backend.c

   1 /*
   2  * fio - the flexible io tester
   3  *
   4  * Copyright (C) 2005 Jens Axboe <axboe@suse.de>
   5  * Copyright (C) 2006-2012 Jens Axboe <axboe@kernel.dk>
   6  *
   7  * The license below covers all files distributed with fio unless otherwise
   8  * noted in the file itself.
   9  *
  10  *  This program is free software; you can redistribute it and/or modify
  11  *  it under the terms of the GNU General Public License version 2 as
  12  *  published by the Free Software Foundation.
  13  *
  14  *  This program is distributed in the hope that it will be useful,
  15  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  *  GNU General Public License for more details.
  18  *
  19  *  You should have received a copy of the GNU General Public License
  20  *  along with this program; if not, write to the Free Software
  21  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  22  *
  23  */
  24 #include <unistd.h>
  25 #include <fcntl.h>
  26 #include <string.h>
  27 #include <limits.h>
  28 #include <signal.h>
  29 #include <time.h>
  30 #include <locale.h>
  31 #include <assert.h>
  32 #include <time.h>
  33 #include <inttypes.h>
  34 #include <sys/stat.h>
  35 #include <sys/wait.h>
  36 #include <sys/ipc.h>
  37 #include <sys/mman.h>
  38 #include <math.h>
  39
  40 #include "fio.h"
  41 #ifndef FIO_NO_HAVE_SHM_H
  42 #include <sys/shm.h>
  43 #endif
  44 #include "hash.h"
  45 #include "smalloc.h"
  46 #include "verify.h"
  47 #include "trim.h"
  48 #include "diskutil.h"
  49 #include "cgroup.h"
  50 #include "profile.h"
  51 #include "lib/rand.h"
  52 #include "lib/memalign.h"
  53 #include "server.h"
  54 #include "lib/getrusage.h"
  55 #include "idletime.h"
  56 #include "err.h"
  57 #include "workqueue.h"
  58 #include "lib/mountcheck.h"
  59 #include "rate-submit.h"
  60 #include "helper_thread.h"
  61
  62 static struct fio_mutex *startup_mutex;
  63 static struct flist_head *cgroup_list;
  64 static char *cgroup_mnt;
  65 static int exit_value;
  66 static volatile int fio_abort;
  67 static unsigned int nr_process = 0;
  68 static unsigned int nr_thread = 0;
  69
  70 struct io_log *agg_io_log[DDIR_RWDIR_CNT];
  71
  72 int groupid = 0;
  73 unsigned int thread_number = 0;
  74 unsigned int stat_number = 0;
  75 int shm_id = 0;
  76 int temp_stall_ts;
  77 unsigned long done_secs = 0;
  78
  79 #define JOB_START_TIMEOUT       (5 * 1000)
  80
  81 static void sig_int(int sig)
  82 {
  83         if (threads) {
  84                 if (is_backend)
  85                         fio_server_got_signal(sig);
  86                 else {
  87                         log_info("\nfio: terminating on signal %d\n", sig);
  88                         log_info_flush();
  89                         exit_value = 128;
  90                 }
  91
  92                 fio_terminate_threads(TERMINATE_ALL);
  93         }
  94 }
  95
  96 void sig_show_status(int sig)
  97 {
  98         show_running_run_stats();
  99 }
 100
 101 static void set_sig_handlers(void)
 102 {
 103         struct sigaction act;
 104
 105         memset(&act, 0, sizeof(act));
 106         act.sa_handler = sig_int;
 107         act.sa_flags = SA_RESTART;
 108         sigaction(SIGINT, &act, NULL);
 109
 110         memset(&act, 0, sizeof(act));
 111         act.sa_handler = sig_int;
 112         act.sa_flags = SA_RESTART;
 113         sigaction(SIGTERM, &act, NULL);
 114
 115 /* Windows uses SIGBREAK as a quit signal from other applications */
 116 #ifdef WIN32
 117         memset(&act, 0, sizeof(act));
 118         act.sa_handler = sig_int;
 119         act.sa_flags = SA_RESTART;
 120         sigaction(SIGBREAK, &act, NULL);
 121 #endif
 122
 123         memset(&act, 0, sizeof(act));
 124         act.sa_handler = sig_show_status;
 125         act.sa_flags = SA_RESTART;
 126         sigaction(SIGUSR1, &act, NULL);
 127
 128         if (is_backend) {
 129                 memset(&act, 0, sizeof(act));
 130                 act.sa_handler = sig_int;
 131                 act.sa_flags = SA_RESTART;
 132                 sigaction(SIGPIPE, &act, NULL);
 133         }
 134 }
 135
 136 /*
 137  * Check if we are above the minimum rate given.
 138  */
 139 static bool __check_min_rate(struct thread_data *td, struct timeval *now,
 140                              enum fio_ddir ddir)
 141 {
 142         unsigned long long bytes = 0;
 143         unsigned long iops = 0;
 144         unsigned long spent;
 145         unsigned long rate;
 146         unsigned int ratemin = 0;
 147         unsigned int rate_iops = 0;
 148         unsigned int rate_iops_min = 0;
 149
 150         assert(ddir_rw(ddir));
 151
 152         if (!td->o.ratemin[ddir] && !td->o.rate_iops_min[ddir])
 153                 return false;
 154
 155         /*
 156          * allow a 2 second settle period in the beginning
 157          */
 158         if (mtime_since(&td->start, now) < 2000)
 159                 return false;
 160
 161         iops += td->this_io_blocks[ddir];
 162         bytes += td->this_io_bytes[ddir];
 163         ratemin += td->o.ratemin[ddir];
 164         rate_iops += td->o.rate_iops[ddir];
 165         rate_iops_min += td->o.rate_iops_min[ddir];
 166
 167         /*
 168          * if rate blocks is set, sample is running
 169          */
 170         if (td->rate_bytes[ddir] || td->rate_blocks[ddir]) {
 171                 spent = mtime_since(&td->lastrate[ddir], now);
 172                 if (spent < td->o.ratecycle)
 173                         return false;
 174
 175                 if (td->o.rate[ddir] || td->o.ratemin[ddir]) {
 176                         /*
 177                          * check bandwidth specified rate
 178                          */
 179                         if (bytes < td->rate_bytes[ddir]) {
 180                                 log_err("%s: rate_min=%uB/s not met, only transferred %lluB\n",
 181                                         td->o.name, ratemin, bytes);
 182                                 return true;
 183                         } else {
 184                                 if (spent)
 185                                         rate = ((bytes - td->rate_bytes[ddir]) * 1000) / spent;
 186                                 else
 187                                         rate = 0;
 188
 189                                 if (rate < ratemin ||
 190                                     bytes < td->rate_bytes[ddir]) {
 191                                         log_err("%s: rate_min=%uB/s not met, got %luB/s\n",
 192                                                 td->o.name, ratemin, rate);
 193                                         return true;
 194                                 }
 195                         }
 196                 } else {
 197                         /*
 198                          * checks iops specified rate
 199                          */
 200                         if (iops < rate_iops) {
 201                                 log_err("%s: rate_iops_min=%u not met, only performed %lu IOs\n",
 202                                                 td->o.name, rate_iops, iops);
 203                                 return true;
 204                         } else {
 205                                 if (spent)
 206                                         rate = ((iops - td->rate_blocks[ddir]) * 1000) / spent;
 207                                 else
 208                                         rate = 0;
 209
 210                                 if (rate < rate_iops_min ||
 211                                     iops < td->rate_blocks[ddir]) {
 212                                         log_err("%s: rate_iops_min=%u not met, got %lu IOPS\n",
 213                                                 td->o.name, rate_iops_min, rate);
 214                                         return true;
 215                                 }
 216                         }
 217                 }
 218         }
 219
 220         td->rate_bytes[ddir] = bytes;
 221         td->rate_blocks[ddir] = iops;
 222         memcpy(&td->lastrate[ddir], now, sizeof(*now));
 223         return false;
 224 }
 225
 226 static bool check_min_rate(struct thread_data *td, struct timeval *now)
 227 {
 228         bool ret = false;
 229
 230         if (td->bytes_done[DDIR_READ])
 231                 ret |= __check_min_rate(td, now, DDIR_READ);
 232         if (td->bytes_done[DDIR_WRITE])
 233                 ret |= __check_min_rate(td, now, DDIR_WRITE);
 234         if (td->bytes_done[DDIR_TRIM])
 235                 ret |= __check_min_rate(td, now, DDIR_TRIM);
 236
 237         return ret;
 238 }
 239
 240 /*
 241  * When job exits, we can cancel the in-flight IO if we are using async
 242  * io. Attempt to do so.
 243  */
 244 static void cleanup_pending_aio(struct thread_data *td)
 245 {
 246         int r;
 247
 248         /*
 249          * get immediately available events, if any
 250          */
 251         r = io_u_queued_complete(td, 0);
 252         if (r < 0)
 253                 return;
 254
 255         /*
 256          * now cancel remaining active events
 257          */
 258         if (td->io_ops->cancel) {
 259                 struct io_u *io_u;
 260                 int i;
 261
 262                 io_u_qiter(&td->io_u_all, io_u, i) {
 263                         if (io_u->flags & IO_U_F_FLIGHT) {
 264                                 r = td->io_ops->cancel(td, io_u);
 265                                 if (!r)
 266                                         put_io_u(td, io_u);
 267                         }
 268                 }
 269         }
 270
 271         if (td->cur_depth)
 272                 r = io_u_queued_complete(td, td->cur_depth);
 273 }
 274
 275 /*
 276  * Helper to handle the final sync of a file. Works just like the normal
 277  * io path, just does everything sync.
 278  */
 279 static bool fio_io_sync(struct thread_data *td, struct fio_file *f)
 280 {
 281         struct io_u *io_u = __get_io_u(td);
 282         int ret;
 283
 284         if (!io_u)
 285                 return true;
 286
 287         io_u->ddir = DDIR_SYNC;
 288         io_u->file = f;
 289
 290         if (td_io_prep(td, io_u)) {
 291                 put_io_u(td, io_u);
 292                 return true;
 293         }
 294
 295 requeue:
 296         ret = td_io_queue(td, io_u);
 297         if (ret < 0) {
 298                 td_verror(td, io_u->error, "td_io_queue");
 299                 put_io_u(td, io_u);
 300                 return true;
 301         } else if (ret == FIO_Q_QUEUED) {
 302                 if (td_io_commit(td))
 303                         return true;
 304                 if (io_u_queued_complete(td, 1) < 0)
 305                         return true;
 306         } else if (ret == FIO_Q_COMPLETED) {
 307                 if (io_u->error) {
 308                         td_verror(td, io_u->error, "td_io_queue");
 309                         return true;
 310                 }
 311
 312                 if (io_u_sync_complete(td, io_u) < 0)
 313                         return true;
 314         } else if (ret == FIO_Q_BUSY) {
 315                 if (td_io_commit(td))
 316                         return true;
 317                 goto requeue;
 318         }
 319
 320         return false;
 321 }
 322
 323 static int fio_file_fsync(struct thread_data *td, struct fio_file *f)
 324 {
 325         int ret;
 326
 327         if (fio_file_open(f))
 328                 return fio_io_sync(td, f);
 329
 330         if (td_io_open_file(td, f))
 331                 return 1;
 332
 333         ret = fio_io_sync(td, f);
 334         td_io_close_file(td, f);
 335         return ret;
 336 }
 337
 338 static inline void __update_tv_cache(struct thread_data *td)
 339 {
 340         fio_gettime(&td->tv_cache, NULL);
 341 }
 342
 343 static inline void update_tv_cache(struct thread_data *td)
 344 {
 345         if ((++td->tv_cache_nr & td->tv_cache_mask) == td->tv_cache_mask)
 346                 __update_tv_cache(td);
 347 }
 348
 349 static inline bool runtime_exceeded(struct thread_data *td, struct timeval *t)
 350 {
 351         if (in_ramp_time(td))
 352                 return false;
 353         if (!td->o.timeout)
 354                 return false;
 355         if (utime_since(&td->epoch, t) >= td->o.timeout)
 356                 return true;
 357
 358         return false;
 359 }
 360
 361 /*
 362  * We need to update the runtime consistently in ms, but keep a running
 363  * tally of the current elapsed time in microseconds for sub millisecond
 364  * updates.
 365  */
 366 static inline void update_runtime(struct thread_data *td,
 367                                   unsigned long long *elapsed_us,
 368                                   const enum fio_ddir ddir)
 369 {
 370         if (ddir == DDIR_WRITE && td_write(td) && td->o.verify_only)
 371                 return;
 372
 373         td->ts.runtime[ddir] -= (elapsed_us[ddir] + 999) / 1000;
 374         elapsed_us[ddir] += utime_since_now(&td->start);
 375         td->ts.runtime[ddir] += (elapsed_us[ddir] + 999) / 1000;
 376 }
 377
 378 static bool break_on_this_error(struct thread_data *td, enum fio_ddir ddir,
 379                                 int *retptr)
 380 {
 381         int ret = *retptr;
 382
 383         if (ret < 0 || td->error) {
 384                 int err = td->error;
 385                 enum error_type_bit eb;
 386
 387                 if (ret < 0)
 388                         err = -ret;
 389
 390                 eb = td_error_type(ddir, err);
 391                 if (!(td->o.continue_on_error & (1 << eb)))
 392                         return true;
 393
 394                 if (td_non_fatal_error(td, eb, err)) {
 395                         /*
 396                          * Continue with the I/Os in case of
 397                          * a non fatal error.
 398                          */
 399                         update_error_count(td, err);
 400                         td_clear_error(td);
 401                         *retptr = 0;
 402                         return false;
 403                 } else if (td->o.fill_device && err == ENOSPC) {
 404                         /*
 405                          * We expect to hit this error if
 406                          * fill_device option is set.
 407                          */
 408                         td_clear_error(td);
 409                         fio_mark_td_terminate(td);
 410                         return true;
 411                 } else {
 412                         /*
 413                          * Stop the I/O in case of a fatal
 414                          * error.
 415                          */
 416                         update_error_count(td, err);
 417                         return true;
 418                 }
 419         }
 420
 421         return false;
 422 }
 423
 424 static void check_update_rusage(struct thread_data *td)
 425 {
 426         if (td->update_rusage) {
 427                 td->update_rusage = 0;
 428                 update_rusage_stat(td);
 429                 fio_mutex_up(td->rusage_sem);
 430         }
 431 }
 432
 433 static int wait_for_completions(struct thread_data *td, struct timeval *time)
 434 {
 435         const int full = queue_full(td);
 436         int min_evts = 0;
 437         int ret;
 438
 439         if (td->flags & TD_F_REGROW_LOGS)
 440                 return io_u_quiesce(td);
 441
 442         /*
 443          * if the queue is full, we MUST reap at least 1 event
 444          */
 445         min_evts = min(td->o.iodepth_batch_complete_min, td->cur_depth);
 446         if ((full && !min_evts) || !td->o.iodepth_batch_complete_min)
 447                 min_evts = 1;
 448
 449         if (time && (__should_check_rate(td, DDIR_READ) ||
 450             __should_check_rate(td, DDIR_WRITE) ||
 451             __should_check_rate(td, DDIR_TRIM)))
 452                 fio_gettime(time, NULL);
 453
 454         do {
 455                 ret = io_u_queued_complete(td, min_evts);
 456                 if (ret < 0)
 457                         break;
 458         } while (full && (td->cur_depth > td->o.iodepth_low));
 459
 460         return ret;
 461 }
 462
 463 int io_queue_event(struct thread_data *td, struct io_u *io_u, int *ret,
 464                    enum fio_ddir ddir, uint64_t *bytes_issued, int from_verify,
 465                    struct timeval *comp_time)
 466 {
 467         int ret2;
 468
 469         switch (*ret) {
 470         case FIO_Q_COMPLETED:
 471                 if (io_u->error) {
 472                         *ret = -io_u->error;
 473                         clear_io_u(td, io_u);
 474                 } else if (io_u->resid) {
 475                         int bytes = io_u->xfer_buflen - io_u->resid;
 476                         struct fio_file *f = io_u->file;
 477
 478                         if (bytes_issued)
 479                                 *bytes_issued += bytes;
 480
 481                         if (!from_verify)
 482                                 trim_io_piece(td, io_u);
 483
 484                         /*
 485                          * zero read, fail
 486                          */
 487                         if (!bytes) {
 488                                 if (!from_verify)
 489                                         unlog_io_piece(td, io_u);
 490                                 td_verror(td, EIO, "full resid");
 491                                 put_io_u(td, io_u);
 492                                 break;
 493                         }
 494
 495                         io_u->xfer_buflen = io_u->resid;
 496                         io_u->xfer_buf += bytes;
 497                         io_u->offset += bytes;
 498
 499                         if (ddir_rw(io_u->ddir))
 500                                 td->ts.short_io_u[io_u->ddir]++;
 501
 502                         f = io_u->file;
 503                         if (io_u->offset == f->real_file_size)
 504                                 goto sync_done;
 505
 506                         requeue_io_u(td, &io_u);
 507                 } else {
 508 sync_done:
 509                         if (comp_time && (__should_check_rate(td, DDIR_READ) ||
 510                             __should_check_rate(td, DDIR_WRITE) ||
 511                             __should_check_rate(td, DDIR_TRIM)))
 512                                 fio_gettime(comp_time, NULL);
 513
 514                         *ret = io_u_sync_complete(td, io_u);
 515                         if (*ret < 0)
 516                                 break;
 517                 }
 518
 519                 if (td->flags & TD_F_REGROW_LOGS)
 520                         regrow_logs(td);
 521
 522                 /*
 523                  * when doing I/O (not when verifying),
 524                  * check for any errors that are to be ignored
 525                  */
 526                 if (!from_verify)
 527                         break;
 528
 529                 return 0;
 530         case FIO_Q_QUEUED:
 531                 /*
 532                  * if the engine doesn't have a commit hook,
 533                  * the io_u is really queued. if it does have such
 534                  * a hook, it has to call io_u_queued() itself.
 535                  */
 536                 if (td->io_ops->commit == NULL)
 537                         io_u_queued(td, io_u);
 538                 if (bytes_issued)
 539                         *bytes_issued += io_u->xfer_buflen;
 540                 break;
 541         case FIO_Q_BUSY:
 542                 if (!from_verify)
 543                         unlog_io_piece(td, io_u);
 544                 requeue_io_u(td, &io_u);
 545                 ret2 = td_io_commit(td);
 546                 if (ret2 < 0)
 547                         *ret = ret2;
 548                 break;
 549         default:
 550                 assert(*ret < 0);
 551                 td_verror(td, -(*ret), "td_io_queue");
 552                 break;
 553         }
 554
 555         if (break_on_this_error(td, ddir, ret))
 556                 return 1;
 557
 558         return 0;
 559 }
 560
 561 static inline bool io_in_polling(struct thread_data *td)
 562 {
 563         return !td->o.iodepth_batch_complete_min &&
 564                    !td->o.iodepth_batch_complete_max;
 565 }
 566 /*
 567  * Unlinks files from thread data fio_file structure
 568  */
 569 static int unlink_all_files(struct thread_data *td)
 570 {
 571         struct fio_file *f;
 572         unsigned int i;
 573         int ret = 0;
 574
 575         for_each_file(td, f, i) {
 576                 if (f->filetype != FIO_TYPE_FILE)
 577                         continue;
 578                 ret = td_io_unlink_file(td, f);
 579                 if (ret)
 580                         break;
 581         }
 582
 583         if (ret)
 584                 td_verror(td, ret, "unlink_all_files");
 585
 586         return ret;
 587 }
 588
 589 /*
 590  * The main verify engine. Runs over the writes we previously submitted,
 591  * reads the blocks back in, and checks the crc/md5 of the data.
 592  */
 593 static void do_verify(struct thread_data *td, uint64_t verify_bytes)
 594 {
 595         struct fio_file *f;
 596         struct io_u *io_u;
 597         int ret, min_events;
 598         unsigned int i;
 599
 600         dprint(FD_VERIFY, "starting loop\n");
 601
 602         /*
 603          * sync io first and invalidate cache, to make sure we really
 604          * read from disk.
 605          */
 606         for_each_file(td, f, i) {
 607                 if (!fio_file_open(f))
 608                         continue;
 609                 if (fio_io_sync(td, f))
 610                         break;
 611                 if (file_invalidate_cache(td, f))
 612                         break;
 613         }
 614
 615         check_update_rusage(td);
 616
 617         if (td->error)
 618                 return;
 619
 620         /*
 621          * verify_state needs to be reset before verification
 622          * proceeds so that expected random seeds match actual
 623          * random seeds in headers. The main loop will reset
 624          * all random number generators if randrepeat is set.
 625          */
 626         if (!td->o.rand_repeatable)
 627                 td_fill_verify_state_seed(td);
 628
 629         td_set_runstate(td, TD_VERIFYING);
 630
 631         io_u = NULL;
 632         while (!td->terminate) {
 633                 enum fio_ddir ddir;
 634                 int full;
 635
 636                 update_tv_cache(td);
 637                 check_update_rusage(td);
 638
 639                 if (runtime_exceeded(td, &td->tv_cache)) {
 640                         __update_tv_cache(td);
 641                         if (runtime_exceeded(td, &td->tv_cache)) {
 642                                 fio_mark_td_terminate(td);
 643                                 break;
 644                         }
 645                 }
 646
 647                 if (flow_threshold_exceeded(td))
 648                         continue;
 649
 650                 if (!td->o.experimental_verify) {
 651                         io_u = __get_io_u(td);
 652                         if (!io_u)
 653                                 break;
 654
 655                         if (get_next_verify(td, io_u)) {
 656                                 put_io_u(td, io_u);
 657                                 break;
 658                         }
 659
 660                         if (td_io_prep(td, io_u)) {
 661                                 put_io_u(td, io_u);
 662                                 break;
 663                         }
 664                 } else {
 665                         if (ddir_rw_sum(td->bytes_done) + td->o.rw_min_bs > verify_bytes)
 666                                 break;
 667
 668                         while ((io_u = get_io_u(td)) != NULL) {
 669                                 if (IS_ERR_OR_NULL(io_u)) {
 670                                         io_u = NULL;
 671                                         ret = FIO_Q_BUSY;
 672                                         goto reap;
 673                                 }
 674
 675                                 /*
 676                                  * We are only interested in the places where
 677                                  * we wrote or trimmed IOs. Turn those into
 678                                  * reads for verification purposes.
 679                                  */
 680                                 if (io_u->ddir == DDIR_READ) {
 681                                         /*
 682                                          * Pretend we issued it for rwmix
 683                                          * accounting
 684                                          */
 685                                         td->io_issues[DDIR_READ]++;
 686                                         put_io_u(td, io_u);
 687                                         continue;
 688                                 } else if (io_u->ddir == DDIR_TRIM) {
 689                                         io_u->ddir = DDIR_READ;
 690                                         io_u_set(td, io_u, IO_U_F_TRIMMED);
 691                                         break;
 692                                 } else if (io_u->ddir == DDIR_WRITE) {
 693                                         io_u->ddir = DDIR_READ;
 694                                         break;
 695                                 } else {
 696                                         put_io_u(td, io_u);
 697                                         continue;
 698                                 }
 699                         }
 700
 701                         if (!io_u)
 702                                 break;
 703                 }
 704
 705                 if (verify_state_should_stop(td, io_u)) {
 706                         put_io_u(td, io_u);
 707                         break;
 708                 }
 709
 710                 if (td->o.verify_async)
 711                         io_u->end_io = verify_io_u_async;
 712                 else
 713                         io_u->end_io = verify_io_u;
 714
 715                 ddir = io_u->ddir;
 716                 if (!td->o.disable_slat)
 717                         fio_gettime(&io_u->start_time, NULL);
 718
 719                 ret = td_io_queue(td, io_u);
 720
 721                 if (io_queue_event(td, io_u, &ret, ddir, NULL, 1, NULL))
 722                         break;
 723
 724                 /*
 725                  * if we can queue more, do so. but check if there are
 726                  * completed io_u's first. Note that we can get BUSY even
 727                  * without IO queued, if the system is resource starved.
 728                  */
 729 reap:
 730                 full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth);
 731                 if (full || io_in_polling(td))
 732                         ret = wait_for_completions(td, NULL);
 733
 734                 if (ret < 0)
 735                         break;
 736         }
 737
 738         check_update_rusage(td);
 739
 740         if (!td->error) {
 741                 min_events = td->cur_depth;
 742
 743                 if (min_events)
 744                         ret = io_u_queued_complete(td, min_events);
 745         } else
 746                 cleanup_pending_aio(td);
 747
 748         td_set_runstate(td, TD_RUNNING);
 749
 750         dprint(FD_VERIFY, "exiting loop\n");
 751 }
 752
 753 static bool exceeds_number_ios(struct thread_data *td)
 754 {
 755         unsigned long long number_ios;
 756
 757         if (!td->o.number_ios)
 758                 return false;
 759
 760         number_ios = ddir_rw_sum(td->io_blocks);
 761         number_ios += td->io_u_queued + td->io_u_in_flight;
 762
 763         return number_ios >= (td->o.number_ios * td->loops);
 764 }
 765
 766 static bool io_bytes_exceeded(struct thread_data *td, uint64_t *this_bytes)
 767 {
 768         unsigned long long bytes, limit;
 769
 770         if (td_rw(td))
 771                 bytes = this_bytes[DDIR_READ] + this_bytes[DDIR_WRITE];
 772         else if (td_write(td))
 773                 bytes = this_bytes[DDIR_WRITE];
 774         else if (td_read(td))
 775                 bytes = this_bytes[DDIR_READ];
 776         else
 777                 bytes = this_bytes[DDIR_TRIM];
 778
 779         if (td->o.io_size)
 780                 limit = td->o.io_size;
 781         else
 782                 limit = td->o.size;
 783
 784         limit *= td->loops;
 785         return bytes >= limit || exceeds_number_ios(td);
 786 }
 787
 788 static bool io_issue_bytes_exceeded(struct thread_data *td)
 789 {
 790         return io_bytes_exceeded(td, td->io_issue_bytes);
 791 }
 792
 793 static bool io_complete_bytes_exceeded(struct thread_data *td)
 794 {
 795         return io_bytes_exceeded(td, td->this_io_bytes);
 796 }
 797
 798 /*
 799  * used to calculate the next io time for rate control
 800  *
 801  */
 802 static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
 803 {
 804         uint64_t secs, remainder, bps, bytes, iops;
 805
 806         assert(!(td->flags & TD_F_CHILD));
 807         bytes = td->rate_io_issue_bytes[ddir];
 808         bps = td->rate_bps[ddir];
 809
 810         if (td->o.rate_process == RATE_PROCESS_POISSON) {
 811                 uint64_t val;
 812                 iops = bps / td->o.bs[ddir];
 813                 val = (int64_t) (1000000 / iops) *
 814                                 -logf(__rand_0_1(&td->poisson_state[ddir]));
 815                 if (val) {
 816                         dprint(FD_RATE, "poisson rate iops=%llu, ddir=%d\n",
 817                                         (unsigned long long) 1000000 / val,
 818                                         ddir);
 819                 }
 820                 td->last_usec[ddir] += val;
 821                 return td->last_usec[ddir];
 822         } else if (bps) {
 823                 secs = bytes / bps;
 824                 remainder = bytes % bps;
 825                 return remainder * 1000000 / bps + secs * 1000000;
 826         }
 827
 828         return 0;
 829 }
 830
 831 /*
 832  * Main IO worker function. It retrieves io_u's to process and queues
 833  * and reaps them, checking for rate and errors along the way.
 834  *
 835  * Returns number of bytes written and trimmed.
 836  */
 837 static void do_io(struct thread_data *td, uint64_t *bytes_done)
 838 {
 839         unsigned int i;
 840         int ret = 0;
 841         uint64_t total_bytes, bytes_issued = 0;
 842
 843         for (i = 0; i < DDIR_RWDIR_CNT; i++)
 844                 bytes_done[i] = td->bytes_done[i];
 845
 846         if (in_ramp_time(td))
 847                 td_set_runstate(td, TD_RAMP);
 848         else
 849                 td_set_runstate(td, TD_RUNNING);
 850
 851         lat_target_init(td);
 852
 853         total_bytes = td->o.size;
 854         /*
 855         * Allow random overwrite workloads to write up to io_size
 856         * before starting verification phase as 'size' doesn't apply.
 857         */
 858         if (td_write(td) && td_random(td) && td->o.norandommap)
 859                 total_bytes = max(total_bytes, (uint64_t) td->o.io_size);
 860         /*
 861          * If verify_backlog is enabled, we'll run the verify in this
 862          * handler as well. For that case, we may need up to twice the
 863          * amount of bytes.
 864          */
 865         if (td->o.verify != VERIFY_NONE &&
 866            (td_write(td) && td->o.verify_backlog))
 867                 total_bytes += td->o.size;
 868
 869         /* In trimwrite mode, each byte is trimmed and then written, so
 870          * allow total_bytes to be twice as big */
 871         if (td_trimwrite(td))
 872                 total_bytes += td->total_io_size;
 873
 874         while ((td->o.read_iolog_file && !flist_empty(&td->io_log_list)) ||
 875                 (!flist_empty(&td->trim_list)) || !io_issue_bytes_exceeded(td) ||
 876                 td->o.time_based) {
 877                 struct timeval comp_time;
 878                 struct io_u *io_u;
 879                 int full;
 880                 enum fio_ddir ddir;
 881
 882                 check_update_rusage(td);
 883
 884                 if (td->terminate || td->done)
 885                         break;
 886
 887                 update_tv_cache(td);
 888
 889                 if (runtime_exceeded(td, &td->tv_cache)) {
 890                         __update_tv_cache(td);
 891                         if (runtime_exceeded(td, &td->tv_cache)) {
 892                                 fio_mark_td_terminate(td);
 893                                 break;
 894                         }
 895                 }
 896
 897                 if (flow_threshold_exceeded(td))
 898                         continue;
 899
 900                 /*
 901                  * Break if we exceeded the bytes. The exception is time
 902                  * based runs, but we still need to break out of the loop
 903                  * for those to run verification, if enabled.
 904                  */
 905                 if (bytes_issued >= total_bytes &&
 906                     (!td->o.time_based ||
 907                      (td->o.time_based && td->o.verify != VERIFY_NONE)))
 908                         break;
 909
 910                 io_u = get_io_u(td);
 911                 if (IS_ERR_OR_NULL(io_u)) {
 912                         int err = PTR_ERR(io_u);
 913
 914                         io_u = NULL;
 915                         if (err == -EBUSY) {
 916                                 ret = FIO_Q_BUSY;
 917                                 goto reap;
 918                         }
 919                         if (td->o.latency_target)
 920                                 goto reap;
 921                         break;
 922                 }
 923
 924                 ddir = io_u->ddir;
 925
 926                 /*
 927                  * Add verification end_io handler if:
 928                  *      - Asked to verify (!td_rw(td))
 929                  *      - Or the io_u is from our verify list (mixed write/ver)
 930                  */
 931                 if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ &&
 932                     ((io_u->flags & IO_U_F_VER_LIST) || !td_rw(td))) {
 933
 934                         if (!td->o.verify_pattern_bytes) {
 935                                 io_u->rand_seed = __rand(&td->verify_state);
 936                                 if (sizeof(int) != sizeof(long *))
 937                                         io_u->rand_seed *= __rand(&td->verify_state);
 938                         }
 939
 940                         if (verify_state_should_stop(td, io_u)) {
 941                                 put_io_u(td, io_u);
 942                                 break;
 943                         }
 944
 945                         if (td->o.verify_async)
 946                                 io_u->end_io = verify_io_u_async;
 947                         else
 948                                 io_u->end_io = verify_io_u;
 949                         td_set_runstate(td, TD_VERIFYING);
 950                 } else if (in_ramp_time(td))
 951                         td_set_runstate(td, TD_RAMP);
 952                 else
 953                         td_set_runstate(td, TD_RUNNING);
 954
 955                 /*
 956                  * Always log IO before it's issued, so we know the specific
 957                  * order of it. The logged unit will track when the IO has
 958                  * completed.
 959                  */
 960                 if (td_write(td) && io_u->ddir == DDIR_WRITE &&
 961                     td->o.do_verify &&
 962                     td->o.verify != VERIFY_NONE &&
 963                     !td->o.experimental_verify)
 964                         log_io_piece(td, io_u);
 965
 966                 if (td->o.io_submit_mode == IO_MODE_OFFLOAD) {
 967                         const unsigned long blen = io_u->xfer_buflen;
 968                         const enum fio_ddir ddir = acct_ddir(io_u);
 969
 970                         if (td->error)
 971                                 break;
 972
 973                         workqueue_enqueue(&td->io_wq, &io_u->work);
 974                         ret = FIO_Q_QUEUED;
 975
 976                         if (ddir_rw(ddir)) {
 977                                 td->io_issues[ddir]++;
 978                                 td->io_issue_bytes[ddir] += blen;
 979                                 td->rate_io_issue_bytes[ddir] += blen;
 980                         }
 981
 982                         if (should_check_rate(td))
 983                                 td->rate_next_io_time[ddir] = usec_for_io(td, ddir);
 984
 985                 } else {
 986                         ret = td_io_queue(td, io_u);
 987
 988                         if (should_check_rate(td))
 989                                 td->rate_next_io_time[ddir] = usec_for_io(td, ddir);
 990
 991                         if (io_queue_event(td, io_u, &ret, ddir, &bytes_issued, 0, &comp_time))
 992                                 break;
 993
 994                         /*
 995                          * See if we need to complete some commands. Note that
 996                          * we can get BUSY even without IO queued, if the
 997                          * system is resource starved.
 998                          */
 999 reap:
1000                         full = queue_full(td) ||
1001                                 (ret == FIO_Q_BUSY && td->cur_depth);
1002                         if (full || io_in_polling(td))
1003                                 ret = wait_for_completions(td, &comp_time);
1004                 }
1005                 if (ret < 0)
1006                         break;
1007                 if (!ddir_rw_sum(td->bytes_done) &&
1008                     !td_ioengine_flagged(td, FIO_NOIO))
1009                         continue;
1010
1011                 if (!in_ramp_time(td) && should_check_rate(td)) {
1012                         if (check_min_rate(td, &comp_time)) {
1013                                 if (exitall_on_terminate || td->o.exitall_error)
1014                                         fio_terminate_threads(td->groupid);
1015                                 td_verror(td, EIO, "check_min_rate");
1016                                 break;
1017                         }
1018                 }
1019                 if (!in_ramp_time(td) && td->o.latency_target)
1020                         lat_target_check(td);
1021
1022                 if (td->o.thinktime) {
1023                         unsigned long long b;
1024
1025                         b = ddir_rw_sum(td->io_blocks);
1026                         if (!(b % td->o.thinktime_blocks)) {
1027                                 int left;
1028
1029                                 io_u_quiesce(td);
1030
1031                                 if (td->o.thinktime_spin)
1032                                         usec_spin(td->o.thinktime_spin);
1033
1034                                 left = td->o.thinktime - td->o.thinktime_spin;
1035                                 if (left)
1036                                         usec_sleep(td, left);
1037                         }
1038                 }
1039         }
1040
1041         check_update_rusage(td);
1042
1043         if (td->trim_entries)
1044                 log_err("fio: %lu trim entries leaked?\n", td->trim_entries);
1045
1046         if (td->o.fill_device && td->error == ENOSPC) {
1047                 td->error = 0;
1048                 fio_mark_td_terminate(td);
1049         }
1050         if (!td->error) {
1051                 struct fio_file *f;
1052
1053                 if (td->o.io_submit_mode == IO_MODE_OFFLOAD) {
1054                         workqueue_flush(&td->io_wq);
1055                         i = 0;
1056                 } else
1057                         i = td->cur_depth;
1058
1059                 if (i) {
1060                         ret = io_u_queued_complete(td, i);
1061                         if (td->o.fill_device && td->error == ENOSPC)
1062                                 td->error = 0;
1063                 }
1064
1065                 if (should_fsync(td) && td->o.end_fsync) {
1066                         td_set_runstate(td, TD_FSYNCING);
1067
1068                         for_each_file(td, f, i) {
1069                                 if (!fio_file_fsync(td, f))
1070                                         continue;
1071
1072                                 log_err("fio: end_fsync failed for file %s\n",
1073                                                                 f->file_name);
1074                         }
1075                 }
1076         } else
1077                 cleanup_pending_aio(td);
1078
1079         /*
1080          * stop job if we failed doing any IO
1081          */
1082         if (!ddir_rw_sum(td->this_io_bytes))
1083                 td->done = 1;
1084
1085         for (i = 0; i < DDIR_RWDIR_CNT; i++)
1086                 bytes_done[i] = td->bytes_done[i] - bytes_done[i];
1087 }
1088
1089 static void free_file_completion_logging(struct thread_data *td)
1090 {
1091         struct fio_file *f;
1092         unsigned int i;
1093
1094         for_each_file(td, f, i) {
1095                 if (!f->last_write_comp)
1096                         break;
1097                 sfree(f->last_write_comp);
1098         }
1099 }
1100
1101 static int init_file_completion_logging(struct thread_data *td,
1102                                         unsigned int depth)
1103 {
1104         struct fio_file *f;
1105         unsigned int i;
1106
1107         if (td->o.verify == VERIFY_NONE || !td->o.verify_state_save)
1108                 return 0;
1109
1110         for_each_file(td, f, i) {
1111                 f->last_write_comp = scalloc(depth, sizeof(uint64_t));
1112                 if (!f->last_write_comp)
1113                         goto cleanup;
1114         }
1115
1116         return 0;
1117
1118 cleanup:
1119         free_file_completion_logging(td);
1120         log_err("fio: failed to alloc write comp data\n");
1121         return 1;
1122 }
1123
1124 static void cleanup_io_u(struct thread_data *td)
1125 {
1126         struct io_u *io_u;
1127
1128         while ((io_u = io_u_qpop(&td->io_u_freelist)) != NULL) {
1129
1130                 if (td->io_ops->io_u_free)
1131                         td->io_ops->io_u_free(td, io_u);
1132
1133                 fio_memfree(io_u, sizeof(*io_u));
1134         }
1135
1136         free_io_mem(td);
1137
1138         io_u_rexit(&td->io_u_requeues);
1139         io_u_qexit(&td->io_u_freelist);
1140         io_u_qexit(&td->io_u_all);
1141
1142         free_file_completion_logging(td);
1143 }
1144
1145 static int init_io_u(struct thread_data *td)
1146 {
1147         struct io_u *io_u;
1148         unsigned int max_bs, min_write;
1149         int cl_align, i, max_units;
1150         int data_xfer = 1, err;
1151         char *p;
1152
1153         max_units = td->o.iodepth;
1154         max_bs = td_max_bs(td);
1155         min_write = td->o.min_bs[DDIR_WRITE];
1156         td->orig_buffer_size = (unsigned long long) max_bs
1157                                         * (unsigned long long) max_units;
1158
1159         if (td_ioengine_flagged(td, FIO_NOIO) || !(td_read(td) || td_write(td)))
1160                 data_xfer = 0;
1161
1162         err = 0;
1163         err += io_u_rinit(&td->io_u_requeues, td->o.iodepth);
1164         err += io_u_qinit(&td->io_u_freelist, td->o.iodepth);
1165         err += io_u_qinit(&td->io_u_all, td->o.iodepth);
1166
1167         if (err) {
1168                 log_err("fio: failed setting up IO queues\n");
1169                 return 1;
1170         }
1171
1172         /*
1173          * if we may later need to do address alignment, then add any
1174          * possible adjustment here so that we don't cause a buffer
1175          * overflow later. this adjustment may be too much if we get
1176          * lucky and the allocator gives us an aligned address.
1177          */
1178         if (td->o.odirect || td->o.mem_align || td->o.oatomic ||
1179             td_ioengine_flagged(td, FIO_RAWIO))
1180                 td->orig_buffer_size += page_mask + td->o.mem_align;
1181
1182         if (td->o.mem_type == MEM_SHMHUGE || td->o.mem_type == MEM_MMAPHUGE) {
1183                 unsigned long bs;
1184
1185                 bs = td->orig_buffer_size + td->o.hugepage_size - 1;
1186                 td->orig_buffer_size = bs & ~(td->o.hugepage_size - 1);
1187         }
1188
1189         if (td->orig_buffer_size != (size_t) td->orig_buffer_size) {
1190                 log_err("fio: IO memory too large. Reduce max_bs or iodepth\n");
1191                 return 1;
1192         }
1193
1194         if (data_xfer && allocate_io_mem(td))
1195                 return 1;
1196
1197         if (td->o.odirect || td->o.mem_align || td->o.oatomic ||
1198             td_ioengine_flagged(td, FIO_RAWIO))
1199                 p = PTR_ALIGN(td->orig_buffer, page_mask) + td->o.mem_align;
1200         else
1201                 p = td->orig_buffer;
1202
1203         cl_align = os_cache_line_size();
1204
1205         for (i = 0; i < max_units; i++) {
1206                 void *ptr;
1207
1208                 if (td->terminate)
1209                         return 1;
1210
1211                 ptr = fio_memalign(cl_align, sizeof(*io_u));
1212                 if (!ptr) {
1213                         log_err("fio: unable to allocate aligned memory\n");
1214                         break;
1215                 }
1216
1217                 io_u = ptr;
1218                 memset(io_u, 0, sizeof(*io_u));
1219                 INIT_FLIST_HEAD(&io_u->verify_list);
1220                 dprint(FD_MEM, "io_u alloc %p, index %u\n", io_u, i);
1221
1222                 if (data_xfer) {
1223                         io_u->buf = p;
1224                         dprint(FD_MEM, "io_u %p, mem %p\n", io_u, io_u->buf);
1225
1226                         if (td_write(td))
1227                                 io_u_fill_buffer(td, io_u, min_write, max_bs);
1228                         if (td_write(td) && td->o.verify_pattern_bytes) {
1229                                 /*
1230                                  * Fill the buffer with the pattern if we are
1231                                  * going to be doing writes.
1232                                  */
1233                                 fill_verify_pattern(td, io_u->buf, max_bs, io_u, 0, 0);
1234                         }
1235                 }
1236
1237                 io_u->index = i;
1238                 io_u->flags = IO_U_F_FREE;
1239                 io_u_qpush(&td->io_u_freelist, io_u);
1240
1241                 /*
1242                  * io_u never leaves this stack, used for iteration of all
1243                  * io_u buffers.
1244                  */
1245                 io_u_qpush(&td->io_u_all, io_u);
1246
1247                 if (td->io_ops->io_u_init) {
1248                         int ret = td->io_ops->io_u_init(td, io_u);
1249
1250                         if (ret) {
1251                                 log_err("fio: failed to init engine data: %d\n", ret);
1252                                 return 1;
1253                         }
1254                 }
1255
1256                 p += max_bs;
1257         }
1258
1259         if (init_file_completion_logging(td, max_units))
1260                 return 1;
1261
1262         return 0;
1263 }
1264
1265 /*
1266  * This function is Linux specific.
1267  * FIO_HAVE_IOSCHED_SWITCH enabled currently means it's Linux.
1268  */
1269 static int switch_ioscheduler(struct thread_data *td)
1270 {
1271 #ifdef FIO_HAVE_IOSCHED_SWITCH
1272         char tmp[256], tmp2[128];
1273         FILE *f;
1274         int ret;
1275
1276         if (td_ioengine_flagged(td, FIO_DISKLESSIO))
1277                 return 0;
1278
1279         assert(td->files && td->files[0]);
1280         sprintf(tmp, "%s/queue/scheduler", td->files[0]->du->sysfs_root);
1281
1282         f = fopen(tmp, "r+");
1283         if (!f) {
1284                 if (errno == ENOENT) {
1285                         log_err("fio: os or kernel doesn't support IO scheduler"
1286                                 " switching\n");
1287                         return 0;
1288                 }
1289                 td_verror(td, errno, "fopen iosched");
1290                 return 1;
1291         }
1292
1293         /*
1294          * Set io scheduler.
1295          */
1296         ret = fwrite(td->o.ioscheduler, strlen(td->o.ioscheduler), 1, f);
1297         if (ferror(f) || ret != 1) {
1298                 td_verror(td, errno, "fwrite");
1299                 fclose(f);
1300                 return 1;
1301         }
1302
1303         rewind(f);
1304
1305         /*
1306          * Read back and check that the selected scheduler is now the default.
1307          */
1308         memset(tmp, 0, sizeof(tmp));
1309         ret = fread(tmp, sizeof(tmp), 1, f);
1310         if (ferror(f) || ret < 0) {
1311                 td_verror(td, errno, "fread");
1312                 fclose(f);
1313                 return 1;
1314         }
1315         /*
1316          * either a list of io schedulers or "none\n" is expected.
1317          */
1318         tmp[strlen(tmp) - 1] = '\0';
1319
1320         /*
1321          * Write to "none" entry doesn't fail, so check the result here.
1322          */
1323         if (!strcmp(tmp, "none")) {
1324                 log_err("fio: io scheduler is not tunable\n");
1325                 fclose(f);
1326                 return 0;
1327         }
1328
1329         sprintf(tmp2, "[%s]", td->o.ioscheduler);
1330         if (!strstr(tmp, tmp2)) {
1331                 log_err("fio: io scheduler %s not found\n", td->o.ioscheduler);
1332                 td_verror(td, EINVAL, "iosched_switch");
1333                 fclose(f);
1334                 return 1;
1335         }
1336
1337         fclose(f);
1338         return 0;
1339 #else
1340         return 0;
1341 #endif
1342 }
1343
1344 static bool keep_running(struct thread_data *td)
1345 {
1346         unsigned long long limit;
1347
1348         if (td->done)
1349                 return false;
1350         if (td->o.time_based)
1351                 return true;
1352         if (td->o.loops) {
1353                 td->o.loops--;
1354                 return true;
1355         }
1356         if (exceeds_number_ios(td))
1357                 return false;
1358
1359         if (td->o.io_size)
1360                 limit = td->o.io_size;
1361         else
1362                 limit = td->o.size;
1363
1364         if (limit != -1ULL && ddir_rw_sum(td->io_bytes) < limit) {
1365                 uint64_t diff;
1366
1367                 /*
1368                  * If the difference is less than the maximum IO size, we
1369                  * are done.
1370                  */
1371                 diff = limit - ddir_rw_sum(td->io_bytes);
1372                 if (diff < td_max_bs(td))
1373                         return false;
1374
1375                 if (fio_files_done(td) && !td->o.io_size)
1376                         return false;
1377
1378                 return true;
1379         }
1380
1381         return false;
1382 }
1383
1384 static int exec_string(struct thread_options *o, const char *string, const char *mode)
1385 {
1386         size_t newlen = strlen(string) + strlen(o->name) + strlen(mode) + 9 + 1;
1387         int ret;
1388         char *str;
1389
1390         str = malloc(newlen);
1391         sprintf(str, "%s &> %s.%s.txt", string, o->name, mode);
1392
1393         log_info("%s : Saving output of %s in %s.%s.txt\n",o->name, mode, o->name, mode);
1394         ret = system(str);
1395         if (ret == -1)
1396                 log_err("fio: exec of cmd <%s> failed\n", str);
1397
1398         free(str);
1399         return ret;
1400 }
1401
1402 /*
1403  * Dry run to compute correct state of numberio for verification.
1404  */
1405 static uint64_t do_dry_run(struct thread_data *td)
1406 {
1407         td_set_runstate(td, TD_RUNNING);
1408
1409         while ((td->o.read_iolog_file && !flist_empty(&td->io_log_list)) ||
1410                 (!flist_empty(&td->trim_list)) || !io_complete_bytes_exceeded(td)) {
1411                 struct io_u *io_u;
1412                 int ret;
1413
1414                 if (td->terminate || td->done)
1415                         break;
1416
1417                 io_u = get_io_u(td);
1418                 if (IS_ERR_OR_NULL(io_u))
1419                         break;
1420
1421                 io_u_set(td, io_u, IO_U_F_FLIGHT);
1422                 io_u->error = 0;
1423                 io_u->resid = 0;
1424                 if (ddir_rw(acct_ddir(io_u)))
1425                         td->io_issues[acct_ddir(io_u)]++;
1426                 if (ddir_rw(io_u->ddir)) {
1427                         io_u_mark_depth(td, 1);
1428                         td->ts.total_io_u[io_u->ddir]++;
1429                 }
1430
1431                 if (td_write(td) && io_u->ddir == DDIR_WRITE &&
1432                     td->o.do_verify &&
1433                     td->o.verify != VERIFY_NONE &&
1434                     !td->o.experimental_verify)
1435                         log_io_piece(td, io_u);
1436
1437                 ret = io_u_sync_complete(td, io_u);
1438                 (void) ret;
1439         }
1440
1441         return td->bytes_done[DDIR_WRITE] + td->bytes_done[DDIR_TRIM];
1442 }
1443
1444 struct fork_data {
1445         struct thread_data *td;
1446         struct sk_out *sk_out;
1447 };
1448
1449 /*
1450  * Entry point for the thread based jobs. The process based jobs end up
1451  * here as well, after a little setup.
1452  */
1453 static void *thread_main(void *data)
1454 {
1455         struct fork_data *fd = data;
1456         unsigned long long elapsed_us[DDIR_RWDIR_CNT] = { 0, };
1457         struct thread_data *td = fd->td;
1458         struct thread_options *o = &td->o;
1459         struct sk_out *sk_out = fd->sk_out;
1460         int deadlock_loop_cnt;
1461         int clear_state;
1462         int ret;
1463
1464         sk_out_assign(sk_out);
1465         free(fd);
1466
1467         if (!o->use_thread) {
1468                 setsid();
1469                 td->pid = getpid();
1470         } else
1471                 td->pid = gettid();
1472
1473         fio_local_clock_init(o->use_thread);
1474
1475         dprint(FD_PROCESS, "jobs pid=%d started\n", (int) td->pid);
1476
1477         if (is_backend)
1478                 fio_server_send_start(td);
1479
1480         INIT_FLIST_HEAD(&td->io_log_list);
1481         INIT_FLIST_HEAD(&td->io_hist_list);
1482         INIT_FLIST_HEAD(&td->verify_list);
1483         INIT_FLIST_HEAD(&td->trim_list);
1484         INIT_FLIST_HEAD(&td->next_rand_list);
1485         td->io_hist_tree = RB_ROOT;
1486
1487         ret = mutex_cond_init_pshared(&td->io_u_lock, &td->free_cond);
1488         if (ret) {
1489                 td_verror(td, ret, "mutex_cond_init_pshared");
1490                 goto err;
1491         }
1492         ret = cond_init_pshared(&td->verify_cond);
1493         if (ret) {
1494                 td_verror(td, ret, "mutex_cond_pshared");
1495                 goto err;
1496         }
1497
1498         td_set_runstate(td, TD_INITIALIZED);
1499         dprint(FD_MUTEX, "up startup_mutex\n");
1500         fio_mutex_up(startup_mutex);
1501         dprint(FD_MUTEX, "wait on td->mutex\n");
1502         fio_mutex_down(td->mutex);
1503         dprint(FD_MUTEX, "done waiting on td->mutex\n");
1504
1505         /*
1506          * A new gid requires privilege, so we need to do this before setting
1507          * the uid.
1508          */
1509         if (o->gid != -1U && setgid(o->gid)) {
1510                 td_verror(td, errno, "setgid");
1511                 goto err;
1512         }
1513         if (o->uid != -1U && setuid(o->uid)) {
1514                 td_verror(td, errno, "setuid");
1515                 goto err;
1516         }
1517
1518         /*
1519          * Do this early, we don't want the compress threads to be limited
1520          * to the same CPUs as the IO workers. So do this before we set
1521          * any potential CPU affinity
1522          */
1523         if (iolog_compress_init(td, sk_out))
1524                 goto err;
1525
1526         /*
1527          * If we have a gettimeofday() thread, make sure we exclude that
1528          * thread from this job
1529          */
1530         if (o->gtod_cpu)
1531                 fio_cpu_clear(&o->cpumask, o->gtod_cpu);
1532
1533         /*
1534          * Set affinity first, in case it has an impact on the memory
1535          * allocations.
1536          */
1537         if (fio_option_is_set(o, cpumask)) {
1538                 if (o->cpus_allowed_policy == FIO_CPUS_SPLIT) {
1539                         ret = fio_cpus_split(&o->cpumask, td->thread_number - 1);
1540                         if (!ret) {
1541                                 log_err("fio: no CPUs set\n");
1542                                 log_err("fio: Try increasing number of available CPUs\n");
1543                                 td_verror(td, EINVAL, "cpus_split");
1544                                 goto err;
1545                         }
1546                 }
1547                 ret = fio_setaffinity(td->pid, o->cpumask);
1548                 if (ret == -1) {
1549                         td_verror(td, errno, "cpu_set_affinity");
1550                         goto err;
1551                 }
1552         }
1553
1554 #ifdef CONFIG_LIBNUMA
1555         /* numa node setup */
1556         if (fio_option_is_set(o, numa_cpunodes) ||
1557             fio_option_is_set(o, numa_memnodes)) {
1558                 struct bitmask *mask;
1559
1560                 if (numa_available() < 0) {
1561                         td_verror(td, errno, "Does not support NUMA API\n");
1562                         goto err;
1563                 }
1564
1565                 if (fio_option_is_set(o, numa_cpunodes)) {
1566                         mask = numa_parse_nodestring(o->numa_cpunodes);
1567                         ret = numa_run_on_node_mask(mask);
1568                         numa_free_nodemask(mask);
1569                         if (ret == -1) {
1570                                 td_verror(td, errno, \
1571                                         "numa_run_on_node_mask failed\n");
1572                                 goto err;
1573                         }
1574                 }
1575
1576                 if (fio_option_is_set(o, numa_memnodes)) {
1577                         mask = NULL;
1578                         if (o->numa_memnodes)
1579                                 mask = numa_parse_nodestring(o->numa_memnodes);
1580
1581                         switch (o->numa_mem_mode) {
1582                         case MPOL_INTERLEAVE:
1583                                 numa_set_interleave_mask(mask);
1584                                 break;
1585                         case MPOL_BIND:
1586                                 numa_set_membind(mask);
1587                                 break;
1588                         case MPOL_LOCAL:
1589                                 numa_set_localalloc();
1590                                 break;
1591                         case MPOL_PREFERRED:
1592                                 numa_set_preferred(o->numa_mem_prefer_node);
1593                                 break;
1594                         case MPOL_DEFAULT:
1595                         default:
1596                                 break;
1597                         }
1598
1599                         if (mask)
1600                                 numa_free_nodemask(mask);
1601
1602                 }
1603         }
1604 #endif
1605
1606         if (fio_pin_memory(td))
1607                 goto err;
1608
1609         /*
1610          * May alter parameters that init_io_u() will use, so we need to
1611          * do this first.
1612          */
1613         if (init_iolog(td))
1614                 goto err;
1615
1616         if (init_io_u(td))
1617                 goto err;
1618
1619         if (o->verify_async && verify_async_init(td))
1620                 goto err;
1621
1622         if (fio_option_is_set(o, ioprio) ||
1623             fio_option_is_set(o, ioprio_class)) {
1624                 ret = ioprio_set(IOPRIO_WHO_PROCESS, 0, o->ioprio_class, o->ioprio);
1625                 if (ret == -1) {
1626                         td_verror(td, errno, "ioprio_set");
1627                         goto err;
1628                 }
1629         }
1630
1631         if (o->cgroup && cgroup_setup(td, cgroup_list, &cgroup_mnt))
1632                 goto err;
1633
1634         errno = 0;
1635         if (nice(o->nice) == -1 && errno != 0) {
1636                 td_verror(td, errno, "nice");
1637                 goto err;
1638         }
1639
1640         if (o->ioscheduler && switch_ioscheduler(td))
1641                 goto err;
1642
1643         if (!o->create_serialize && setup_files(td))
1644                 goto err;
1645
1646         if (td_io_init(td))
1647                 goto err;
1648
1649         if (init_random_map(td))
1650                 goto err;
1651
1652         if (o->exec_prerun && exec_string(o, o->exec_prerun, (const char *)"prerun"))
1653                 goto err;
1654
1655         if (o->pre_read) {
1656                 if (pre_read_files(td) < 0)
1657                         goto err;
1658         }
1659
1660         fio_verify_init(td);
1661
1662         if (rate_submit_init(td, sk_out))
1663                 goto err;
1664
1665         set_epoch_time(td, o->log_unix_epoch);
1666         fio_getrusage(&td->ru_start);
1667         memcpy(&td->bw_sample_time, &td->epoch, sizeof(td->epoch));
1668         memcpy(&td->iops_sample_time, &td->epoch, sizeof(td->epoch));
1669         memcpy(&td->ss.prev_time, &td->epoch, sizeof(td->epoch));
1670
1671         if (o->ratemin[DDIR_READ] || o->ratemin[DDIR_WRITE] ||
1672                         o->ratemin[DDIR_TRIM]) {
1673                 memcpy(&td->lastrate[DDIR_READ], &td->bw_sample_time,
1674                                         sizeof(td->bw_sample_time));
1675                 memcpy(&td->lastrate[DDIR_WRITE], &td->bw_sample_time,
1676                                         sizeof(td->bw_sample_time));
1677                 memcpy(&td->lastrate[DDIR_TRIM], &td->bw_sample_time,
1678                                         sizeof(td->bw_sample_time));
1679         }
1680
1681         clear_state = 0;
1682         while (keep_running(td)) {
1683                 uint64_t verify_bytes;
1684
1685                 fio_gettime(&td->start, NULL);
1686                 memcpy(&td->tv_cache, &td->start, sizeof(td->start));
1687
1688                 if (clear_state) {
1689                         clear_io_state(td, 0);
1690
1691                         if (o->unlink_each_loop && unlink_all_files(td))
1692                                 break;
1693                 }
1694
1695                 prune_io_piece_log(td);
1696
1697                 if (td->o.verify_only && td_write(td))
1698                         verify_bytes = do_dry_run(td);
1699                 else {
1700                         uint64_t bytes_done[DDIR_RWDIR_CNT];
1701
1702                         do_io(td, bytes_done);
1703
1704                         if (!ddir_rw_sum(bytes_done)) {
1705                                 fio_mark_td_terminate(td);
1706                                 verify_bytes = 0;
1707                         } else {
1708                                 verify_bytes = bytes_done[DDIR_WRITE] +
1709                                                 bytes_done[DDIR_TRIM];
1710                         }
1711                 }
1712
1713                 /*
1714                  * If we took too long to shut down, the main thread could
1715                  * already consider us reaped/exited. If that happens, break
1716                  * out and clean up.
1717                  */
1718                 if (td->runstate >= TD_EXITED)
1719                         break;
1720
1721                 clear_state = 1;
1722
1723                 /*
1724                  * Make sure we've successfully updated the rusage stats
1725                  * before waiting on the stat mutex. Otherwise we could have
1726                  * the stat thread holding stat mutex and waiting for
1727                  * the rusage_sem, which would never get upped because
1728                  * this thread is waiting for the stat mutex.
1729                  */
1730                 deadlock_loop_cnt = 0;
1731                 do {
1732                         check_update_rusage(td);
1733                         if (!fio_mutex_down_trylock(stat_mutex))
1734                                 break;
1735                         usleep(1000);
1736                         if (deadlock_loop_cnt++ > 5000) {
1737                                 log_err("fio seems to be stuck grabbing stat_mutex, forcibly exiting\n");
1738                                 td->error = EDEADLK;
1739                                 goto err;
1740                         }
1741                 } while (1);
1742
1743                 if (td_read(td) && td->io_bytes[DDIR_READ])
1744                         update_runtime(td, elapsed_us, DDIR_READ);
1745                 if (td_write(td) && td->io_bytes[DDIR_WRITE])
1746                         update_runtime(td, elapsed_us, DDIR_WRITE);
1747                 if (td_trim(td) && td->io_bytes[DDIR_TRIM])
1748                         update_runtime(td, elapsed_us, DDIR_TRIM);
1749                 fio_gettime(&td->start, NULL);
1750                 fio_mutex_up(stat_mutex);
1751
1752                 if (td->error || td->terminate)
1753                         break;
1754
1755                 if (!o->do_verify ||
1756                     o->verify == VERIFY_NONE ||
1757                     td_ioengine_flagged(td, FIO_UNIDIR))
1758                         continue;
1759
1760                 clear_io_state(td, 0);
1761
1762                 fio_gettime(&td->start, NULL);
1763
1764                 do_verify(td, verify_bytes);
1765
1766                 /*
1767                  * See comment further up for why this is done here.
1768                  */
1769                 check_update_rusage(td);
1770
1771                 fio_mutex_down(stat_mutex);
1772                 update_runtime(td, elapsed_us, DDIR_READ);
1773                 fio_gettime(&td->start, NULL);
1774                 fio_mutex_up(stat_mutex);
1775
1776                 if (td->error || td->terminate)
1777                         break;
1778         }
1779
1780         td_set_runstate(td, TD_FINISHING);
1781
1782         update_rusage_stat(td);
1783         td->ts.total_run_time = mtime_since_now(&td->epoch);
1784         td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ];
1785         td->ts.io_bytes[DDIR_WRITE] = td->io_bytes[DDIR_WRITE];
1786         td->ts.io_bytes[DDIR_TRIM] = td->io_bytes[DDIR_TRIM];
1787
1788         if (td->o.verify_state_save && !(td->flags & TD_F_VSTATE_SAVED) &&
1789             (td->o.verify != VERIFY_NONE && td_write(td)))
1790                 verify_save_state(td->thread_number);
1791
1792         fio_unpin_memory(td);
1793
1794         td_writeout_logs(td, true);
1795
1796         iolog_compress_exit(td);
1797         rate_submit_exit(td);
1798
1799         if (o->exec_postrun)
1800                 exec_string(o, o->exec_postrun, (const char *)"postrun");
1801
1802         if (exitall_on_terminate || (o->exitall_error && td->error))
1803                 fio_terminate_threads(td->groupid);
1804
1805 err:
1806         if (td->error)
1807                 log_info("fio: pid=%d, err=%d/%s\n", (int) td->pid, td->error,
1808                                                         td->verror);
1809
1810         if (o->verify_async)
1811                 verify_async_exit(td);
1812
1813         close_and_free_files(td);
1814         cleanup_io_u(td);
1815         close_ioengine(td);
1816         cgroup_shutdown(td, &cgroup_mnt);
1817         verify_free_state(td);
1818
1819         if (td->zone_state_index) {
1820                 int i;
1821
1822                 for (i = 0; i < DDIR_RWDIR_CNT; i++)
1823                         free(td->zone_state_index[i]);
1824                 free(td->zone_state_index);
1825                 td->zone_state_index = NULL;
1826         }
1827
1828         if (fio_option_is_set(o, cpumask)) {
1829                 ret = fio_cpuset_exit(&o->cpumask);
1830                 if (ret)
1831                         td_verror(td, ret, "fio_cpuset_exit");
1832         }
1833
1834         /*
1835          * do this very late, it will log file closing as well
1836          */
1837         if (o->write_iolog_file)
1838                 write_iolog_close(td);
1839
1840         td_set_runstate(td, TD_EXITED);
1841
1842         /*
1843          * Do this last after setting our runstate to exited, so we
1844          * know that the stat thread is signaled.
1845          */
1846         check_update_rusage(td);
1847
1848         sk_out_drop();
1849         return (void *) (uintptr_t) td->error;
1850 }
1851
1852 static void dump_td_info(struct thread_data *td)
1853 {
1854         log_err("fio: job '%s' (state=%d) hasn't exited in %lu seconds, it "
1855                 "appears to be stuck. Doing forceful exit of this job.\n",
1856                         td->o.name, td->runstate,
1857                         (unsigned long) time_since_now(&td->terminate_time));
1858 }
1859
1860 /*
1861  * Run over the job map and reap the threads that have exited, if any.
1862  */
1863 static void reap_threads(unsigned int *nr_running, uint64_t *t_rate,
1864                          uint64_t *m_rate)
1865 {
1866         struct thread_data *td;
1867         unsigned int cputhreads, realthreads, pending;
1868         int i, status, ret;
1869
1870         /*
1871          * reap exited threads (TD_EXITED -> TD_REAPED)
1872          */
1873         realthreads = pending = cputhreads = 0;
1874         for_each_td(td, i) {
1875                 int flags = 0;
1876
1877                 /*
1878                  * ->io_ops is NULL for a thread that has closed its
1879                  * io engine
1880                  */
1881                 if (td->io_ops && !strcmp(td->io_ops->name, "cpuio"))
1882                         cputhreads++;
1883                 else
1884                         realthreads++;
1885
1886                 if (!td->pid) {
1887                         pending++;
1888                         continue;
1889                 }
1890                 if (td->runstate == TD_REAPED)
1891                         continue;
1892                 if (td->o.use_thread) {
1893                         if (td->runstate == TD_EXITED) {
1894                                 td_set_runstate(td, TD_REAPED);
1895                                 goto reaped;
1896                         }
1897                         continue;
1898                 }
1899
1900                 flags = WNOHANG;
1901                 if (td->runstate == TD_EXITED)
1902                         flags = 0;
1903
1904                 /*
1905                  * check if someone quit or got killed in an unusual way
1906                  */
1907                 ret = waitpid(td->pid, &status, flags);
1908                 if (ret < 0) {
1909                         if (errno == ECHILD) {
1910                                 log_err("fio: pid=%d disappeared %d\n",
1911                                                 (int) td->pid, td->runstate);
1912                                 td->sig = ECHILD;
1913                                 td_set_runstate(td, TD_REAPED);
1914                                 goto reaped;
1915                         }
1916                         perror("waitpid");
1917                 } else if (ret == td->pid) {
1918                         if (WIFSIGNALED(status)) {
1919                                 int sig = WTERMSIG(status);
1920
1921                                 if (sig != SIGTERM && sig != SIGUSR2)
1922                                         log_err("fio: pid=%d, got signal=%d\n",
1923                                                         (int) td->pid, sig);
1924                                 td->sig = sig;
1925                                 td_set_runstate(td, TD_REAPED);
1926                                 goto reaped;
1927                         }
1928                         if (WIFEXITED(status)) {
1929                                 if (WEXITSTATUS(status) && !td->error)
1930                                         td->error = WEXITSTATUS(status);
1931
1932                                 td_set_runstate(td, TD_REAPED);
1933                                 goto reaped;
1934                         }
1935                 }
1936
1937                 /*
1938                  * If the job is stuck, do a forceful timeout of it and
1939                  * move on.
1940                  */
1941                 if (td->terminate &&
1942                     td->runstate < TD_FSYNCING &&
1943                     time_since_now(&td->terminate_time) >= FIO_REAP_TIMEOUT) {
1944                         dump_td_info(td);
1945                         td_set_runstate(td, TD_REAPED);
1946                         goto reaped;
1947                 }
1948
1949                 /*
1950                  * thread is not dead, continue
1951                  */
1952                 pending++;
1953                 continue;
1954 reaped:
1955                 (*nr_running)--;
1956                 (*m_rate) -= ddir_rw_sum(td->o.ratemin);
1957                 (*t_rate) -= ddir_rw_sum(td->o.rate);
1958                 if (!td->pid)
1959                         pending--;
1960
1961                 if (td->error)
1962                         exit_value++;
1963
1964                 done_secs += mtime_since_now(&td->epoch) / 1000;
1965                 profile_td_exit(td);
1966         }
1967
1968         if (*nr_running == cputhreads && !pending && realthreads)
1969                 fio_terminate_threads(TERMINATE_ALL);
1970 }
1971
1972 static bool __check_trigger_file(void)
1973 {
1974         struct stat sb;
1975
1976         if (!trigger_file)
1977                 return false;
1978
1979         if (stat(trigger_file, &sb))
1980                 return false;
1981
1982         if (unlink(trigger_file) < 0)
1983                 log_err("fio: failed to unlink %s: %s\n", trigger_file,
1984                                                         strerror(errno));
1985
1986         return true;
1987 }
1988
1989 static bool trigger_timedout(void)
1990 {
1991         if (trigger_timeout)
1992                 return time_since_genesis() >= trigger_timeout;
1993
1994         return false;
1995 }
1996
1997 void exec_trigger(const char *cmd)
1998 {
1999         int ret;
2000
2001         if (!cmd)
2002                 return;
2003
2004         ret = system(cmd);
2005         if (ret == -1)
2006                 log_err("fio: failed executing %s trigger\n", cmd);
2007 }
2008
2009 void check_trigger_file(void)
2010 {
2011         if (__check_trigger_file() || trigger_timedout()) {
2012                 if (nr_clients)
2013                         fio_clients_send_trigger(trigger_remote_cmd);
2014                 else {
2015                         verify_save_state(IO_LIST_ALL);
2016                         fio_terminate_threads(TERMINATE_ALL);
2017                         exec_trigger(trigger_cmd);
2018                 }
2019         }
2020 }
2021
2022 static int fio_verify_load_state(struct thread_data *td)
2023 {
2024         int ret;
2025
2026         if (!td->o.verify_state)
2027                 return 0;
2028
2029         if (is_backend) {
2030                 void *data;
2031
2032                 ret = fio_server_get_verify_state(td->o.name,
2033                                         td->thread_number - 1, &data);
2034                 if (!ret)
2035                         verify_assign_state(td, data);
2036         } else
2037                 ret = verify_load_state(td, "local");
2038
2039         return ret;
2040 }
2041
2042 static void do_usleep(unsigned int usecs)
2043 {
2044         check_for_running_stats();
2045         check_trigger_file();
2046         usleep(usecs);
2047 }
2048
2049 static bool check_mount_writes(struct thread_data *td)
2050 {
2051         struct fio_file *f;
2052         unsigned int i;
2053
2054         if (!td_write(td) || td->o.allow_mounted_write)
2055                 return false;
2056
2057         /*
2058          * If FIO_HAVE_CHARDEV_SIZE is defined, it's likely that chrdevs
2059          * are mkfs'd and mounted.
2060          */
2061         for_each_file(td, f, i) {
2062 #ifdef FIO_HAVE_CHARDEV_SIZE
2063                 if (f->filetype != FIO_TYPE_BLOCK && f->filetype != FIO_TYPE_CHAR)
2064 #else
2065                 if (f->filetype != FIO_TYPE_BLOCK)
2066 #endif
2067                         continue;
2068                 if (device_is_mounted(f->file_name))
2069                         goto mounted;
2070         }
2071
2072         return false;
2073 mounted:
2074         log_err("fio: %s appears mounted, and 'allow_mounted_write' isn't set. Aborting.\n", f->file_name);
2075         return true;
2076 }
2077
2078 static bool waitee_running(struct thread_data *me)
2079 {
2080         const char *waitee = me->o.wait_for;
2081         const char *self = me->o.name;
2082         struct thread_data *td;
2083         int i;
2084
2085         if (!waitee)
2086                 return false;
2087
2088         for_each_td(td, i) {
2089                 if (!strcmp(td->o.name, self) || strcmp(td->o.name, waitee))
2090                         continue;
2091
2092                 if (td->runstate < TD_EXITED) {
2093                         dprint(FD_PROCESS, "%s fenced by %s(%s)\n",
2094                                         self, td->o.name,
2095                                         runstate_to_name(td->runstate));
2096                         return true;
2097                 }
2098         }
2099
2100         dprint(FD_PROCESS, "%s: %s completed, can run\n", self, waitee);
2101         return false;
2102 }
2103
2104 /*
2105  * Main function for kicking off and reaping jobs, as needed.
2106  */
2107 static void run_threads(struct sk_out *sk_out)
2108 {
2109         struct thread_data *td;
2110         unsigned int i, todo, nr_running, nr_started;
2111         uint64_t m_rate, t_rate;
2112         uint64_t spent;
2113
2114         if (fio_gtod_offload && fio_start_gtod_thread())
2115                 return;
2116
2117         fio_idle_prof_init();
2118
2119         set_sig_handlers();
2120
2121         nr_thread = nr_process = 0;
2122         for_each_td(td, i) {
2123                 if (check_mount_writes(td))
2124                         return;
2125                 if (td->o.use_thread)
2126                         nr_thread++;
2127                 else
2128                         nr_process++;
2129         }
2130
2131         if (output_format & FIO_OUTPUT_NORMAL) {
2132                 log_info("Starting ");
2133                 if (nr_thread)
2134                         log_info("%d thread%s", nr_thread,
2135                                                 nr_thread > 1 ? "s" : "");
2136                 if (nr_process) {
2137                         if (nr_thread)
2138                                 log_info(" and ");
2139                         log_info("%d process%s", nr_process,
2140                                                 nr_process > 1 ? "es" : "");
2141                 }
2142                 log_info("\n");
2143                 log_info_flush();
2144         }
2145
2146         todo = thread_number;
2147         nr_running = 0;
2148         nr_started = 0;
2149         m_rate = t_rate = 0;
2150
2151         for_each_td(td, i) {
2152                 print_status_init(td->thread_number - 1);
2153
2154                 if (!td->o.create_serialize)
2155                         continue;
2156
2157                 if (fio_verify_load_state(td))
2158                         goto reap;
2159
2160                 /*
2161                  * do file setup here so it happens sequentially,
2162                  * we don't want X number of threads getting their
2163                  * client data interspersed on disk
2164                  */
2165                 if (setup_files(td)) {
2166 reap:
2167                         exit_value++;
2168                         if (td->error)
2169                                 log_err("fio: pid=%d, err=%d/%s\n",
2170                                         (int) td->pid, td->error, td->verror);
2171                         td_set_runstate(td, TD_REAPED);
2172                         todo--;
2173                 } else {
2174                         struct fio_file *f;
2175                         unsigned int j;
2176
2177                         /*
2178                          * for sharing to work, each job must always open
2179                          * its own files. so close them, if we opened them
2180                          * for creation
2181                          */
2182                         for_each_file(td, f, j) {
2183                                 if (fio_file_open(f))
2184                                         td_io_close_file(td, f);
2185                         }
2186                 }
2187         }
2188
2189         /* start idle threads before io threads start to run */
2190         fio_idle_prof_start();
2191
2192         set_genesis_time();
2193
2194         while (todo) {
2195                 struct thread_data *map[REAL_MAX_JOBS];
2196                 struct timeval this_start;
2197                 int this_jobs = 0, left;
2198                 struct fork_data *fd;
2199
2200                 /*
2201                  * create threads (TD_NOT_CREATED -> TD_CREATED)
2202                  */
2203                 for_each_td(td, i) {
2204                         if (td->runstate != TD_NOT_CREATED)
2205                                 continue;
2206
2207                         /*
2208                          * never got a chance to start, killed by other
2209                          * thread for some reason
2210                          */
2211                         if (td->terminate) {
2212                                 todo--;
2213                                 continue;
2214                         }
2215
2216                         if (td->o.start_delay) {
2217                                 spent = utime_since_genesis();
2218
2219                                 if (td->o.start_delay > spent)
2220                                         continue;
2221                         }
2222
2223                         if (td->o.stonewall && (nr_started || nr_running)) {
2224                                 dprint(FD_PROCESS, "%s: stonewall wait\n",
2225                                                         td->o.name);
2226                                 break;
2227                         }
2228
2229                         if (waitee_running(td)) {
2230                                 dprint(FD_PROCESS, "%s: waiting for %s\n",
2231                                                 td->o.name, td->o.wait_for);
2232                                 continue;
2233                         }
2234
2235                         init_disk_util(td);
2236
2237                         td->rusage_sem = fio_mutex_init(FIO_MUTEX_LOCKED);
2238                         td->update_rusage = 0;
2239
2240                         /*
2241                          * Set state to created. Thread will transition
2242                          * to TD_INITIALIZED when it's done setting up.
2243                          */
2244                         td_set_runstate(td, TD_CREATED);
2245                         map[this_jobs++] = td;
2246                         nr_started++;
2247
2248                         fd = calloc(1, sizeof(*fd));
2249                         fd->td = td;
2250                         fd->sk_out = sk_out;
2251
2252                         if (td->o.use_thread) {
2253                                 int ret;
2254
2255                                 dprint(FD_PROCESS, "will pthread_create\n");
2256                                 ret = pthread_create(&td->thread, NULL,
2257                                                         thread_main, fd);
2258                                 if (ret) {
2259                                         log_err("pthread_create: %s\n",
2260                                                         strerror(ret));
2261                                         free(fd);
2262                                         nr_started--;
2263                                         break;
2264                                 }
2265                                 ret = pthread_detach(td->thread);
2266                                 if (ret)
2267                                         log_err("pthread_detach: %s",
2268                                                         strerror(ret));
2269                         } else {
2270                                 pid_t pid;
2271                                 dprint(FD_PROCESS, "will fork\n");
2272                                 pid = fork();
2273                                 if (!pid) {
2274                                         int ret;
2275
2276                                         ret = (int)(uintptr_t)thread_main(fd);
2277                                         _exit(ret);
2278                                 } else if (i == fio_debug_jobno)
2279                                         *fio_debug_jobp = pid;
2280                         }
2281                         dprint(FD_MUTEX, "wait on startup_mutex\n");
2282                         if (fio_mutex_down_timeout(startup_mutex, 10000)) {
2283                                 log_err("fio: job startup hung? exiting.\n");
2284                                 fio_terminate_threads(TERMINATE_ALL);
2285                                 fio_abort = 1;
2286                                 nr_started--;
2287                                 break;
2288                         }
2289                         dprint(FD_MUTEX, "done waiting on startup_mutex\n");
2290                 }
2291
2292                 /*
2293                  * Wait for the started threads to transition to
2294                  * TD_INITIALIZED.
2295                  */
2296                 fio_gettime(&this_start, NULL);
2297                 left = this_jobs;
2298                 while (left && !fio_abort) {
2299                         if (mtime_since_now(&this_start) > JOB_START_TIMEOUT)
2300                                 break;
2301
2302                         do_usleep(100000);
2303
2304                         for (i = 0; i < this_jobs; i++) {
2305                                 td = map[i];
2306                                 if (!td)
2307                                         continue;
2308                                 if (td->runstate == TD_INITIALIZED) {
2309                                         map[i] = NULL;
2310                                         left--;
2311                                 } else if (td->runstate >= TD_EXITED) {
2312                                         map[i] = NULL;
2313                                         left--;
2314                                         todo--;
2315                                         nr_running++; /* work-around... */
2316                                 }
2317                         }
2318                 }
2319
2320                 if (left) {
2321                         log_err("fio: %d job%s failed to start\n", left,
2322                                         left > 1 ? "s" : "");
2323                         for (i = 0; i < this_jobs; i++) {
2324                                 td = map[i];
2325                                 if (!td)
2326                                         continue;
2327                                 kill(td->pid, SIGTERM);
2328                         }
2329                         break;
2330                 }
2331
2332                 /*
2333                  * start created threads (TD_INITIALIZED -> TD_RUNNING).
2334                  */
2335                 for_each_td(td, i) {
2336                         if (td->runstate != TD_INITIALIZED)
2337                                 continue;
2338
2339                         if (in_ramp_time(td))
2340                                 td_set_runstate(td, TD_RAMP);
2341                         else
2342                                 td_set_runstate(td, TD_RUNNING);
2343                         nr_running++;
2344                         nr_started--;
2345                         m_rate += ddir_rw_sum(td->o.ratemin);
2346                         t_rate += ddir_rw_sum(td->o.rate);
2347                         todo--;
2348                         fio_mutex_up(td->mutex);
2349                 }
2350
2351                 reap_threads(&nr_running, &t_rate, &m_rate);
2352
2353                 if (todo)
2354                         do_usleep(100000);
2355         }
2356
2357         while (nr_running) {
2358                 reap_threads(&nr_running, &t_rate, &m_rate);
2359                 do_usleep(10000);
2360         }
2361
2362         fio_idle_prof_stop();
2363
2364         update_io_ticks();
2365 }
2366
2367 static void free_disk_util(void)
2368 {
2369         disk_util_prune_entries();
2370         helper_thread_destroy();
2371 }
2372
2373 int fio_backend(struct sk_out *sk_out)
2374 {
2375         struct thread_data *td;
2376         int i;
2377
2378         if (exec_profile) {
2379                 if (load_profile(exec_profile))
2380                         return 1;
2381                 free(exec_profile);
2382                 exec_profile = NULL;
2383         }
2384         if (!thread_number)
2385                 return 0;
2386
2387         if (write_bw_log) {
2388                 struct log_params p = {
2389                         .log_type = IO_LOG_TYPE_BW,
2390                 };
2391
2392                 setup_log(&agg_io_log[DDIR_READ], &p, "agg-read_bw.log");
2393                 setup_log(&agg_io_log[DDIR_WRITE], &p, "agg-write_bw.log");
2394                 setup_log(&agg_io_log[DDIR_TRIM], &p, "agg-trim_bw.log");
2395         }
2396
2397         startup_mutex = fio_mutex_init(FIO_MUTEX_LOCKED);
2398         if (startup_mutex == NULL)
2399                 return 1;
2400
2401         set_genesis_time();
2402         stat_init();
2403         helper_thread_create(startup_mutex, sk_out);
2404
2405         cgroup_list = smalloc(sizeof(*cgroup_list));
2406         INIT_FLIST_HEAD(cgroup_list);
2407
2408         run_threads(sk_out);
2409
2410         helper_thread_exit();
2411
2412         if (!fio_abort) {
2413                 __show_run_stats();
2414                 if (write_bw_log) {
2415                         for (i = 0; i < DDIR_RWDIR_CNT; i++) {
2416                                 struct io_log *log = agg_io_log[i];
2417
2418                                 flush_log(log, false);
2419                                 free_log(log);
2420                         }
2421                 }
2422         }
2423
2424         for_each_td(td, i) {
2425                 if (td->ss.dur) {
2426                         if (td->ss.iops_data != NULL) {
2427                                 free(td->ss.iops_data);
2428                                 free(td->ss.bw_data);
2429                         }
2430                 }
2431                 fio_options_free(td);
2432                 if (td->rusage_sem) {
2433                         fio_mutex_remove(td->rusage_sem);
2434                         td->rusage_sem = NULL;
2435                 }
2436                 fio_mutex_remove(td->mutex);
2437                 td->mutex = NULL;
2438         }
2439
2440         free_disk_util();
2441         cgroup_kill(cgroup_list);
2442         sfree(cgroup_list);
2443         sfree(cgroup_mnt);
2444
2445         fio_mutex_remove(startup_mutex);
2446         stat_exit();
2447         return exit_value;
2448 }