src/lttngtop.c

   1 /*
   2  * Copyright (C) 2011-2012 Julien Desfossez
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License Version 2 as
   6  * published by the Free Software Foundation;
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  11  * GNU General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along
  14  * with this program; if not, write to the Free Software Foundation, Inc.,
  15  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  16  */
  17
  18 #define _GNU_SOURCE
  19 #include <config.h>
  20 #include <stdio.h>
  21 #include <stdint.h>
  22 #include <babeltrace/babeltrace.h>
  23 #include <babeltrace/ctf/events.h>
  24 #include <babeltrace/ctf/callbacks.h>
  25 #include <babeltrace/ctf/iterator.h>
  26 #include <fcntl.h>
  27 #include <pthread.h>
  28 #include <popt.h>
  29 #include <stdlib.h>
  30 #include <ftw.h>
  31 #include <dirent.h>
  32 #include <ctype.h>
  33 #include <sys/stat.h>
  34 #include <unistd.h>
  35 #include <string.h>
  36 #include <errno.h>
  37 #include <sys/types.h>
  38 #include <fts.h>
  39 #include <assert.h>
  40
  41 #include "lttngtoptypes.h"
  42 #include "cputop.h"
  43 #include "iostreamtop.h"
  44 #include "cursesdisplay.h"
  45 #include "common.h"
  46
  47 #define DEFAULT_FILE_ARRAY_SIZE 1
  48
  49 const char *opt_input_path;
  50
  51 struct lttngtop *copy;
  52 pthread_t display_thread;
  53 pthread_t timer_thread;
  54
  55 unsigned long refresh_display = 1 * NSEC_PER_SEC;
  56 unsigned long last_display_update = 0;
  57 int quit = 0;
  58
  59 enum {
  60         OPT_NONE = 0,
  61         OPT_HELP,
  62         OPT_LIST,
  63         OPT_VERBOSE,
  64         OPT_DEBUG,
  65         OPT_NAMES,
  66 };
  67
  68 static struct poptOption long_options[] = {
  69         /* longName, shortName, argInfo, argPtr, value, descrip, argDesc */
  70         { "help", 'h', POPT_ARG_NONE, NULL, OPT_HELP, NULL, NULL },
  71         { NULL, 0, 0, NULL, 0, NULL, NULL },
  72 };
  73
  74 void *refresh_thread(void *p)
  75 {
  76         while (1) {
  77                 if (quit)
  78                         return NULL;
  79                 sem_wait(&pause_sem);
  80                 sem_post(&pause_sem);
  81                 sem_post(&timer);
  82                 sleep(refresh_display/NSEC_PER_SEC);
  83         }
  84 }
  85
  86 void *ncurses_display(void *p)
  87 {
  88         unsigned int current_display_index = 0;
  89
  90         sem_wait(&bootstrap);
  91         /*
  92          * Prevent the 1 second delay when we hit ESC
  93          */
  94         ESCDELAY = 0;
  95         init_ncurses();
  96
  97         while (1) {
  98                 sem_wait(&timer);
  99                 sem_wait(&goodtodisplay);
 100                 sem_wait(&pause_sem);
 101
 102                 copy = g_ptr_array_index(copies, current_display_index);
 103                 assert(copy);
 104                 display(current_display_index++);
 105
 106                 sem_post(&goodtoupdate);
 107                 sem_post(&pause_sem);
 108
 109                 if (quit) {
 110                         reset_ncurses();
 111                         pthread_exit(0);
 112                 }
 113         }
 114 }
 115
 116 /*
 117  * hook on each event to check the timestamp and refresh the display if
 118  * necessary
 119  */
 120 enum bt_cb_ret check_timestamp(struct bt_ctf_event *call_data, void *private_data)
 121 {
 122         unsigned long timestamp;
 123
 124         timestamp = bt_ctf_get_timestamp(call_data);
 125         if (timestamp == -1ULL)
 126                 goto error;
 127
 128         if (last_display_update == 0)
 129                 last_display_update = timestamp;
 130
 131         if (timestamp - last_display_update >= refresh_display) {
 132                 sem_wait(&goodtoupdate);
 133                 g_ptr_array_add(copies, get_copy_lttngtop(last_display_update,
 134                                         timestamp));
 135                 sem_post(&goodtodisplay);
 136                 sem_post(&bootstrap);
 137                 last_display_update = timestamp;
 138         }
 139         return BT_CB_OK;
 140
 141 error:
 142         fprintf(stderr, "check_timestamp callback error\n");
 143         return BT_CB_ERROR_STOP;
 144 }
 145
 146 /*
 147  * get_perf_counter : get or create and return a perf_counter struct for
 148  * either a process or a cpu (only one of the 2 parameters mandatory)
 149  */
 150 struct perfcounter *get_perf_counter(const char *name, struct processtop *proc,
 151                 struct cputime *cpu)
 152 {
 153         struct perfcounter *ret;
 154         GHashTable *table;
 155
 156         if (proc)
 157                 table = proc->perf;
 158         else if (cpu)
 159                 table = cpu->perf;
 160         else
 161                 goto error;
 162
 163         ret = g_hash_table_lookup(table, (gpointer) name);
 164         if (ret)
 165                 goto end;
 166
 167         ret = g_new0(struct perfcounter, 1);
 168         /* by default, make it visible in the UI */
 169         ret->visible = 1;
 170         g_hash_table_insert(table, (gpointer) strdup(name), ret);
 171
 172 end:
 173         return ret;
 174
 175 error:
 176         return NULL;
 177 }
 178
 179 void update_perf_value(struct processtop *proc, struct cputime *cpu,
 180                 const char *name, int value)
 181 {
 182         struct perfcounter *cpu_perf, *process_perf;
 183
 184         cpu_perf = get_perf_counter(name, NULL, cpu);
 185         if (cpu_perf->count < value) {
 186                 process_perf = get_perf_counter(name, proc, NULL);
 187                 process_perf->count += value - cpu_perf->count;
 188                 cpu_perf->count = value;
 189         }
 190 }
 191
 192 void extract_perf_counter_scope(const struct bt_ctf_event *event,
 193                 const struct bt_definition *scope,
 194                 struct processtop *proc,
 195                 struct cputime *cpu)
 196 {
 197         struct bt_definition const * const *list = NULL;
 198         const struct bt_definition *field;
 199         unsigned int count;
 200         struct perfcounter *perfcounter;
 201         GHashTableIter iter;
 202         gpointer key;
 203         int ret;
 204
 205         if (!scope)
 206                 goto end;
 207
 208         ret = bt_ctf_get_field_list(event, scope, &list, &count);
 209         if (ret < 0)
 210                 goto end;
 211
 212         if (count == 0)
 213                 goto end;
 214
 215         g_hash_table_iter_init(&iter, global_perf_liszt);
 216         while (g_hash_table_iter_next (&iter, &key, (gpointer) &perfcounter)) {
 217                 field = bt_ctf_get_field(event, scope, (char *) key);
 218                 if (field) {
 219                         int value = bt_ctf_get_uint64(field);
 220                         if (bt_ctf_field_get_error())
 221                                 continue;
 222                         update_perf_value(proc, cpu, (char *) key, value);
 223                 }
 224         }
 225
 226 end:
 227         return;
 228 }
 229
 230 void update_perf_counter(struct processtop *proc, const struct bt_ctf_event *event)
 231 {
 232         struct cputime *cpu;
 233         const struct bt_definition *scope;
 234
 235         cpu = get_cpu(get_cpu_id(event));
 236
 237         scope = bt_ctf_get_top_level_scope(event, BT_STREAM_EVENT_CONTEXT);
 238         extract_perf_counter_scope(event, scope, proc, cpu);
 239
 240         scope = bt_ctf_get_top_level_scope(event, BT_STREAM_PACKET_CONTEXT);
 241         extract_perf_counter_scope(event, scope, proc, cpu);
 242
 243         scope = bt_ctf_get_top_level_scope(event, BT_EVENT_CONTEXT);
 244         extract_perf_counter_scope(event, scope, proc, cpu);
 245 }
 246
 247 enum bt_cb_ret fix_process_table(struct bt_ctf_event *call_data,
 248                 void *private_data)
 249 {
 250         int pid, tid, ppid;
 251         char *comm;
 252         struct processtop *parent, *child;
 253         unsigned long timestamp;
 254
 255         timestamp = bt_ctf_get_timestamp(call_data);
 256         if (timestamp == -1ULL)
 257                 goto error;
 258
 259         pid = get_context_pid(call_data);
 260         if (pid == -1ULL) {
 261                 goto error;
 262         }
 263         tid = get_context_tid(call_data);
 264         if (tid == -1ULL) {
 265                 goto error;
 266         }
 267         ppid = get_context_ppid(call_data);
 268         if (ppid == -1ULL) {
 269                 goto error;
 270         }
 271         comm = get_context_comm(call_data);
 272         if (!comm) {
 273                 goto error;
 274         }
 275
 276         /* find or create the current process */
 277         child = find_process_tid(&lttngtop, tid, comm);
 278         if (!child)
 279                 child = add_proc(&lttngtop, tid, comm, timestamp);
 280         update_proc(child, pid, tid, ppid, comm);
 281
 282         if (pid != tid) {
 283                 /* find or create the parent */
 284                 parent = find_process_tid(&lttngtop, pid, comm);
 285                 if (!parent) {
 286                         parent = add_proc(&lttngtop, pid, comm, timestamp);
 287                         parent->pid = pid;
 288                 }
 289
 290                 /* attach the parent to the current process */
 291                 child->threadparent = parent;
 292                 add_thread(parent, child);
 293         }
 294
 295         update_perf_counter(child, call_data);
 296
 297         return BT_CB_OK;
 298
 299 error:
 300         return BT_CB_ERROR_STOP;
 301 }
 302
 303 void init_lttngtop()
 304 {
 305         copies = g_ptr_array_new();
 306         global_perf_liszt = g_hash_table_new(g_str_hash, g_str_equal);
 307
 308         sem_init(&goodtodisplay, 0, 0);
 309         sem_init(&goodtoupdate, 0, 1);
 310         sem_init(&timer, 0, 1);
 311         sem_init(&bootstrap, 0, 0);
 312         sem_init(&pause_sem, 0, 1);
 313         sem_init(&end_trace_sem, 0, 0);
 314
 315         reset_global_counters();
 316         lttngtop.nbproc = 0;
 317         lttngtop.nbthreads = 0;
 318         lttngtop.nbfiles = 0;
 319
 320         lttngtop.process_table = g_ptr_array_new();
 321         lttngtop.files_table = g_ptr_array_new();
 322         lttngtop.cpu_table = g_ptr_array_new();
 323 }
 324
 325 void usage(FILE *fp)
 326 {
 327         fprintf(fp, "LTTngTop %s\n\n", VERSION);
 328         fprintf(fp, "Usage : lttngtop /path/to/trace\n");
 329 }
 330
 331 /*
 332  * Return 0 if caller should continue, < 0 if caller should return
 333  * error, > 0 if caller should exit without reporting error.
 334  */
 335 static int parse_options(int argc, char **argv)
 336 {
 337         poptContext pc;
 338         int opt, ret = 0;
 339
 340         if (argc == 1) {
 341                 usage(stdout);
 342                 return 1;   /* exit cleanly */
 343         }
 344
 345         pc = poptGetContext(NULL, argc, (const char **) argv, long_options, 0);
 346         poptReadDefaultConfig(pc, 0);
 347
 348         while ((opt = poptGetNextOpt(pc)) != -1) {
 349                 switch (opt) {
 350                         case OPT_HELP:
 351                                 usage(stdout);
 352                                 ret = 1;    /* exit cleanly */
 353                                 goto end;
 354                         default:
 355                                 ret = -EINVAL;
 356                                 goto end;
 357                 }
 358         }
 359
 360         opt_input_path = poptGetArg(pc);
 361         if (!opt_input_path) {
 362                 ret = -EINVAL;
 363                 goto end;
 364         }
 365 end:
 366         if (pc) {
 367                 poptFreeContext(pc);
 368         }
 369         return ret;
 370 }
 371
 372 void iter_trace(struct bt_context *bt_ctx)
 373 {
 374         struct bt_ctf_iter *iter;
 375         struct bt_iter_pos begin_pos;
 376         const struct bt_ctf_event *event;
 377         int ret = 0;
 378
 379         begin_pos.type = BT_SEEK_BEGIN;
 380         iter = bt_ctf_iter_create(bt_ctx, &begin_pos, NULL);
 381
 382         /* at each event check if we need to refresh */
 383         bt_ctf_iter_add_callback(iter, 0, NULL, 0,
 384                         check_timestamp,
 385                         NULL, NULL, NULL);
 386         /* at each event, verify the status of the process table */
 387         bt_ctf_iter_add_callback(iter, 0, NULL, 0,
 388                         fix_process_table,
 389                         NULL, NULL, NULL);
 390         /* to handle the scheduling events */
 391         bt_ctf_iter_add_callback(iter,
 392                         g_quark_from_static_string("sched_switch"),
 393                         NULL, 0, handle_sched_switch, NULL, NULL, NULL);
 394         /* to clean up the process table */
 395         bt_ctf_iter_add_callback(iter,
 396                         g_quark_from_static_string("sched_process_free"),
 397                         NULL, 0, handle_sched_process_free, NULL, NULL, NULL);
 398         /* to get all the process from the statedumps */
 399         bt_ctf_iter_add_callback(iter,
 400                         g_quark_from_static_string(
 401                                 "lttng_statedump_process_state"),
 402                         NULL, 0, handle_statedump_process_state,
 403                         NULL, NULL, NULL);
 404
 405         /* for IO top */
 406         bt_ctf_iter_add_callback(iter,
 407                         g_quark_from_static_string("exit_syscall"),
 408                         NULL, 0, handle_exit_syscall, NULL, NULL, NULL);
 409         bt_ctf_iter_add_callback(iter,
 410                         g_quark_from_static_string("sys_write"),
 411                         NULL, 0, handle_sys_write, NULL, NULL, NULL);
 412         bt_ctf_iter_add_callback(iter,
 413                         g_quark_from_static_string("sys_read"),
 414                         NULL, 0, handle_sys_read, NULL, NULL, NULL);
 415         bt_ctf_iter_add_callback(iter,
 416                         g_quark_from_static_string("sys_open"),
 417                         NULL, 0, handle_sys_open, NULL, NULL, NULL);
 418         bt_ctf_iter_add_callback(iter,
 419                         g_quark_from_static_string("sys_close"),
 420                         NULL, 0, handle_sys_close, NULL, NULL, NULL);
 421         bt_ctf_iter_add_callback(iter,
 422                         g_quark_from_static_string(
 423                                         "lttng_statedump_file_descriptor"),
 424                         NULL, 0, handle_statedump_file_descriptor,
 425                         NULL, NULL, NULL);
 426
 427         while ((event = bt_ctf_iter_read_event(iter)) != NULL) {
 428                 ret = bt_iter_next(bt_ctf_get_iter(iter));
 429                 if (ret < 0)
 430                         goto end_iter;
 431         }
 432
 433         /* block until quit, we reached the end of the trace */
 434         sem_wait(&end_trace_sem);
 435
 436 end_iter:
 437         bt_ctf_iter_destroy(iter);
 438 }
 439
 440 /*
 441  * bt_context_add_traces_recursive: Open a trace recursively
 442  * (copied from BSD code in converter/babeltrace.c)
 443  *
 444  * Find each trace present in the subdirectory starting from the given
 445  * path, and add them to the context. The packet_seek parameter can be
 446  * NULL: this specify to use the default format packet_seek.
 447  *
 448  * Return: 0 on success, nonzero on failure.
 449  * Unable to open toplevel: failure.
 450  * Unable to open some subdirectory or file: warn and continue;
 451  */
 452 int bt_context_add_traces_recursive(struct bt_context *ctx, const char *path,
 453                 const char *format_str,
 454                 void (*packet_seek)(struct bt_stream_pos *pos,
 455                         size_t offset, int whence))
 456 {
 457         FTS *tree;
 458         FTSENT *node;
 459         GArray *trace_ids;
 460         char lpath[PATH_MAX];
 461         char * const paths[2] = { lpath, NULL };
 462         int ret = -1;
 463
 464         /*
 465          * Need to copy path, because fts_open can change it.
 466          * It is the pointer array, not the strings, that are constant.
 467          */
 468         strncpy(lpath, path, PATH_MAX);
 469         lpath[PATH_MAX - 1] = '\0';
 470
 471         tree = fts_open(paths, FTS_NOCHDIR | FTS_LOGICAL, 0);
 472         if (tree == NULL) {
 473                 fprintf(stderr, "[error] [Context] Cannot traverse \"%s\" for reading.\n",
 474                                 path);
 475                 return -EINVAL;
 476         }
 477
 478         trace_ids = g_array_new(FALSE, TRUE, sizeof(int));
 479
 480         while ((node = fts_read(tree))) {
 481                 int dirfd, metafd;
 482
 483                 if (!(node->fts_info & FTS_D))
 484                         continue;
 485
 486                 dirfd = open(node->fts_accpath, 0);
 487                 if (dirfd < 0) {
 488                         fprintf(stderr, "[error] [Context] Unable to open trace "
 489                                 "directory file descriptor.\n");
 490                         ret = dirfd;
 491                         goto error;
 492                 }
 493                 metafd = openat(dirfd, "metadata", O_RDONLY);
 494                 if (metafd < 0) {
 495                         close(dirfd);
 496                         ret = -1;
 497                         continue;
 498                 } else {
 499                         int trace_id;
 500
 501                         ret = close(metafd);
 502                         if (ret < 0) {
 503                                 perror("close");
 504                                 goto error;
 505                         }
 506                         ret = close(dirfd);
 507                         if (ret < 0) {
 508                                 perror("close");
 509                                 goto error;
 510                         }
 511
 512                         trace_id = bt_context_add_trace(ctx,
 513                                 node->fts_accpath, format_str,
 514                                 packet_seek, NULL, NULL);
 515                         if (trace_id < 0) {
 516                                 fprintf(stderr, "[warning] [Context] opening trace \"%s\" from %s "
 517                                         "for reading.\n", node->fts_accpath, path);
 518                                 /* Allow to skip erroneous traces. */
 519                                 continue;
 520                         }
 521                         g_array_append_val(trace_ids, trace_id);
 522                 }
 523         }
 524
 525         g_array_free(trace_ids, TRUE);
 526         return ret;
 527
 528 error:
 529         return ret;
 530 }
 531
 532 static int check_field_requirements(const struct bt_ctf_field_decl *const * field_list,
 533                 int field_cnt, int *tid_check, int *pid_check,
 534                 int *procname_check, int *ppid_check)
 535 {
 536         int j;
 537         struct perfcounter *global;
 538         const char *name;
 539
 540         for (j = 0; j < field_cnt; j++) {
 541                 name = bt_ctf_get_decl_field_name(field_list[j]);
 542                 if (*tid_check == 0) {
 543                         if (strncmp(name, "tid", 3) == 0)
 544                                 (*tid_check)++;
 545                 }
 546                 if (*pid_check == 0) {
 547                         if (strncmp(name, "pid", 3) == 0)
 548                                 (*pid_check)++;
 549                 }
 550                 if (*ppid_check == 0) {
 551                         if (strncmp(name, "ppid", 4) == 0)
 552                                 (*ppid_check)++;
 553                 }
 554                 if (*procname_check == 0) {
 555                         if (strncmp(name, "procname", 8) == 0)
 556                                 (*procname_check)++;
 557                 }
 558                 if (strncmp(name, "perf_", 5) == 0) {
 559                         global = g_hash_table_lookup(global_perf_liszt, (gpointer) name);
 560                         if (!global) {
 561                                 global = g_new0(struct perfcounter, 1);
 562                                 /* by default, sort on the first perf context */
 563                                 if (g_hash_table_size(global_perf_liszt) == 0)
 564                                         global->sort = 1;
 565                                 global->visible = 1;
 566                                 g_hash_table_insert(global_perf_liszt, (gpointer) strdup(name), global);
 567                         }
 568                 }
 569         }
 570
 571         if (*tid_check == 1 && *pid_check == 1 && *ppid_check == 1 &&
 572                         *procname_check == 1)
 573                 return 0;
 574
 575         return -1;
 576 }
 577
 578 /*
 579  * check_requirements: check if the required context informations are available
 580  *
 581  * If each mandatory context information is available for at least in one
 582  * event, return 0 otherwise return -1.
 583  */
 584 int check_requirements(struct bt_context *ctx)
 585 {
 586         unsigned int i, evt_cnt, field_cnt;
 587         struct bt_ctf_event_decl *const * evt_list;
 588         const struct bt_ctf_field_decl *const * field_list;
 589         int tid_check = 0;
 590         int pid_check = 0;
 591         int procname_check = 0;
 592         int ppid_check = 0;
 593         int ret = 0;
 594
 595         bt_ctf_get_event_decl_list(0, ctx, &evt_list, &evt_cnt);
 596         for (i = 0; i < evt_cnt; i++) {
 597                 bt_ctf_get_decl_fields(evt_list[i], BT_STREAM_EVENT_CONTEXT,
 598                                 &field_list, &field_cnt);
 599                 ret = check_field_requirements(field_list, field_cnt,
 600                                 &tid_check, &pid_check, &procname_check,
 601                                 &ppid_check);
 602
 603                 bt_ctf_get_decl_fields(evt_list[i], BT_EVENT_CONTEXT,
 604                                 &field_list, &field_cnt);
 605                 ret = check_field_requirements(field_list, field_cnt,
 606                                 &tid_check, &pid_check, &procname_check,
 607                                 &ppid_check);
 608
 609                 bt_ctf_get_decl_fields(evt_list[i], BT_STREAM_PACKET_CONTEXT,
 610                                 &field_list, &field_cnt);
 611                 ret = check_field_requirements(field_list, field_cnt,
 612                                 &tid_check, &pid_check, &procname_check,
 613                                 &ppid_check);
 614         }
 615
 616         if (tid_check == 0) {
 617                 ret = -1;
 618                 fprintf(stderr, "[error] missing tid context information\n");
 619         }
 620         if (pid_check == 0) {
 621                 ret = -1;
 622                 fprintf(stderr, "[error] missing pid context information\n");
 623         }
 624         if (ppid_check == 0) {
 625                 ret = -1;
 626                 fprintf(stderr, "[error] missing ppid context information\n");
 627         }
 628         if (procname_check == 0) {
 629                 ret = -1;
 630                 fprintf(stderr, "[error] missing procname context information\n");
 631         }
 632
 633         return ret;
 634 }
 635
 636 int main(int argc, char **argv)
 637 {
 638         int ret;
 639         struct bt_context *bt_ctx = NULL;
 640
 641         ret = parse_options(argc, argv);
 642         if (ret < 0) {
 643                 fprintf(stdout, "Error parsing options.\n\n");
 644                 usage(stdout);
 645                 exit(EXIT_FAILURE);
 646         } else if (ret > 0) {
 647                 exit(EXIT_SUCCESS);
 648         }
 649
 650         init_lttngtop();
 651
 652         bt_ctx = bt_context_create();
 653         ret = bt_context_add_traces_recursive(bt_ctx, opt_input_path, "ctf", NULL);
 654         if (ret < 0) {
 655                 fprintf(stderr, "[error] Opening the trace\n");
 656                 goto end;
 657         }
 658
 659         ret = check_requirements(bt_ctx);
 660         if (ret < 0) {
 661                 fprintf(stderr, "[error] some mandatory contexts were missing, exiting.\n");
 662                 goto end;
 663         }
 664
 665         pthread_create(&display_thread, NULL, ncurses_display, (void *) NULL);
 666         pthread_create(&timer_thread, NULL, refresh_thread, (void *) NULL);
 667
 668         iter_trace(bt_ctx);
 669
 670         quit = 1;
 671         pthread_join(display_thread, NULL);
 672         pthread_join(timer_thread, NULL);
 673
 674 end:
 675         bt_context_put(bt_ctx);
 676         return 0;
 677 }