backend.c revision de54cfd8b8e93d2a32a02961f1587b83f0763aa8
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 39#include "fio.h" 40#ifndef FIO_NO_HAVE_SHM_H 41#include <sys/shm.h> 42#endif 43#include "hash.h" 44#include "smalloc.h" 45#include "verify.h" 46#include "trim.h" 47#include "diskutil.h" 48#include "cgroup.h" 49#include "profile.h" 50#include "lib/rand.h" 51#include "memalign.h" 52#include "server.h" 53#include "lib/getrusage.h" 54#include "idletime.h" 55#include "err.h" 56#include "lib/tp.h" 57 58static pthread_t helper_thread; 59static pthread_mutex_t helper_lock; 60pthread_cond_t helper_cond; 61int helper_do_stat = 0; 62 63static struct fio_mutex *startup_mutex; 64static struct flist_head *cgroup_list; 65static char *cgroup_mnt; 66static int exit_value; 67static volatile int fio_abort; 68static unsigned int nr_process = 0; 69static unsigned int nr_thread = 0; 70 71struct io_log *agg_io_log[DDIR_RWDIR_CNT]; 72 73int groupid = 0; 74unsigned int thread_number = 0; 75unsigned int stat_number = 0; 76int shm_id = 0; 77int temp_stall_ts; 78unsigned long done_secs = 0; 79volatile int helper_exit = 0; 80 81#define PAGE_ALIGN(buf) \ 82 (char *) (((uintptr_t) (buf) + page_mask) & ~page_mask) 83 84#define JOB_START_TIMEOUT (5 * 1000) 85 86static void sig_int(int sig) 87{ 88 if (threads) { 89 if (is_backend) 90 fio_server_got_signal(sig); 91 else { 92 log_info("\nfio: terminating on signal %d\n", sig); 93 log_info_flush(); 94 exit_value = 128; 95 } 96 97 fio_terminate_threads(TERMINATE_ALL); 98 } 99} 100 101static void sig_show_status(int sig) 102{ 103 show_running_run_stats(); 104} 105 106static void set_sig_handlers(void) 107{ 108 struct sigaction act; 109 110 memset(&act, 0, sizeof(act)); 111 act.sa_handler = sig_int; 112 act.sa_flags = SA_RESTART; 113 sigaction(SIGINT, &act, NULL); 114 115 memset(&act, 0, sizeof(act)); 116 act.sa_handler = sig_int; 117 act.sa_flags = SA_RESTART; 118 sigaction(SIGTERM, &act, NULL); 119 120/* Windows uses SIGBREAK as a quit signal from other applications */ 121#ifdef WIN32 122 memset(&act, 0, sizeof(act)); 123 act.sa_handler = sig_int; 124 act.sa_flags = SA_RESTART; 125 sigaction(SIGBREAK, &act, NULL); 126#endif 127 128 memset(&act, 0, sizeof(act)); 129 act.sa_handler = sig_show_status; 130 act.sa_flags = SA_RESTART; 131 sigaction(SIGUSR1, &act, NULL); 132 133 if (is_backend) { 134 memset(&act, 0, sizeof(act)); 135 act.sa_handler = sig_int; 136 act.sa_flags = SA_RESTART; 137 sigaction(SIGPIPE, &act, NULL); 138 } 139} 140 141/* 142 * Check if we are above the minimum rate given. 143 */ 144static int __check_min_rate(struct thread_data *td, struct timeval *now, 145 enum fio_ddir ddir) 146{ 147 unsigned long long bytes = 0; 148 unsigned long iops = 0; 149 unsigned long spent; 150 unsigned long rate; 151 unsigned int ratemin = 0; 152 unsigned int rate_iops = 0; 153 unsigned int rate_iops_min = 0; 154 155 assert(ddir_rw(ddir)); 156 157 if (!td->o.ratemin[ddir] && !td->o.rate_iops_min[ddir]) 158 return 0; 159 160 /* 161 * allow a 2 second settle period in the beginning 162 */ 163 if (mtime_since(&td->start, now) < 2000) 164 return 0; 165 166 iops += td->this_io_blocks[ddir]; 167 bytes += td->this_io_bytes[ddir]; 168 ratemin += td->o.ratemin[ddir]; 169 rate_iops += td->o.rate_iops[ddir]; 170 rate_iops_min += td->o.rate_iops_min[ddir]; 171 172 /* 173 * if rate blocks is set, sample is running 174 */ 175 if (td->rate_bytes[ddir] || td->rate_blocks[ddir]) { 176 spent = mtime_since(&td->lastrate[ddir], now); 177 if (spent < td->o.ratecycle) 178 return 0; 179 180 if (td->o.rate[ddir]) { 181 /* 182 * check bandwidth specified rate 183 */ 184 if (bytes < td->rate_bytes[ddir]) { 185 log_err("%s: min rate %u not met\n", td->o.name, 186 ratemin); 187 return 1; 188 } else { 189 if (spent) 190 rate = ((bytes - td->rate_bytes[ddir]) * 1000) / spent; 191 else 192 rate = 0; 193 194 if (rate < ratemin || 195 bytes < td->rate_bytes[ddir]) { 196 log_err("%s: min rate %u not met, got" 197 " %luKB/sec\n", td->o.name, 198 ratemin, rate); 199 return 1; 200 } 201 } 202 } else { 203 /* 204 * checks iops specified rate 205 */ 206 if (iops < rate_iops) { 207 log_err("%s: min iops rate %u not met\n", 208 td->o.name, rate_iops); 209 return 1; 210 } else { 211 if (spent) 212 rate = ((iops - td->rate_blocks[ddir]) * 1000) / spent; 213 else 214 rate = 0; 215 216 if (rate < rate_iops_min || 217 iops < td->rate_blocks[ddir]) { 218 log_err("%s: min iops rate %u not met," 219 " got %lu\n", td->o.name, 220 rate_iops_min, rate); 221 } 222 } 223 } 224 } 225 226 td->rate_bytes[ddir] = bytes; 227 td->rate_blocks[ddir] = iops; 228 memcpy(&td->lastrate[ddir], now, sizeof(*now)); 229 return 0; 230} 231 232static int check_min_rate(struct thread_data *td, struct timeval *now, 233 uint64_t *bytes_done) 234{ 235 int ret = 0; 236 237 if (bytes_done[DDIR_READ]) 238 ret |= __check_min_rate(td, now, DDIR_READ); 239 if (bytes_done[DDIR_WRITE]) 240 ret |= __check_min_rate(td, now, DDIR_WRITE); 241 if (bytes_done[DDIR_TRIM]) 242 ret |= __check_min_rate(td, now, DDIR_TRIM); 243 244 return ret; 245} 246 247/* 248 * When job exits, we can cancel the in-flight IO if we are using async 249 * io. Attempt to do so. 250 */ 251static void cleanup_pending_aio(struct thread_data *td) 252{ 253 int r; 254 255 /* 256 * get immediately available events, if any 257 */ 258 r = io_u_queued_complete(td, 0, NULL); 259 if (r < 0) 260 return; 261 262 /* 263 * now cancel remaining active events 264 */ 265 if (td->io_ops->cancel) { 266 struct io_u *io_u; 267 int i; 268 269 io_u_qiter(&td->io_u_all, io_u, i) { 270 if (io_u->flags & IO_U_F_FLIGHT) { 271 r = td->io_ops->cancel(td, io_u); 272 if (!r) 273 put_io_u(td, io_u); 274 } 275 } 276 } 277 278 if (td->cur_depth) 279 r = io_u_queued_complete(td, td->cur_depth, NULL); 280} 281 282/* 283 * Helper to handle the final sync of a file. Works just like the normal 284 * io path, just does everything sync. 285 */ 286static int fio_io_sync(struct thread_data *td, struct fio_file *f) 287{ 288 struct io_u *io_u = __get_io_u(td); 289 int ret; 290 291 if (!io_u) 292 return 1; 293 294 io_u->ddir = DDIR_SYNC; 295 io_u->file = f; 296 297 if (td_io_prep(td, io_u)) { 298 put_io_u(td, io_u); 299 return 1; 300 } 301 302requeue: 303 ret = td_io_queue(td, io_u); 304 if (ret < 0) { 305 td_verror(td, io_u->error, "td_io_queue"); 306 put_io_u(td, io_u); 307 return 1; 308 } else if (ret == FIO_Q_QUEUED) { 309 if (io_u_queued_complete(td, 1, NULL) < 0) 310 return 1; 311 } else if (ret == FIO_Q_COMPLETED) { 312 if (io_u->error) { 313 td_verror(td, io_u->error, "td_io_queue"); 314 return 1; 315 } 316 317 if (io_u_sync_complete(td, io_u, NULL) < 0) 318 return 1; 319 } else if (ret == FIO_Q_BUSY) { 320 if (td_io_commit(td)) 321 return 1; 322 goto requeue; 323 } 324 325 return 0; 326} 327 328static int fio_file_fsync(struct thread_data *td, struct fio_file *f) 329{ 330 int ret; 331 332 if (fio_file_open(f)) 333 return fio_io_sync(td, f); 334 335 if (td_io_open_file(td, f)) 336 return 1; 337 338 ret = fio_io_sync(td, f); 339 td_io_close_file(td, f); 340 return ret; 341} 342 343static inline void __update_tv_cache(struct thread_data *td) 344{ 345 fio_gettime(&td->tv_cache, NULL); 346} 347 348static inline void update_tv_cache(struct thread_data *td) 349{ 350 if ((++td->tv_cache_nr & td->tv_cache_mask) == td->tv_cache_mask) 351 __update_tv_cache(td); 352} 353 354static inline int runtime_exceeded(struct thread_data *td, struct timeval *t) 355{ 356 if (in_ramp_time(td)) 357 return 0; 358 if (!td->o.timeout) 359 return 0; 360 if (utime_since(&td->epoch, t) >= td->o.timeout) 361 return 1; 362 363 return 0; 364} 365 366static int break_on_this_error(struct thread_data *td, enum fio_ddir ddir, 367 int *retptr) 368{ 369 int ret = *retptr; 370 371 if (ret < 0 || td->error) { 372 int err = td->error; 373 enum error_type_bit eb; 374 375 if (ret < 0) 376 err = -ret; 377 378 eb = td_error_type(ddir, err); 379 if (!(td->o.continue_on_error & (1 << eb))) 380 return 1; 381 382 if (td_non_fatal_error(td, eb, err)) { 383 /* 384 * Continue with the I/Os in case of 385 * a non fatal error. 386 */ 387 update_error_count(td, err); 388 td_clear_error(td); 389 *retptr = 0; 390 return 0; 391 } else if (td->o.fill_device && err == ENOSPC) { 392 /* 393 * We expect to hit this error if 394 * fill_device option is set. 395 */ 396 td_clear_error(td); 397 fio_mark_td_terminate(td); 398 return 1; 399 } else { 400 /* 401 * Stop the I/O in case of a fatal 402 * error. 403 */ 404 update_error_count(td, err); 405 return 1; 406 } 407 } 408 409 return 0; 410} 411 412static void check_update_rusage(struct thread_data *td) 413{ 414 if (td->update_rusage) { 415 td->update_rusage = 0; 416 update_rusage_stat(td); 417 fio_mutex_up(td->rusage_sem); 418 } 419} 420 421/* 422 * The main verify engine. Runs over the writes we previously submitted, 423 * reads the blocks back in, and checks the crc/md5 of the data. 424 */ 425static void do_verify(struct thread_data *td, uint64_t verify_bytes) 426{ 427 uint64_t bytes_done[DDIR_RWDIR_CNT] = { 0, 0, 0 }; 428 struct fio_file *f; 429 struct io_u *io_u; 430 int ret, min_events; 431 unsigned int i; 432 433 dprint(FD_VERIFY, "starting loop\n"); 434 435 /* 436 * sync io first and invalidate cache, to make sure we really 437 * read from disk. 438 */ 439 for_each_file(td, f, i) { 440 if (!fio_file_open(f)) 441 continue; 442 if (fio_io_sync(td, f)) 443 break; 444 if (file_invalidate_cache(td, f)) 445 break; 446 } 447 448 check_update_rusage(td); 449 450 if (td->error) 451 return; 452 453 td_set_runstate(td, TD_VERIFYING); 454 455 io_u = NULL; 456 while (!td->terminate) { 457 enum fio_ddir ddir; 458 int ret2, full; 459 460 update_tv_cache(td); 461 check_update_rusage(td); 462 463 if (runtime_exceeded(td, &td->tv_cache)) { 464 __update_tv_cache(td); 465 if (runtime_exceeded(td, &td->tv_cache)) { 466 fio_mark_td_terminate(td); 467 break; 468 } 469 } 470 471 if (flow_threshold_exceeded(td)) 472 continue; 473 474 if (!td->o.experimental_verify) { 475 io_u = __get_io_u(td); 476 if (!io_u) 477 break; 478 479 if (get_next_verify(td, io_u)) { 480 put_io_u(td, io_u); 481 break; 482 } 483 484 if (td_io_prep(td, io_u)) { 485 put_io_u(td, io_u); 486 break; 487 } 488 } else { 489 if (ddir_rw_sum(bytes_done) + td->o.rw_min_bs > verify_bytes) 490 break; 491 492 while ((io_u = get_io_u(td)) != NULL) { 493 if (IS_ERR(io_u)) { 494 io_u = NULL; 495 ret = FIO_Q_BUSY; 496 goto reap; 497 } 498 499 /* 500 * We are only interested in the places where 501 * we wrote or trimmed IOs. Turn those into 502 * reads for verification purposes. 503 */ 504 if (io_u->ddir == DDIR_READ) { 505 /* 506 * Pretend we issued it for rwmix 507 * accounting 508 */ 509 td->io_issues[DDIR_READ]++; 510 put_io_u(td, io_u); 511 continue; 512 } else if (io_u->ddir == DDIR_TRIM) { 513 io_u->ddir = DDIR_READ; 514 io_u->flags |= IO_U_F_TRIMMED; 515 break; 516 } else if (io_u->ddir == DDIR_WRITE) { 517 io_u->ddir = DDIR_READ; 518 break; 519 } else { 520 put_io_u(td, io_u); 521 continue; 522 } 523 } 524 525 if (!io_u) 526 break; 527 } 528 529 if (verify_state_should_stop(td, io_u)) { 530 put_io_u(td, io_u); 531 break; 532 } 533 534 if (td->o.verify_async) 535 io_u->end_io = verify_io_u_async; 536 else 537 io_u->end_io = verify_io_u; 538 539 ddir = io_u->ddir; 540 541 ret = td_io_queue(td, io_u); 542 switch (ret) { 543 case FIO_Q_COMPLETED: 544 if (io_u->error) { 545 ret = -io_u->error; 546 clear_io_u(td, io_u); 547 } else if (io_u->resid) { 548 int bytes = io_u->xfer_buflen - io_u->resid; 549 550 /* 551 * zero read, fail 552 */ 553 if (!bytes) { 554 td_verror(td, EIO, "full resid"); 555 put_io_u(td, io_u); 556 break; 557 } 558 559 io_u->xfer_buflen = io_u->resid; 560 io_u->xfer_buf += bytes; 561 io_u->offset += bytes; 562 563 if (ddir_rw(io_u->ddir)) 564 td->ts.short_io_u[io_u->ddir]++; 565 566 f = io_u->file; 567 if (io_u->offset == f->real_file_size) 568 goto sync_done; 569 570 requeue_io_u(td, &io_u); 571 } else { 572sync_done: 573 ret = io_u_sync_complete(td, io_u, bytes_done); 574 if (ret < 0) 575 break; 576 } 577 continue; 578 case FIO_Q_QUEUED: 579 break; 580 case FIO_Q_BUSY: 581 requeue_io_u(td, &io_u); 582 ret2 = td_io_commit(td); 583 if (ret2 < 0) 584 ret = ret2; 585 break; 586 default: 587 assert(ret < 0); 588 td_verror(td, -ret, "td_io_queue"); 589 break; 590 } 591 592 if (break_on_this_error(td, ddir, &ret)) 593 break; 594 595 /* 596 * if we can queue more, do so. but check if there are 597 * completed io_u's first. Note that we can get BUSY even 598 * without IO queued, if the system is resource starved. 599 */ 600reap: 601 full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth); 602 if (full || !td->o.iodepth_batch_complete) { 603 min_events = min(td->o.iodepth_batch_complete, 604 td->cur_depth); 605 /* 606 * if the queue is full, we MUST reap at least 1 event 607 */ 608 if (full && !min_events) 609 min_events = 1; 610 611 do { 612 /* 613 * Reap required number of io units, if any, 614 * and do the verification on them through 615 * the callback handler 616 */ 617 if (io_u_queued_complete(td, min_events, bytes_done) < 0) { 618 ret = -1; 619 break; 620 } 621 } while (full && (td->cur_depth > td->o.iodepth_low)); 622 } 623 if (ret < 0) 624 break; 625 } 626 627 check_update_rusage(td); 628 629 if (!td->error) { 630 min_events = td->cur_depth; 631 632 if (min_events) 633 ret = io_u_queued_complete(td, min_events, NULL); 634 } else 635 cleanup_pending_aio(td); 636 637 td_set_runstate(td, TD_RUNNING); 638 639 dprint(FD_VERIFY, "exiting loop\n"); 640} 641 642static unsigned int exceeds_number_ios(struct thread_data *td) 643{ 644 unsigned long long number_ios; 645 646 if (!td->o.number_ios) 647 return 0; 648 649 number_ios = ddir_rw_sum(td->this_io_blocks); 650 number_ios += td->io_u_queued + td->io_u_in_flight; 651 652 return number_ios >= td->o.number_ios; 653} 654 655static int io_bytes_exceeded(struct thread_data *td) 656{ 657 unsigned long long bytes, limit; 658 659 if (td_rw(td)) 660 bytes = td->this_io_bytes[DDIR_READ] + td->this_io_bytes[DDIR_WRITE]; 661 else if (td_write(td)) 662 bytes = td->this_io_bytes[DDIR_WRITE]; 663 else if (td_read(td)) 664 bytes = td->this_io_bytes[DDIR_READ]; 665 else 666 bytes = td->this_io_bytes[DDIR_TRIM]; 667 668 if (td->o.io_limit) 669 limit = td->o.io_limit; 670 else 671 limit = td->o.size; 672 673 return bytes >= limit || exceeds_number_ios(td); 674} 675 676/* 677 * Main IO worker function. It retrieves io_u's to process and queues 678 * and reaps them, checking for rate and errors along the way. 679 * 680 * Returns number of bytes written and trimmed. 681 */ 682static uint64_t do_io(struct thread_data *td) 683{ 684 uint64_t bytes_done[DDIR_RWDIR_CNT] = { 0, 0, 0 }; 685 unsigned int i; 686 int ret = 0; 687 uint64_t total_bytes, bytes_issued = 0; 688 689 if (in_ramp_time(td)) 690 td_set_runstate(td, TD_RAMP); 691 else 692 td_set_runstate(td, TD_RUNNING); 693 694 lat_target_init(td); 695 696 /* 697 * If verify_backlog is enabled, we'll run the verify in this 698 * handler as well. For that case, we may need up to twice the 699 * amount of bytes. 700 */ 701 total_bytes = td->o.size; 702 if (td->o.verify != VERIFY_NONE && 703 (td_write(td) && td->o.verify_backlog)) 704 total_bytes += td->o.size; 705 706 while ((td->o.read_iolog_file && !flist_empty(&td->io_log_list)) || 707 (!flist_empty(&td->trim_list)) || !io_bytes_exceeded(td) || 708 td->o.time_based) { 709 struct timeval comp_time; 710 int min_evts = 0; 711 struct io_u *io_u; 712 int ret2, full; 713 enum fio_ddir ddir; 714 715 check_update_rusage(td); 716 717 if (td->terminate || td->done) 718 break; 719 720 update_tv_cache(td); 721 722 if (runtime_exceeded(td, &td->tv_cache)) { 723 __update_tv_cache(td); 724 if (runtime_exceeded(td, &td->tv_cache)) { 725 fio_mark_td_terminate(td); 726 break; 727 } 728 } 729 730 if (flow_threshold_exceeded(td)) 731 continue; 732 733 if (bytes_issued >= total_bytes) 734 break; 735 736 io_u = get_io_u(td); 737 if (IS_ERR_OR_NULL(io_u)) { 738 int err = PTR_ERR(io_u); 739 740 io_u = NULL; 741 if (err == -EBUSY) { 742 ret = FIO_Q_BUSY; 743 goto reap; 744 } 745 if (td->o.latency_target) 746 goto reap; 747 break; 748 } 749 750 ddir = io_u->ddir; 751 752 /* 753 * Add verification end_io handler if: 754 * - Asked to verify (!td_rw(td)) 755 * - Or the io_u is from our verify list (mixed write/ver) 756 */ 757 if (td->o.verify != VERIFY_NONE && io_u->ddir == DDIR_READ && 758 ((io_u->flags & IO_U_F_VER_LIST) || !td_rw(td))) { 759 760 if (!td->o.verify_pattern_bytes) { 761 io_u->rand_seed = __rand(&td->verify_state); 762 if (sizeof(int) != sizeof(long *)) 763 io_u->rand_seed *= __rand(&td->verify_state); 764 } 765 766 if (verify_state_should_stop(td, io_u)) { 767 put_io_u(td, io_u); 768 break; 769 } 770 771 if (td->o.verify_async) 772 io_u->end_io = verify_io_u_async; 773 else 774 io_u->end_io = verify_io_u; 775 td_set_runstate(td, TD_VERIFYING); 776 } else if (in_ramp_time(td)) 777 td_set_runstate(td, TD_RAMP); 778 else 779 td_set_runstate(td, TD_RUNNING); 780 781 /* 782 * Always log IO before it's issued, so we know the specific 783 * order of it. The logged unit will track when the IO has 784 * completed. 785 */ 786 if (td_write(td) && io_u->ddir == DDIR_WRITE && 787 td->o.do_verify && 788 td->o.verify != VERIFY_NONE && 789 !td->o.experimental_verify) 790 log_io_piece(td, io_u); 791 792 ret = td_io_queue(td, io_u); 793 switch (ret) { 794 case FIO_Q_COMPLETED: 795 if (io_u->error) { 796 ret = -io_u->error; 797 unlog_io_piece(td, io_u); 798 clear_io_u(td, io_u); 799 } else if (io_u->resid) { 800 int bytes = io_u->xfer_buflen - io_u->resid; 801 struct fio_file *f = io_u->file; 802 803 bytes_issued += bytes; 804 805 trim_io_piece(td, io_u); 806 807 /* 808 * zero read, fail 809 */ 810 if (!bytes) { 811 unlog_io_piece(td, io_u); 812 td_verror(td, EIO, "full resid"); 813 put_io_u(td, io_u); 814 break; 815 } 816 817 io_u->xfer_buflen = io_u->resid; 818 io_u->xfer_buf += bytes; 819 io_u->offset += bytes; 820 821 if (ddir_rw(io_u->ddir)) 822 td->ts.short_io_u[io_u->ddir]++; 823 824 if (io_u->offset == f->real_file_size) 825 goto sync_done; 826 827 requeue_io_u(td, &io_u); 828 } else { 829sync_done: 830 if (__should_check_rate(td, DDIR_READ) || 831 __should_check_rate(td, DDIR_WRITE) || 832 __should_check_rate(td, DDIR_TRIM)) 833 fio_gettime(&comp_time, NULL); 834 835 ret = io_u_sync_complete(td, io_u, bytes_done); 836 if (ret < 0) 837 break; 838 bytes_issued += io_u->xfer_buflen; 839 } 840 break; 841 case FIO_Q_QUEUED: 842 /* 843 * if the engine doesn't have a commit hook, 844 * the io_u is really queued. if it does have such 845 * a hook, it has to call io_u_queued() itself. 846 */ 847 if (td->io_ops->commit == NULL) 848 io_u_queued(td, io_u); 849 bytes_issued += io_u->xfer_buflen; 850 break; 851 case FIO_Q_BUSY: 852 unlog_io_piece(td, io_u); 853 requeue_io_u(td, &io_u); 854 ret2 = td_io_commit(td); 855 if (ret2 < 0) 856 ret = ret2; 857 break; 858 default: 859 assert(ret < 0); 860 put_io_u(td, io_u); 861 break; 862 } 863 864 if (break_on_this_error(td, ddir, &ret)) 865 break; 866 867 /* 868 * See if we need to complete some commands. Note that we 869 * can get BUSY even without IO queued, if the system is 870 * resource starved. 871 */ 872reap: 873 full = queue_full(td) || (ret == FIO_Q_BUSY && td->cur_depth); 874 if (full || !td->o.iodepth_batch_complete) { 875 min_evts = min(td->o.iodepth_batch_complete, 876 td->cur_depth); 877 /* 878 * if the queue is full, we MUST reap at least 1 event 879 */ 880 if (full && !min_evts) 881 min_evts = 1; 882 883 if (__should_check_rate(td, DDIR_READ) || 884 __should_check_rate(td, DDIR_WRITE) || 885 __should_check_rate(td, DDIR_TRIM)) 886 fio_gettime(&comp_time, NULL); 887 888 do { 889 ret = io_u_queued_complete(td, min_evts, bytes_done); 890 if (ret < 0) 891 break; 892 893 } while (full && (td->cur_depth > td->o.iodepth_low)); 894 } 895 896 if (ret < 0) 897 break; 898 if (!ddir_rw_sum(bytes_done) && !(td->io_ops->flags & FIO_NOIO)) 899 continue; 900 901 if (!in_ramp_time(td) && should_check_rate(td, bytes_done)) { 902 if (check_min_rate(td, &comp_time, bytes_done)) { 903 if (exitall_on_terminate) 904 fio_terminate_threads(td->groupid); 905 td_verror(td, EIO, "check_min_rate"); 906 break; 907 } 908 } 909 if (!in_ramp_time(td) && td->o.latency_target) 910 lat_target_check(td); 911 912 if (td->o.thinktime) { 913 unsigned long long b; 914 915 b = ddir_rw_sum(td->io_blocks); 916 if (!(b % td->o.thinktime_blocks)) { 917 int left; 918 919 io_u_quiesce(td); 920 921 if (td->o.thinktime_spin) 922 usec_spin(td->o.thinktime_spin); 923 924 left = td->o.thinktime - td->o.thinktime_spin; 925 if (left) 926 usec_sleep(td, left); 927 } 928 } 929 } 930 931 check_update_rusage(td); 932 933 if (td->trim_entries) 934 log_err("fio: %lu trim entries leaked?\n", td->trim_entries); 935 936 if (td->o.fill_device && td->error == ENOSPC) { 937 td->error = 0; 938 fio_mark_td_terminate(td); 939 } 940 if (!td->error) { 941 struct fio_file *f; 942 943 i = td->cur_depth; 944 if (i) { 945 ret = io_u_queued_complete(td, i, bytes_done); 946 if (td->o.fill_device && td->error == ENOSPC) 947 td->error = 0; 948 } 949 950 if (should_fsync(td) && td->o.end_fsync) { 951 td_set_runstate(td, TD_FSYNCING); 952 953 for_each_file(td, f, i) { 954 if (!fio_file_fsync(td, f)) 955 continue; 956 957 log_err("fio: end_fsync failed for file %s\n", 958 f->file_name); 959 } 960 } 961 } else 962 cleanup_pending_aio(td); 963 964 /* 965 * stop job if we failed doing any IO 966 */ 967 if (!ddir_rw_sum(td->this_io_bytes)) 968 td->done = 1; 969 970 return bytes_done[DDIR_WRITE] + bytes_done[DDIR_TRIM]; 971} 972 973static void cleanup_io_u(struct thread_data *td) 974{ 975 struct io_u *io_u; 976 977 while ((io_u = io_u_qpop(&td->io_u_freelist)) != NULL) { 978 979 if (td->io_ops->io_u_free) 980 td->io_ops->io_u_free(td, io_u); 981 982 fio_memfree(io_u, sizeof(*io_u)); 983 } 984 985 free_io_mem(td); 986 987 io_u_rexit(&td->io_u_requeues); 988 io_u_qexit(&td->io_u_freelist); 989 io_u_qexit(&td->io_u_all); 990 991 if (td->last_write_comp) 992 sfree(td->last_write_comp); 993} 994 995static int init_io_u(struct thread_data *td) 996{ 997 struct io_u *io_u; 998 unsigned int max_bs, min_write; 999 int cl_align, i, max_units; 1000 int data_xfer = 1, err; 1001 char *p; 1002 1003 max_units = td->o.iodepth; 1004 max_bs = td_max_bs(td); 1005 min_write = td->o.min_bs[DDIR_WRITE]; 1006 td->orig_buffer_size = (unsigned long long) max_bs 1007 * (unsigned long long) max_units; 1008 1009 if ((td->io_ops->flags & FIO_NOIO) || !(td_read(td) || td_write(td))) 1010 data_xfer = 0; 1011 1012 err = 0; 1013 err += io_u_rinit(&td->io_u_requeues, td->o.iodepth); 1014 err += io_u_qinit(&td->io_u_freelist, td->o.iodepth); 1015 err += io_u_qinit(&td->io_u_all, td->o.iodepth); 1016 1017 if (err) { 1018 log_err("fio: failed setting up IO queues\n"); 1019 return 1; 1020 } 1021 1022 /* 1023 * if we may later need to do address alignment, then add any 1024 * possible adjustment here so that we don't cause a buffer 1025 * overflow later. this adjustment may be too much if we get 1026 * lucky and the allocator gives us an aligned address. 1027 */ 1028 if (td->o.odirect || td->o.mem_align || td->o.oatomic || 1029 (td->io_ops->flags & FIO_RAWIO)) 1030 td->orig_buffer_size += page_mask + td->o.mem_align; 1031 1032 if (td->o.mem_type == MEM_SHMHUGE || td->o.mem_type == MEM_MMAPHUGE) { 1033 unsigned long bs; 1034 1035 bs = td->orig_buffer_size + td->o.hugepage_size - 1; 1036 td->orig_buffer_size = bs & ~(td->o.hugepage_size - 1); 1037 } 1038 1039 if (td->orig_buffer_size != (size_t) td->orig_buffer_size) { 1040 log_err("fio: IO memory too large. Reduce max_bs or iodepth\n"); 1041 return 1; 1042 } 1043 1044 if (data_xfer && allocate_io_mem(td)) 1045 return 1; 1046 1047 if (td->o.odirect || td->o.mem_align || td->o.oatomic || 1048 (td->io_ops->flags & FIO_RAWIO)) 1049 p = PAGE_ALIGN(td->orig_buffer) + td->o.mem_align; 1050 else 1051 p = td->orig_buffer; 1052 1053 cl_align = os_cache_line_size(); 1054 1055 for (i = 0; i < max_units; i++) { 1056 void *ptr; 1057 1058 if (td->terminate) 1059 return 1; 1060 1061 ptr = fio_memalign(cl_align, sizeof(*io_u)); 1062 if (!ptr) { 1063 log_err("fio: unable to allocate aligned memory\n"); 1064 break; 1065 } 1066 1067 io_u = ptr; 1068 memset(io_u, 0, sizeof(*io_u)); 1069 INIT_FLIST_HEAD(&io_u->verify_list); 1070 dprint(FD_MEM, "io_u alloc %p, index %u\n", io_u, i); 1071 1072 if (data_xfer) { 1073 io_u->buf = p; 1074 dprint(FD_MEM, "io_u %p, mem %p\n", io_u, io_u->buf); 1075 1076 if (td_write(td)) 1077 io_u_fill_buffer(td, io_u, min_write, max_bs); 1078 if (td_write(td) && td->o.verify_pattern_bytes) { 1079 /* 1080 * Fill the buffer with the pattern if we are 1081 * going to be doing writes. 1082 */ 1083 fill_verify_pattern(td, io_u->buf, max_bs, io_u, 0, 0); 1084 } 1085 } 1086 1087 io_u->index = i; 1088 io_u->flags = IO_U_F_FREE; 1089 io_u_qpush(&td->io_u_freelist, io_u); 1090 1091 /* 1092 * io_u never leaves this stack, used for iteration of all 1093 * io_u buffers. 1094 */ 1095 io_u_qpush(&td->io_u_all, io_u); 1096 1097 if (td->io_ops->io_u_init) { 1098 int ret = td->io_ops->io_u_init(td, io_u); 1099 1100 if (ret) { 1101 log_err("fio: failed to init engine data: %d\n", ret); 1102 return 1; 1103 } 1104 } 1105 1106 p += max_bs; 1107 } 1108 1109 if (td->o.verify != VERIFY_NONE) { 1110 td->last_write_comp = scalloc(max_units, sizeof(uint64_t)); 1111 if (!td->last_write_comp) { 1112 log_err("fio: failed to alloc write comp data\n"); 1113 return 1; 1114 } 1115 } 1116 1117 return 0; 1118} 1119 1120static int switch_ioscheduler(struct thread_data *td) 1121{ 1122 char tmp[256], tmp2[128]; 1123 FILE *f; 1124 int ret; 1125 1126 if (td->io_ops->flags & FIO_DISKLESSIO) 1127 return 0; 1128 1129 sprintf(tmp, "%s/queue/scheduler", td->sysfs_root); 1130 1131 f = fopen(tmp, "r+"); 1132 if (!f) { 1133 if (errno == ENOENT) { 1134 log_err("fio: os or kernel doesn't support IO scheduler" 1135 " switching\n"); 1136 return 0; 1137 } 1138 td_verror(td, errno, "fopen iosched"); 1139 return 1; 1140 } 1141 1142 /* 1143 * Set io scheduler. 1144 */ 1145 ret = fwrite(td->o.ioscheduler, strlen(td->o.ioscheduler), 1, f); 1146 if (ferror(f) || ret != 1) { 1147 td_verror(td, errno, "fwrite"); 1148 fclose(f); 1149 return 1; 1150 } 1151 1152 rewind(f); 1153 1154 /* 1155 * Read back and check that the selected scheduler is now the default. 1156 */ 1157 ret = fread(tmp, sizeof(tmp), 1, f); 1158 if (ferror(f) || ret < 0) { 1159 td_verror(td, errno, "fread"); 1160 fclose(f); 1161 return 1; 1162 } 1163 tmp[sizeof(tmp) - 1] = '\0'; 1164 1165 1166 sprintf(tmp2, "[%s]", td->o.ioscheduler); 1167 if (!strstr(tmp, tmp2)) { 1168 log_err("fio: io scheduler %s not found\n", td->o.ioscheduler); 1169 td_verror(td, EINVAL, "iosched_switch"); 1170 fclose(f); 1171 return 1; 1172 } 1173 1174 fclose(f); 1175 return 0; 1176} 1177 1178static int keep_running(struct thread_data *td) 1179{ 1180 unsigned long long limit; 1181 1182 if (td->done) 1183 return 0; 1184 if (td->o.time_based) 1185 return 1; 1186 if (td->o.loops) { 1187 td->o.loops--; 1188 return 1; 1189 } 1190 if (exceeds_number_ios(td)) 1191 return 0; 1192 1193 if (td->o.io_limit) 1194 limit = td->o.io_limit; 1195 else 1196 limit = td->o.size; 1197 1198 if (limit != -1ULL && ddir_rw_sum(td->io_bytes) < limit) { 1199 uint64_t diff; 1200 1201 /* 1202 * If the difference is less than the minimum IO size, we 1203 * are done. 1204 */ 1205 diff = limit - ddir_rw_sum(td->io_bytes); 1206 if (diff < td_max_bs(td)) 1207 return 0; 1208 1209 if (fio_files_done(td)) 1210 return 0; 1211 1212 return 1; 1213 } 1214 1215 return 0; 1216} 1217 1218static int exec_string(struct thread_options *o, const char *string, const char *mode) 1219{ 1220 int ret, newlen = strlen(string) + strlen(o->name) + strlen(mode) + 9 + 1; 1221 char *str; 1222 1223 str = malloc(newlen); 1224 sprintf(str, "%s &> %s.%s.txt", string, o->name, mode); 1225 1226 log_info("%s : Saving output of %s in %s.%s.txt\n",o->name, mode, o->name, mode); 1227 ret = system(str); 1228 if (ret == -1) 1229 log_err("fio: exec of cmd <%s> failed\n", str); 1230 1231 free(str); 1232 return ret; 1233} 1234 1235/* 1236 * Dry run to compute correct state of numberio for verification. 1237 */ 1238static uint64_t do_dry_run(struct thread_data *td) 1239{ 1240 uint64_t bytes_done[DDIR_RWDIR_CNT] = { 0, 0, 0 }; 1241 1242 td_set_runstate(td, TD_RUNNING); 1243 1244 while ((td->o.read_iolog_file && !flist_empty(&td->io_log_list)) || 1245 (!flist_empty(&td->trim_list)) || !io_bytes_exceeded(td)) { 1246 struct io_u *io_u; 1247 int ret; 1248 1249 if (td->terminate || td->done) 1250 break; 1251 1252 io_u = get_io_u(td); 1253 if (!io_u) 1254 break; 1255 1256 io_u->flags |= IO_U_F_FLIGHT; 1257 io_u->error = 0; 1258 io_u->resid = 0; 1259 if (ddir_rw(acct_ddir(io_u))) 1260 td->io_issues[acct_ddir(io_u)]++; 1261 if (ddir_rw(io_u->ddir)) { 1262 io_u_mark_depth(td, 1); 1263 td->ts.total_io_u[io_u->ddir]++; 1264 } 1265 1266 if (td_write(td) && io_u->ddir == DDIR_WRITE && 1267 td->o.do_verify && 1268 td->o.verify != VERIFY_NONE && 1269 !td->o.experimental_verify) 1270 log_io_piece(td, io_u); 1271 1272 ret = io_u_sync_complete(td, io_u, bytes_done); 1273 (void) ret; 1274 } 1275 1276 return bytes_done[DDIR_WRITE] + bytes_done[DDIR_TRIM]; 1277} 1278 1279/* 1280 * Entry point for the thread based jobs. The process based jobs end up 1281 * here as well, after a little setup. 1282 */ 1283static void *thread_main(void *data) 1284{ 1285 unsigned long long elapsed; 1286 struct thread_data *td = data; 1287 struct thread_options *o = &td->o; 1288 pthread_condattr_t attr; 1289 int clear_state; 1290 int ret; 1291 1292 if (!o->use_thread) { 1293 setsid(); 1294 td->pid = getpid(); 1295 } else 1296 td->pid = gettid(); 1297 1298 fio_local_clock_init(o->use_thread); 1299 1300 dprint(FD_PROCESS, "jobs pid=%d started\n", (int) td->pid); 1301 1302 if (is_backend) 1303 fio_server_send_start(td); 1304 1305 INIT_FLIST_HEAD(&td->io_log_list); 1306 INIT_FLIST_HEAD(&td->io_hist_list); 1307 INIT_FLIST_HEAD(&td->verify_list); 1308 INIT_FLIST_HEAD(&td->trim_list); 1309 INIT_FLIST_HEAD(&td->next_rand_list); 1310 pthread_mutex_init(&td->io_u_lock, NULL); 1311 td->io_hist_tree = RB_ROOT; 1312 1313 pthread_condattr_init(&attr); 1314 pthread_cond_init(&td->verify_cond, &attr); 1315 pthread_cond_init(&td->free_cond, &attr); 1316 1317 td_set_runstate(td, TD_INITIALIZED); 1318 dprint(FD_MUTEX, "up startup_mutex\n"); 1319 fio_mutex_up(startup_mutex); 1320 dprint(FD_MUTEX, "wait on td->mutex\n"); 1321 fio_mutex_down(td->mutex); 1322 dprint(FD_MUTEX, "done waiting on td->mutex\n"); 1323 1324 /* 1325 * A new gid requires privilege, so we need to do this before setting 1326 * the uid. 1327 */ 1328 if (o->gid != -1U && setgid(o->gid)) { 1329 td_verror(td, errno, "setgid"); 1330 goto err; 1331 } 1332 if (o->uid != -1U && setuid(o->uid)) { 1333 td_verror(td, errno, "setuid"); 1334 goto err; 1335 } 1336 1337 /* 1338 * If we have a gettimeofday() thread, make sure we exclude that 1339 * thread from this job 1340 */ 1341 if (o->gtod_cpu) 1342 fio_cpu_clear(&o->cpumask, o->gtod_cpu); 1343 1344 /* 1345 * Set affinity first, in case it has an impact on the memory 1346 * allocations. 1347 */ 1348 if (o->cpumask_set) { 1349 if (o->cpus_allowed_policy == FIO_CPUS_SPLIT) { 1350 ret = fio_cpus_split(&o->cpumask, td->thread_number - 1); 1351 if (!ret) { 1352 log_err("fio: no CPUs set\n"); 1353 log_err("fio: Try increasing number of available CPUs\n"); 1354 td_verror(td, EINVAL, "cpus_split"); 1355 goto err; 1356 } 1357 } 1358 ret = fio_setaffinity(td->pid, o->cpumask); 1359 if (ret == -1) { 1360 td_verror(td, errno, "cpu_set_affinity"); 1361 goto err; 1362 } 1363 } 1364 1365#ifdef CONFIG_LIBNUMA 1366 /* numa node setup */ 1367 if (o->numa_cpumask_set || o->numa_memmask_set) { 1368 struct bitmask *mask; 1369 1370 if (numa_available() < 0) { 1371 td_verror(td, errno, "Does not support NUMA API\n"); 1372 goto err; 1373 } 1374 1375 if (o->numa_cpumask_set) { 1376 mask = numa_parse_nodestring(o->numa_cpunodes); 1377 ret = numa_run_on_node_mask(mask); 1378 numa_free_nodemask(mask); 1379 if (ret == -1) { 1380 td_verror(td, errno, \ 1381 "numa_run_on_node_mask failed\n"); 1382 goto err; 1383 } 1384 } 1385 1386 if (o->numa_memmask_set) { 1387 1388 mask = NULL; 1389 if (o->numa_memnodes) 1390 mask = numa_parse_nodestring(o->numa_memnodes); 1391 1392 switch (o->numa_mem_mode) { 1393 case MPOL_INTERLEAVE: 1394 numa_set_interleave_mask(mask); 1395 break; 1396 case MPOL_BIND: 1397 numa_set_membind(mask); 1398 break; 1399 case MPOL_LOCAL: 1400 numa_set_localalloc(); 1401 break; 1402 case MPOL_PREFERRED: 1403 numa_set_preferred(o->numa_mem_prefer_node); 1404 break; 1405 case MPOL_DEFAULT: 1406 default: 1407 break; 1408 } 1409 1410 if (mask) 1411 numa_free_nodemask(mask); 1412 1413 } 1414 } 1415#endif 1416 1417 if (fio_pin_memory(td)) 1418 goto err; 1419 1420 /* 1421 * May alter parameters that init_io_u() will use, so we need to 1422 * do this first. 1423 */ 1424 if (init_iolog(td)) 1425 goto err; 1426 1427 if (init_io_u(td)) 1428 goto err; 1429 1430 if (o->verify_async && verify_async_init(td)) 1431 goto err; 1432 1433 if (o->ioprio) { 1434 ret = ioprio_set(IOPRIO_WHO_PROCESS, 0, o->ioprio_class, o->ioprio); 1435 if (ret == -1) { 1436 td_verror(td, errno, "ioprio_set"); 1437 goto err; 1438 } 1439 } 1440 1441 if (o->cgroup && cgroup_setup(td, cgroup_list, &cgroup_mnt)) 1442 goto err; 1443 1444 errno = 0; 1445 if (nice(o->nice) == -1 && errno != 0) { 1446 td_verror(td, errno, "nice"); 1447 goto err; 1448 } 1449 1450 if (o->ioscheduler && switch_ioscheduler(td)) 1451 goto err; 1452 1453 if (!o->create_serialize && setup_files(td)) 1454 goto err; 1455 1456 if (td_io_init(td)) 1457 goto err; 1458 1459 if (init_random_map(td)) 1460 goto err; 1461 1462 if (o->exec_prerun && exec_string(o, o->exec_prerun, (const char *)"prerun")) 1463 goto err; 1464 1465 if (o->pre_read) { 1466 if (pre_read_files(td) < 0) 1467 goto err; 1468 } 1469 1470 if (td->flags & TD_F_COMPRESS_LOG) 1471 tp_init(&td->tp_data); 1472 1473 fio_verify_init(td); 1474 1475 fio_gettime(&td->epoch, NULL); 1476 fio_getrusage(&td->ru_start); 1477 clear_state = 0; 1478 while (keep_running(td)) { 1479 uint64_t verify_bytes; 1480 1481 fio_gettime(&td->start, NULL); 1482 memcpy(&td->bw_sample_time, &td->start, sizeof(td->start)); 1483 memcpy(&td->iops_sample_time, &td->start, sizeof(td->start)); 1484 memcpy(&td->tv_cache, &td->start, sizeof(td->start)); 1485 1486 if (o->ratemin[DDIR_READ] || o->ratemin[DDIR_WRITE] || 1487 o->ratemin[DDIR_TRIM]) { 1488 memcpy(&td->lastrate[DDIR_READ], &td->bw_sample_time, 1489 sizeof(td->bw_sample_time)); 1490 memcpy(&td->lastrate[DDIR_WRITE], &td->bw_sample_time, 1491 sizeof(td->bw_sample_time)); 1492 memcpy(&td->lastrate[DDIR_TRIM], &td->bw_sample_time, 1493 sizeof(td->bw_sample_time)); 1494 } 1495 1496 if (clear_state) 1497 clear_io_state(td); 1498 1499 prune_io_piece_log(td); 1500 1501 if (td->o.verify_only && (td_write(td) || td_rw(td))) 1502 verify_bytes = do_dry_run(td); 1503 else 1504 verify_bytes = do_io(td); 1505 1506 clear_state = 1; 1507 1508 fio_mutex_down(stat_mutex); 1509 if (td_read(td) && td->io_bytes[DDIR_READ]) { 1510 elapsed = mtime_since_now(&td->start); 1511 td->ts.runtime[DDIR_READ] += elapsed; 1512 } 1513 if (td_write(td) && td->io_bytes[DDIR_WRITE]) { 1514 elapsed = mtime_since_now(&td->start); 1515 td->ts.runtime[DDIR_WRITE] += elapsed; 1516 } 1517 if (td_trim(td) && td->io_bytes[DDIR_TRIM]) { 1518 elapsed = mtime_since_now(&td->start); 1519 td->ts.runtime[DDIR_TRIM] += elapsed; 1520 } 1521 fio_gettime(&td->start, NULL); 1522 fio_mutex_up(stat_mutex); 1523 1524 if (td->error || td->terminate) 1525 break; 1526 1527 if (!o->do_verify || 1528 o->verify == VERIFY_NONE || 1529 (td->io_ops->flags & FIO_UNIDIR)) 1530 continue; 1531 1532 clear_io_state(td); 1533 1534 fio_gettime(&td->start, NULL); 1535 1536 do_verify(td, verify_bytes); 1537 1538 fio_mutex_down(stat_mutex); 1539 td->ts.runtime[DDIR_READ] += mtime_since_now(&td->start); 1540 fio_gettime(&td->start, NULL); 1541 fio_mutex_up(stat_mutex); 1542 1543 if (td->error || td->terminate) 1544 break; 1545 } 1546 1547 update_rusage_stat(td); 1548 td->ts.total_run_time = mtime_since_now(&td->epoch); 1549 td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ]; 1550 td->ts.io_bytes[DDIR_WRITE] = td->io_bytes[DDIR_WRITE]; 1551 td->ts.io_bytes[DDIR_TRIM] = td->io_bytes[DDIR_TRIM]; 1552 1553 if (td->o.verify_state_save && !(td->flags & TD_F_VSTATE_SAVED) && 1554 (td->o.verify != VERIFY_NONE && td_write(td))) { 1555 struct all_io_list *state; 1556 size_t sz; 1557 1558 state = get_all_io_list(td->thread_number, &sz); 1559 if (state) { 1560 __verify_save_state(state, "local"); 1561 free(state); 1562 } 1563 } 1564 1565 fio_unpin_memory(td); 1566 1567 fio_writeout_logs(td); 1568 1569 if (td->flags & TD_F_COMPRESS_LOG) 1570 tp_exit(&td->tp_data); 1571 1572 if (o->exec_postrun) 1573 exec_string(o, o->exec_postrun, (const char *)"postrun"); 1574 1575 if (exitall_on_terminate) 1576 fio_terminate_threads(td->groupid); 1577 1578err: 1579 if (td->error) 1580 log_info("fio: pid=%d, err=%d/%s\n", (int) td->pid, td->error, 1581 td->verror); 1582 1583 if (o->verify_async) 1584 verify_async_exit(td); 1585 1586 close_and_free_files(td); 1587 cleanup_io_u(td); 1588 close_ioengine(td); 1589 cgroup_shutdown(td, &cgroup_mnt); 1590 verify_free_state(td); 1591 1592 if (o->cpumask_set) { 1593 ret = fio_cpuset_exit(&o->cpumask); 1594 if (ret) 1595 td_verror(td, ret, "fio_cpuset_exit"); 1596 } 1597 1598 /* 1599 * do this very late, it will log file closing as well 1600 */ 1601 if (o->write_iolog_file) 1602 write_iolog_close(td); 1603 1604 fio_mutex_remove(td->mutex); 1605 td->mutex = NULL; 1606 1607 td_set_runstate(td, TD_EXITED); 1608 1609 /* 1610 * Do this last after setting our runstate to exited, so we 1611 * know that the stat thread is signaled. 1612 */ 1613 check_update_rusage(td); 1614 1615 return (void *) (uintptr_t) td->error; 1616} 1617 1618 1619/* 1620 * We cannot pass the td data into a forked process, so attach the td and 1621 * pass it to the thread worker. 1622 */ 1623static int fork_main(int shmid, int offset) 1624{ 1625 struct thread_data *td; 1626 void *data, *ret; 1627 1628#if !defined(__hpux) && !defined(CONFIG_NO_SHM) 1629 data = shmat(shmid, NULL, 0); 1630 if (data == (void *) -1) { 1631 int __err = errno; 1632 1633 perror("shmat"); 1634 return __err; 1635 } 1636#else 1637 /* 1638 * HP-UX inherits shm mappings? 1639 */ 1640 data = threads; 1641#endif 1642 1643 td = data + offset * sizeof(struct thread_data); 1644 ret = thread_main(td); 1645 shmdt(data); 1646 return (int) (uintptr_t) ret; 1647} 1648 1649static void dump_td_info(struct thread_data *td) 1650{ 1651 log_err("fio: job '%s' hasn't exited in %lu seconds, it appears to " 1652 "be stuck. Doing forceful exit of this job.\n", td->o.name, 1653 (unsigned long) time_since_now(&td->terminate_time)); 1654} 1655 1656/* 1657 * Run over the job map and reap the threads that have exited, if any. 1658 */ 1659static void reap_threads(unsigned int *nr_running, unsigned int *t_rate, 1660 unsigned int *m_rate) 1661{ 1662 struct thread_data *td; 1663 unsigned int cputhreads, realthreads, pending; 1664 int i, status, ret; 1665 1666 /* 1667 * reap exited threads (TD_EXITED -> TD_REAPED) 1668 */ 1669 realthreads = pending = cputhreads = 0; 1670 for_each_td(td, i) { 1671 int flags = 0; 1672 1673 /* 1674 * ->io_ops is NULL for a thread that has closed its 1675 * io engine 1676 */ 1677 if (td->io_ops && !strcmp(td->io_ops->name, "cpuio")) 1678 cputhreads++; 1679 else 1680 realthreads++; 1681 1682 if (!td->pid) { 1683 pending++; 1684 continue; 1685 } 1686 if (td->runstate == TD_REAPED) 1687 continue; 1688 if (td->o.use_thread) { 1689 if (td->runstate == TD_EXITED) { 1690 td_set_runstate(td, TD_REAPED); 1691 goto reaped; 1692 } 1693 continue; 1694 } 1695 1696 flags = WNOHANG; 1697 if (td->runstate == TD_EXITED) 1698 flags = 0; 1699 1700 /* 1701 * check if someone quit or got killed in an unusual way 1702 */ 1703 ret = waitpid(td->pid, &status, flags); 1704 if (ret < 0) { 1705 if (errno == ECHILD) { 1706 log_err("fio: pid=%d disappeared %d\n", 1707 (int) td->pid, td->runstate); 1708 td->sig = ECHILD; 1709 td_set_runstate(td, TD_REAPED); 1710 goto reaped; 1711 } 1712 perror("waitpid"); 1713 } else if (ret == td->pid) { 1714 if (WIFSIGNALED(status)) { 1715 int sig = WTERMSIG(status); 1716 1717 if (sig != SIGTERM && sig != SIGUSR2) 1718 log_err("fio: pid=%d, got signal=%d\n", 1719 (int) td->pid, sig); 1720 td->sig = sig; 1721 td_set_runstate(td, TD_REAPED); 1722 goto reaped; 1723 } 1724 if (WIFEXITED(status)) { 1725 if (WEXITSTATUS(status) && !td->error) 1726 td->error = WEXITSTATUS(status); 1727 1728 td_set_runstate(td, TD_REAPED); 1729 goto reaped; 1730 } 1731 } 1732 1733 /* 1734 * If the job is stuck, do a forceful timeout of it and 1735 * move on. 1736 */ 1737 if (td->terminate && 1738 time_since_now(&td->terminate_time) >= FIO_REAP_TIMEOUT) { 1739 dump_td_info(td); 1740 td_set_runstate(td, TD_REAPED); 1741 goto reaped; 1742 } 1743 1744 /* 1745 * thread is not dead, continue 1746 */ 1747 pending++; 1748 continue; 1749reaped: 1750 (*nr_running)--; 1751 (*m_rate) -= ddir_rw_sum(td->o.ratemin); 1752 (*t_rate) -= ddir_rw_sum(td->o.rate); 1753 if (!td->pid) 1754 pending--; 1755 1756 if (td->error) 1757 exit_value++; 1758 1759 done_secs += mtime_since_now(&td->epoch) / 1000; 1760 profile_td_exit(td); 1761 } 1762 1763 if (*nr_running == cputhreads && !pending && realthreads) 1764 fio_terminate_threads(TERMINATE_ALL); 1765} 1766 1767static int __check_trigger_file(void) 1768{ 1769 struct stat sb; 1770 1771 if (!trigger_file) 1772 return 0; 1773 1774 if (stat(trigger_file, &sb)) 1775 return 0; 1776 1777 if (unlink(trigger_file) < 0) 1778 log_err("fio: failed to unlink %s: %s\n", trigger_file, 1779 strerror(errno)); 1780 1781 return 1; 1782} 1783 1784static int trigger_timedout(void) 1785{ 1786 if (trigger_timeout) 1787 return time_since_genesis() >= trigger_timeout; 1788 1789 return 0; 1790} 1791 1792void exec_trigger(const char *cmd) 1793{ 1794 int ret; 1795 1796 if (!cmd) 1797 return; 1798 1799 ret = system(cmd); 1800 if (ret == -1) 1801 log_err("fio: failed executing %s trigger\n", cmd); 1802} 1803 1804void check_trigger_file(void) 1805{ 1806 if (__check_trigger_file() || trigger_timedout()) { 1807 if (nr_clients) 1808 fio_clients_send_trigger(trigger_cmd); 1809 else { 1810 verify_save_state(); 1811 fio_terminate_threads(TERMINATE_ALL); 1812 exec_trigger(trigger_cmd); 1813 } 1814 } 1815} 1816 1817static int fio_verify_load_state(struct thread_data *td) 1818{ 1819 int ret; 1820 1821 if (!td->o.verify_state) 1822 return 0; 1823 1824 if (is_backend) { 1825 void *data; 1826 1827 ret = fio_server_get_verify_state(td->o.name, 1828 td->thread_number - 1, &data); 1829 if (!ret) 1830 verify_convert_assign_state(td, data); 1831 } else 1832 ret = verify_load_state(td, "local"); 1833 1834 return ret; 1835} 1836 1837static void do_usleep(unsigned int usecs) 1838{ 1839 check_for_running_stats(); 1840 check_trigger_file(); 1841 usleep(usecs); 1842} 1843 1844/* 1845 * Main function for kicking off and reaping jobs, as needed. 1846 */ 1847static void run_threads(void) 1848{ 1849 struct thread_data *td; 1850 unsigned int i, todo, nr_running, m_rate, t_rate, nr_started; 1851 uint64_t spent; 1852 1853 if (fio_gtod_offload && fio_start_gtod_thread()) 1854 return; 1855 1856 fio_idle_prof_init(); 1857 1858 set_sig_handlers(); 1859 1860 nr_thread = nr_process = 0; 1861 for_each_td(td, i) { 1862 if (td->o.use_thread) 1863 nr_thread++; 1864 else 1865 nr_process++; 1866 } 1867 1868 if (output_format == FIO_OUTPUT_NORMAL) { 1869 log_info("Starting "); 1870 if (nr_thread) 1871 log_info("%d thread%s", nr_thread, 1872 nr_thread > 1 ? "s" : ""); 1873 if (nr_process) { 1874 if (nr_thread) 1875 log_info(" and "); 1876 log_info("%d process%s", nr_process, 1877 nr_process > 1 ? "es" : ""); 1878 } 1879 log_info("\n"); 1880 log_info_flush(); 1881 } 1882 1883 todo = thread_number; 1884 nr_running = 0; 1885 nr_started = 0; 1886 m_rate = t_rate = 0; 1887 1888 for_each_td(td, i) { 1889 print_status_init(td->thread_number - 1); 1890 1891 if (!td->o.create_serialize) 1892 continue; 1893 1894 if (fio_verify_load_state(td)) 1895 goto reap; 1896 1897 /* 1898 * do file setup here so it happens sequentially, 1899 * we don't want X number of threads getting their 1900 * client data interspersed on disk 1901 */ 1902 if (setup_files(td)) { 1903reap: 1904 exit_value++; 1905 if (td->error) 1906 log_err("fio: pid=%d, err=%d/%s\n", 1907 (int) td->pid, td->error, td->verror); 1908 td_set_runstate(td, TD_REAPED); 1909 todo--; 1910 } else { 1911 struct fio_file *f; 1912 unsigned int j; 1913 1914 /* 1915 * for sharing to work, each job must always open 1916 * its own files. so close them, if we opened them 1917 * for creation 1918 */ 1919 for_each_file(td, f, j) { 1920 if (fio_file_open(f)) 1921 td_io_close_file(td, f); 1922 } 1923 } 1924 } 1925 1926 /* start idle threads before io threads start to run */ 1927 fio_idle_prof_start(); 1928 1929 set_genesis_time(); 1930 1931 while (todo) { 1932 struct thread_data *map[REAL_MAX_JOBS]; 1933 struct timeval this_start; 1934 int this_jobs = 0, left; 1935 1936 /* 1937 * create threads (TD_NOT_CREATED -> TD_CREATED) 1938 */ 1939 for_each_td(td, i) { 1940 if (td->runstate != TD_NOT_CREATED) 1941 continue; 1942 1943 /* 1944 * never got a chance to start, killed by other 1945 * thread for some reason 1946 */ 1947 if (td->terminate) { 1948 todo--; 1949 continue; 1950 } 1951 1952 if (td->o.start_delay) { 1953 spent = utime_since_genesis(); 1954 1955 if (td->o.start_delay > spent) 1956 continue; 1957 } 1958 1959 if (td->o.stonewall && (nr_started || nr_running)) { 1960 dprint(FD_PROCESS, "%s: stonewall wait\n", 1961 td->o.name); 1962 break; 1963 } 1964 1965 init_disk_util(td); 1966 1967 td->rusage_sem = fio_mutex_init(FIO_MUTEX_LOCKED); 1968 td->update_rusage = 0; 1969 1970 /* 1971 * Set state to created. Thread will transition 1972 * to TD_INITIALIZED when it's done setting up. 1973 */ 1974 td_set_runstate(td, TD_CREATED); 1975 map[this_jobs++] = td; 1976 nr_started++; 1977 1978 if (td->o.use_thread) { 1979 int ret; 1980 1981 dprint(FD_PROCESS, "will pthread_create\n"); 1982 ret = pthread_create(&td->thread, NULL, 1983 thread_main, td); 1984 if (ret) { 1985 log_err("pthread_create: %s\n", 1986 strerror(ret)); 1987 nr_started--; 1988 break; 1989 } 1990 ret = pthread_detach(td->thread); 1991 if (ret) 1992 log_err("pthread_detach: %s", 1993 strerror(ret)); 1994 } else { 1995 pid_t pid; 1996 dprint(FD_PROCESS, "will fork\n"); 1997 pid = fork(); 1998 if (!pid) { 1999 int ret = fork_main(shm_id, i); 2000 2001 _exit(ret); 2002 } else if (i == fio_debug_jobno) 2003 *fio_debug_jobp = pid; 2004 } 2005 dprint(FD_MUTEX, "wait on startup_mutex\n"); 2006 if (fio_mutex_down_timeout(startup_mutex, 10)) { 2007 log_err("fio: job startup hung? exiting.\n"); 2008 fio_terminate_threads(TERMINATE_ALL); 2009 fio_abort = 1; 2010 nr_started--; 2011 break; 2012 } 2013 dprint(FD_MUTEX, "done waiting on startup_mutex\n"); 2014 } 2015 2016 /* 2017 * Wait for the started threads to transition to 2018 * TD_INITIALIZED. 2019 */ 2020 fio_gettime(&this_start, NULL); 2021 left = this_jobs; 2022 while (left && !fio_abort) { 2023 if (mtime_since_now(&this_start) > JOB_START_TIMEOUT) 2024 break; 2025 2026 do_usleep(100000); 2027 2028 for (i = 0; i < this_jobs; i++) { 2029 td = map[i]; 2030 if (!td) 2031 continue; 2032 if (td->runstate == TD_INITIALIZED) { 2033 map[i] = NULL; 2034 left--; 2035 } else if (td->runstate >= TD_EXITED) { 2036 map[i] = NULL; 2037 left--; 2038 todo--; 2039 nr_running++; /* work-around... */ 2040 } 2041 } 2042 } 2043 2044 if (left) { 2045 log_err("fio: %d job%s failed to start\n", left, 2046 left > 1 ? "s" : ""); 2047 for (i = 0; i < this_jobs; i++) { 2048 td = map[i]; 2049 if (!td) 2050 continue; 2051 kill(td->pid, SIGTERM); 2052 } 2053 break; 2054 } 2055 2056 /* 2057 * start created threads (TD_INITIALIZED -> TD_RUNNING). 2058 */ 2059 for_each_td(td, i) { 2060 if (td->runstate != TD_INITIALIZED) 2061 continue; 2062 2063 if (in_ramp_time(td)) 2064 td_set_runstate(td, TD_RAMP); 2065 else 2066 td_set_runstate(td, TD_RUNNING); 2067 nr_running++; 2068 nr_started--; 2069 m_rate += ddir_rw_sum(td->o.ratemin); 2070 t_rate += ddir_rw_sum(td->o.rate); 2071 todo--; 2072 fio_mutex_up(td->mutex); 2073 } 2074 2075 reap_threads(&nr_running, &t_rate, &m_rate); 2076 2077 if (todo) 2078 do_usleep(100000); 2079 } 2080 2081 while (nr_running) { 2082 reap_threads(&nr_running, &t_rate, &m_rate); 2083 do_usleep(10000); 2084 } 2085 2086 fio_idle_prof_stop(); 2087 2088 update_io_ticks(); 2089} 2090 2091static void wait_for_helper_thread_exit(void) 2092{ 2093 void *ret; 2094 2095 helper_exit = 1; 2096 pthread_cond_signal(&helper_cond); 2097 pthread_join(helper_thread, &ret); 2098} 2099 2100static void free_disk_util(void) 2101{ 2102 disk_util_prune_entries(); 2103 2104 pthread_cond_destroy(&helper_cond); 2105} 2106 2107static void *helper_thread_main(void *data) 2108{ 2109 int ret = 0; 2110 2111 fio_mutex_up(startup_mutex); 2112 2113 while (!ret) { 2114 uint64_t sec = DISK_UTIL_MSEC / 1000; 2115 uint64_t nsec = (DISK_UTIL_MSEC % 1000) * 1000000; 2116 struct timespec ts; 2117 struct timeval tv; 2118 2119 gettimeofday(&tv, NULL); 2120 ts.tv_sec = tv.tv_sec + sec; 2121 ts.tv_nsec = (tv.tv_usec * 1000) + nsec; 2122 if (ts.tv_nsec > 1000000000ULL) { 2123 ts.tv_nsec -= 1000000000ULL; 2124 ts.tv_sec++; 2125 } 2126 2127 pthread_cond_timedwait(&helper_cond, &helper_lock, &ts); 2128 2129 ret = update_io_ticks(); 2130 2131 if (helper_do_stat) { 2132 helper_do_stat = 0; 2133 __show_running_run_stats(); 2134 } 2135 2136 if (!is_backend) 2137 print_thread_status(); 2138 } 2139 2140 return NULL; 2141} 2142 2143static int create_helper_thread(void) 2144{ 2145 int ret; 2146 2147 setup_disk_util(); 2148 2149 pthread_cond_init(&helper_cond, NULL); 2150 pthread_mutex_init(&helper_lock, NULL); 2151 2152 ret = pthread_create(&helper_thread, NULL, helper_thread_main, NULL); 2153 if (ret) { 2154 log_err("Can't create helper thread: %s\n", strerror(ret)); 2155 return 1; 2156 } 2157 2158 dprint(FD_MUTEX, "wait on startup_mutex\n"); 2159 fio_mutex_down(startup_mutex); 2160 dprint(FD_MUTEX, "done waiting on startup_mutex\n"); 2161 return 0; 2162} 2163 2164int fio_backend(void) 2165{ 2166 struct thread_data *td; 2167 int i; 2168 2169 if (exec_profile) { 2170 if (load_profile(exec_profile)) 2171 return 1; 2172 free(exec_profile); 2173 exec_profile = NULL; 2174 } 2175 if (!thread_number) 2176 return 0; 2177 2178 if (write_bw_log) { 2179 struct log_params p = { 2180 .log_type = IO_LOG_TYPE_BW, 2181 }; 2182 2183 setup_log(&agg_io_log[DDIR_READ], &p, "agg-read_bw.log"); 2184 setup_log(&agg_io_log[DDIR_WRITE], &p, "agg-write_bw.log"); 2185 setup_log(&agg_io_log[DDIR_TRIM], &p, "agg-trim_bw.log"); 2186 } 2187 2188 startup_mutex = fio_mutex_init(FIO_MUTEX_LOCKED); 2189 if (startup_mutex == NULL) 2190 return 1; 2191 2192 set_genesis_time(); 2193 stat_init(); 2194 create_helper_thread(); 2195 2196 cgroup_list = smalloc(sizeof(*cgroup_list)); 2197 INIT_FLIST_HEAD(cgroup_list); 2198 2199 run_threads(); 2200 2201 wait_for_helper_thread_exit(); 2202 2203 if (!fio_abort) { 2204 __show_run_stats(); 2205 if (write_bw_log) { 2206 for (i = 0; i < DDIR_RWDIR_CNT; i++) { 2207 struct io_log *log = agg_io_log[i]; 2208 2209 flush_log(log); 2210 free_log(log); 2211 } 2212 } 2213 } 2214 2215 for_each_td(td, i) { 2216 fio_options_free(td); 2217 if (td->rusage_sem) { 2218 fio_mutex_remove(td->rusage_sem); 2219 td->rusage_sem = NULL; 2220 } 2221 } 2222 2223 free_disk_util(); 2224 cgroup_kill(cgroup_list); 2225 sfree(cgroup_list); 2226 sfree(cgroup_mnt); 2227 2228 fio_mutex_remove(startup_mutex); 2229 stat_exit(); 2230 return exit_value; 2231} 2232