns), or, in the future, 918 // than the min latency desired. 919 if (num > 0) // 1st entry: [ 1 unit .. bucket_range units ] 920 i = num / ftrace->bucket_range + 1; 921 if (num >= max_latency - min_latency) 922 i = bucket_num -1; 923 } 924 if ((unsigned)i >= bucket_num) 925 i = bucket_num - 1; 926 927 num += min_latency; 928 do_inc: 929 buckets[i]++; 930 update_stats(&latency_stats, num); 931 932 next: 933 /* empty the line buffer for the next output */ 934 linebuf[0] = '\0'; 935 } 936 937 /* preserve any remaining output (before newline) */ 938 strcat(linebuf, p); 939 } 940 941 static void display_histogram(struct perf_ftrace *ftrace, int buckets[]) 942 { 943 int min_latency = ftrace->min_latency; 944 bool use_nsec = ftrace->use_nsec; 945 unsigned int bucket_num = ftrace->bucket_num; 946 unsigned int i; 947 int total = 0; 948 int bar_total = 46; /* to fit in 80 column */ 949 char bar[] = "###############################################"; 950 int bar_len; 951 952 for (i = 0; i < bucket_num; i++) 953 total += buckets[i]; 954 955 if (total == 0) { 956 printf("No data found\n"); 957 return; 958 } 959 960 printf("# %14s | %10s | %-*s |\n", 961 " DURATION ", "COUNT", bar_total, "GRAPH"); 962 963 bar_len = buckets[0] * bar_total / total; 964 965 if (!ftrace->hide_empty || buckets[0]) 966 printf(" %4d - %4d %s | %10d | %.*s%*s |\n", 967 0, min_latency ?: 1, use_nsec ? "ns" : "us", 968 buckets[0], bar_len, bar, bar_total - bar_len, ""); 969 970 for (i = 1; i < bucket_num - 1; i++) { 971 unsigned int start, stop; 972 const char *unit = use_nsec ? "ns" : "us"; 973 974 if (ftrace->hide_empty && !buckets[i]) 975 continue; 976 if (!ftrace->bucket_range) { 977 start = (1 << (i - 1)); 978 stop = 1 << i; 979 980 if (start >= 1024) { 981 start >>= 10; 982 stop >>= 10; 983 unit = use_nsec ? "us" : "ms"; 984 } 985 } else { 986 start = (i - 1) * ftrace->bucket_range + min_latency; 987 stop = i * ftrace->bucket_range + min_latency; 988 989 if (start >= ftrace->max_latency) 990 break; 991 if (stop > ftrace->max_latency) 992 stop = ftrace->max_latency; 993 994 if (start >= 1000) { 995 double dstart = start / 1000.0, 996 dstop = stop / 1000.0; 997 printf(" %4.2f - %-4.2f", dstart, dstop); 998 unit = use_nsec ? "us" : "ms"; 999 goto print_bucket_info; 1000 } 1001 } 1002 1003 printf(" %4d - %4d", start, stop); 1004 print_bucket_info: 1005 bar_len = buckets[i] * bar_total / total; 1006 printf(" %s | %10d | %.*s%*s |\n", unit, buckets[i], bar_len, bar, 1007 bar_total - bar_len, ""); 1008 } 1009 1010 bar_len = buckets[bucket_num - 1] * bar_total / total; 1011 if (ftrace->hide_empty && !buckets[bucket_num - 1]) 1012 goto print_stats; 1013 if (!ftrace->bucket_range) { 1014 printf(" %4d - %-4s %s", 1, "...", use_nsec ? "ms" : "s "); 1015 } else { 1016 unsigned int upper_outlier = (bucket_num - 2) * ftrace->bucket_range + min_latency; 1017 if (upper_outlier > ftrace->max_latency) 1018 upper_outlier = ftrace->max_latency; 1019 1020 if (upper_outlier >= 1000) { 1021 double dstart = upper_outlier / 1000.0; 1022 1023 printf(" %4.2f - %-4s %s", dstart, "...", use_nsec ? "us" : "ms"); 1024 } else { 1025 printf(" %4d - %4s %s", upper_outlier, "...", use_nsec ? "ns" : "us"); 1026 } 1027 } 1028 printf(" | %10d | %.*s%*s |\n", buckets[bucket_num - 1], 1029 bar_len, bar, bar_total - bar_len, ""); 1030 1031 print_stats: 1032 printf("\n# statistics (in %s)\n", ftrace->use_nsec ? "nsec" : "usec"); 1033 printf(" total time: %20.0f\n", latency_stats.mean * latency_stats.n); 1034 printf(" avg time: %20.0f\n", latency_stats.mean); 1035 printf(" max time: %20"PRIu64"\n", latency_stats.max); 1036 printf(" min time: %20"PRIu64"\n", latency_stats.min); 1037 printf(" count: %20.0f\n", latency_stats.n); 1038 } 1039 1040 static int prepare_func_latency(struct perf_ftrace *ftrace) 1041 { 1042 char *trace_file; 1043 int fd; 1044 1045 if (ftrace->target.use_bpf) 1046 return perf_ftrace__latency_prepare_bpf(ftrace); 1047 1048 if (init_tracing_instance() < 0) 1049 return -1; 1050 1051 if (reset_tracing_files(ftrace) < 0) { 1052 pr_err("failed to reset ftrace\n"); 1053 return -1; 1054 } 1055 1056 /* reset ftrace buffer */ 1057 if (write_tracing_file("trace", "0") < 0) 1058 return -1; 1059 1060 if (set_tracing_options(ftrace) < 0) 1061 return -1; 1062 1063 /* force to use the function_graph tracer to track duration */ 1064 if (write_tracing_file("current_tracer", "function_graph") < 0) { 1065 pr_err("failed to set current_tracer to function_graph\n"); 1066 return -1; 1067 } 1068 1069 trace_file = get_tracing_instance_file("trace_pipe"); 1070 if (!trace_file) { 1071 pr_err("failed to open trace_pipe\n"); 1072 return -1; 1073 } 1074 1075 fd = open(trace_file, O_RDONLY); 1076 if (fd < 0) 1077 pr_err("failed to open trace_pipe\n"); 1078 1079 init_stats(&latency_stats); 1080 1081 put_tracing_file(trace_file); 1082 return fd; 1083 } 1084 1085 static int start_func_latency(struct perf_ftrace *ftrace) 1086 { 1087 if (ftrace->target.use_bpf) 1088 return perf_ftrace__latency_start_bpf(ftrace); 1089 1090 if (write_tracing_file("tracing_on", "1") < 0) { 1091 pr_err("can't enable tracing\n"); 1092 return -1; 1093 } 1094 1095 return 0; 1096 } 1097 1098 static int stop_func_latency(struct perf_ftrace *ftrace) 1099 { 1100 if (ftrace->target.use_bpf) 1101 return perf_ftrace__latency_stop_bpf(ftrace); 1102 1103 write_tracing_file("tracing_on", "0"); 1104 return 0; 1105 } 1106 1107 static int read_func_latency(struct perf_ftrace *ftrace, int buckets[]) 1108 { 1109 if (ftrace->target.use_bpf) 1110 return perf_ftrace__latency_read_bpf(ftrace, buckets, &latency_stats); 1111 1112 return 0; 1113 } 1114 1115 static int cleanup_func_latency(struct perf_ftrace *ftrace) 1116 { 1117 if (ftrace->target.use_bpf) 1118 return perf_ftrace__latency_cleanup_bpf(ftrace); 1119 1120 exit_tracing_instance(); 1121 return 0; 1122 } 1123 1124 static int __cmd_latency(struct perf_ftrace *ftrace) 1125 { 1126 int trace_fd; 1127 char buf[4096]; 1128 char line[256]; 1129 struct pollfd pollfd = { 1130 .events = POLLIN, 1131 }; 1132 int *buckets; 1133 1134 trace_fd = prepare_func_latency(ftrace); 1135 if (trace_fd < 0) 1136 goto out; 1137 1138 fcntl(trace_fd, F_SETFL, O_NONBLOCK); 1139 pollfd.fd = trace_fd; 1140 1141 if (start_func_latency(ftrace) < 0) 1142 goto out; 1143 1144 evlist__start_workload(ftrace->evlist); 1145 1146 buckets = calloc(ftrace->bucket_num, sizeof(*buckets)); 1147 if (buckets == NULL) { 1148 pr_err("failed to allocate memory for the buckets\n"); 1149 goto out; 1150 } 1151 1152 line[0] = '\0'; 1153 while (!done) { 1154 if (poll(&pollfd, 1, -1) < 0) 1155 break; 1156 1157 if (pollfd.revents & POLLIN) { 1158 int n = read(trace_fd, buf, sizeof(buf) - 1); 1159 if (n < 0) 1160 break; 1161 1162 make_histogram(ftrace, buckets, buf, n, line); 1163 } 1164 } 1165 1166 stop_func_latency(ftrace); 1167 1168 if (workload_exec_errno) { 1169 const char *emsg = str_error_r(workload_exec_errno, buf, sizeof(buf)); 1170 pr_err("workload failed: %s\n", emsg); 1171 goto out_free_buckets; 1172 } 1173 1174 /* read remaining buffer contents */ 1175 while (!ftrace->target.use_bpf) { 1176 int n = read(trace_fd, buf, sizeof(buf) - 1); 1177 if (n <= 0) 1178 break; 1179 make_histogram(ftrace, buckets, buf, n, line); 1180 } 1181 1182 read_func_latency(ftrace, buckets); 1183 1184 display_histogram(ftrace, buckets); 1185 1186 out_free_buckets: 1187 free(buckets); 1188 out: 1189 close(trace_fd); 1190 cleanup_func_latency(ftrace); 1191 1192 return (done && !workload_exec_errno) ? 0 : -1; 1193 } 1194 1195 static size_t profile_hash(long func, void *ctx __maybe_unused) 1196 { 1197 return str_hash((char *)func); 1198 } 1199 1200 static bool profile_equal(long func1, long func2, void *ctx __maybe_unused) 1201 { 1202 return !strcmp((char *)func1, (char *)func2); 1203 } 1204 1205 static int prepare_func_profile(struct perf_ftrace *ftrace) 1206 { 1207 ftrace->tracer = "function_graph"; 1208 ftrace->graph_tail = 1; 1209 ftrace->graph_verbose = 0; 1210 1211 ftrace->profile_hash = hashmap__new(profile_hash, profile_equal, NULL); 1212 if (ftrace->profile_hash == NULL) 1213 return -ENOMEM; 1214 1215 return 0; 1216 } 1217 1218 /* This is saved in a hashmap keyed by the function name */ 1219 struct ftrace_profile_data { 1220 struct stats st; 1221 }; 1222 1223 static int add_func_duration(struct perf_ftrace *ftrace, char *func, double time_ns) 1224 { 1225 struct ftrace_profile_data *prof = NULL; 1226 1227 if (!hashmap__find(ftrace->profile_hash, func, &prof)) { 1228 char *key = strdup(func); 1229 1230 if (key == NULL) 1231 return -ENOMEM; 1232 1233 prof = zalloc(sizeof(*prof)); 1234 if (prof == NULL) { 1235 free(key); 1236 return -ENOMEM; 1237 } 1238 1239 init_stats(&prof->st); 1240 hashmap__add(ftrace->profile_hash, key, prof); 1241 } 1242 1243 update_stats(&prof->st, time_ns); 1244 return 0; 1245 } 1246 1247 /* 1248 * The ftrace function_graph text output normally looks like below: 1249 * 1250 * CPU DURATION FUNCTION 1251 * 1252 * 0) | syscall_trace_enter.isra.0() { 1253 * 0) | __audit_syscall_entry() { 1254 * 0) | auditd_test_task() { 1255 * 0) 0.271 us | __rcu_read_lock(); 1256 * 0) 0.275 us | __rcu_read_unlock(); 1257 * 0) 1.254 us | } /\* auditd_test_task *\/ 1258 * 0) 0.279 us | ktime_get_coarse_real_ts64(); 1259 * 0) 2.227 us | } /\* __audit_syscall_entry *\/ 1260 * 0) 2.713 us | } /\* syscall_trace_enter.isra.0 *\/ 1261 * 1262 * Parse the line and get the duration and function name. 1263 */ 1264 static int parse_func_duration(struct perf_ftrace *ftrace, char *line, size_t len) 1265 { 1266 char *p; 1267 char *func; 1268 double duration; 1269 1270 /* skip CPU */ 1271 p = strchr(line, ')'); 1272 if (p == NULL) 1273 return 0; 1274 1275 /* get duration */ 1276 p = skip_spaces(p + 1); 1277 1278 /* no duration? */ 1279 if (p == NULL || *p == '|') 1280 return 0; 1281 1282 /* skip markers like '*' or '!' for longer than ms */ 1283 if (!isdigit(*p)) 1284 p++; 1285 1286 duration = strtod(p, &p); 1287 1288 if (strncmp(p, " us", 3)) { 1289 pr_debug("non-usec time found.. ignoring\n"); 1290 return 0; 1291 } 1292 1293 /* 1294 * profile stat keeps the max and min values as integer, 1295 * convert to nsec time so that we can have accurate max. 1296 */ 1297 duration *= 1000; 1298 1299 /* skip to the pipe */ 1300 while (p < line + len && *p != '|') 1301 p++; 1302 1303 if (*p++ != '|') 1304 return -EINVAL; 1305 1306 /* get function name */ 1307 func = skip_spaces(p); 1308 1309 /* skip the closing bracket and the start of comment */ 1310 if (*func == '}') 1311 func += 5; 1312 1313 /* remove semi-colon or end of comment at the end */ 1314 p = line + len - 1; 1315 while (!isalnum(*p) && *p != ']') { 1316 *p = '\0'; 1317 --p; 1318 } 1319 1320 return add_func_duration(ftrace, func, duration); 1321 } 1322 1323 enum perf_ftrace_profile_sort_key { 1324 PFP_SORT_TOTAL = 0, 1325 PFP_SORT_AVG, 1326 PFP_SORT_MAX, 1327 PFP_SORT_COUNT, 1328 PFP_SORT_NAME, 1329 }; 1330 1331 static enum perf_ftrace_profile_sort_key profile_sort = PFP_SORT_TOTAL; 1332 1333 static int cmp_profile_data(const void *a, const void *b) 1334 { 1335 const struct hashmap_entry *e1 = *(const struct hashmap_entry **)a; 1336 const struct hashmap_entry *e2 = *(const struct hashmap_entry **)b; 1337 struct ftrace_profile_data *p1 = e1->pvalue; 1338 struct ftrace_profile_data *p2 = e2->pvalue; 1339 double v1, v2; 1340 1341 switch (profile_sort) { 1342 case PFP_SORT_NAME: 1343 return strcmp(e1->pkey, e2->pkey); 1344 case PFP_SORT_AVG: 1345 v1 = p1->st.mean; 1346 v2 = p2->st.mean; 1347 break; 1348 case PFP_SORT_MAX: 1349 v1 = p1->st.max; 1350 v2 = p2->st.max; 1351 break; 1352 case PFP_SORT_COUNT: 1353 v1 = p1->st.n; 1354 v2 = p2->st.n; 1355 break; 1356 case PFP_SORT_TOTAL: 1357 default: 1358 v1 = p1->st.n * p1->st.mean; 1359 v2 = p2->st.n * p2->st.mean; 1360 break; 1361 } 1362 1363 if (v1 > v2) 1364 return -1; 1365 if (v1 < v2) 1366 return 1; 1367 return 0; 1368 } 1369 1370 static void print_profile_result(struct perf_ftrace *ftrace) 1371 { 1372 struct hashmap_entry *entry, **profile; 1373 size_t i, nr, bkt; 1374 1375 nr = hashmap__size(ftrace->profile_hash); 1376 if (nr == 0) 1377 return; 1378 1379 profile = calloc(nr, sizeof(*profile)); 1380 if (profile == NULL) { 1381 pr_err("failed to allocate memory for the result\n"); 1382 return; 1383 } 1384 1385 i = 0; 1386 hashmap__for_each_entry(ftrace->profile_hash, entry, bkt) 1387 profile[i++] = entry; 1388 1389 assert(i == nr); 1390 1391 //cmp_profile_data(profile[0], profile[1]); 1392 qsort(profile, nr, sizeof(*profile), cmp_profile_data); 1393 1394 printf("# %10s %10s %10s %10s %s\n", 1395 "Total (us)", "Avg (us)", "Max (us)", "Count", "Function"); 1396 1397 for (i = 0; i < nr; i++) { 1398 const char *name = profile[i]->pkey; 1399 struct ftrace_profile_data *p = profile[i]->pvalue; 1400 1401 printf("%12.3f %10.3f %6"PRIu64".%03"PRIu64" %10.0f %s\n", 1402 p->st.n * p->st.mean / 1000, p->st.mean / 1000, 1403 p->st.max / 1000, p->st.max % 1000, p->st.n, name); 1404 } 1405 1406 free(profile); 1407 1408 hashmap__for_each_entry(ftrace->profile_hash, entry, bkt) { 1409 free((char *)entry->pkey); 1410 free(entry->pvalue); 1411 } 1412 1413 hashmap__free(ftrace->profile_hash); 1414 ftrace->profile_hash = NULL; 1415 } 1416 1417 static int __cmd_profile(struct perf_ftrace *ftrace) 1418 { 1419 char *trace_file; 1420 int trace_fd; 1421 char buf[4096]; 1422 struct io io; 1423 char *line = NULL; 1424 size_t line_len = 0; 1425 1426 if (prepare_func_profile(ftrace) < 0) { 1427 pr_err("failed to prepare func profiler\n"); 1428 goto out; 1429 } 1430 1431 if (init_tracing_instance() < 0) 1432 goto out; 1433 1434 if (reset_tracing_files(ftrace) < 0) { 1435 pr_err("failed to reset ftrace\n"); 1436 goto out_reset; 1437 } 1438 1439 /* reset ftrace buffer */ 1440 if (write_tracing_file("trace", "0") < 0) 1441 goto out_reset; 1442 1443 if (set_tracing_options(ftrace) < 0) 1444 goto out_reset; 1445 1446 if (write_tracing_file("current_tracer", ftrace->tracer) < 0) { 1447 pr_err("failed to set current_tracer to %s\n", ftrace->tracer); 1448 goto out_reset; 1449 } 1450 1451 setup_pager(); 1452 1453 trace_file = get_tracing_instance_file("trace_pipe"); 1454 if (!trace_file) { 1455 pr_err("failed to open trace_pipe\n"); 1456 goto out_reset; 1457 } 1458 1459 trace_fd = open(trace_file, O_RDONLY); 1460 1461 put_tracing_file(trace_file); 1462 1463 if (trace_fd < 0) { 1464 pr_err("failed to open trace_pipe\n"); 1465 goto out_reset; 1466 } 1467 1468 fcntl(trace_fd, F_SETFL, O_NONBLOCK); 1469 1470 if (write_tracing_file("tracing_on", "1") < 0) { 1471 pr_err("can't enable tracing\n"); 1472 goto out_close_fd; 1473 } 1474 1475 evlist__start_workload(ftrace->evlist); 1476 1477 io__init(&io, trace_fd, buf, sizeof(buf)); 1478 io.timeout_ms = -1; 1479 1480 while (!done && !io.eof) { 1481 if (io__getline(&io, &line, &line_len) < 0) 1482 break; 1483 1484 if (parse_func_duration(ftrace, line, line_len) < 0) 1485 break; 1486 } 1487 1488 write_tracing_file("tracing_on", "0"); 1489 1490 if (workload_exec_errno) { 1491 const char *emsg = str_error_r(workload_exec_errno, buf, sizeof(buf)); 1492 /* flush stdout first so below error msg appears at the end. */ 1493 fflush(stdout); 1494 pr_err("workload failed: %s\n", emsg); 1495 goto out_free_line; 1496 } 1497 1498 /* read remaining buffer contents */ 1499 io.timeout_ms = 0; 1500 while (!io.eof) { 1501 if (io__getline(&io, &line, &line_len) < 0) 1502 break; 1503 1504 if (parse_func_duration(ftrace, line, line_len) < 0) 1505 break; 1506 } 1507 1508 print_profile_result(ftrace); 1509 1510 out_free_line: 1511 free(line); 1512 out_close_fd: 1513 close(trace_fd); 1514 out_reset: 1515 exit_tracing_instance(); 1516 out: 1517 return (done && !workload_exec_errno) ? 0 : -1; 1518 } 1519 1520 static int perf_ftrace_config(const char *var, const char *value, void *cb) 1521 { 1522 struct perf_ftrace *ftrace = cb; 1523 1524 if (!strstarts(var, "ftrace.")) 1525 return 0; 1526 1527 if (strcmp(var, "ftrace.tracer")) 1528 return -1; 1529 1530 if (!strcmp(value, "function_graph") || 1531 !strcmp(value, "function")) { 1532 ftrace->tracer = value; 1533 return 0; 1534 } 1535 1536 pr_err("Please select \"function_graph\" (default) or \"function\"\n"); 1537 return -1; 1538 } 1539 1540 static void list_function_cb(char *str, void *arg) 1541 { 1542 struct strfilter *filter = (struct strfilter *)arg; 1543 1544 if (strfilter__compare(filter, str)) 1545 printf("%s", str); 1546 } 1547 1548 static int opt_list_avail_functions(const struct option *opt __maybe_unused, 1549 const char *str, int unset) 1550 { 1551 struct strfilter *filter; 1552 const char *err = NULL; 1553 int ret; 1554 1555 if (unset || !str) 1556 return -1; 1557 1558 filter = strfilter__new(str, &err); 1559 if (!filter) 1560 return err ? -EINVAL : -ENOMEM; 1561 1562 ret = strfilter__or(filter, str, &err); 1563 if (ret == -EINVAL) { 1564 pr_err("Filter parse error at %td.\n", err - str + 1); 1565 pr_err("Source: \"%s\"\n", str); 1566 pr_err(" %*c\n", (int)(err - str + 1), '^'); 1567 strfilter__delete(filter); 1568 return ret; 1569 } 1570 1571 ret = read_tracing_file_by_line("available_filter_functions", 1572 list_function_cb, filter); 1573 strfilter__delete(filter); 1574 if (ret < 0) 1575 return ret; 1576 1577 exit(0); 1578 } 1579 1580 static int parse_filter_func(const struct option *opt, const char *str, 1581 int unset __maybe_unused) 1582 { 1583 struct list_head *head = opt->value; 1584 struct filter_entry *entry; 1585 1586 entry = malloc(sizeof(*entry) + strlen(str) + 1); 1587 if (entry == NULL) 1588 return -ENOMEM; 1589 1590 strcpy(entry->name, str); 1591 list_add_tail(&entry->list, head); 1592 1593 return 0; 1594 } 1595 1596 static void delete_filter_func(struct list_head *head) 1597 { 1598 struct filter_entry *pos, *tmp; 1599 1600 list_for_each_entry_safe(pos, tmp, head, list) { 1601 list_del_init(&pos->list); 1602 free(pos); 1603 } 1604 } 1605 1606 static int parse_filter_event(const struct option *opt, const char *str, 1607 int unset __maybe_unused) 1608 { 1609 struct list_head *head = opt->value; 1610 struct filter_entry *entry; 1611 char *s, *p; 1612 int ret = -ENOMEM; 1613 1614 s = strdup(str); 1615 if (s == NULL) 1616 return -ENOMEM; 1617 1618 while ((p = strsep(&s, ",")) != NULL) { 1619 entry = malloc(sizeof(*entry) + strlen(p) + 1); 1620 if (entry == NULL) 1621 goto out; 1622 1623 strcpy(entry->name, p); 1624 list_add_tail(&entry->list, head); 1625 } 1626 ret = 0; 1627 1628 out: 1629 free(s); 1630 return ret; 1631 } 1632 1633 static int parse_buffer_size(const struct option *opt, 1634 const char *str, int unset) 1635 { 1636 unsigned long *s = (unsigned long *)opt->value; 1637 static struct parse_tag tags_size[] = { 1638 { .tag = 'B', .mult = 1 }, 1639 { .tag = 'K', .mult = 1 << 10 }, 1640 { .tag = 'M', .mult = 1 << 20 }, 1641 { .tag = 'G', .mult = 1 << 30 }, 1642 { .tag = 0 }, 1643 }; 1644 unsigned long val; 1645 1646 if (unset) { 1647 *s = 0; 1648 return 0; 1649 } 1650 1651 val = parse_tag_value(str, tags_size); 1652 if (val != (unsigned long) -1) { 1653 if (val < 1024) { 1654 pr_err("buffer size too small, must larger than 1KB."); 1655 return -1; 1656 } 1657 *s = val; 1658 return 0; 1659 } 1660 1661 return -1; 1662 } 1663 1664 static int parse_func_tracer_opts(const struct option *opt, 1665 const char *str, int unset) 1666 { 1667 int ret; 1668 struct perf_ftrace *ftrace = (struct perf_ftrace *) opt->value; 1669 struct sublevel_option func_tracer_opts[] = { 1670 { .name = "call-graph", .value_ptr = &ftrace->func_stack_trace }, 1671 { .name = "irq-info", .value_ptr = &ftrace->func_irq_info }, 1672 { .name = NULL, } 1673 }; 1674 1675 if (unset) 1676 return 0; 1677 1678 ret = perf_parse_sublevel_options(str, func_tracer_opts); 1679 if (ret) 1680 return ret; 1681 1682 return 0; 1683 } 1684 1685 static int parse_graph_tracer_opts(const struct option *opt, 1686 const char *str, int unset) 1687 { 1688 int ret; 1689 struct perf_ftrace *ftrace = (struct perf_ftrace *) opt->value; 1690 struct sublevel_option graph_tracer_opts[] = { 1691 { .name = "args", .value_ptr = &ftrace->graph_args }, 1692 { .name = "retval", .value_ptr = &ftrace->graph_retval }, 1693 { .name = "retval-hex", .value_ptr = &ftrace->graph_retval_hex }, 1694 { .name = "retaddr", .value_ptr = &ftrace->graph_retaddr }, 1695 { .name = "nosleep-time", .value_ptr = &ftrace->graph_nosleep_time }, 1696 { .name = "noirqs", .value_ptr = &ftrace->graph_noirqs }, 1697 { .name = "verbose", .value_ptr = &ftrace->graph_verbose }, 1698 { .name = "thresh", .value_ptr = &ftrace->graph_thresh }, 1699 { .name = "depth", .value_ptr = &ftrace->graph_depth }, 1700 { .name = "tail", .value_ptr = &ftrace->graph_tail }, 1701 { .name = NULL, } 1702 }; 1703 1704 if (unset) 1705 return 0; 1706 1707 ret = perf_parse_sublevel_options(str, graph_tracer_opts); 1708 if (ret) 1709 return ret; 1710 1711 return 0; 1712 } 1713 1714 static int parse_sort_key(const struct option *opt, const char *str, int unset) 1715 { 1716 enum perf_ftrace_profile_sort_key *key = (void *)opt->value; 1717 1718 if (unset) 1719 return 0; 1720 1721 if (!strcmp(str, "total")) 1722 *key = PFP_SORT_TOTAL; 1723 else if (!strcmp(str, "avg")) 1724 *key = PFP_SORT_AVG; 1725 else if (!strcmp(str, "max")) 1726 *key = PFP_SORT_MAX; 1727 else if (!strcmp(str, "count")) 1728 *key = PFP_SORT_COUNT; 1729 else if (!strcmp(str, "name")) 1730 *key = PFP_SORT_NAME; 1731 else { 1732 pr_err("Unknown sort key: %s\n", str); 1733 return -1; 1734 } 1735 return 0; 1736 } 1737 1738 enum perf_ftrace_subcommand { 1739 PERF_FTRACE_NONE, 1740 PERF_FTRACE_TRACE, 1741 PERF_FTRACE_LATENCY, 1742 PERF_FTRACE_PROFILE, 1743 }; 1744 1745 int cmd_ftrace(int argc, const char **argv) 1746 { 1747 int ret; 1748 int (*cmd_func)(struct perf_ftrace *) = NULL; 1749 struct perf_ftrace ftrace = { 1750 .tracer = DEFAULT_TRACER, 1751 }; 1752 const struct option common_options[] = { 1753 OPT_STRING('p', "pid", &ftrace.target.pid, "pid", 1754 "Trace on existing process id"), 1755 /* TODO: Add short option -t after -t/--tracer can be removed. */ 1756 OPT_STRING(0, "tid", &ftrace.target.tid, "tid", 1757 "Trace on existing thread id (exclusive to --pid)"), 1758 OPT_INCR('v', "verbose", &verbose, 1759 "Be more verbose"), 1760 OPT_BOOLEAN('a', "all-cpus", &ftrace.target.system_wide, 1761 "System-wide collection from all CPUs"), 1762 OPT_STRING('C', "cpu", &ftrace.target.cpu_list, "cpu", 1763 "List of cpus to monitor"), 1764 OPT_END() 1765 }; 1766 const struct option ftrace_options[] = { 1767 OPT_STRING('t', "tracer", &ftrace.tracer, "tracer", 1768 "Tracer to use: function_graph(default) or function"), 1769 OPT_CALLBACK_DEFAULT('F', "funcs", NULL, "[FILTER]", 1770 "Show available functions to filter", 1771 opt_list_avail_functions, "*"), 1772 OPT_CALLBACK('T', "trace-funcs", &ftrace.filters, "func", 1773 "Trace given functions using function tracer", 1774 parse_filter_func), 1775 OPT_CALLBACK('N', "notrace-funcs", &ftrace.notrace, "func", 1776 "Do not trace given functions", parse_filter_func), 1777 OPT_CALLBACK(0, "func-opts", &ftrace, "options", 1778 "Function tracer options, available options: call-graph,irq-info", 1779 parse_func_tracer_opts), 1780 OPT_CALLBACK('G', "graph-funcs", &ftrace.graph_funcs, "func", 1781 "Trace given functions using function_graph tracer", 1782 parse_filter_func), 1783 OPT_CALLBACK('g', "nograph-funcs", &ftrace.nograph_funcs, "func", 1784 "Set nograph filter on given functions", parse_filter_func), 1785 OPT_CALLBACK(0, "graph-opts", &ftrace, "options", 1786 "Graph tracer options, available options: args,retval,retval-hex,retaddr,nosleep-time,noirqs,verbose,thresh=<n>,depth=<n>", 1787 parse_graph_tracer_opts), 1788 OPT_CALLBACK('m', "buffer-size", &ftrace.percpu_buffer_size, "size", 1789 "Size of per cpu buffer, needs to use a B, K, M or G suffix.", parse_buffer_size), 1790 OPT_BOOLEAN(0, "inherit", &ftrace.inherit, 1791 "Trace children processes"), 1792 OPT_INTEGER('D', "delay", &ftrace.target.initial_delay, 1793 "Number of milliseconds to wait before starting tracing after program start"), 1794 OPT_PARENT(common_options), 1795 }; 1796 const struct option latency_options[] = { 1797 OPT_CALLBACK('T', "trace-funcs", &ftrace.filters, "func", 1798 "Show latency of given function", parse_filter_func), 1799 OPT_CALLBACK('e', "events", &ftrace.event_pair, "event1,event2", 1800 "Show latency between the two events", parse_filter_event), 1801 #ifdef HAVE_BPF_SKEL 1802 OPT_BOOLEAN('b', "use-bpf", &ftrace.target.use_bpf, 1803 "Use BPF to measure function latency"), 1804 #endif 1805 OPT_BOOLEAN('n', "use-nsec", &ftrace.use_nsec, 1806 "Use nano-second histogram"), 1807 OPT_UINTEGER(0, "bucket-range", &ftrace.bucket_range, 1808 "Bucket range in ms or ns (-n/--use-nsec), default is log2() mode"), 1809 OPT_UINTEGER(0, "min-latency", &ftrace.min_latency, 1810 "Minimum latency (1st bucket). Works only with --bucket-range."), 1811 OPT_UINTEGER(0, "max-latency", &ftrace.max_latency, 1812 "Maximum latency (last bucket). Works only with --bucket-range."), 1813 OPT_BOOLEAN(0, "hide-empty", &ftrace.hide_empty, 1814 "Hide empty buckets in the histogram"), 1815 OPT_PARENT(common_options), 1816 }; 1817 const struct option profile_options[] = { 1818 OPT_CALLBACK('T', "trace-funcs", &ftrace.filters, "func", 1819 "Trace given functions using function tracer", 1820 parse_filter_func), 1821 OPT_CALLBACK('N', "notrace-funcs", &ftrace.notrace, "func", 1822 "Do not trace given functions", parse_filter_func), 1823 OPT_CALLBACK('G', "graph-funcs", &ftrace.graph_funcs, "func", 1824 "Trace given functions using function_graph tracer", 1825 parse_filter_func), 1826 OPT_CALLBACK('g', "nograph-funcs", &ftrace.nograph_funcs, "func", 1827 "Set nograph filter on given functions", parse_filter_func), 1828 OPT_CALLBACK('m', "buffer-size", &ftrace.percpu_buffer_size, "size", 1829 "Size of per cpu buffer, needs to use a B, K, M or G suffix.", parse_buffer_size), 1830 OPT_CALLBACK('s', "sort", &profile_sort, "key", 1831 "Sort result by key: total (default), avg, max, count, name.", 1832 parse_sort_key), 1833 OPT_CALLBACK(0, "graph-opts", &ftrace, "options", 1834 "Graph tracer options, available options: nosleep-time,noirqs,thresh=<n>,depth=<n>", 1835 parse_graph_tracer_opts), 1836 OPT_PARENT(common_options), 1837 }; 1838 const struct option *options = ftrace_options; 1839 1840 const char * const ftrace_usage[] = { 1841 "perf ftrace [<options>] [<command>]", 1842 "perf ftrace [<options>] -- [<command>] [<options>]", 1843 "perf ftrace {trace|latency|profile} [<options>] [<command>]", 1844 "perf ftrace {trace|latency|profile} [<options>] -- [<command>] [<options>]", 1845 NULL 1846 }; 1847 enum perf_ftrace_subcommand subcmd = PERF_FTRACE_NONE; 1848 1849 INIT_LIST_HEAD(&ftrace.filters); 1850 INIT_LIST_HEAD(&ftrace.notrace); 1851 INIT_LIST_HEAD(&ftrace.graph_funcs); 1852 INIT_LIST_HEAD(&ftrace.nograph_funcs); 1853 INIT_LIST_HEAD(&ftrace.event_pair); 1854 1855 signal(SIGINT, sig_handler); 1856 signal(SIGUSR1, sig_handler); 1857 signal(SIGCHLD, sig_handler); 1858 signal(SIGPIPE, sig_handler); 1859 1860 if (!check_ftrace_capable()) 1861 return -1; 1862 1863 if (!is_ftrace_supported()) { 1864 pr_err("ftrace is not supported on this system\n"); 1865 return -ENOTSUP; 1866 } 1867 1868 ret = perf_config(perf_ftrace_config, &ftrace); 1869 if (ret < 0) 1870 return -1; 1871 1872 if (argc > 1) { 1873 if (!strcmp(argv[1], "trace")) { 1874 subcmd = PERF_FTRACE_TRACE; 1875 } else if (!strcmp(argv[1], "latency")) { 1876 subcmd = PERF_FTRACE_LATENCY; 1877 options = latency_options; 1878 } else if (!strcmp(argv[1], "profile")) { 1879 subcmd = PERF_FTRACE_PROFILE; 1880 options = profile_options; 1881 } 1882 1883 if (subcmd != PERF_FTRACE_NONE) { 1884 argc--; 1885 argv++; 1886 } 1887 } 1888 /* for backward compatibility */ 1889 if (subcmd == PERF_FTRACE_NONE) 1890 subcmd = PERF_FTRACE_TRACE; 1891 1892 argc = parse_options(argc, argv, options, ftrace_usage, 1893 PARSE_OPT_STOP_AT_NON_OPTION); 1894 if (argc < 0) { 1895 ret = -EINVAL; 1896 goto out_delete_filters; 1897 } 1898 1899 /* Make system wide (-a) the default target. */ 1900 if (!argc && target__none(&ftrace.target)) 1901 ftrace.target.system_wide = true; 1902 1903 switch (subcmd) { 1904 case PERF_FTRACE_TRACE: 1905 cmd_func = __cmd_ftrace; 1906 break; 1907 case PERF_FTRACE_LATENCY: 1908 if (list_empty(&ftrace.filters) && list_empty(&ftrace.event_pair)) { 1909 pr_err("Should provide a function or events to measure\n"); 1910 parse_options_usage(ftrace_usage, options, "T", 1); 1911 parse_options_usage(NULL, options, "e", 1); 1912 ret = -EINVAL; 1913 goto out_delete_filters; 1914 } 1915 if (!list_empty(&ftrace.filters) && !list_empty(&ftrace.event_pair)) { 1916 pr_err("Please specify either of function or events\n"); 1917 parse_options_usage(ftrace_usage, options, "T", 1); 1918 parse_options_usage(NULL, options, "e", 1); 1919 ret = -EINVAL; 1920 goto out_delete_filters; 1921 } 1922 if (!list_empty(&ftrace.event_pair) && !ftrace.target.use_bpf) { 1923 pr_err("Event processing needs BPF\n"); 1924 parse_options_usage(ftrace_usage, options, "b", 1); 1925 parse_options_usage(NULL, options, "e", 1); 1926 ret = -EINVAL; 1927 goto out_delete_filters; 1928 } 1929 if (!ftrace.bucket_range && ftrace.min_latency) { 1930 pr_err("--min-latency works only with --bucket-range\n"); 1931 parse_options_usage(ftrace_usage, options, 1932 "min-latency", /*short_opt=*/false); 1933 ret = -EINVAL; 1934 goto out_delete_filters; 1935 } 1936 if (ftrace.bucket_range && !ftrace.min_latency) { 1937 /* default min latency should be the bucket range */ 1938 ftrace.min_latency = ftrace.bucket_range; 1939 } 1940 if (!ftrace.bucket_range && ftrace.max_latency) { 1941 pr_err("--max-latency works only with --bucket-range\n"); 1942 parse_options_usage(ftrace_usage, options, 1943 "max-latency", /*short_opt=*/false); 1944 ret = -EINVAL; 1945 goto out_delete_filters; 1946 } 1947 if (ftrace.bucket_range && ftrace.max_latency && 1948 ftrace.max_latency < ftrace.min_latency + ftrace.bucket_range) { 1949 /* we need at least 1 bucket excluding min and max buckets */ 1950 pr_err("--max-latency must be larger than min-latency + bucket-range\n"); 1951 parse_options_usage(ftrace_usage, options, 1952 "max-latency", /*short_opt=*/false); 1953 ret = -EINVAL; 1954 goto out_delete_filters; 1955 } 1956 /* set default unless max_latency is set and valid */ 1957 ftrace.bucket_num = NUM_BUCKET; 1958 if (ftrace.bucket_range) { 1959 if (ftrace.max_latency) 1960 ftrace.bucket_num = (ftrace.max_latency - ftrace.min_latency) / 1961 ftrace.bucket_range + 2; 1962 else 1963 /* default max latency should depend on bucket range and num_buckets */ 1964 ftrace.max_latency = (NUM_BUCKET - 2) * ftrace.bucket_range + 1965 ftrace.min_latency; 1966 } 1967 cmd_func = __cmd_latency; 1968 break; 1969 case PERF_FTRACE_PROFILE: 1970 cmd_func = __cmd_profile; 1971 break; 1972 case PERF_FTRACE_NONE: 1973 default: 1974 pr_err("Invalid subcommand\n"); 1975 ret = -EINVAL; 1976 goto out_delete_filters; 1977 } 1978 1979 ret = target__validate(&ftrace.target); 1980 if (ret) { 1981 char errbuf[512]; 1982 1983 target__strerror(&ftrace.target, ret, errbuf, 512); 1984 pr_err("%s\n", errbuf); 1985 goto out_delete_filters; 1986 } 1987 1988 ftrace.evlist = evlist__new(); 1989 if (ftrace.evlist == NULL) { 1990 ret = -ENOMEM; 1991 goto out_delete_filters; 1992 } 1993 1994 ret = evlist__create_maps(ftrace.evlist, &ftrace.target); 1995 if (ret < 0) 1996 goto out_delete_evlist; 1997 1998 if (argc) { 1999 ret = evlist__prepare_workload(ftrace.evlist, &ftrace.target, 2000 argv, false, 2001 ftrace__workload_exec_failed_signal); 2002 if (ret < 0) 2003 goto out_delete_evlist; 2004 } 2005 2006 ret = cmd_func(&ftrace); 2007 2008 out_delete_evlist: 2009 evlist__delete(ftrace.evlist); 2010 2011 out_delete_filters: 2012 delete_filter_func(&ftrace.filters); 2013 delete_filter_func(&ftrace.notrace); 2014 delete_filter_func(&ftrace.graph_funcs); 2015 delete_filter_func(&ftrace.nograph_funcs); 2016 delete_filter_func(&ftrace.event_pair); 2017 2018 return ret; 2019 } 2020