1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * builtin-kwork.c 4 * 5 * Copyright (c) 2022 Huawei Inc, Yang Jihong <yangjihong1@huawei.com> 6 */ 7 8 #include "builtin.h" 9 10 #include "util/data.h" 11 #include "util/evlist.h" 12 #include "util/evsel.h" 13 #include "util/header.h" 14 #include "util/kwork.h" 15 #include "util/debug.h" 16 #include "util/session.h" 17 #include "util/symbol.h" 18 #include "util/thread.h" 19 #include "util/string2.h" 20 #include "util/callchain.h" 21 #include "util/evsel_fprintf.h" 22 #include "util/util.h" 23 24 #include <subcmd/pager.h> 25 #include <subcmd/parse-options.h> 26 #include <traceevent/event-parse.h> 27 28 #include <errno.h> 29 #include <inttypes.h> 30 #include <signal.h> 31 #include <linux/err.h> 32 #include <linux/time64.h> 33 #include <linux/zalloc.h> 34 35 /* 36 * report header elements width 37 */ 38 #define PRINT_CPU_WIDTH 4 39 #define PRINT_COUNT_WIDTH 9 40 #define PRINT_RUNTIME_WIDTH 10 41 #define PRINT_LATENCY_WIDTH 10 42 #define PRINT_TIMESTAMP_WIDTH 17 43 #define PRINT_KWORK_NAME_WIDTH 30 44 #define RPINT_DECIMAL_WIDTH 3 45 #define PRINT_BRACKETPAIR_WIDTH 2 46 #define PRINT_TIME_UNIT_SEC_WIDTH 2 47 #define PRINT_TIME_UNIT_MESC_WIDTH 3 48 #define PRINT_RUNTIME_HEADER_WIDTH (PRINT_RUNTIME_WIDTH + PRINT_TIME_UNIT_MESC_WIDTH) 49 #define PRINT_LATENCY_HEADER_WIDTH (PRINT_LATENCY_WIDTH + PRINT_TIME_UNIT_MESC_WIDTH) 50 #define PRINT_TIMEHIST_CPU_WIDTH (PRINT_CPU_WIDTH + PRINT_BRACKETPAIR_WIDTH) 51 #define PRINT_TIMESTAMP_HEADER_WIDTH (PRINT_TIMESTAMP_WIDTH + PRINT_TIME_UNIT_SEC_WIDTH) 52 53 struct sort_dimension { 54 const char *name; 55 int (*cmp)(struct kwork_work *l, struct kwork_work *r); 56 struct list_head list; 57 }; 58 59 static int id_cmp(struct kwork_work *l, struct kwork_work *r) 60 { 61 if (l->cpu > r->cpu) 62 return 1; 63 if (l->cpu < r->cpu) 64 return -1; 65 66 if (l->id > r->id) 67 return 1; 68 if (l->id < r->id) 69 return -1; 70 71 return 0; 72 } 73 74 static int count_cmp(struct kwork_work *l, struct kwork_work *r) 75 { 76 if (l->nr_atoms > r->nr_atoms) 77 return 1; 78 if (l->nr_atoms < r->nr_atoms) 79 return -1; 80 81 return 0; 82 } 83 84 static int runtime_cmp(struct kwork_work *l, struct kwork_work *r) 85 { 86 if (l->total_runtime > r->total_runtime) 87 return 1; 88 if (l->total_runtime < r->total_runtime) 89 return -1; 90 91 return 0; 92 } 93 94 static int max_runtime_cmp(struct kwork_work *l, struct kwork_work *r) 95 { 96 if (l->max_runtime > r->max_runtime) 97 return 1; 98 if (l->max_runtime < r->max_runtime) 99 return -1; 100 101 return 0; 102 } 103 104 static int avg_latency_cmp(struct kwork_work *l, struct kwork_work *r) 105 { 106 u64 avgl, avgr; 107 108 if (!r->nr_atoms) 109 return 1; 110 if (!l->nr_atoms) 111 return -1; 112 113 avgl = l->total_latency / l->nr_atoms; 114 avgr = r->total_latency / r->nr_atoms; 115 116 if (avgl > avgr) 117 return 1; 118 if (avgl < avgr) 119 return -1; 120 121 return 0; 122 } 123 124 static int max_latency_cmp(struct kwork_work *l, struct kwork_work *r) 125 { 126 if (l->max_latency > r->max_latency) 127 return 1; 128 if (l->max_latency < r->max_latency) 129 return -1; 130 131 return 0; 132 } 133 134 static int sort_dimension__add(struct perf_kwork *kwork __maybe_unused, 135 const char *tok, struct list_head *list) 136 { 137 size_t i; 138 static struct sort_dimension max_sort_dimension = { 139 .name = "max", 140 .cmp = max_runtime_cmp, 141 }; 142 static struct sort_dimension id_sort_dimension = { 143 .name = "id", 144 .cmp = id_cmp, 145 }; 146 static struct sort_dimension runtime_sort_dimension = { 147 .name = "runtime", 148 .cmp = runtime_cmp, 149 }; 150 static struct sort_dimension count_sort_dimension = { 151 .name = "count", 152 .cmp = count_cmp, 153 }; 154 static struct sort_dimension avg_sort_dimension = { 155 .name = "avg", 156 .cmp = avg_latency_cmp, 157 }; 158 struct sort_dimension *available_sorts[] = { 159 &id_sort_dimension, 160 &max_sort_dimension, 161 &count_sort_dimension, 162 &runtime_sort_dimension, 163 &avg_sort_dimension, 164 }; 165 166 if (kwork->report == KWORK_REPORT_LATENCY) 167 max_sort_dimension.cmp = max_latency_cmp; 168 169 for (i = 0; i < ARRAY_SIZE(available_sorts); i++) { 170 if (!strcmp(available_sorts[i]->name, tok)) { 171 list_add_tail(&available_sorts[i]->list, list); 172 return 0; 173 } 174 } 175 176 return -1; 177 } 178 179 static void setup_sorting(struct perf_kwork *kwork, 180 const struct option *options, 181 const char * const usage_msg[]) 182 { 183 char *tmp, *tok, *str = strdup(kwork->sort_order); 184 185 for (tok = strtok_r(str, ", ", &tmp); 186 tok; tok = strtok_r(NULL, ", ", &tmp)) { 187 if (sort_dimension__add(kwork, tok, &kwork->sort_list) < 0) 188 usage_with_options_msg(usage_msg, options, 189 "Unknown --sort key: `%s'", tok); 190 } 191 192 pr_debug("Sort order: %s\n", kwork->sort_order); 193 free(str); 194 } 195 196 static struct kwork_atom *atom_new(struct perf_kwork *kwork, 197 struct perf_sample *sample) 198 { 199 unsigned long i; 200 struct kwork_atom_page *page; 201 struct kwork_atom *atom = NULL; 202 203 list_for_each_entry(page, &kwork->atom_page_list, list) { 204 if (!bitmap_full(page->bitmap, NR_ATOM_PER_PAGE)) { 205 i = find_first_zero_bit(page->bitmap, NR_ATOM_PER_PAGE); 206 BUG_ON(i >= NR_ATOM_PER_PAGE); 207 atom = &page->atoms[i]; 208 goto found_atom; 209 } 210 } 211 212 /* 213 * new page 214 */ 215 page = zalloc(sizeof(*page)); 216 if (page == NULL) { 217 pr_err("Failed to zalloc kwork atom page\n"); 218 return NULL; 219 } 220 221 i = 0; 222 atom = &page->atoms[0]; 223 list_add_tail(&page->list, &kwork->atom_page_list); 224 225 found_atom: 226 __set_bit(i, page->bitmap); 227 atom->time = sample->time; 228 atom->prev = NULL; 229 atom->page_addr = page; 230 atom->bit_inpage = i; 231 return atom; 232 } 233 234 static void atom_free(struct kwork_atom *atom) 235 { 236 if (atom->prev != NULL) 237 atom_free(atom->prev); 238 239 __clear_bit(atom->bit_inpage, 240 ((struct kwork_atom_page *)atom->page_addr)->bitmap); 241 } 242 243 static void atom_del(struct kwork_atom *atom) 244 { 245 list_del(&atom->list); 246 atom_free(atom); 247 } 248 249 static int work_cmp(struct list_head *list, 250 struct kwork_work *l, struct kwork_work *r) 251 { 252 int ret = 0; 253 struct sort_dimension *sort; 254 255 BUG_ON(list_empty(list)); 256 257 list_for_each_entry(sort, list, list) { 258 ret = sort->cmp(l, r); 259 if (ret) 260 return ret; 261 } 262 263 return ret; 264 } 265 266 static struct kwork_work *work_search(struct rb_root_cached *root, 267 struct kwork_work *key, 268 struct list_head *sort_list) 269 { 270 int cmp; 271 struct kwork_work *work; 272 struct rb_node *node = root->rb_root.rb_node; 273 274 while (node) { 275 work = container_of(node, struct kwork_work, node); 276 cmp = work_cmp(sort_list, key, work); 277 if (cmp > 0) 278 node = node->rb_left; 279 else if (cmp < 0) 280 node = node->rb_right; 281 else { 282 if (work->name == NULL) 283 work->name = key->name; 284 return work; 285 } 286 } 287 return NULL; 288 } 289 290 static void work_insert(struct rb_root_cached *root, 291 struct kwork_work *key, struct list_head *sort_list) 292 { 293 int cmp; 294 bool leftmost = true; 295 struct kwork_work *cur; 296 struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL; 297 298 while (*new) { 299 cur = container_of(*new, struct kwork_work, node); 300 parent = *new; 301 cmp = work_cmp(sort_list, key, cur); 302 303 if (cmp > 0) 304 new = &((*new)->rb_left); 305 else { 306 new = &((*new)->rb_right); 307 leftmost = false; 308 } 309 } 310 311 rb_link_node(&key->node, parent, new); 312 rb_insert_color_cached(&key->node, root, leftmost); 313 } 314 315 static struct kwork_work *work_new(struct kwork_work *key) 316 { 317 int i; 318 struct kwork_work *work = zalloc(sizeof(*work)); 319 320 if (work == NULL) { 321 pr_err("Failed to zalloc kwork work\n"); 322 return NULL; 323 } 324 325 for (i = 0; i < KWORK_TRACE_MAX; i++) 326 INIT_LIST_HEAD(&work->atom_list[i]); 327 328 work->id = key->id; 329 work->cpu = key->cpu; 330 work->name = key->name; 331 work->class = key->class; 332 return work; 333 } 334 335 static struct kwork_work *work_findnew(struct rb_root_cached *root, 336 struct kwork_work *key, 337 struct list_head *sort_list) 338 { 339 struct kwork_work *work = work_search(root, key, sort_list); 340 341 if (work != NULL) 342 return work; 343 344 work = work_new(key); 345 if (work) 346 work_insert(root, work, sort_list); 347 348 return work; 349 } 350 351 static void profile_update_timespan(struct perf_kwork *kwork, 352 struct perf_sample *sample) 353 { 354 if (!kwork->summary) 355 return; 356 357 if ((kwork->timestart == 0) || (kwork->timestart > sample->time)) 358 kwork->timestart = sample->time; 359 360 if (kwork->timeend < sample->time) 361 kwork->timeend = sample->time; 362 } 363 364 static bool profile_event_match(struct perf_kwork *kwork, 365 struct kwork_work *work, 366 struct perf_sample *sample) 367 { 368 int cpu = work->cpu; 369 u64 time = sample->time; 370 struct perf_time_interval *ptime = &kwork->ptime; 371 372 if ((kwork->cpu_list != NULL) && !test_bit(cpu, kwork->cpu_bitmap)) 373 return false; 374 375 if (((ptime->start != 0) && (ptime->start > time)) || 376 ((ptime->end != 0) && (ptime->end < time))) 377 return false; 378 379 if ((kwork->profile_name != NULL) && 380 (work->name != NULL) && 381 (strcmp(work->name, kwork->profile_name) != 0)) 382 return false; 383 384 profile_update_timespan(kwork, sample); 385 return true; 386 } 387 388 static int work_push_atom(struct perf_kwork *kwork, 389 struct kwork_class *class, 390 enum kwork_trace_type src_type, 391 enum kwork_trace_type dst_type, 392 struct evsel *evsel, 393 struct perf_sample *sample, 394 struct machine *machine, 395 struct kwork_work **ret_work) 396 { 397 struct kwork_atom *atom, *dst_atom; 398 struct kwork_work *work, key; 399 400 BUG_ON(class->work_init == NULL); 401 class->work_init(class, &key, evsel, sample, machine); 402 403 atom = atom_new(kwork, sample); 404 if (atom == NULL) 405 return -1; 406 407 work = work_findnew(&class->work_root, &key, &kwork->cmp_id); 408 if (work == NULL) { 409 free(atom); 410 return -1; 411 } 412 413 if (!profile_event_match(kwork, work, sample)) 414 return 0; 415 416 if (dst_type < KWORK_TRACE_MAX) { 417 dst_atom = list_last_entry_or_null(&work->atom_list[dst_type], 418 struct kwork_atom, list); 419 if (dst_atom != NULL) { 420 atom->prev = dst_atom; 421 list_del(&dst_atom->list); 422 } 423 } 424 425 if (ret_work != NULL) 426 *ret_work = work; 427 428 list_add_tail(&atom->list, &work->atom_list[src_type]); 429 430 return 0; 431 } 432 433 static struct kwork_atom *work_pop_atom(struct perf_kwork *kwork, 434 struct kwork_class *class, 435 enum kwork_trace_type src_type, 436 enum kwork_trace_type dst_type, 437 struct evsel *evsel, 438 struct perf_sample *sample, 439 struct machine *machine, 440 struct kwork_work **ret_work) 441 { 442 struct kwork_atom *atom, *src_atom; 443 struct kwork_work *work, key; 444 445 BUG_ON(class->work_init == NULL); 446 class->work_init(class, &key, evsel, sample, machine); 447 448 work = work_findnew(&class->work_root, &key, &kwork->cmp_id); 449 if (ret_work != NULL) 450 *ret_work = work; 451 452 if (work == NULL) 453 return NULL; 454 455 if (!profile_event_match(kwork, work, sample)) 456 return NULL; 457 458 atom = list_last_entry_or_null(&work->atom_list[dst_type], 459 struct kwork_atom, list); 460 if (atom != NULL) 461 return atom; 462 463 src_atom = atom_new(kwork, sample); 464 if (src_atom != NULL) 465 list_add_tail(&src_atom->list, &work->atom_list[src_type]); 466 else { 467 if (ret_work != NULL) 468 *ret_work = NULL; 469 } 470 471 return NULL; 472 } 473 474 static void report_update_exit_event(struct kwork_work *work, 475 struct kwork_atom *atom, 476 struct perf_sample *sample) 477 { 478 u64 delta; 479 u64 exit_time = sample->time; 480 u64 entry_time = atom->time; 481 482 if ((entry_time != 0) && (exit_time >= entry_time)) { 483 delta = exit_time - entry_time; 484 if ((delta > work->max_runtime) || 485 (work->max_runtime == 0)) { 486 work->max_runtime = delta; 487 work->max_runtime_start = entry_time; 488 work->max_runtime_end = exit_time; 489 } 490 work->total_runtime += delta; 491 work->nr_atoms++; 492 } 493 } 494 495 static int report_entry_event(struct perf_kwork *kwork, 496 struct kwork_class *class, 497 struct evsel *evsel, 498 struct perf_sample *sample, 499 struct machine *machine) 500 { 501 return work_push_atom(kwork, class, KWORK_TRACE_ENTRY, 502 KWORK_TRACE_MAX, evsel, sample, 503 machine, NULL); 504 } 505 506 static int report_exit_event(struct perf_kwork *kwork, 507 struct kwork_class *class, 508 struct evsel *evsel, 509 struct perf_sample *sample, 510 struct machine *machine) 511 { 512 struct kwork_atom *atom = NULL; 513 struct kwork_work *work = NULL; 514 515 atom = work_pop_atom(kwork, class, KWORK_TRACE_EXIT, 516 KWORK_TRACE_ENTRY, evsel, sample, 517 machine, &work); 518 if (work == NULL) 519 return -1; 520 521 if (atom != NULL) { 522 report_update_exit_event(work, atom, sample); 523 atom_del(atom); 524 } 525 526 return 0; 527 } 528 529 static void latency_update_entry_event(struct kwork_work *work, 530 struct kwork_atom *atom, 531 struct perf_sample *sample) 532 { 533 u64 delta; 534 u64 entry_time = sample->time; 535 u64 raise_time = atom->time; 536 537 if ((raise_time != 0) && (entry_time >= raise_time)) { 538 delta = entry_time - raise_time; 539 if ((delta > work->max_latency) || 540 (work->max_latency == 0)) { 541 work->max_latency = delta; 542 work->max_latency_start = raise_time; 543 work->max_latency_end = entry_time; 544 } 545 work->total_latency += delta; 546 work->nr_atoms++; 547 } 548 } 549 550 static int latency_raise_event(struct perf_kwork *kwork, 551 struct kwork_class *class, 552 struct evsel *evsel, 553 struct perf_sample *sample, 554 struct machine *machine) 555 { 556 return work_push_atom(kwork, class, KWORK_TRACE_RAISE, 557 KWORK_TRACE_MAX, evsel, sample, 558 machine, NULL); 559 } 560 561 static int latency_entry_event(struct perf_kwork *kwork, 562 struct kwork_class *class, 563 struct evsel *evsel, 564 struct perf_sample *sample, 565 struct machine *machine) 566 { 567 struct kwork_atom *atom = NULL; 568 struct kwork_work *work = NULL; 569 570 atom = work_pop_atom(kwork, class, KWORK_TRACE_ENTRY, 571 KWORK_TRACE_RAISE, evsel, sample, 572 machine, &work); 573 if (work == NULL) 574 return -1; 575 576 if (atom != NULL) { 577 latency_update_entry_event(work, atom, sample); 578 atom_del(atom); 579 } 580 581 return 0; 582 } 583 584 static void timehist_save_callchain(struct perf_kwork *kwork, 585 struct perf_sample *sample, 586 struct evsel *evsel, 587 struct machine *machine) 588 { 589 struct symbol *sym; 590 struct thread *thread; 591 struct callchain_cursor_node *node; 592 struct callchain_cursor *cursor; 593 594 if (!kwork->show_callchain || sample->callchain == NULL) 595 return; 596 597 /* want main thread for process - has maps */ 598 thread = machine__findnew_thread(machine, sample->pid, sample->pid); 599 if (thread == NULL) { 600 pr_debug("Failed to get thread for pid %d\n", sample->pid); 601 return; 602 } 603 604 cursor = get_tls_callchain_cursor(); 605 606 if (thread__resolve_callchain(thread, cursor, evsel, sample, 607 NULL, NULL, kwork->max_stack + 2) != 0) { 608 pr_debug("Failed to resolve callchain, skipping\n"); 609 goto out_put; 610 } 611 612 callchain_cursor_commit(cursor); 613 614 while (true) { 615 node = callchain_cursor_current(cursor); 616 if (node == NULL) 617 break; 618 619 sym = node->ms.sym; 620 if (sym) { 621 if (!strcmp(sym->name, "__softirqentry_text_start") || 622 !strcmp(sym->name, "__do_softirq")) 623 sym->ignore = 1; 624 } 625 626 callchain_cursor_advance(cursor); 627 } 628 629 out_put: 630 thread__put(thread); 631 } 632 633 static void timehist_print_event(struct perf_kwork *kwork, 634 struct kwork_work *work, 635 struct kwork_atom *atom, 636 struct perf_sample *sample, 637 struct addr_location *al) 638 { 639 char entrytime[32], exittime[32]; 640 char kwork_name[PRINT_KWORK_NAME_WIDTH]; 641 642 /* 643 * runtime start 644 */ 645 timestamp__scnprintf_usec(atom->time, 646 entrytime, sizeof(entrytime)); 647 printf(" %*s ", PRINT_TIMESTAMP_WIDTH, entrytime); 648 649 /* 650 * runtime end 651 */ 652 timestamp__scnprintf_usec(sample->time, 653 exittime, sizeof(exittime)); 654 printf(" %*s ", PRINT_TIMESTAMP_WIDTH, exittime); 655 656 /* 657 * cpu 658 */ 659 printf(" [%0*d] ", PRINT_CPU_WIDTH, work->cpu); 660 661 /* 662 * kwork name 663 */ 664 if (work->class && work->class->work_name) { 665 work->class->work_name(work, kwork_name, 666 PRINT_KWORK_NAME_WIDTH); 667 printf(" %-*s ", PRINT_KWORK_NAME_WIDTH, kwork_name); 668 } else 669 printf(" %-*s ", PRINT_KWORK_NAME_WIDTH, ""); 670 671 /* 672 *runtime 673 */ 674 printf(" %*.*f ", 675 PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH, 676 (double)(sample->time - atom->time) / NSEC_PER_MSEC); 677 678 /* 679 * delaytime 680 */ 681 if (atom->prev != NULL) 682 printf(" %*.*f ", PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH, 683 (double)(atom->time - atom->prev->time) / NSEC_PER_MSEC); 684 else 685 printf(" %*s ", PRINT_LATENCY_WIDTH, " "); 686 687 /* 688 * callchain 689 */ 690 if (kwork->show_callchain) { 691 struct callchain_cursor *cursor = get_tls_callchain_cursor(); 692 693 if (cursor == NULL) 694 return; 695 696 printf(" "); 697 698 sample__fprintf_sym(sample, al, 0, 699 EVSEL__PRINT_SYM | EVSEL__PRINT_ONELINE | 700 EVSEL__PRINT_CALLCHAIN_ARROW | 701 EVSEL__PRINT_SKIP_IGNORED, 702 cursor, symbol_conf.bt_stop_list, 703 stdout); 704 } 705 706 printf("\n"); 707 } 708 709 static int timehist_raise_event(struct perf_kwork *kwork, 710 struct kwork_class *class, 711 struct evsel *evsel, 712 struct perf_sample *sample, 713 struct machine *machine) 714 { 715 return work_push_atom(kwork, class, KWORK_TRACE_RAISE, 716 KWORK_TRACE_MAX, evsel, sample, 717 machine, NULL); 718 } 719 720 static int timehist_entry_event(struct perf_kwork *kwork, 721 struct kwork_class *class, 722 struct evsel *evsel, 723 struct perf_sample *sample, 724 struct machine *machine) 725 { 726 int ret; 727 struct kwork_work *work = NULL; 728 729 ret = work_push_atom(kwork, class, KWORK_TRACE_ENTRY, 730 KWORK_TRACE_RAISE, evsel, sample, 731 machine, &work); 732 if (ret) 733 return ret; 734 735 if (work != NULL) 736 timehist_save_callchain(kwork, sample, evsel, machine); 737 738 return 0; 739 } 740 741 static int timehist_exit_event(struct perf_kwork *kwork, 742 struct kwork_class *class, 743 struct evsel *evsel, 744 struct perf_sample *sample, 745 struct machine *machine) 746 { 747 struct kwork_atom *atom = NULL; 748 struct kwork_work *work = NULL; 749 struct addr_location al; 750 int ret = 0; 751 752 addr_location__init(&al); 753 if (machine__resolve(machine, &al, sample) < 0) { 754 pr_debug("Problem processing event, skipping it\n"); 755 ret = -1; 756 goto out; 757 } 758 759 atom = work_pop_atom(kwork, class, KWORK_TRACE_EXIT, 760 KWORK_TRACE_ENTRY, evsel, sample, 761 machine, &work); 762 if (work == NULL) { 763 ret = -1; 764 goto out; 765 } 766 767 if (atom != NULL) { 768 work->nr_atoms++; 769 timehist_print_event(kwork, work, atom, sample, &al); 770 atom_del(atom); 771 } 772 773 out: 774 addr_location__exit(&al); 775 return ret; 776 } 777 778 static struct kwork_class kwork_irq; 779 static int process_irq_handler_entry_event(struct perf_tool *tool, 780 struct evsel *evsel, 781 struct perf_sample *sample, 782 struct machine *machine) 783 { 784 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 785 786 if (kwork->tp_handler->entry_event) 787 return kwork->tp_handler->entry_event(kwork, &kwork_irq, 788 evsel, sample, machine); 789 return 0; 790 } 791 792 static int process_irq_handler_exit_event(struct perf_tool *tool, 793 struct evsel *evsel, 794 struct perf_sample *sample, 795 struct machine *machine) 796 { 797 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 798 799 if (kwork->tp_handler->exit_event) 800 return kwork->tp_handler->exit_event(kwork, &kwork_irq, 801 evsel, sample, machine); 802 return 0; 803 } 804 805 const struct evsel_str_handler irq_tp_handlers[] = { 806 { "irq:irq_handler_entry", process_irq_handler_entry_event, }, 807 { "irq:irq_handler_exit", process_irq_handler_exit_event, }, 808 }; 809 810 static int irq_class_init(struct kwork_class *class, 811 struct perf_session *session) 812 { 813 if (perf_session__set_tracepoints_handlers(session, irq_tp_handlers)) { 814 pr_err("Failed to set irq tracepoints handlers\n"); 815 return -1; 816 } 817 818 class->work_root = RB_ROOT_CACHED; 819 return 0; 820 } 821 822 static void irq_work_init(struct kwork_class *class, 823 struct kwork_work *work, 824 struct evsel *evsel, 825 struct perf_sample *sample, 826 struct machine *machine __maybe_unused) 827 { 828 work->class = class; 829 work->cpu = sample->cpu; 830 work->id = evsel__intval(evsel, sample, "irq"); 831 work->name = evsel__strval(evsel, sample, "name"); 832 } 833 834 static void irq_work_name(struct kwork_work *work, char *buf, int len) 835 { 836 snprintf(buf, len, "%s:%" PRIu64 "", work->name, work->id); 837 } 838 839 static struct kwork_class kwork_irq = { 840 .name = "irq", 841 .type = KWORK_CLASS_IRQ, 842 .nr_tracepoints = 2, 843 .tp_handlers = irq_tp_handlers, 844 .class_init = irq_class_init, 845 .work_init = irq_work_init, 846 .work_name = irq_work_name, 847 }; 848 849 static struct kwork_class kwork_softirq; 850 static int process_softirq_raise_event(struct perf_tool *tool, 851 struct evsel *evsel, 852 struct perf_sample *sample, 853 struct machine *machine) 854 { 855 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 856 857 if (kwork->tp_handler->raise_event) 858 return kwork->tp_handler->raise_event(kwork, &kwork_softirq, 859 evsel, sample, machine); 860 861 return 0; 862 } 863 864 static int process_softirq_entry_event(struct perf_tool *tool, 865 struct evsel *evsel, 866 struct perf_sample *sample, 867 struct machine *machine) 868 { 869 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 870 871 if (kwork->tp_handler->entry_event) 872 return kwork->tp_handler->entry_event(kwork, &kwork_softirq, 873 evsel, sample, machine); 874 875 return 0; 876 } 877 878 static int process_softirq_exit_event(struct perf_tool *tool, 879 struct evsel *evsel, 880 struct perf_sample *sample, 881 struct machine *machine) 882 { 883 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 884 885 if (kwork->tp_handler->exit_event) 886 return kwork->tp_handler->exit_event(kwork, &kwork_softirq, 887 evsel, sample, machine); 888 889 return 0; 890 } 891 892 const struct evsel_str_handler softirq_tp_handlers[] = { 893 { "irq:softirq_raise", process_softirq_raise_event, }, 894 { "irq:softirq_entry", process_softirq_entry_event, }, 895 { "irq:softirq_exit", process_softirq_exit_event, }, 896 }; 897 898 static int softirq_class_init(struct kwork_class *class, 899 struct perf_session *session) 900 { 901 if (perf_session__set_tracepoints_handlers(session, 902 softirq_tp_handlers)) { 903 pr_err("Failed to set softirq tracepoints handlers\n"); 904 return -1; 905 } 906 907 class->work_root = RB_ROOT_CACHED; 908 return 0; 909 } 910 911 static char *evsel__softirq_name(struct evsel *evsel, u64 num) 912 { 913 char *name = NULL; 914 bool found = false; 915 struct tep_print_flag_sym *sym = NULL; 916 struct tep_print_arg *args = evsel->tp_format->print_fmt.args; 917 918 if ((args == NULL) || (args->next == NULL)) 919 return NULL; 920 921 /* skip softirq field: "REC->vec" */ 922 for (sym = args->next->symbol.symbols; sym != NULL; sym = sym->next) { 923 if ((eval_flag(sym->value) == (unsigned long long)num) && 924 (strlen(sym->str) != 0)) { 925 found = true; 926 break; 927 } 928 } 929 930 if (!found) 931 return NULL; 932 933 name = strdup(sym->str); 934 if (name == NULL) { 935 pr_err("Failed to copy symbol name\n"); 936 return NULL; 937 } 938 return name; 939 } 940 941 static void softirq_work_init(struct kwork_class *class, 942 struct kwork_work *work, 943 struct evsel *evsel, 944 struct perf_sample *sample, 945 struct machine *machine __maybe_unused) 946 { 947 u64 num = evsel__intval(evsel, sample, "vec"); 948 949 work->id = num; 950 work->class = class; 951 work->cpu = sample->cpu; 952 work->name = evsel__softirq_name(evsel, num); 953 } 954 955 static void softirq_work_name(struct kwork_work *work, char *buf, int len) 956 { 957 snprintf(buf, len, "(s)%s:%" PRIu64 "", work->name, work->id); 958 } 959 960 static struct kwork_class kwork_softirq = { 961 .name = "softirq", 962 .type = KWORK_CLASS_SOFTIRQ, 963 .nr_tracepoints = 3, 964 .tp_handlers = softirq_tp_handlers, 965 .class_init = softirq_class_init, 966 .work_init = softirq_work_init, 967 .work_name = softirq_work_name, 968 }; 969 970 static struct kwork_class kwork_workqueue; 971 static int process_workqueue_activate_work_event(struct perf_tool *tool, 972 struct evsel *evsel, 973 struct perf_sample *sample, 974 struct machine *machine) 975 { 976 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 977 978 if (kwork->tp_handler->raise_event) 979 return kwork->tp_handler->raise_event(kwork, &kwork_workqueue, 980 evsel, sample, machine); 981 982 return 0; 983 } 984 985 static int process_workqueue_execute_start_event(struct perf_tool *tool, 986 struct evsel *evsel, 987 struct perf_sample *sample, 988 struct machine *machine) 989 { 990 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 991 992 if (kwork->tp_handler->entry_event) 993 return kwork->tp_handler->entry_event(kwork, &kwork_workqueue, 994 evsel, sample, machine); 995 996 return 0; 997 } 998 999 static int process_workqueue_execute_end_event(struct perf_tool *tool, 1000 struct evsel *evsel, 1001 struct perf_sample *sample, 1002 struct machine *machine) 1003 { 1004 struct perf_kwork *kwork = container_of(tool, struct perf_kwork, tool); 1005 1006 if (kwork->tp_handler->exit_event) 1007 return kwork->tp_handler->exit_event(kwork, &kwork_workqueue, 1008 evsel, sample, machine); 1009 1010 return 0; 1011 } 1012 1013 const struct evsel_str_handler workqueue_tp_handlers[] = { 1014 { "workqueue:workqueue_activate_work", process_workqueue_activate_work_event, }, 1015 { "workqueue:workqueue_execute_start", process_workqueue_execute_start_event, }, 1016 { "workqueue:workqueue_execute_end", process_workqueue_execute_end_event, }, 1017 }; 1018 1019 static int workqueue_class_init(struct kwork_class *class, 1020 struct perf_session *session) 1021 { 1022 if (perf_session__set_tracepoints_handlers(session, 1023 workqueue_tp_handlers)) { 1024 pr_err("Failed to set workqueue tracepoints handlers\n"); 1025 return -1; 1026 } 1027 1028 class->work_root = RB_ROOT_CACHED; 1029 return 0; 1030 } 1031 1032 static void workqueue_work_init(struct kwork_class *class, 1033 struct kwork_work *work, 1034 struct evsel *evsel, 1035 struct perf_sample *sample, 1036 struct machine *machine) 1037 { 1038 char *modp = NULL; 1039 unsigned long long function_addr = evsel__intval(evsel, 1040 sample, "function"); 1041 1042 work->class = class; 1043 work->cpu = sample->cpu; 1044 work->id = evsel__intval(evsel, sample, "work"); 1045 work->name = function_addr == 0 ? NULL : 1046 machine__resolve_kernel_addr(machine, &function_addr, &modp); 1047 } 1048 1049 static void workqueue_work_name(struct kwork_work *work, char *buf, int len) 1050 { 1051 if (work->name != NULL) 1052 snprintf(buf, len, "(w)%s", work->name); 1053 else 1054 snprintf(buf, len, "(w)0x%" PRIx64, work->id); 1055 } 1056 1057 static struct kwork_class kwork_workqueue = { 1058 .name = "workqueue", 1059 .type = KWORK_CLASS_WORKQUEUE, 1060 .nr_tracepoints = 3, 1061 .tp_handlers = workqueue_tp_handlers, 1062 .class_init = workqueue_class_init, 1063 .work_init = workqueue_work_init, 1064 .work_name = workqueue_work_name, 1065 }; 1066 1067 static struct kwork_class *kwork_class_supported_list[KWORK_CLASS_MAX] = { 1068 [KWORK_CLASS_IRQ] = &kwork_irq, 1069 [KWORK_CLASS_SOFTIRQ] = &kwork_softirq, 1070 [KWORK_CLASS_WORKQUEUE] = &kwork_workqueue, 1071 }; 1072 1073 static void print_separator(int len) 1074 { 1075 printf(" %.*s\n", len, graph_dotted_line); 1076 } 1077 1078 static int report_print_work(struct perf_kwork *kwork, struct kwork_work *work) 1079 { 1080 int ret = 0; 1081 char kwork_name[PRINT_KWORK_NAME_WIDTH]; 1082 char max_runtime_start[32], max_runtime_end[32]; 1083 char max_latency_start[32], max_latency_end[32]; 1084 1085 printf(" "); 1086 1087 /* 1088 * kwork name 1089 */ 1090 if (work->class && work->class->work_name) { 1091 work->class->work_name(work, kwork_name, 1092 PRINT_KWORK_NAME_WIDTH); 1093 ret += printf(" %-*s |", PRINT_KWORK_NAME_WIDTH, kwork_name); 1094 } else { 1095 ret += printf(" %-*s |", PRINT_KWORK_NAME_WIDTH, ""); 1096 } 1097 1098 /* 1099 * cpu 1100 */ 1101 ret += printf(" %0*d |", PRINT_CPU_WIDTH, work->cpu); 1102 1103 /* 1104 * total runtime 1105 */ 1106 if (kwork->report == KWORK_REPORT_RUNTIME) { 1107 ret += printf(" %*.*f ms |", 1108 PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH, 1109 (double)work->total_runtime / NSEC_PER_MSEC); 1110 } else if (kwork->report == KWORK_REPORT_LATENCY) { // avg delay 1111 ret += printf(" %*.*f ms |", 1112 PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH, 1113 (double)work->total_latency / 1114 work->nr_atoms / NSEC_PER_MSEC); 1115 } 1116 1117 /* 1118 * count 1119 */ 1120 ret += printf(" %*" PRIu64 " |", PRINT_COUNT_WIDTH, work->nr_atoms); 1121 1122 /* 1123 * max runtime, max runtime start, max runtime end 1124 */ 1125 if (kwork->report == KWORK_REPORT_RUNTIME) { 1126 timestamp__scnprintf_usec(work->max_runtime_start, 1127 max_runtime_start, 1128 sizeof(max_runtime_start)); 1129 timestamp__scnprintf_usec(work->max_runtime_end, 1130 max_runtime_end, 1131 sizeof(max_runtime_end)); 1132 ret += printf(" %*.*f ms | %*s s | %*s s |", 1133 PRINT_RUNTIME_WIDTH, RPINT_DECIMAL_WIDTH, 1134 (double)work->max_runtime / NSEC_PER_MSEC, 1135 PRINT_TIMESTAMP_WIDTH, max_runtime_start, 1136 PRINT_TIMESTAMP_WIDTH, max_runtime_end); 1137 } 1138 /* 1139 * max delay, max delay start, max delay end 1140 */ 1141 else if (kwork->report == KWORK_REPORT_LATENCY) { 1142 timestamp__scnprintf_usec(work->max_latency_start, 1143 max_latency_start, 1144 sizeof(max_latency_start)); 1145 timestamp__scnprintf_usec(work->max_latency_end, 1146 max_latency_end, 1147 sizeof(max_latency_end)); 1148 ret += printf(" %*.*f ms | %*s s | %*s s |", 1149 PRINT_LATENCY_WIDTH, RPINT_DECIMAL_WIDTH, 1150 (double)work->max_latency / NSEC_PER_MSEC, 1151 PRINT_TIMESTAMP_WIDTH, max_latency_start, 1152 PRINT_TIMESTAMP_WIDTH, max_latency_end); 1153 } 1154 1155 printf("\n"); 1156 return ret; 1157 } 1158 1159 static int report_print_header(struct perf_kwork *kwork) 1160 { 1161 int ret; 1162 1163 printf("\n "); 1164 ret = printf(" %-*s | %-*s |", 1165 PRINT_KWORK_NAME_WIDTH, "Kwork Name", 1166 PRINT_CPU_WIDTH, "Cpu"); 1167 1168 if (kwork->report == KWORK_REPORT_RUNTIME) { 1169 ret += printf(" %-*s |", 1170 PRINT_RUNTIME_HEADER_WIDTH, "Total Runtime"); 1171 } else if (kwork->report == KWORK_REPORT_LATENCY) { 1172 ret += printf(" %-*s |", 1173 PRINT_LATENCY_HEADER_WIDTH, "Avg delay"); 1174 } 1175 1176 ret += printf(" %-*s |", PRINT_COUNT_WIDTH, "Count"); 1177 1178 if (kwork->report == KWORK_REPORT_RUNTIME) { 1179 ret += printf(" %-*s | %-*s | %-*s |", 1180 PRINT_RUNTIME_HEADER_WIDTH, "Max runtime", 1181 PRINT_TIMESTAMP_HEADER_WIDTH, "Max runtime start", 1182 PRINT_TIMESTAMP_HEADER_WIDTH, "Max runtime end"); 1183 } else if (kwork->report == KWORK_REPORT_LATENCY) { 1184 ret += printf(" %-*s | %-*s | %-*s |", 1185 PRINT_LATENCY_HEADER_WIDTH, "Max delay", 1186 PRINT_TIMESTAMP_HEADER_WIDTH, "Max delay start", 1187 PRINT_TIMESTAMP_HEADER_WIDTH, "Max delay end"); 1188 } 1189 1190 printf("\n"); 1191 print_separator(ret); 1192 return ret; 1193 } 1194 1195 static void timehist_print_header(void) 1196 { 1197 /* 1198 * header row 1199 */ 1200 printf(" %-*s %-*s %-*s %-*s %-*s %-*s\n", 1201 PRINT_TIMESTAMP_WIDTH, "Runtime start", 1202 PRINT_TIMESTAMP_WIDTH, "Runtime end", 1203 PRINT_TIMEHIST_CPU_WIDTH, "Cpu", 1204 PRINT_KWORK_NAME_WIDTH, "Kwork name", 1205 PRINT_RUNTIME_WIDTH, "Runtime", 1206 PRINT_RUNTIME_WIDTH, "Delaytime"); 1207 1208 /* 1209 * units row 1210 */ 1211 printf(" %-*s %-*s %-*s %-*s %-*s %-*s\n", 1212 PRINT_TIMESTAMP_WIDTH, "", 1213 PRINT_TIMESTAMP_WIDTH, "", 1214 PRINT_TIMEHIST_CPU_WIDTH, "", 1215 PRINT_KWORK_NAME_WIDTH, "(TYPE)NAME:NUM", 1216 PRINT_RUNTIME_WIDTH, "(msec)", 1217 PRINT_RUNTIME_WIDTH, "(msec)"); 1218 1219 /* 1220 * separator 1221 */ 1222 printf(" %.*s %.*s %.*s %.*s %.*s %.*s\n", 1223 PRINT_TIMESTAMP_WIDTH, graph_dotted_line, 1224 PRINT_TIMESTAMP_WIDTH, graph_dotted_line, 1225 PRINT_TIMEHIST_CPU_WIDTH, graph_dotted_line, 1226 PRINT_KWORK_NAME_WIDTH, graph_dotted_line, 1227 PRINT_RUNTIME_WIDTH, graph_dotted_line, 1228 PRINT_RUNTIME_WIDTH, graph_dotted_line); 1229 } 1230 1231 static void print_summary(struct perf_kwork *kwork) 1232 { 1233 u64 time = kwork->timeend - kwork->timestart; 1234 1235 printf(" Total count : %9" PRIu64 "\n", kwork->all_count); 1236 printf(" Total runtime (msec) : %9.3f (%.3f%% load average)\n", 1237 (double)kwork->all_runtime / NSEC_PER_MSEC, 1238 time == 0 ? 0 : (double)kwork->all_runtime / time); 1239 printf(" Total time span (msec) : %9.3f\n", 1240 (double)time / NSEC_PER_MSEC); 1241 } 1242 1243 static unsigned long long nr_list_entry(struct list_head *head) 1244 { 1245 struct list_head *pos; 1246 unsigned long long n = 0; 1247 1248 list_for_each(pos, head) 1249 n++; 1250 1251 return n; 1252 } 1253 1254 static void print_skipped_events(struct perf_kwork *kwork) 1255 { 1256 int i; 1257 const char *const kwork_event_str[] = { 1258 [KWORK_TRACE_RAISE] = "raise", 1259 [KWORK_TRACE_ENTRY] = "entry", 1260 [KWORK_TRACE_EXIT] = "exit", 1261 }; 1262 1263 if ((kwork->nr_skipped_events[KWORK_TRACE_MAX] != 0) && 1264 (kwork->nr_events != 0)) { 1265 printf(" INFO: %.3f%% skipped events (%" PRIu64 " including ", 1266 (double)kwork->nr_skipped_events[KWORK_TRACE_MAX] / 1267 (double)kwork->nr_events * 100.0, 1268 kwork->nr_skipped_events[KWORK_TRACE_MAX]); 1269 1270 for (i = 0; i < KWORK_TRACE_MAX; i++) { 1271 printf("%" PRIu64 " %s%s", 1272 kwork->nr_skipped_events[i], 1273 kwork_event_str[i], 1274 (i == KWORK_TRACE_MAX - 1) ? ")\n" : ", "); 1275 } 1276 } 1277 1278 if (verbose > 0) 1279 printf(" INFO: use %lld atom pages\n", 1280 nr_list_entry(&kwork->atom_page_list)); 1281 } 1282 1283 static void print_bad_events(struct perf_kwork *kwork) 1284 { 1285 if ((kwork->nr_lost_events != 0) && (kwork->nr_events != 0)) { 1286 printf(" INFO: %.3f%% lost events (%ld out of %ld, in %ld chunks)\n", 1287 (double)kwork->nr_lost_events / 1288 (double)kwork->nr_events * 100.0, 1289 kwork->nr_lost_events, kwork->nr_events, 1290 kwork->nr_lost_chunks); 1291 } 1292 } 1293 1294 static void work_sort(struct perf_kwork *kwork, struct kwork_class *class) 1295 { 1296 struct rb_node *node; 1297 struct kwork_work *data; 1298 struct rb_root_cached *root = &class->work_root; 1299 1300 pr_debug("Sorting %s ...\n", class->name); 1301 for (;;) { 1302 node = rb_first_cached(root); 1303 if (!node) 1304 break; 1305 1306 rb_erase_cached(node, root); 1307 data = rb_entry(node, struct kwork_work, node); 1308 work_insert(&kwork->sorted_work_root, 1309 data, &kwork->sort_list); 1310 } 1311 } 1312 1313 static void perf_kwork__sort(struct perf_kwork *kwork) 1314 { 1315 struct kwork_class *class; 1316 1317 list_for_each_entry(class, &kwork->class_list, list) 1318 work_sort(kwork, class); 1319 } 1320 1321 static int perf_kwork__check_config(struct perf_kwork *kwork, 1322 struct perf_session *session) 1323 { 1324 int ret; 1325 struct evsel *evsel; 1326 struct kwork_class *class; 1327 1328 static struct trace_kwork_handler report_ops = { 1329 .entry_event = report_entry_event, 1330 .exit_event = report_exit_event, 1331 }; 1332 static struct trace_kwork_handler latency_ops = { 1333 .raise_event = latency_raise_event, 1334 .entry_event = latency_entry_event, 1335 }; 1336 static struct trace_kwork_handler timehist_ops = { 1337 .raise_event = timehist_raise_event, 1338 .entry_event = timehist_entry_event, 1339 .exit_event = timehist_exit_event, 1340 }; 1341 1342 switch (kwork->report) { 1343 case KWORK_REPORT_RUNTIME: 1344 kwork->tp_handler = &report_ops; 1345 break; 1346 case KWORK_REPORT_LATENCY: 1347 kwork->tp_handler = &latency_ops; 1348 break; 1349 case KWORK_REPORT_TIMEHIST: 1350 kwork->tp_handler = &timehist_ops; 1351 break; 1352 default: 1353 pr_debug("Invalid report type %d\n", kwork->report); 1354 return -1; 1355 } 1356 1357 list_for_each_entry(class, &kwork->class_list, list) 1358 if ((class->class_init != NULL) && 1359 (class->class_init(class, session) != 0)) 1360 return -1; 1361 1362 if (kwork->cpu_list != NULL) { 1363 ret = perf_session__cpu_bitmap(session, 1364 kwork->cpu_list, 1365 kwork->cpu_bitmap); 1366 if (ret < 0) { 1367 pr_err("Invalid cpu bitmap\n"); 1368 return -1; 1369 } 1370 } 1371 1372 if (kwork->time_str != NULL) { 1373 ret = perf_time__parse_str(&kwork->ptime, kwork->time_str); 1374 if (ret != 0) { 1375 pr_err("Invalid time span\n"); 1376 return -1; 1377 } 1378 } 1379 1380 list_for_each_entry(evsel, &session->evlist->core.entries, core.node) { 1381 if (kwork->show_callchain && !evsel__has_callchain(evsel)) { 1382 pr_debug("Samples do not have callchains\n"); 1383 kwork->show_callchain = 0; 1384 symbol_conf.use_callchain = 0; 1385 } 1386 } 1387 1388 return 0; 1389 } 1390 1391 static int perf_kwork__read_events(struct perf_kwork *kwork) 1392 { 1393 int ret = -1; 1394 struct perf_session *session = NULL; 1395 1396 struct perf_data data = { 1397 .path = input_name, 1398 .mode = PERF_DATA_MODE_READ, 1399 .force = kwork->force, 1400 }; 1401 1402 session = perf_session__new(&data, &kwork->tool); 1403 if (IS_ERR(session)) { 1404 pr_debug("Error creating perf session\n"); 1405 return PTR_ERR(session); 1406 } 1407 1408 symbol__init(&session->header.env); 1409 1410 if (perf_kwork__check_config(kwork, session) != 0) 1411 goto out_delete; 1412 1413 if (session->tevent.pevent && 1414 tep_set_function_resolver(session->tevent.pevent, 1415 machine__resolve_kernel_addr, 1416 &session->machines.host) < 0) { 1417 pr_err("Failed to set libtraceevent function resolver\n"); 1418 goto out_delete; 1419 } 1420 1421 if (kwork->report == KWORK_REPORT_TIMEHIST) 1422 timehist_print_header(); 1423 1424 ret = perf_session__process_events(session); 1425 if (ret) { 1426 pr_debug("Failed to process events, error %d\n", ret); 1427 goto out_delete; 1428 } 1429 1430 kwork->nr_events = session->evlist->stats.nr_events[0]; 1431 kwork->nr_lost_events = session->evlist->stats.total_lost; 1432 kwork->nr_lost_chunks = session->evlist->stats.nr_events[PERF_RECORD_LOST]; 1433 1434 out_delete: 1435 perf_session__delete(session); 1436 return ret; 1437 } 1438 1439 static void process_skipped_events(struct perf_kwork *kwork, 1440 struct kwork_work *work) 1441 { 1442 int i; 1443 unsigned long long count; 1444 1445 for (i = 0; i < KWORK_TRACE_MAX; i++) { 1446 count = nr_list_entry(&work->atom_list[i]); 1447 kwork->nr_skipped_events[i] += count; 1448 kwork->nr_skipped_events[KWORK_TRACE_MAX] += count; 1449 } 1450 } 1451 1452 struct kwork_work *perf_kwork_add_work(struct perf_kwork *kwork, 1453 struct kwork_class *class, 1454 struct kwork_work *key) 1455 { 1456 struct kwork_work *work = NULL; 1457 1458 work = work_new(key); 1459 if (work == NULL) 1460 return NULL; 1461 1462 work_insert(&class->work_root, work, &kwork->cmp_id); 1463 return work; 1464 } 1465 1466 static void sig_handler(int sig) 1467 { 1468 /* 1469 * Simply capture termination signal so that 1470 * the program can continue after pause returns 1471 */ 1472 pr_debug("Captuer signal %d\n", sig); 1473 } 1474 1475 static int perf_kwork__report_bpf(struct perf_kwork *kwork) 1476 { 1477 int ret; 1478 1479 signal(SIGINT, sig_handler); 1480 signal(SIGTERM, sig_handler); 1481 1482 ret = perf_kwork__trace_prepare_bpf(kwork); 1483 if (ret) 1484 return -1; 1485 1486 printf("Starting trace, Hit <Ctrl+C> to stop and report\n"); 1487 1488 perf_kwork__trace_start(); 1489 1490 /* 1491 * a simple pause, wait here for stop signal 1492 */ 1493 pause(); 1494 1495 perf_kwork__trace_finish(); 1496 1497 perf_kwork__report_read_bpf(kwork); 1498 1499 perf_kwork__report_cleanup_bpf(); 1500 1501 return 0; 1502 } 1503 1504 static int perf_kwork__report(struct perf_kwork *kwork) 1505 { 1506 int ret; 1507 struct rb_node *next; 1508 struct kwork_work *work; 1509 1510 if (kwork->use_bpf) 1511 ret = perf_kwork__report_bpf(kwork); 1512 else 1513 ret = perf_kwork__read_events(kwork); 1514 1515 if (ret != 0) 1516 return -1; 1517 1518 perf_kwork__sort(kwork); 1519 1520 setup_pager(); 1521 1522 ret = report_print_header(kwork); 1523 next = rb_first_cached(&kwork->sorted_work_root); 1524 while (next) { 1525 work = rb_entry(next, struct kwork_work, node); 1526 process_skipped_events(kwork, work); 1527 1528 if (work->nr_atoms != 0) { 1529 report_print_work(kwork, work); 1530 if (kwork->summary) { 1531 kwork->all_runtime += work->total_runtime; 1532 kwork->all_count += work->nr_atoms; 1533 } 1534 } 1535 next = rb_next(next); 1536 } 1537 print_separator(ret); 1538 1539 if (kwork->summary) { 1540 print_summary(kwork); 1541 print_separator(ret); 1542 } 1543 1544 print_bad_events(kwork); 1545 print_skipped_events(kwork); 1546 printf("\n"); 1547 1548 return 0; 1549 } 1550 1551 typedef int (*tracepoint_handler)(struct perf_tool *tool, 1552 struct evsel *evsel, 1553 struct perf_sample *sample, 1554 struct machine *machine); 1555 1556 static int perf_kwork__process_tracepoint_sample(struct perf_tool *tool, 1557 union perf_event *event __maybe_unused, 1558 struct perf_sample *sample, 1559 struct evsel *evsel, 1560 struct machine *machine) 1561 { 1562 int err = 0; 1563 1564 if (evsel->handler != NULL) { 1565 tracepoint_handler f = evsel->handler; 1566 1567 err = f(tool, evsel, sample, machine); 1568 } 1569 1570 return err; 1571 } 1572 1573 static int perf_kwork__timehist(struct perf_kwork *kwork) 1574 { 1575 /* 1576 * event handlers for timehist option 1577 */ 1578 kwork->tool.comm = perf_event__process_comm; 1579 kwork->tool.exit = perf_event__process_exit; 1580 kwork->tool.fork = perf_event__process_fork; 1581 kwork->tool.attr = perf_event__process_attr; 1582 kwork->tool.tracing_data = perf_event__process_tracing_data; 1583 kwork->tool.build_id = perf_event__process_build_id; 1584 kwork->tool.ordered_events = true; 1585 kwork->tool.ordering_requires_timestamps = true; 1586 symbol_conf.use_callchain = kwork->show_callchain; 1587 1588 if (symbol__validate_sym_arguments()) { 1589 pr_err("Failed to validate sym arguments\n"); 1590 return -1; 1591 } 1592 1593 setup_pager(); 1594 1595 return perf_kwork__read_events(kwork); 1596 } 1597 1598 static void setup_event_list(struct perf_kwork *kwork, 1599 const struct option *options, 1600 const char * const usage_msg[]) 1601 { 1602 int i; 1603 struct kwork_class *class; 1604 char *tmp, *tok, *str; 1605 1606 if (kwork->event_list_str == NULL) 1607 goto null_event_list_str; 1608 1609 str = strdup(kwork->event_list_str); 1610 for (tok = strtok_r(str, ", ", &tmp); 1611 tok; tok = strtok_r(NULL, ", ", &tmp)) { 1612 for (i = 0; i < KWORK_CLASS_MAX; i++) { 1613 class = kwork_class_supported_list[i]; 1614 if (strcmp(tok, class->name) == 0) { 1615 list_add_tail(&class->list, &kwork->class_list); 1616 break; 1617 } 1618 } 1619 if (i == KWORK_CLASS_MAX) { 1620 usage_with_options_msg(usage_msg, options, 1621 "Unknown --event key: `%s'", tok); 1622 } 1623 } 1624 free(str); 1625 1626 null_event_list_str: 1627 /* 1628 * config all kwork events if not specified 1629 */ 1630 if (list_empty(&kwork->class_list)) { 1631 for (i = 0; i < KWORK_CLASS_MAX; i++) { 1632 list_add_tail(&kwork_class_supported_list[i]->list, 1633 &kwork->class_list); 1634 } 1635 } 1636 1637 pr_debug("Config event list:"); 1638 list_for_each_entry(class, &kwork->class_list, list) 1639 pr_debug(" %s", class->name); 1640 pr_debug("\n"); 1641 } 1642 1643 static int perf_kwork__record(struct perf_kwork *kwork, 1644 int argc, const char **argv) 1645 { 1646 const char **rec_argv; 1647 unsigned int rec_argc, i, j; 1648 struct kwork_class *class; 1649 1650 const char *const record_args[] = { 1651 "record", 1652 "-a", 1653 "-R", 1654 "-m", "1024", 1655 "-c", "1", 1656 }; 1657 1658 rec_argc = ARRAY_SIZE(record_args) + argc - 1; 1659 1660 list_for_each_entry(class, &kwork->class_list, list) 1661 rec_argc += 2 * class->nr_tracepoints; 1662 1663 rec_argv = calloc(rec_argc + 1, sizeof(char *)); 1664 if (rec_argv == NULL) 1665 return -ENOMEM; 1666 1667 for (i = 0; i < ARRAY_SIZE(record_args); i++) 1668 rec_argv[i] = strdup(record_args[i]); 1669 1670 list_for_each_entry(class, &kwork->class_list, list) { 1671 for (j = 0; j < class->nr_tracepoints; j++) { 1672 rec_argv[i++] = strdup("-e"); 1673 rec_argv[i++] = strdup(class->tp_handlers[j].name); 1674 } 1675 } 1676 1677 for (j = 1; j < (unsigned int)argc; j++, i++) 1678 rec_argv[i] = argv[j]; 1679 1680 BUG_ON(i != rec_argc); 1681 1682 pr_debug("record comm: "); 1683 for (j = 0; j < rec_argc; j++) 1684 pr_debug("%s ", rec_argv[j]); 1685 pr_debug("\n"); 1686 1687 return cmd_record(i, rec_argv); 1688 } 1689 1690 int cmd_kwork(int argc, const char **argv) 1691 { 1692 static struct perf_kwork kwork = { 1693 .class_list = LIST_HEAD_INIT(kwork.class_list), 1694 .tool = { 1695 .mmap = perf_event__process_mmap, 1696 .mmap2 = perf_event__process_mmap2, 1697 .sample = perf_kwork__process_tracepoint_sample, 1698 }, 1699 .atom_page_list = LIST_HEAD_INIT(kwork.atom_page_list), 1700 .sort_list = LIST_HEAD_INIT(kwork.sort_list), 1701 .cmp_id = LIST_HEAD_INIT(kwork.cmp_id), 1702 .sorted_work_root = RB_ROOT_CACHED, 1703 .tp_handler = NULL, 1704 .profile_name = NULL, 1705 .cpu_list = NULL, 1706 .time_str = NULL, 1707 .force = false, 1708 .event_list_str = NULL, 1709 .summary = false, 1710 .sort_order = NULL, 1711 .show_callchain = false, 1712 .max_stack = 5, 1713 .timestart = 0, 1714 .timeend = 0, 1715 .nr_events = 0, 1716 .nr_lost_chunks = 0, 1717 .nr_lost_events = 0, 1718 .all_runtime = 0, 1719 .all_count = 0, 1720 .nr_skipped_events = { 0 }, 1721 }; 1722 static const char default_report_sort_order[] = "runtime, max, count"; 1723 static const char default_latency_sort_order[] = "avg, max, count"; 1724 const struct option kwork_options[] = { 1725 OPT_INCR('v', "verbose", &verbose, 1726 "be more verbose (show symbol address, etc)"), 1727 OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, 1728 "dump raw trace in ASCII"), 1729 OPT_STRING('k', "kwork", &kwork.event_list_str, "kwork", 1730 "list of kwork to profile (irq, softirq, workqueue, etc)"), 1731 OPT_BOOLEAN('f', "force", &kwork.force, "don't complain, do it"), 1732 OPT_END() 1733 }; 1734 const struct option report_options[] = { 1735 OPT_STRING('s', "sort", &kwork.sort_order, "key[,key2...]", 1736 "sort by key(s): runtime, max, count"), 1737 OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu", 1738 "list of cpus to profile"), 1739 OPT_STRING('n', "name", &kwork.profile_name, "name", 1740 "event name to profile"), 1741 OPT_STRING(0, "time", &kwork.time_str, "str", 1742 "Time span for analysis (start,stop)"), 1743 OPT_STRING('i', "input", &input_name, "file", 1744 "input file name"), 1745 OPT_BOOLEAN('S', "with-summary", &kwork.summary, 1746 "Show summary with statistics"), 1747 #ifdef HAVE_BPF_SKEL 1748 OPT_BOOLEAN('b', "use-bpf", &kwork.use_bpf, 1749 "Use BPF to measure kwork runtime"), 1750 #endif 1751 OPT_PARENT(kwork_options) 1752 }; 1753 const struct option latency_options[] = { 1754 OPT_STRING('s', "sort", &kwork.sort_order, "key[,key2...]", 1755 "sort by key(s): avg, max, count"), 1756 OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu", 1757 "list of cpus to profile"), 1758 OPT_STRING('n', "name", &kwork.profile_name, "name", 1759 "event name to profile"), 1760 OPT_STRING(0, "time", &kwork.time_str, "str", 1761 "Time span for analysis (start,stop)"), 1762 OPT_STRING('i', "input", &input_name, "file", 1763 "input file name"), 1764 #ifdef HAVE_BPF_SKEL 1765 OPT_BOOLEAN('b', "use-bpf", &kwork.use_bpf, 1766 "Use BPF to measure kwork latency"), 1767 #endif 1768 OPT_PARENT(kwork_options) 1769 }; 1770 const struct option timehist_options[] = { 1771 OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name, 1772 "file", "vmlinux pathname"), 1773 OPT_STRING(0, "kallsyms", &symbol_conf.kallsyms_name, 1774 "file", "kallsyms pathname"), 1775 OPT_BOOLEAN('g', "call-graph", &kwork.show_callchain, 1776 "Display call chains if present"), 1777 OPT_UINTEGER(0, "max-stack", &kwork.max_stack, 1778 "Maximum number of functions to display backtrace."), 1779 OPT_STRING(0, "symfs", &symbol_conf.symfs, "directory", 1780 "Look for files with symbols relative to this directory"), 1781 OPT_STRING(0, "time", &kwork.time_str, "str", 1782 "Time span for analysis (start,stop)"), 1783 OPT_STRING('C', "cpu", &kwork.cpu_list, "cpu", 1784 "list of cpus to profile"), 1785 OPT_STRING('n', "name", &kwork.profile_name, "name", 1786 "event name to profile"), 1787 OPT_STRING('i', "input", &input_name, "file", 1788 "input file name"), 1789 OPT_PARENT(kwork_options) 1790 }; 1791 const char *kwork_usage[] = { 1792 NULL, 1793 NULL 1794 }; 1795 const char * const report_usage[] = { 1796 "perf kwork report [<options>]", 1797 NULL 1798 }; 1799 const char * const latency_usage[] = { 1800 "perf kwork latency [<options>]", 1801 NULL 1802 }; 1803 const char * const timehist_usage[] = { 1804 "perf kwork timehist [<options>]", 1805 NULL 1806 }; 1807 const char *const kwork_subcommands[] = { 1808 "record", "report", "latency", "timehist", NULL 1809 }; 1810 1811 argc = parse_options_subcommand(argc, argv, kwork_options, 1812 kwork_subcommands, kwork_usage, 1813 PARSE_OPT_STOP_AT_NON_OPTION); 1814 if (!argc) 1815 usage_with_options(kwork_usage, kwork_options); 1816 1817 setup_event_list(&kwork, kwork_options, kwork_usage); 1818 sort_dimension__add(&kwork, "id", &kwork.cmp_id); 1819 1820 if (strlen(argv[0]) > 2 && strstarts("record", argv[0])) 1821 return perf_kwork__record(&kwork, argc, argv); 1822 else if (strlen(argv[0]) > 2 && strstarts("report", argv[0])) { 1823 kwork.sort_order = default_report_sort_order; 1824 if (argc > 1) { 1825 argc = parse_options(argc, argv, report_options, report_usage, 0); 1826 if (argc) 1827 usage_with_options(report_usage, report_options); 1828 } 1829 kwork.report = KWORK_REPORT_RUNTIME; 1830 setup_sorting(&kwork, report_options, report_usage); 1831 return perf_kwork__report(&kwork); 1832 } else if (strlen(argv[0]) > 2 && strstarts("latency", argv[0])) { 1833 kwork.sort_order = default_latency_sort_order; 1834 if (argc > 1) { 1835 argc = parse_options(argc, argv, latency_options, latency_usage, 0); 1836 if (argc) 1837 usage_with_options(latency_usage, latency_options); 1838 } 1839 kwork.report = KWORK_REPORT_LATENCY; 1840 setup_sorting(&kwork, latency_options, latency_usage); 1841 return perf_kwork__report(&kwork); 1842 } else if (strlen(argv[0]) > 2 && strstarts("timehist", argv[0])) { 1843 if (argc > 1) { 1844 argc = parse_options(argc, argv, timehist_options, timehist_usage, 0); 1845 if (argc) 1846 usage_with_options(timehist_usage, timehist_options); 1847 } 1848 kwork.report = KWORK_REPORT_TIMEHIST; 1849 return perf_kwork__timehist(&kwork); 1850 } else 1851 usage_with_options(kwork_usage, kwork_options); 1852 1853 return 0; 1854 } 1855