tools/perf/builtin-lock.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 #include <errno.h>
   3 #include <inttypes.h>
   4 #include "builtin.h"
   5 #include "perf.h"
   6
   7 #include "util/evlist.h" // for struct evsel_str_handler
   8 #include "util/evsel.h"
   9 #include "util/symbol.h"
  10 #include "util/thread.h"
  11 #include "util/header.h"
  12
  13 #include <subcmd/pager.h>
  14 #include <subcmd/parse-options.h>
  15 #include "util/trace-event.h"
  16
  17 #include "util/debug.h"
  18 #include "util/session.h"
  19 #include "util/tool.h"
  20 #include "util/data.h"
  21
  22 #include <sys/types.h>
  23 #include <sys/prctl.h>
  24 #include <semaphore.h>
  25 #include <pthread.h>
  26 #include <math.h>
  27 #include <limits.h>
  28
  29 #include <linux/list.h>
  30 #include <linux/hash.h>
  31 #include <linux/kernel.h>
  32 #include <linux/zalloc.h>
  33 #include <linux/err.h>
  34
  35 static struct perf_session *session;
  36
  37 /* based on kernel/lockdep.c */
  38 #define LOCKHASH_BITS           12
  39 #define LOCKHASH_SIZE           (1UL << LOCKHASH_BITS)
  40
  41 static struct list_head lockhash_table[LOCKHASH_SIZE];
  42
  43 #define __lockhashfn(key)       hash_long((unsigned long)key, LOCKHASH_BITS)
  44 #define lockhashentry(key)      (lockhash_table + __lockhashfn((key)))
  45
  46 struct lock_stat {
  47         struct list_head        hash_entry;
  48         struct rb_node          rb;             /* used for sorting */
  49
  50         /*
  51          * FIXME: evsel__intval() returns u64,
  52          * so address of lockdep_map should be dealed as 64bit.
  53          * Is there more better solution?
  54          */
  55         void                    *addr;          /* address of lockdep_map, used as ID */
  56         char                    *name;          /* for strcpy(), we cannot use const */
  57
  58         unsigned int            nr_acquire;
  59         unsigned int            nr_acquired;
  60         unsigned int            nr_contended;
  61         unsigned int            nr_release;
  62
  63         unsigned int            nr_readlock;
  64         unsigned int            nr_trylock;
  65
  66         /* these times are in nano sec. */
  67         u64                     avg_wait_time;
  68         u64                     wait_time_total;
  69         u64                     wait_time_min;
  70         u64                     wait_time_max;
  71
  72         int                     discard; /* flag of blacklist */
  73 };
  74
  75 /*
  76  * States of lock_seq_stat
  77  *
  78  * UNINITIALIZED is required for detecting first event of acquire.
  79  * As the nature of lock events, there is no guarantee
  80  * that the first event for the locks are acquire,
  81  * it can be acquired, contended or release.
  82  */
  83 #define SEQ_STATE_UNINITIALIZED      0         /* initial state */
  84 #define SEQ_STATE_RELEASED      1
  85 #define SEQ_STATE_ACQUIRING     2
  86 #define SEQ_STATE_ACQUIRED      3
  87 #define SEQ_STATE_READ_ACQUIRED 4
  88 #define SEQ_STATE_CONTENDED     5
  89
  90 /*
  91  * MAX_LOCK_DEPTH
  92  * Imported from include/linux/sched.h.
  93  * Should this be synchronized?
  94  */
  95 #define MAX_LOCK_DEPTH 48
  96
  97 /*
  98  * struct lock_seq_stat:
  99  * Place to put on state of one lock sequence
 100  * 1) acquire -> acquired -> release
 101  * 2) acquire -> contended -> acquired -> release
 102  * 3) acquire (with read or try) -> release
 103  * 4) Are there other patterns?
 104  */
 105 struct lock_seq_stat {
 106         struct list_head        list;
 107         int                     state;
 108         u64                     prev_event_time;
 109         void                    *addr;
 110
 111         int                     read_count;
 112 };
 113
 114 struct thread_stat {
 115         struct rb_node          rb;
 116
 117         u32                     tid;
 118         struct list_head        seq_list;
 119 };
 120
 121 static struct rb_root           thread_stats;
 122
 123 static struct thread_stat *thread_stat_find(u32 tid)
 124 {
 125         struct rb_node *node;
 126         struct thread_stat *st;
 127
 128         node = thread_stats.rb_node;
 129         while (node) {
 130                 st = container_of(node, struct thread_stat, rb);
 131                 if (st->tid == tid)
 132                         return st;
 133                 else if (tid < st->tid)
 134                         node = node->rb_left;
 135                 else
 136                         node = node->rb_right;
 137         }
 138
 139         return NULL;
 140 }
 141
 142 static void thread_stat_insert(struct thread_stat *new)
 143 {
 144         struct rb_node **rb = &thread_stats.rb_node;
 145         struct rb_node *parent = NULL;
 146         struct thread_stat *p;
 147
 148         while (*rb) {
 149                 p = container_of(*rb, struct thread_stat, rb);
 150                 parent = *rb;
 151
 152                 if (new->tid < p->tid)
 153                         rb = &(*rb)->rb_left;
 154                 else if (new->tid > p->tid)
 155                         rb = &(*rb)->rb_right;
 156                 else
 157                         BUG_ON("inserting invalid thread_stat\n");
 158         }
 159
 160         rb_link_node(&new->rb, parent, rb);
 161         rb_insert_color(&new->rb, &thread_stats);
 162 }
 163
 164 static struct thread_stat *thread_stat_findnew_after_first(u32 tid)
 165 {
 166         struct thread_stat *st;
 167
 168         st = thread_stat_find(tid);
 169         if (st)
 170                 return st;
 171
 172         st = zalloc(sizeof(struct thread_stat));
 173         if (!st) {
 174                 pr_err("memory allocation failed\n");
 175                 return NULL;
 176         }
 177
 178         st->tid = tid;
 179         INIT_LIST_HEAD(&st->seq_list);
 180
 181         thread_stat_insert(st);
 182
 183         return st;
 184 }
 185
 186 static struct thread_stat *thread_stat_findnew_first(u32 tid);
 187 static struct thread_stat *(*thread_stat_findnew)(u32 tid) =
 188         thread_stat_findnew_first;
 189
 190 static struct thread_stat *thread_stat_findnew_first(u32 tid)
 191 {
 192         struct thread_stat *st;
 193
 194         st = zalloc(sizeof(struct thread_stat));
 195         if (!st) {
 196                 pr_err("memory allocation failed\n");
 197                 return NULL;
 198         }
 199         st->tid = tid;
 200         INIT_LIST_HEAD(&st->seq_list);
 201
 202         rb_link_node(&st->rb, NULL, &thread_stats.rb_node);
 203         rb_insert_color(&st->rb, &thread_stats);
 204
 205         thread_stat_findnew = thread_stat_findnew_after_first;
 206         return st;
 207 }
 208
 209 /* build simple key function one is bigger than two */
 210 #define SINGLE_KEY(member)                                              \
 211         static int lock_stat_key_ ## member(struct lock_stat *one,      \
 212                                          struct lock_stat *two)         \
 213         {                                                               \
 214                 return one->member > two->member;                       \
 215         }
 216
 217 SINGLE_KEY(nr_acquired)
 218 SINGLE_KEY(nr_contended)
 219 SINGLE_KEY(avg_wait_time)
 220 SINGLE_KEY(wait_time_total)
 221 SINGLE_KEY(wait_time_max)
 222
 223 static int lock_stat_key_wait_time_min(struct lock_stat *one,
 224                                         struct lock_stat *two)
 225 {
 226         u64 s1 = one->wait_time_min;
 227         u64 s2 = two->wait_time_min;
 228         if (s1 == ULLONG_MAX)
 229                 s1 = 0;
 230         if (s2 == ULLONG_MAX)
 231                 s2 = 0;
 232         return s1 > s2;
 233 }
 234
 235 struct lock_key {
 236         /*
 237          * name: the value for specify by user
 238          * this should be simpler than raw name of member
 239          * e.g. nr_acquired -> acquired, wait_time_total -> wait_total
 240          */
 241         const char              *name;
 242         int                     (*key)(struct lock_stat*, struct lock_stat*);
 243 };
 244
 245 static const char               *sort_key = "acquired";
 246
 247 static int                      (*compare)(struct lock_stat *, struct lock_stat *);
 248
 249 static struct rb_root           result; /* place to store sorted data */
 250
 251 #define DEF_KEY_LOCK(name, fn_suffix)   \
 252         { #name, lock_stat_key_ ## fn_suffix }
 253 struct lock_key keys[] = {
 254         DEF_KEY_LOCK(acquired, nr_acquired),
 255         DEF_KEY_LOCK(contended, nr_contended),
 256         DEF_KEY_LOCK(avg_wait, avg_wait_time),
 257         DEF_KEY_LOCK(wait_total, wait_time_total),
 258         DEF_KEY_LOCK(wait_min, wait_time_min),
 259         DEF_KEY_LOCK(wait_max, wait_time_max),
 260
 261         /* extra comparisons much complicated should be here */
 262
 263         { NULL, NULL }
 264 };
 265
 266 static int select_key(void)
 267 {
 268         int i;
 269
 270         for (i = 0; keys[i].name; i++) {
 271                 if (!strcmp(keys[i].name, sort_key)) {
 272                         compare = keys[i].key;
 273                         return 0;
 274                 }
 275         }
 276
 277         pr_err("Unknown compare key: %s\n", sort_key);
 278
 279         return -1;
 280 }
 281
 282 static void insert_to_result(struct lock_stat *st,
 283                              int (*bigger)(struct lock_stat *, struct lock_stat *))
 284 {
 285         struct rb_node **rb = &result.rb_node;
 286         struct rb_node *parent = NULL;
 287         struct lock_stat *p;
 288
 289         while (*rb) {
 290                 p = container_of(*rb, struct lock_stat, rb);
 291                 parent = *rb;
 292
 293                 if (bigger(st, p))
 294                         rb = &(*rb)->rb_left;
 295                 else
 296                         rb = &(*rb)->rb_right;
 297         }
 298
 299         rb_link_node(&st->rb, parent, rb);
 300         rb_insert_color(&st->rb, &result);
 301 }
 302
 303 /* returns left most element of result, and erase it */
 304 static struct lock_stat *pop_from_result(void)
 305 {
 306         struct rb_node *node = result.rb_node;
 307
 308         if (!node)
 309                 return NULL;
 310
 311         while (node->rb_left)
 312                 node = node->rb_left;
 313
 314         rb_erase(node, &result);
 315         return container_of(node, struct lock_stat, rb);
 316 }
 317
 318 static struct lock_stat *lock_stat_findnew(void *addr, const char *name)
 319 {
 320         struct list_head *entry = lockhashentry(addr);
 321         struct lock_stat *ret, *new;
 322
 323         list_for_each_entry(ret, entry, hash_entry) {
 324                 if (ret->addr == addr)
 325                         return ret;
 326         }
 327
 328         new = zalloc(sizeof(struct lock_stat));
 329         if (!new)
 330                 goto alloc_failed;
 331
 332         new->addr = addr;
 333         new->name = zalloc(sizeof(char) * strlen(name) + 1);
 334         if (!new->name) {
 335                 free(new);
 336                 goto alloc_failed;
 337         }
 338
 339         strcpy(new->name, name);
 340         new->wait_time_min = ULLONG_MAX;
 341
 342         list_add(&new->hash_entry, entry);
 343         return new;
 344
 345 alloc_failed:
 346         pr_err("memory allocation failed\n");
 347         return NULL;
 348 }
 349
 350 struct trace_lock_handler {
 351         int (*acquire_event)(struct evsel *evsel,
 352                              struct perf_sample *sample);
 353
 354         int (*acquired_event)(struct evsel *evsel,
 355                               struct perf_sample *sample);
 356
 357         int (*contended_event)(struct evsel *evsel,
 358                                struct perf_sample *sample);
 359
 360         int (*release_event)(struct evsel *evsel,
 361                              struct perf_sample *sample);
 362 };
 363
 364 static struct lock_seq_stat *get_seq(struct thread_stat *ts, void *addr)
 365 {
 366         struct lock_seq_stat *seq;
 367
 368         list_for_each_entry(seq, &ts->seq_list, list) {
 369                 if (seq->addr == addr)
 370                         return seq;
 371         }
 372
 373         seq = zalloc(sizeof(struct lock_seq_stat));
 374         if (!seq) {
 375                 pr_err("memory allocation failed\n");
 376                 return NULL;
 377         }
 378         seq->state = SEQ_STATE_UNINITIALIZED;
 379         seq->addr = addr;
 380
 381         list_add(&seq->list, &ts->seq_list);
 382         return seq;
 383 }
 384
 385 enum broken_state {
 386         BROKEN_ACQUIRE,
 387         BROKEN_ACQUIRED,
 388         BROKEN_CONTENDED,
 389         BROKEN_RELEASE,
 390         BROKEN_MAX,
 391 };
 392
 393 static int bad_hist[BROKEN_MAX];
 394
 395 enum acquire_flags {
 396         TRY_LOCK = 1,
 397         READ_LOCK = 2,
 398 };
 399
 400 static int report_lock_acquire_event(struct evsel *evsel,
 401                                      struct perf_sample *sample)
 402 {
 403         void *addr;
 404         struct lock_stat *ls;
 405         struct thread_stat *ts;
 406         struct lock_seq_stat *seq;
 407         const char *name = evsel__strval(evsel, sample, "name");
 408         u64 tmp  = evsel__intval(evsel, sample, "lockdep_addr");
 409         int flag = evsel__intval(evsel, sample, "flag");
 410
 411         memcpy(&addr, &tmp, sizeof(void *));
 412
 413         ls = lock_stat_findnew(addr, name);
 414         if (!ls)
 415                 return -ENOMEM;
 416         if (ls->discard)
 417                 return 0;
 418
 419         ts = thread_stat_findnew(sample->tid);
 420         if (!ts)
 421                 return -ENOMEM;
 422
 423         seq = get_seq(ts, addr);
 424         if (!seq)
 425                 return -ENOMEM;
 426
 427         switch (seq->state) {
 428         case SEQ_STATE_UNINITIALIZED:
 429         case SEQ_STATE_RELEASED:
 430                 if (!flag) {
 431                         seq->state = SEQ_STATE_ACQUIRING;
 432                 } else {
 433                         if (flag & TRY_LOCK)
 434                                 ls->nr_trylock++;
 435                         if (flag & READ_LOCK)
 436                                 ls->nr_readlock++;
 437                         seq->state = SEQ_STATE_READ_ACQUIRED;
 438                         seq->read_count = 1;
 439                         ls->nr_acquired++;
 440                 }
 441                 break;
 442         case SEQ_STATE_READ_ACQUIRED:
 443                 if (flag & READ_LOCK) {
 444                         seq->read_count++;
 445                         ls->nr_acquired++;
 446                         goto end;
 447                 } else {
 448                         goto broken;
 449                 }
 450                 break;
 451         case SEQ_STATE_ACQUIRED:
 452         case SEQ_STATE_ACQUIRING:
 453         case SEQ_STATE_CONTENDED:
 454 broken:
 455                 /* broken lock sequence, discard it */
 456                 ls->discard = 1;
 457                 bad_hist[BROKEN_ACQUIRE]++;
 458                 list_del_init(&seq->list);
 459                 free(seq);
 460                 goto end;
 461         default:
 462                 BUG_ON("Unknown state of lock sequence found!\n");
 463                 break;
 464         }
 465
 466         ls->nr_acquire++;
 467         seq->prev_event_time = sample->time;
 468 end:
 469         return 0;
 470 }
 471
 472 static int report_lock_acquired_event(struct evsel *evsel,
 473                                       struct perf_sample *sample)
 474 {
 475         void *addr;
 476         struct lock_stat *ls;
 477         struct thread_stat *ts;
 478         struct lock_seq_stat *seq;
 479         u64 contended_term;
 480         const char *name = evsel__strval(evsel, sample, "name");
 481         u64 tmp = evsel__intval(evsel, sample, "lockdep_addr");
 482
 483         memcpy(&addr, &tmp, sizeof(void *));
 484
 485         ls = lock_stat_findnew(addr, name);
 486         if (!ls)
 487                 return -ENOMEM;
 488         if (ls->discard)
 489                 return 0;
 490
 491         ts = thread_stat_findnew(sample->tid);
 492         if (!ts)
 493                 return -ENOMEM;
 494
 495         seq = get_seq(ts, addr);
 496         if (!seq)
 497                 return -ENOMEM;
 498
 499         switch (seq->state) {
 500         case SEQ_STATE_UNINITIALIZED:
 501                 /* orphan event, do nothing */
 502                 return 0;
 503         case SEQ_STATE_ACQUIRING:
 504                 break;
 505         case SEQ_STATE_CONTENDED:
 506                 contended_term = sample->time - seq->prev_event_time;
 507                 ls->wait_time_total += contended_term;
 508                 if (contended_term < ls->wait_time_min)
 509                         ls->wait_time_min = contended_term;
 510                 if (ls->wait_time_max < contended_term)
 511                         ls->wait_time_max = contended_term;
 512                 break;
 513         case SEQ_STATE_RELEASED:
 514         case SEQ_STATE_ACQUIRED:
 515         case SEQ_STATE_READ_ACQUIRED:
 516                 /* broken lock sequence, discard it */
 517                 ls->discard = 1;
 518                 bad_hist[BROKEN_ACQUIRED]++;
 519                 list_del_init(&seq->list);
 520                 free(seq);
 521                 goto end;
 522         default:
 523                 BUG_ON("Unknown state of lock sequence found!\n");
 524                 break;
 525         }
 526
 527         seq->state = SEQ_STATE_ACQUIRED;
 528         ls->nr_acquired++;
 529         ls->avg_wait_time = ls->nr_contended ? ls->wait_time_total/ls->nr_contended : 0;
 530         seq->prev_event_time = sample->time;
 531 end:
 532         return 0;
 533 }
 534
 535 static int report_lock_contended_event(struct evsel *evsel,
 536                                        struct perf_sample *sample)
 537 {
 538         void *addr;
 539         struct lock_stat *ls;
 540         struct thread_stat *ts;
 541         struct lock_seq_stat *seq;
 542         const char *name = evsel__strval(evsel, sample, "name");
 543         u64 tmp = evsel__intval(evsel, sample, "lockdep_addr");
 544
 545         memcpy(&addr, &tmp, sizeof(void *));
 546
 547         ls = lock_stat_findnew(addr, name);
 548         if (!ls)
 549                 return -ENOMEM;
 550         if (ls->discard)
 551                 return 0;
 552
 553         ts = thread_stat_findnew(sample->tid);
 554         if (!ts)
 555                 return -ENOMEM;
 556
 557         seq = get_seq(ts, addr);
 558         if (!seq)
 559                 return -ENOMEM;
 560
 561         switch (seq->state) {
 562         case SEQ_STATE_UNINITIALIZED:
 563                 /* orphan event, do nothing */
 564                 return 0;
 565         case SEQ_STATE_ACQUIRING:
 566                 break;
 567         case SEQ_STATE_RELEASED:
 568         case SEQ_STATE_ACQUIRED:
 569         case SEQ_STATE_READ_ACQUIRED:
 570         case SEQ_STATE_CONTENDED:
 571                 /* broken lock sequence, discard it */
 572                 ls->discard = 1;
 573                 bad_hist[BROKEN_CONTENDED]++;
 574                 list_del_init(&seq->list);
 575                 free(seq);
 576                 goto end;
 577         default:
 578                 BUG_ON("Unknown state of lock sequence found!\n");
 579                 break;
 580         }
 581
 582         seq->state = SEQ_STATE_CONTENDED;
 583         ls->nr_contended++;
 584         ls->avg_wait_time = ls->wait_time_total/ls->nr_contended;
 585         seq->prev_event_time = sample->time;
 586 end:
 587         return 0;
 588 }
 589
 590 static int report_lock_release_event(struct evsel *evsel,
 591                                      struct perf_sample *sample)
 592 {
 593         void *addr;
 594         struct lock_stat *ls;
 595         struct thread_stat *ts;
 596         struct lock_seq_stat *seq;
 597         const char *name = evsel__strval(evsel, sample, "name");
 598         u64 tmp = evsel__intval(evsel, sample, "lockdep_addr");
 599
 600         memcpy(&addr, &tmp, sizeof(void *));
 601
 602         ls = lock_stat_findnew(addr, name);
 603         if (!ls)
 604                 return -ENOMEM;
 605         if (ls->discard)
 606                 return 0;
 607
 608         ts = thread_stat_findnew(sample->tid);
 609         if (!ts)
 610                 return -ENOMEM;
 611
 612         seq = get_seq(ts, addr);
 613         if (!seq)
 614                 return -ENOMEM;
 615
 616         switch (seq->state) {
 617         case SEQ_STATE_UNINITIALIZED:
 618                 goto end;
 619         case SEQ_STATE_ACQUIRED:
 620                 break;
 621         case SEQ_STATE_READ_ACQUIRED:
 622                 seq->read_count--;
 623                 BUG_ON(seq->read_count < 0);
 624                 if (!seq->read_count) {
 625                         ls->nr_release++;
 626                         goto end;
 627                 }
 628                 break;
 629         case SEQ_STATE_ACQUIRING:
 630         case SEQ_STATE_CONTENDED:
 631         case SEQ_STATE_RELEASED:
 632                 /* broken lock sequence, discard it */
 633                 ls->discard = 1;
 634                 bad_hist[BROKEN_RELEASE]++;
 635                 goto free_seq;
 636         default:
 637                 BUG_ON("Unknown state of lock sequence found!\n");
 638                 break;
 639         }
 640
 641         ls->nr_release++;
 642 free_seq:
 643         list_del_init(&seq->list);
 644         free(seq);
 645 end:
 646         return 0;
 647 }
 648
 649 /* lock oriented handlers */
 650 /* TODO: handlers for CPU oriented, thread oriented */
 651 static struct trace_lock_handler report_lock_ops  = {
 652         .acquire_event          = report_lock_acquire_event,
 653         .acquired_event         = report_lock_acquired_event,
 654         .contended_event        = report_lock_contended_event,
 655         .release_event          = report_lock_release_event,
 656 };
 657
 658 static struct trace_lock_handler *trace_handler;
 659
 660 static int evsel__process_lock_acquire(struct evsel *evsel, struct perf_sample *sample)
 661 {
 662         if (trace_handler->acquire_event)
 663                 return trace_handler->acquire_event(evsel, sample);
 664         return 0;
 665 }
 666
 667 static int evsel__process_lock_acquired(struct evsel *evsel, struct perf_sample *sample)
 668 {
 669         if (trace_handler->acquired_event)
 670                 return trace_handler->acquired_event(evsel, sample);
 671         return 0;
 672 }
 673
 674 static int evsel__process_lock_contended(struct evsel *evsel, struct perf_sample *sample)
 675 {
 676         if (trace_handler->contended_event)
 677                 return trace_handler->contended_event(evsel, sample);
 678         return 0;
 679 }
 680
 681 static int evsel__process_lock_release(struct evsel *evsel, struct perf_sample *sample)
 682 {
 683         if (trace_handler->release_event)
 684                 return trace_handler->release_event(evsel, sample);
 685         return 0;
 686 }
 687
 688 static void print_bad_events(int bad, int total)
 689 {
 690         /* Output for debug, this have to be removed */
 691         int i;
 692         const char *name[4] =
 693                 { "acquire", "acquired", "contended", "release" };
 694
 695         pr_info("\n=== output for debug===\n\n");
 696         pr_info("bad: %d, total: %d\n", bad, total);
 697         pr_info("bad rate: %.2f %%\n", (double)bad / (double)total * 100);
 698         pr_info("histogram of events caused bad sequence\n");
 699         for (i = 0; i < BROKEN_MAX; i++)
 700                 pr_info(" %10s: %d\n", name[i], bad_hist[i]);
 701 }
 702
 703 /* TODO: various way to print, coloring, nano or milli sec */
 704 static void print_result(void)
 705 {
 706         struct lock_stat *st;
 707         char cut_name[20];
 708         int bad, total;
 709
 710         pr_info("%20s ", "Name");
 711         pr_info("%10s ", "acquired");
 712         pr_info("%10s ", "contended");
 713
 714         pr_info("%15s ", "avg wait (ns)");
 715         pr_info("%15s ", "total wait (ns)");
 716         pr_info("%15s ", "max wait (ns)");
 717         pr_info("%15s ", "min wait (ns)");
 718
 719         pr_info("\n\n");
 720
 721         bad = total = 0;
 722         while ((st = pop_from_result())) {
 723                 total++;
 724                 if (st->discard) {
 725                         bad++;
 726                         continue;
 727                 }
 728                 bzero(cut_name, 20);
 729
 730                 if (strlen(st->name) < 16) {
 731                         /* output raw name */
 732                         pr_info("%20s ", st->name);
 733                 } else {
 734                         strncpy(cut_name, st->name, 16);
 735                         cut_name[16] = '.';
 736                         cut_name[17] = '.';
 737                         cut_name[18] = '.';
 738                         cut_name[19] = '\0';
 739                         /* cut off name for saving output style */
 740                         pr_info("%20s ", cut_name);
 741                 }
 742
 743                 pr_info("%10u ", st->nr_acquired);
 744                 pr_info("%10u ", st->nr_contended);
 745
 746                 pr_info("%15" PRIu64 " ", st->avg_wait_time);
 747                 pr_info("%15" PRIu64 " ", st->wait_time_total);
 748                 pr_info("%15" PRIu64 " ", st->wait_time_max);
 749                 pr_info("%15" PRIu64 " ", st->wait_time_min == ULLONG_MAX ?
 750                        0 : st->wait_time_min);
 751                 pr_info("\n");
 752         }
 753
 754         print_bad_events(bad, total);
 755 }
 756
 757 static bool info_threads, info_map;
 758
 759 static void dump_threads(void)
 760 {
 761         struct thread_stat *st;
 762         struct rb_node *node;
 763         struct thread *t;
 764
 765         pr_info("%10s: comm\n", "Thread ID");
 766
 767         node = rb_first(&thread_stats);
 768         while (node) {
 769                 st = container_of(node, struct thread_stat, rb);
 770                 t = perf_session__findnew(session, st->tid);
 771                 pr_info("%10d: %s\n", st->tid, thread__comm_str(t));
 772                 node = rb_next(node);
 773                 thread__put(t);
 774         }
 775 }
 776
 777 static void dump_map(void)
 778 {
 779         unsigned int i;
 780         struct lock_stat *st;
 781
 782         pr_info("Address of instance: name of class\n");
 783         for (i = 0; i < LOCKHASH_SIZE; i++) {
 784                 list_for_each_entry(st, &lockhash_table[i], hash_entry) {
 785                         pr_info(" %p: %s\n", st->addr, st->name);
 786                 }
 787         }
 788 }
 789
 790 static int dump_info(void)
 791 {
 792         int rc = 0;
 793
 794         if (info_threads)
 795                 dump_threads();
 796         else if (info_map)
 797                 dump_map();
 798         else {
 799                 rc = -1;
 800                 pr_err("Unknown type of information\n");
 801         }
 802
 803         return rc;
 804 }
 805
 806 typedef int (*tracepoint_handler)(struct evsel *evsel,
 807                                   struct perf_sample *sample);
 808
 809 static int process_sample_event(struct perf_tool *tool __maybe_unused,
 810                                 union perf_event *event,
 811                                 struct perf_sample *sample,
 812                                 struct evsel *evsel,
 813                                 struct machine *machine)
 814 {
 815         int err = 0;
 816         struct thread *thread = machine__findnew_thread(machine, sample->pid,
 817                                                         sample->tid);
 818
 819         if (thread == NULL) {
 820                 pr_debug("problem processing %d event, skipping it.\n",
 821                         event->header.type);
 822                 return -1;
 823         }
 824
 825         if (evsel->handler != NULL) {
 826                 tracepoint_handler f = evsel->handler;
 827                 err = f(evsel, sample);
 828         }
 829
 830         thread__put(thread);
 831
 832         return err;
 833 }
 834
 835 static void sort_result(void)
 836 {
 837         unsigned int i;
 838         struct lock_stat *st;
 839
 840         for (i = 0; i < LOCKHASH_SIZE; i++) {
 841                 list_for_each_entry(st, &lockhash_table[i], hash_entry) {
 842                         insert_to_result(st, compare);
 843                 }
 844         }
 845 }
 846
 847 static const struct evsel_str_handler lock_tracepoints[] = {
 848         { "lock:lock_acquire",   evsel__process_lock_acquire,   }, /* CONFIG_LOCKDEP */
 849         { "lock:lock_acquired",  evsel__process_lock_acquired,  }, /* CONFIG_LOCKDEP, CONFIG_LOCK_STAT */
 850         { "lock:lock_contended", evsel__process_lock_contended, }, /* CONFIG_LOCKDEP, CONFIG_LOCK_STAT */
 851         { "lock:lock_release",   evsel__process_lock_release,   }, /* CONFIG_LOCKDEP */
 852 };
 853
 854 static bool force;
 855
 856 static int __cmd_report(bool display_info)
 857 {
 858         int err = -EINVAL;
 859         struct perf_tool eops = {
 860                 .sample          = process_sample_event,
 861                 .comm            = perf_event__process_comm,
 862                 .namespaces      = perf_event__process_namespaces,
 863                 .ordered_events  = true,
 864         };
 865         struct perf_data data = {
 866                 .path  = input_name,
 867                 .mode  = PERF_DATA_MODE_READ,
 868                 .force = force,
 869         };
 870
 871         session = perf_session__new(&data, false, &eops);
 872         if (IS_ERR(session)) {
 873                 pr_err("Initializing perf session failed\n");
 874                 return PTR_ERR(session);
 875         }
 876
 877         symbol__init(&session->header.env);
 878
 879         if (!perf_session__has_traces(session, "lock record"))
 880                 goto out_delete;
 881
 882         if (perf_session__set_tracepoints_handlers(session, lock_tracepoints)) {
 883                 pr_err("Initializing perf session tracepoint handlers failed\n");
 884                 goto out_delete;
 885         }
 886
 887         if (select_key())
 888                 goto out_delete;
 889
 890         err = perf_session__process_events(session);
 891         if (err)
 892                 goto out_delete;
 893
 894         setup_pager();
 895         if (display_info) /* used for info subcommand */
 896                 err = dump_info();
 897         else {
 898                 sort_result();
 899                 print_result();
 900         }
 901
 902 out_delete:
 903         perf_session__delete(session);
 904         return err;
 905 }
 906
 907 static int __cmd_record(int argc, const char **argv)
 908 {
 909         const char *record_args[] = {
 910                 "record", "-R", "-m", "1024", "-c", "1",
 911         };
 912         unsigned int rec_argc, i, j, ret;
 913         const char **rec_argv;
 914
 915         for (i = 0; i < ARRAY_SIZE(lock_tracepoints); i++) {
 916                 if (!is_valid_tracepoint(lock_tracepoints[i].name)) {
 917                                 pr_err("tracepoint %s is not enabled. "
 918                                        "Are CONFIG_LOCKDEP and CONFIG_LOCK_STAT enabled?\n",
 919                                        lock_tracepoints[i].name);
 920                                 return 1;
 921                 }
 922         }
 923
 924         rec_argc = ARRAY_SIZE(record_args) + argc - 1;
 925         /* factor of 2 is for -e in front of each tracepoint */
 926         rec_argc += 2 * ARRAY_SIZE(lock_tracepoints);
 927
 928         rec_argv = calloc(rec_argc + 1, sizeof(char *));
 929         if (!rec_argv)
 930                 return -ENOMEM;
 931
 932         for (i = 0; i < ARRAY_SIZE(record_args); i++)
 933                 rec_argv[i] = strdup(record_args[i]);
 934
 935         for (j = 0; j < ARRAY_SIZE(lock_tracepoints); j++) {
 936                 rec_argv[i++] = "-e";
 937                 rec_argv[i++] = strdup(lock_tracepoints[j].name);
 938         }
 939
 940         for (j = 1; j < (unsigned int)argc; j++, i++)
 941                 rec_argv[i] = argv[j];
 942
 943         BUG_ON(i != rec_argc);
 944
 945         ret = cmd_record(i, rec_argv);
 946         free(rec_argv);
 947         return ret;
 948 }
 949
 950 int cmd_lock(int argc, const char **argv)
 951 {
 952         const struct option lock_options[] = {
 953         OPT_STRING('i', "input", &input_name, "file", "input file name"),
 954         OPT_INCR('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"),
 955         OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"),
 956         OPT_BOOLEAN('f', "force", &force, "don't complain, do it"),
 957         OPT_END()
 958         };
 959
 960         const struct option info_options[] = {
 961         OPT_BOOLEAN('t', "threads", &info_threads,
 962                     "dump thread list in perf.data"),
 963         OPT_BOOLEAN('m', "map", &info_map,
 964                     "map of lock instances (address:name table)"),
 965         OPT_PARENT(lock_options)
 966         };
 967
 968         const struct option report_options[] = {
 969         OPT_STRING('k', "key", &sort_key, "acquired",
 970                     "key for sorting (acquired / contended / avg_wait / wait_total / wait_max / wait_min)"),
 971         /* TODO: type */
 972         OPT_PARENT(lock_options)
 973         };
 974
 975         const char * const info_usage[] = {
 976                 "perf lock info [<options>]",
 977                 NULL
 978         };
 979         const char *const lock_subcommands[] = { "record", "report", "script",
 980                                                  "info", NULL };
 981         const char *lock_usage[] = {
 982                 NULL,
 983                 NULL
 984         };
 985         const char * const report_usage[] = {
 986                 "perf lock report [<options>]",
 987                 NULL
 988         };
 989         unsigned int i;
 990         int rc = 0;
 991
 992         for (i = 0; i < LOCKHASH_SIZE; i++)
 993                 INIT_LIST_HEAD(lockhash_table + i);
 994
 995         argc = parse_options_subcommand(argc, argv, lock_options, lock_subcommands,
 996                                         lock_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 997         if (!argc)
 998                 usage_with_options(lock_usage, lock_options);
 999
1000         if (!strncmp(argv[0], "rec", 3)) {
1001                 return __cmd_record(argc, argv);
1002         } else if (!strncmp(argv[0], "report", 6)) {
1003                 trace_handler = &report_lock_ops;
1004                 if (argc) {
1005                         argc = parse_options(argc, argv,
1006                                              report_options, report_usage, 0);
1007                         if (argc)
1008                                 usage_with_options(report_usage, report_options);
1009                 }
1010                 rc = __cmd_report(false);
1011         } else if (!strcmp(argv[0], "script")) {
1012                 /* Aliased to 'perf script' */
1013                 return cmd_script(argc, argv);
1014         } else if (!strcmp(argv[0], "info")) {
1015                 if (argc) {
1016                         argc = parse_options(argc, argv,
1017                                              info_options, info_usage, 0);
1018                         if (argc)
1019                                 usage_with_options(info_usage, info_options);
1020                 }
1021                 /* recycling report_lock_ops */
1022                 trace_handler = &report_lock_ops;
1023                 rc = __cmd_report(true);
1024         } else {
1025                 usage_with_options(lock_usage, lock_options);
1026         }
1027
1028         return rc;
1029 }