| 1 | /* |
| 2 | * Code related to writing an iolog of what a thread is doing, and to |
| 3 | * later read that back and replay |
| 4 | */ |
| 5 | #include <stdio.h> |
| 6 | #include <stdlib.h> |
| 7 | #include <assert.h> |
| 8 | #include <sys/types.h> |
| 9 | #include <sys/stat.h> |
| 10 | #include <unistd.h> |
| 11 | #ifdef CONFIG_ZLIB |
| 12 | #include <zlib.h> |
| 13 | #endif |
| 14 | |
| 15 | #include "flist.h" |
| 16 | #include "fio.h" |
| 17 | #include "trim.h" |
| 18 | #include "filelock.h" |
| 19 | #include "smalloc.h" |
| 20 | #include "blktrace.h" |
| 21 | #include "pshared.h" |
| 22 | |
| 23 | static int iolog_flush(struct io_log *log); |
| 24 | |
| 25 | static const char iolog_ver2[] = "fio version 2 iolog"; |
| 26 | |
| 27 | void queue_io_piece(struct thread_data *td, struct io_piece *ipo) |
| 28 | { |
| 29 | flist_add_tail(&ipo->list, &td->io_log_list); |
| 30 | td->total_io_size += ipo->len; |
| 31 | } |
| 32 | |
| 33 | void log_io_u(const struct thread_data *td, const struct io_u *io_u) |
| 34 | { |
| 35 | if (!td->o.write_iolog_file) |
| 36 | return; |
| 37 | |
| 38 | fprintf(td->iolog_f, "%s %s %llu %lu\n", io_u->file->file_name, |
| 39 | io_ddir_name(io_u->ddir), |
| 40 | io_u->offset, io_u->buflen); |
| 41 | } |
| 42 | |
| 43 | void log_file(struct thread_data *td, struct fio_file *f, |
| 44 | enum file_log_act what) |
| 45 | { |
| 46 | const char *act[] = { "add", "open", "close" }; |
| 47 | |
| 48 | assert(what < 3); |
| 49 | |
| 50 | if (!td->o.write_iolog_file) |
| 51 | return; |
| 52 | |
| 53 | |
| 54 | /* |
| 55 | * this happens on the pre-open/close done before the job starts |
| 56 | */ |
| 57 | if (!td->iolog_f) |
| 58 | return; |
| 59 | |
| 60 | fprintf(td->iolog_f, "%s %s\n", f->file_name, act[what]); |
| 61 | } |
| 62 | |
| 63 | static void iolog_delay(struct thread_data *td, unsigned long delay) |
| 64 | { |
| 65 | uint64_t usec = utime_since_now(&td->last_issue); |
| 66 | uint64_t this_delay; |
| 67 | struct timespec ts; |
| 68 | |
| 69 | if (delay < td->time_offset) { |
| 70 | td->time_offset = 0; |
| 71 | return; |
| 72 | } |
| 73 | |
| 74 | delay -= td->time_offset; |
| 75 | if (delay < usec) |
| 76 | return; |
| 77 | |
| 78 | delay -= usec; |
| 79 | |
| 80 | fio_gettime(&ts, NULL); |
| 81 | while (delay && !td->terminate) { |
| 82 | this_delay = delay; |
| 83 | if (this_delay > 500000) |
| 84 | this_delay = 500000; |
| 85 | |
| 86 | usec_sleep(td, this_delay); |
| 87 | delay -= this_delay; |
| 88 | } |
| 89 | |
| 90 | usec = utime_since_now(&ts); |
| 91 | if (usec > delay) |
| 92 | td->time_offset = usec - delay; |
| 93 | else |
| 94 | td->time_offset = 0; |
| 95 | } |
| 96 | |
| 97 | static int ipo_special(struct thread_data *td, struct io_piece *ipo) |
| 98 | { |
| 99 | struct fio_file *f; |
| 100 | int ret; |
| 101 | |
| 102 | /* |
| 103 | * Not a special ipo |
| 104 | */ |
| 105 | if (ipo->ddir != DDIR_INVAL) |
| 106 | return 0; |
| 107 | |
| 108 | f = td->files[ipo->fileno]; |
| 109 | |
| 110 | switch (ipo->file_action) { |
| 111 | case FIO_LOG_OPEN_FILE: |
| 112 | if (td->o.replay_redirect && fio_file_open(f)) { |
| 113 | dprint(FD_FILE, "iolog: ignoring re-open of file %s\n", |
| 114 | f->file_name); |
| 115 | break; |
| 116 | } |
| 117 | ret = td_io_open_file(td, f); |
| 118 | if (!ret) |
| 119 | break; |
| 120 | td_verror(td, ret, "iolog open file"); |
| 121 | return -1; |
| 122 | case FIO_LOG_CLOSE_FILE: |
| 123 | td_io_close_file(td, f); |
| 124 | break; |
| 125 | case FIO_LOG_UNLINK_FILE: |
| 126 | td_io_unlink_file(td, f); |
| 127 | break; |
| 128 | default: |
| 129 | log_err("fio: bad file action %d\n", ipo->file_action); |
| 130 | break; |
| 131 | } |
| 132 | |
| 133 | return 1; |
| 134 | } |
| 135 | |
| 136 | int read_iolog_get(struct thread_data *td, struct io_u *io_u) |
| 137 | { |
| 138 | struct io_piece *ipo; |
| 139 | unsigned long elapsed; |
| 140 | |
| 141 | while (!flist_empty(&td->io_log_list)) { |
| 142 | int ret; |
| 143 | |
| 144 | ipo = flist_first_entry(&td->io_log_list, struct io_piece, list); |
| 145 | flist_del(&ipo->list); |
| 146 | remove_trim_entry(td, ipo); |
| 147 | |
| 148 | ret = ipo_special(td, ipo); |
| 149 | if (ret < 0) { |
| 150 | free(ipo); |
| 151 | break; |
| 152 | } else if (ret > 0) { |
| 153 | free(ipo); |
| 154 | continue; |
| 155 | } |
| 156 | |
| 157 | io_u->ddir = ipo->ddir; |
| 158 | if (ipo->ddir != DDIR_WAIT) { |
| 159 | io_u->offset = ipo->offset; |
| 160 | io_u->buflen = ipo->len; |
| 161 | io_u->file = td->files[ipo->fileno]; |
| 162 | get_file(io_u->file); |
| 163 | dprint(FD_IO, "iolog: get %llu/%lu/%s\n", io_u->offset, |
| 164 | io_u->buflen, io_u->file->file_name); |
| 165 | if (ipo->delay) |
| 166 | iolog_delay(td, ipo->delay); |
| 167 | } else { |
| 168 | elapsed = mtime_since_genesis(); |
| 169 | if (ipo->delay > elapsed) |
| 170 | usec_sleep(td, (ipo->delay - elapsed) * 1000); |
| 171 | } |
| 172 | |
| 173 | free(ipo); |
| 174 | |
| 175 | if (io_u->ddir != DDIR_WAIT) |
| 176 | return 0; |
| 177 | } |
| 178 | |
| 179 | td->done = 1; |
| 180 | return 1; |
| 181 | } |
| 182 | |
| 183 | void prune_io_piece_log(struct thread_data *td) |
| 184 | { |
| 185 | struct io_piece *ipo; |
| 186 | struct fio_rb_node *n; |
| 187 | |
| 188 | while ((n = rb_first(&td->io_hist_tree)) != NULL) { |
| 189 | ipo = rb_entry(n, struct io_piece, rb_node); |
| 190 | rb_erase(n, &td->io_hist_tree); |
| 191 | remove_trim_entry(td, ipo); |
| 192 | td->io_hist_len--; |
| 193 | free(ipo); |
| 194 | } |
| 195 | |
| 196 | while (!flist_empty(&td->io_hist_list)) { |
| 197 | ipo = flist_first_entry(&td->io_hist_list, struct io_piece, list); |
| 198 | flist_del(&ipo->list); |
| 199 | remove_trim_entry(td, ipo); |
| 200 | td->io_hist_len--; |
| 201 | free(ipo); |
| 202 | } |
| 203 | } |
| 204 | |
| 205 | /* |
| 206 | * log a successful write, so we can unwind the log for verify |
| 207 | */ |
| 208 | void log_io_piece(struct thread_data *td, struct io_u *io_u) |
| 209 | { |
| 210 | struct fio_rb_node **p, *parent; |
| 211 | struct io_piece *ipo, *__ipo; |
| 212 | |
| 213 | ipo = malloc(sizeof(struct io_piece)); |
| 214 | init_ipo(ipo); |
| 215 | ipo->file = io_u->file; |
| 216 | ipo->offset = io_u->offset; |
| 217 | ipo->len = io_u->buflen; |
| 218 | ipo->numberio = io_u->numberio; |
| 219 | ipo->flags = IP_F_IN_FLIGHT; |
| 220 | |
| 221 | io_u->ipo = ipo; |
| 222 | |
| 223 | if (io_u_should_trim(td, io_u)) { |
| 224 | flist_add_tail(&ipo->trim_list, &td->trim_list); |
| 225 | td->trim_entries++; |
| 226 | } |
| 227 | |
| 228 | /* |
| 229 | * We don't need to sort the entries if we only performed sequential |
| 230 | * writes. In this case, just reading back data in the order we wrote |
| 231 | * it out is the faster but still safe. |
| 232 | * |
| 233 | * One exception is if we don't have a random map in which case we need |
| 234 | * to check for duplicate blocks and drop the old one, which we rely on |
| 235 | * the rb insert/lookup for handling. |
| 236 | */ |
| 237 | if (((!td->o.verifysort) || !td_random(td)) && |
| 238 | file_randommap(td, ipo->file)) { |
| 239 | INIT_FLIST_HEAD(&ipo->list); |
| 240 | flist_add_tail(&ipo->list, &td->io_hist_list); |
| 241 | ipo->flags |= IP_F_ONLIST; |
| 242 | td->io_hist_len++; |
| 243 | return; |
| 244 | } |
| 245 | |
| 246 | RB_CLEAR_NODE(&ipo->rb_node); |
| 247 | |
| 248 | /* |
| 249 | * Sort the entry into the verification list |
| 250 | */ |
| 251 | restart: |
| 252 | p = &td->io_hist_tree.rb_node; |
| 253 | parent = NULL; |
| 254 | while (*p) { |
| 255 | int overlap = 0; |
| 256 | parent = *p; |
| 257 | |
| 258 | __ipo = rb_entry(parent, struct io_piece, rb_node); |
| 259 | if (ipo->file < __ipo->file) |
| 260 | p = &(*p)->rb_left; |
| 261 | else if (ipo->file > __ipo->file) |
| 262 | p = &(*p)->rb_right; |
| 263 | else if (ipo->offset < __ipo->offset) { |
| 264 | p = &(*p)->rb_left; |
| 265 | overlap = ipo->offset + ipo->len > __ipo->offset; |
| 266 | } |
| 267 | else if (ipo->offset > __ipo->offset) { |
| 268 | p = &(*p)->rb_right; |
| 269 | overlap = __ipo->offset + __ipo->len > ipo->offset; |
| 270 | } |
| 271 | else |
| 272 | overlap = 1; |
| 273 | |
| 274 | if (overlap) { |
| 275 | dprint(FD_IO, "iolog: overlap %llu/%lu, %llu/%lu\n", |
| 276 | __ipo->offset, __ipo->len, |
| 277 | ipo->offset, ipo->len); |
| 278 | td->io_hist_len--; |
| 279 | rb_erase(parent, &td->io_hist_tree); |
| 280 | remove_trim_entry(td, __ipo); |
| 281 | if (!(__ipo->flags & IP_F_IN_FLIGHT)) |
| 282 | free(__ipo); |
| 283 | goto restart; |
| 284 | } |
| 285 | } |
| 286 | |
| 287 | rb_link_node(&ipo->rb_node, parent, p); |
| 288 | rb_insert_color(&ipo->rb_node, &td->io_hist_tree); |
| 289 | ipo->flags |= IP_F_ONRB; |
| 290 | td->io_hist_len++; |
| 291 | } |
| 292 | |
| 293 | void unlog_io_piece(struct thread_data *td, struct io_u *io_u) |
| 294 | { |
| 295 | struct io_piece *ipo = io_u->ipo; |
| 296 | |
| 297 | if (td->ts.nr_block_infos) { |
| 298 | uint32_t *info = io_u_block_info(td, io_u); |
| 299 | if (BLOCK_INFO_STATE(*info) < BLOCK_STATE_TRIM_FAILURE) { |
| 300 | if (io_u->ddir == DDIR_TRIM) |
| 301 | *info = BLOCK_INFO_SET_STATE(*info, |
| 302 | BLOCK_STATE_TRIM_FAILURE); |
| 303 | else if (io_u->ddir == DDIR_WRITE) |
| 304 | *info = BLOCK_INFO_SET_STATE(*info, |
| 305 | BLOCK_STATE_WRITE_FAILURE); |
| 306 | } |
| 307 | } |
| 308 | |
| 309 | if (!ipo) |
| 310 | return; |
| 311 | |
| 312 | if (ipo->flags & IP_F_ONRB) |
| 313 | rb_erase(&ipo->rb_node, &td->io_hist_tree); |
| 314 | else if (ipo->flags & IP_F_ONLIST) |
| 315 | flist_del(&ipo->list); |
| 316 | |
| 317 | free(ipo); |
| 318 | io_u->ipo = NULL; |
| 319 | td->io_hist_len--; |
| 320 | } |
| 321 | |
| 322 | void trim_io_piece(struct thread_data *td, const struct io_u *io_u) |
| 323 | { |
| 324 | struct io_piece *ipo = io_u->ipo; |
| 325 | |
| 326 | if (!ipo) |
| 327 | return; |
| 328 | |
| 329 | ipo->len = io_u->xfer_buflen - io_u->resid; |
| 330 | } |
| 331 | |
| 332 | void write_iolog_close(struct thread_data *td) |
| 333 | { |
| 334 | fflush(td->iolog_f); |
| 335 | fclose(td->iolog_f); |
| 336 | free(td->iolog_buf); |
| 337 | td->iolog_f = NULL; |
| 338 | td->iolog_buf = NULL; |
| 339 | } |
| 340 | |
| 341 | /* |
| 342 | * Read version 2 iolog data. It is enhanced to include per-file logging, |
| 343 | * syncs, etc. |
| 344 | */ |
| 345 | static int read_iolog2(struct thread_data *td, FILE *f) |
| 346 | { |
| 347 | unsigned long long offset; |
| 348 | unsigned int bytes; |
| 349 | int reads, writes, waits, fileno = 0, file_action = 0; /* stupid gcc */ |
| 350 | char *rfname, *fname, *act; |
| 351 | char *str, *p; |
| 352 | enum fio_ddir rw; |
| 353 | |
| 354 | free_release_files(td); |
| 355 | |
| 356 | /* |
| 357 | * Read in the read iolog and store it, reuse the infrastructure |
| 358 | * for doing verifications. |
| 359 | */ |
| 360 | str = malloc(4096); |
| 361 | rfname = fname = malloc(256+16); |
| 362 | act = malloc(256+16); |
| 363 | |
| 364 | reads = writes = waits = 0; |
| 365 | while ((p = fgets(str, 4096, f)) != NULL) { |
| 366 | struct io_piece *ipo; |
| 367 | int r; |
| 368 | |
| 369 | r = sscanf(p, "%256s %256s %llu %u", rfname, act, &offset, |
| 370 | &bytes); |
| 371 | |
| 372 | if (td->o.replay_redirect) |
| 373 | fname = td->o.replay_redirect; |
| 374 | |
| 375 | if (r == 4) { |
| 376 | /* |
| 377 | * Check action first |
| 378 | */ |
| 379 | if (!strcmp(act, "wait")) |
| 380 | rw = DDIR_WAIT; |
| 381 | else if (!strcmp(act, "read")) |
| 382 | rw = DDIR_READ; |
| 383 | else if (!strcmp(act, "write")) |
| 384 | rw = DDIR_WRITE; |
| 385 | else if (!strcmp(act, "sync")) |
| 386 | rw = DDIR_SYNC; |
| 387 | else if (!strcmp(act, "datasync")) |
| 388 | rw = DDIR_DATASYNC; |
| 389 | else if (!strcmp(act, "trim")) |
| 390 | rw = DDIR_TRIM; |
| 391 | else { |
| 392 | log_err("fio: bad iolog file action: %s\n", |
| 393 | act); |
| 394 | continue; |
| 395 | } |
| 396 | fileno = get_fileno(td, fname); |
| 397 | } else if (r == 2) { |
| 398 | rw = DDIR_INVAL; |
| 399 | if (!strcmp(act, "add")) { |
| 400 | if (td->o.replay_redirect && |
| 401 | get_fileno(td, fname) != -1) { |
| 402 | dprint(FD_FILE, "iolog: ignoring" |
| 403 | " re-add of file %s\n", fname); |
| 404 | } else { |
| 405 | fileno = add_file(td, fname, 0, 1); |
| 406 | file_action = FIO_LOG_ADD_FILE; |
| 407 | } |
| 408 | continue; |
| 409 | } else if (!strcmp(act, "open")) { |
| 410 | fileno = get_fileno(td, fname); |
| 411 | file_action = FIO_LOG_OPEN_FILE; |
| 412 | } else if (!strcmp(act, "close")) { |
| 413 | fileno = get_fileno(td, fname); |
| 414 | file_action = FIO_LOG_CLOSE_FILE; |
| 415 | } else { |
| 416 | log_err("fio: bad iolog file action: %s\n", |
| 417 | act); |
| 418 | continue; |
| 419 | } |
| 420 | } else { |
| 421 | log_err("bad iolog2: %s\n", p); |
| 422 | continue; |
| 423 | } |
| 424 | |
| 425 | if (rw == DDIR_READ) |
| 426 | reads++; |
| 427 | else if (rw == DDIR_WRITE) { |
| 428 | /* |
| 429 | * Don't add a write for ro mode |
| 430 | */ |
| 431 | if (read_only) |
| 432 | continue; |
| 433 | writes++; |
| 434 | } else if (rw == DDIR_WAIT) { |
| 435 | if (td->o.no_stall) |
| 436 | continue; |
| 437 | waits++; |
| 438 | } else if (rw == DDIR_INVAL) { |
| 439 | } else if (!ddir_sync(rw)) { |
| 440 | log_err("bad ddir: %d\n", rw); |
| 441 | continue; |
| 442 | } |
| 443 | |
| 444 | /* |
| 445 | * Make note of file |
| 446 | */ |
| 447 | ipo = malloc(sizeof(*ipo)); |
| 448 | init_ipo(ipo); |
| 449 | ipo->ddir = rw; |
| 450 | if (rw == DDIR_WAIT) { |
| 451 | ipo->delay = offset; |
| 452 | } else { |
| 453 | if (td->o.replay_scale) |
| 454 | ipo->offset = offset / td->o.replay_scale; |
| 455 | else |
| 456 | ipo->offset = offset; |
| 457 | ipo_bytes_align(td->o.replay_align, ipo); |
| 458 | |
| 459 | ipo->len = bytes; |
| 460 | if (rw != DDIR_INVAL && bytes > td->o.max_bs[rw]) |
| 461 | td->o.max_bs[rw] = bytes; |
| 462 | ipo->fileno = fileno; |
| 463 | ipo->file_action = file_action; |
| 464 | td->o.size += bytes; |
| 465 | } |
| 466 | |
| 467 | queue_io_piece(td, ipo); |
| 468 | } |
| 469 | |
| 470 | free(str); |
| 471 | free(act); |
| 472 | free(rfname); |
| 473 | |
| 474 | if (writes && read_only) { |
| 475 | log_err("fio: <%s> skips replay of %d writes due to" |
| 476 | " read-only\n", td->o.name, writes); |
| 477 | writes = 0; |
| 478 | } |
| 479 | |
| 480 | if (!reads && !writes && !waits) |
| 481 | return 1; |
| 482 | else if (reads && !writes) |
| 483 | td->o.td_ddir = TD_DDIR_READ; |
| 484 | else if (!reads && writes) |
| 485 | td->o.td_ddir = TD_DDIR_WRITE; |
| 486 | else |
| 487 | td->o.td_ddir = TD_DDIR_RW; |
| 488 | |
| 489 | return 0; |
| 490 | } |
| 491 | |
| 492 | /* |
| 493 | * open iolog, check version, and call appropriate parser |
| 494 | */ |
| 495 | static int init_iolog_read(struct thread_data *td) |
| 496 | { |
| 497 | char buffer[256], *p; |
| 498 | FILE *f; |
| 499 | int ret; |
| 500 | |
| 501 | f = fopen(td->o.read_iolog_file, "r"); |
| 502 | if (!f) { |
| 503 | perror("fopen read iolog"); |
| 504 | return 1; |
| 505 | } |
| 506 | |
| 507 | p = fgets(buffer, sizeof(buffer), f); |
| 508 | if (!p) { |
| 509 | td_verror(td, errno, "iolog read"); |
| 510 | log_err("fio: unable to read iolog\n"); |
| 511 | fclose(f); |
| 512 | return 1; |
| 513 | } |
| 514 | |
| 515 | /* |
| 516 | * version 2 of the iolog stores a specific string as the |
| 517 | * first line, check for that |
| 518 | */ |
| 519 | if (!strncmp(iolog_ver2, buffer, strlen(iolog_ver2))) |
| 520 | ret = read_iolog2(td, f); |
| 521 | else { |
| 522 | log_err("fio: iolog version 1 is no longer supported\n"); |
| 523 | ret = 1; |
| 524 | } |
| 525 | |
| 526 | fclose(f); |
| 527 | return ret; |
| 528 | } |
| 529 | |
| 530 | /* |
| 531 | * Set up a log for storing io patterns. |
| 532 | */ |
| 533 | static int init_iolog_write(struct thread_data *td) |
| 534 | { |
| 535 | struct fio_file *ff; |
| 536 | FILE *f; |
| 537 | unsigned int i; |
| 538 | |
| 539 | f = fopen(td->o.write_iolog_file, "a"); |
| 540 | if (!f) { |
| 541 | perror("fopen write iolog"); |
| 542 | return 1; |
| 543 | } |
| 544 | |
| 545 | /* |
| 546 | * That's it for writing, setup a log buffer and we're done. |
| 547 | */ |
| 548 | td->iolog_f = f; |
| 549 | td->iolog_buf = malloc(8192); |
| 550 | setvbuf(f, td->iolog_buf, _IOFBF, 8192); |
| 551 | |
| 552 | /* |
| 553 | * write our version line |
| 554 | */ |
| 555 | if (fprintf(f, "%s\n", iolog_ver2) < 0) { |
| 556 | perror("iolog init\n"); |
| 557 | return 1; |
| 558 | } |
| 559 | |
| 560 | /* |
| 561 | * add all known files |
| 562 | */ |
| 563 | for_each_file(td, ff, i) |
| 564 | log_file(td, ff, FIO_LOG_ADD_FILE); |
| 565 | |
| 566 | return 0; |
| 567 | } |
| 568 | |
| 569 | int init_iolog(struct thread_data *td) |
| 570 | { |
| 571 | int ret = 0; |
| 572 | |
| 573 | if (td->o.read_iolog_file) { |
| 574 | int need_swap; |
| 575 | |
| 576 | /* |
| 577 | * Check if it's a blktrace file and load that if possible. |
| 578 | * Otherwise assume it's a normal log file and load that. |
| 579 | */ |
| 580 | if (is_blktrace(td->o.read_iolog_file, &need_swap)) |
| 581 | ret = load_blktrace(td, td->o.read_iolog_file, need_swap); |
| 582 | else |
| 583 | ret = init_iolog_read(td); |
| 584 | } else if (td->o.write_iolog_file) |
| 585 | ret = init_iolog_write(td); |
| 586 | |
| 587 | if (ret) |
| 588 | td_verror(td, EINVAL, "failed initializing iolog"); |
| 589 | |
| 590 | return ret; |
| 591 | } |
| 592 | |
| 593 | void setup_log(struct io_log **log, struct log_params *p, |
| 594 | const char *filename) |
| 595 | { |
| 596 | struct io_log *l; |
| 597 | int i; |
| 598 | struct io_u_plat_entry *entry; |
| 599 | struct flist_head *list; |
| 600 | |
| 601 | l = scalloc(1, sizeof(*l)); |
| 602 | INIT_FLIST_HEAD(&l->io_logs); |
| 603 | l->log_type = p->log_type; |
| 604 | l->log_offset = p->log_offset; |
| 605 | l->log_gz = p->log_gz; |
| 606 | l->log_gz_store = p->log_gz_store; |
| 607 | l->avg_msec = p->avg_msec; |
| 608 | l->hist_msec = p->hist_msec; |
| 609 | l->hist_coarseness = p->hist_coarseness; |
| 610 | l->filename = strdup(filename); |
| 611 | l->td = p->td; |
| 612 | |
| 613 | /* Initialize histogram lists for each r/w direction, |
| 614 | * with initial io_u_plat of all zeros: |
| 615 | */ |
| 616 | for (i = 0; i < DDIR_RWDIR_CNT; i++) { |
| 617 | list = &l->hist_window[i].list; |
| 618 | INIT_FLIST_HEAD(list); |
| 619 | entry = calloc(1, sizeof(struct io_u_plat_entry)); |
| 620 | flist_add(&entry->list, list); |
| 621 | } |
| 622 | |
| 623 | if (l->td && l->td->o.io_submit_mode != IO_MODE_OFFLOAD) { |
| 624 | struct io_logs *p; |
| 625 | |
| 626 | p = calloc(1, sizeof(*l->pending)); |
| 627 | p->max_samples = DEF_LOG_ENTRIES; |
| 628 | p->log = calloc(p->max_samples, log_entry_sz(l)); |
| 629 | l->pending = p; |
| 630 | } |
| 631 | |
| 632 | if (l->log_offset) |
| 633 | l->log_ddir_mask = LOG_OFFSET_SAMPLE_BIT; |
| 634 | |
| 635 | INIT_FLIST_HEAD(&l->chunk_list); |
| 636 | |
| 637 | if (l->log_gz && !p->td) |
| 638 | l->log_gz = 0; |
| 639 | else if (l->log_gz || l->log_gz_store) { |
| 640 | mutex_init_pshared(&l->chunk_lock); |
| 641 | mutex_init_pshared(&l->deferred_free_lock); |
| 642 | p->td->flags |= TD_F_COMPRESS_LOG; |
| 643 | } |
| 644 | |
| 645 | *log = l; |
| 646 | } |
| 647 | |
| 648 | #ifdef CONFIG_SETVBUF |
| 649 | static void *set_file_buffer(FILE *f) |
| 650 | { |
| 651 | size_t size = 1048576; |
| 652 | void *buf; |
| 653 | |
| 654 | buf = malloc(size); |
| 655 | setvbuf(f, buf, _IOFBF, size); |
| 656 | return buf; |
| 657 | } |
| 658 | |
| 659 | static void clear_file_buffer(void *buf) |
| 660 | { |
| 661 | free(buf); |
| 662 | } |
| 663 | #else |
| 664 | static void *set_file_buffer(FILE *f) |
| 665 | { |
| 666 | return NULL; |
| 667 | } |
| 668 | |
| 669 | static void clear_file_buffer(void *buf) |
| 670 | { |
| 671 | } |
| 672 | #endif |
| 673 | |
| 674 | void free_log(struct io_log *log) |
| 675 | { |
| 676 | while (!flist_empty(&log->io_logs)) { |
| 677 | struct io_logs *cur_log; |
| 678 | |
| 679 | cur_log = flist_first_entry(&log->io_logs, struct io_logs, list); |
| 680 | flist_del_init(&cur_log->list); |
| 681 | free(cur_log->log); |
| 682 | sfree(cur_log); |
| 683 | } |
| 684 | |
| 685 | if (log->pending) { |
| 686 | free(log->pending->log); |
| 687 | free(log->pending); |
| 688 | log->pending = NULL; |
| 689 | } |
| 690 | |
| 691 | free(log->pending); |
| 692 | free(log->filename); |
| 693 | sfree(log); |
| 694 | } |
| 695 | |
| 696 | uint64_t hist_sum(int j, int stride, uint64_t *io_u_plat, |
| 697 | uint64_t *io_u_plat_last) |
| 698 | { |
| 699 | uint64_t sum; |
| 700 | int k; |
| 701 | |
| 702 | if (io_u_plat_last) { |
| 703 | for (k = sum = 0; k < stride; k++) |
| 704 | sum += io_u_plat[j + k] - io_u_plat_last[j + k]; |
| 705 | } else { |
| 706 | for (k = sum = 0; k < stride; k++) |
| 707 | sum += io_u_plat[j + k]; |
| 708 | } |
| 709 | |
| 710 | return sum; |
| 711 | } |
| 712 | |
| 713 | static void flush_hist_samples(FILE *f, int hist_coarseness, void *samples, |
| 714 | uint64_t sample_size) |
| 715 | { |
| 716 | struct io_sample *s; |
| 717 | int log_offset; |
| 718 | uint64_t i, j, nr_samples; |
| 719 | struct io_u_plat_entry *entry, *entry_before; |
| 720 | uint64_t *io_u_plat; |
| 721 | uint64_t *io_u_plat_before; |
| 722 | |
| 723 | int stride = 1 << hist_coarseness; |
| 724 | |
| 725 | if (!sample_size) |
| 726 | return; |
| 727 | |
| 728 | s = __get_sample(samples, 0, 0); |
| 729 | log_offset = (s->__ddir & LOG_OFFSET_SAMPLE_BIT) != 0; |
| 730 | |
| 731 | nr_samples = sample_size / __log_entry_sz(log_offset); |
| 732 | |
| 733 | for (i = 0; i < nr_samples; i++) { |
| 734 | s = __get_sample(samples, log_offset, i); |
| 735 | |
| 736 | entry = s->data.plat_entry; |
| 737 | io_u_plat = entry->io_u_plat; |
| 738 | |
| 739 | entry_before = flist_first_entry(&entry->list, struct io_u_plat_entry, list); |
| 740 | io_u_plat_before = entry_before->io_u_plat; |
| 741 | |
| 742 | fprintf(f, "%lu, %u, %u, ", (unsigned long) s->time, |
| 743 | io_sample_ddir(s), s->bs); |
| 744 | for (j = 0; j < FIO_IO_U_PLAT_NR - stride; j += stride) { |
| 745 | fprintf(f, "%llu, ", (unsigned long long) |
| 746 | hist_sum(j, stride, io_u_plat, io_u_plat_before)); |
| 747 | } |
| 748 | fprintf(f, "%llu\n", (unsigned long long) |
| 749 | hist_sum(FIO_IO_U_PLAT_NR - stride, stride, io_u_plat, |
| 750 | io_u_plat_before)); |
| 751 | |
| 752 | flist_del(&entry_before->list); |
| 753 | free(entry_before); |
| 754 | } |
| 755 | } |
| 756 | |
| 757 | void flush_samples(FILE *f, void *samples, uint64_t sample_size) |
| 758 | { |
| 759 | struct io_sample *s; |
| 760 | int log_offset; |
| 761 | uint64_t i, nr_samples; |
| 762 | |
| 763 | if (!sample_size) |
| 764 | return; |
| 765 | |
| 766 | s = __get_sample(samples, 0, 0); |
| 767 | log_offset = (s->__ddir & LOG_OFFSET_SAMPLE_BIT) != 0; |
| 768 | |
| 769 | nr_samples = sample_size / __log_entry_sz(log_offset); |
| 770 | |
| 771 | for (i = 0; i < nr_samples; i++) { |
| 772 | s = __get_sample(samples, log_offset, i); |
| 773 | |
| 774 | if (!log_offset) { |
| 775 | fprintf(f, "%lu, %" PRId64 ", %u, %u\n", |
| 776 | (unsigned long) s->time, |
| 777 | s->data.val, |
| 778 | io_sample_ddir(s), s->bs); |
| 779 | } else { |
| 780 | struct io_sample_offset *so = (void *) s; |
| 781 | |
| 782 | fprintf(f, "%lu, %" PRId64 ", %u, %u, %llu\n", |
| 783 | (unsigned long) s->time, |
| 784 | s->data.val, |
| 785 | io_sample_ddir(s), s->bs, |
| 786 | (unsigned long long) so->offset); |
| 787 | } |
| 788 | } |
| 789 | } |
| 790 | |
| 791 | #ifdef CONFIG_ZLIB |
| 792 | |
| 793 | struct iolog_flush_data { |
| 794 | struct workqueue_work work; |
| 795 | struct io_log *log; |
| 796 | void *samples; |
| 797 | uint32_t nr_samples; |
| 798 | bool free; |
| 799 | }; |
| 800 | |
| 801 | #define GZ_CHUNK 131072 |
| 802 | |
| 803 | static struct iolog_compress *get_new_chunk(unsigned int seq) |
| 804 | { |
| 805 | struct iolog_compress *c; |
| 806 | |
| 807 | c = malloc(sizeof(*c)); |
| 808 | INIT_FLIST_HEAD(&c->list); |
| 809 | c->buf = malloc(GZ_CHUNK); |
| 810 | c->len = 0; |
| 811 | c->seq = seq; |
| 812 | return c; |
| 813 | } |
| 814 | |
| 815 | static void free_chunk(struct iolog_compress *ic) |
| 816 | { |
| 817 | free(ic->buf); |
| 818 | free(ic); |
| 819 | } |
| 820 | |
| 821 | static int z_stream_init(z_stream *stream, int gz_hdr) |
| 822 | { |
| 823 | int wbits = 15; |
| 824 | |
| 825 | memset(stream, 0, sizeof(*stream)); |
| 826 | stream->zalloc = Z_NULL; |
| 827 | stream->zfree = Z_NULL; |
| 828 | stream->opaque = Z_NULL; |
| 829 | stream->next_in = Z_NULL; |
| 830 | |
| 831 | /* |
| 832 | * zlib magic - add 32 for auto-detection of gz header or not, |
| 833 | * if we decide to store files in a gzip friendly format. |
| 834 | */ |
| 835 | if (gz_hdr) |
| 836 | wbits += 32; |
| 837 | |
| 838 | if (inflateInit2(stream, wbits) != Z_OK) |
| 839 | return 1; |
| 840 | |
| 841 | return 0; |
| 842 | } |
| 843 | |
| 844 | struct inflate_chunk_iter { |
| 845 | unsigned int seq; |
| 846 | int err; |
| 847 | void *buf; |
| 848 | size_t buf_size; |
| 849 | size_t buf_used; |
| 850 | size_t chunk_sz; |
| 851 | }; |
| 852 | |
| 853 | static void finish_chunk(z_stream *stream, FILE *f, |
| 854 | struct inflate_chunk_iter *iter) |
| 855 | { |
| 856 | int ret; |
| 857 | |
| 858 | ret = inflateEnd(stream); |
| 859 | if (ret != Z_OK) |
| 860 | log_err("fio: failed to end log inflation seq %d (%d)\n", |
| 861 | iter->seq, ret); |
| 862 | |
| 863 | flush_samples(f, iter->buf, iter->buf_used); |
| 864 | free(iter->buf); |
| 865 | iter->buf = NULL; |
| 866 | iter->buf_size = iter->buf_used = 0; |
| 867 | } |
| 868 | |
| 869 | /* |
| 870 | * Iterative chunk inflation. Handles cases where we cross into a new |
| 871 | * sequence, doing flush finish of previous chunk if needed. |
| 872 | */ |
| 873 | static size_t inflate_chunk(struct iolog_compress *ic, int gz_hdr, FILE *f, |
| 874 | z_stream *stream, struct inflate_chunk_iter *iter) |
| 875 | { |
| 876 | size_t ret; |
| 877 | |
| 878 | dprint(FD_COMPRESS, "inflate chunk size=%lu, seq=%u\n", |
| 879 | (unsigned long) ic->len, ic->seq); |
| 880 | |
| 881 | if (ic->seq != iter->seq) { |
| 882 | if (iter->seq) |
| 883 | finish_chunk(stream, f, iter); |
| 884 | |
| 885 | z_stream_init(stream, gz_hdr); |
| 886 | iter->seq = ic->seq; |
| 887 | } |
| 888 | |
| 889 | stream->avail_in = ic->len; |
| 890 | stream->next_in = ic->buf; |
| 891 | |
| 892 | if (!iter->buf_size) { |
| 893 | iter->buf_size = iter->chunk_sz; |
| 894 | iter->buf = malloc(iter->buf_size); |
| 895 | } |
| 896 | |
| 897 | while (stream->avail_in) { |
| 898 | size_t this_out = iter->buf_size - iter->buf_used; |
| 899 | int err; |
| 900 | |
| 901 | stream->avail_out = this_out; |
| 902 | stream->next_out = iter->buf + iter->buf_used; |
| 903 | |
| 904 | err = inflate(stream, Z_NO_FLUSH); |
| 905 | if (err < 0) { |
| 906 | log_err("fio: failed inflating log: %d\n", err); |
| 907 | iter->err = err; |
| 908 | break; |
| 909 | } |
| 910 | |
| 911 | iter->buf_used += this_out - stream->avail_out; |
| 912 | |
| 913 | if (!stream->avail_out) { |
| 914 | iter->buf_size += iter->chunk_sz; |
| 915 | iter->buf = realloc(iter->buf, iter->buf_size); |
| 916 | continue; |
| 917 | } |
| 918 | |
| 919 | if (err == Z_STREAM_END) |
| 920 | break; |
| 921 | } |
| 922 | |
| 923 | ret = (void *) stream->next_in - ic->buf; |
| 924 | |
| 925 | dprint(FD_COMPRESS, "inflated to size=%lu\n", (unsigned long) iter->buf_size); |
| 926 | |
| 927 | return ret; |
| 928 | } |
| 929 | |
| 930 | /* |
| 931 | * Inflate stored compressed chunks, or write them directly to the log |
| 932 | * file if so instructed. |
| 933 | */ |
| 934 | static int inflate_gz_chunks(struct io_log *log, FILE *f) |
| 935 | { |
| 936 | struct inflate_chunk_iter iter = { .chunk_sz = log->log_gz, }; |
| 937 | z_stream stream; |
| 938 | |
| 939 | while (!flist_empty(&log->chunk_list)) { |
| 940 | struct iolog_compress *ic; |
| 941 | |
| 942 | ic = flist_first_entry(&log->chunk_list, struct iolog_compress, list); |
| 943 | flist_del(&ic->list); |
| 944 | |
| 945 | if (log->log_gz_store) { |
| 946 | size_t ret; |
| 947 | |
| 948 | dprint(FD_COMPRESS, "log write chunk size=%lu, " |
| 949 | "seq=%u\n", (unsigned long) ic->len, ic->seq); |
| 950 | |
| 951 | ret = fwrite(ic->buf, ic->len, 1, f); |
| 952 | if (ret != 1 || ferror(f)) { |
| 953 | iter.err = errno; |
| 954 | log_err("fio: error writing compressed log\n"); |
| 955 | } |
| 956 | } else |
| 957 | inflate_chunk(ic, log->log_gz_store, f, &stream, &iter); |
| 958 | |
| 959 | free_chunk(ic); |
| 960 | } |
| 961 | |
| 962 | if (iter.seq) { |
| 963 | finish_chunk(&stream, f, &iter); |
| 964 | free(iter.buf); |
| 965 | } |
| 966 | |
| 967 | return iter.err; |
| 968 | } |
| 969 | |
| 970 | /* |
| 971 | * Open compressed log file and decompress the stored chunks and |
| 972 | * write them to stdout. The chunks are stored sequentially in the |
| 973 | * file, so we iterate over them and do them one-by-one. |
| 974 | */ |
| 975 | int iolog_file_inflate(const char *file) |
| 976 | { |
| 977 | struct inflate_chunk_iter iter = { .chunk_sz = 64 * 1024 * 1024, }; |
| 978 | struct iolog_compress ic; |
| 979 | z_stream stream; |
| 980 | struct stat sb; |
| 981 | ssize_t ret; |
| 982 | size_t total; |
| 983 | void *buf; |
| 984 | FILE *f; |
| 985 | |
| 986 | f = fopen(file, "r"); |
| 987 | if (!f) { |
| 988 | perror("fopen"); |
| 989 | return 1; |
| 990 | } |
| 991 | |
| 992 | if (stat(file, &sb) < 0) { |
| 993 | fclose(f); |
| 994 | perror("stat"); |
| 995 | return 1; |
| 996 | } |
| 997 | |
| 998 | ic.buf = buf = malloc(sb.st_size); |
| 999 | ic.len = sb.st_size; |
| 1000 | ic.seq = 1; |
| 1001 | |
| 1002 | ret = fread(ic.buf, ic.len, 1, f); |
| 1003 | if (ret < 0) { |
| 1004 | perror("fread"); |
| 1005 | fclose(f); |
| 1006 | free(buf); |
| 1007 | return 1; |
| 1008 | } else if (ret != 1) { |
| 1009 | log_err("fio: short read on reading log\n"); |
| 1010 | fclose(f); |
| 1011 | free(buf); |
| 1012 | return 1; |
| 1013 | } |
| 1014 | |
| 1015 | fclose(f); |
| 1016 | |
| 1017 | /* |
| 1018 | * Each chunk will return Z_STREAM_END. We don't know how many |
| 1019 | * chunks are in the file, so we just keep looping and incrementing |
| 1020 | * the sequence number until we have consumed the whole compressed |
| 1021 | * file. |
| 1022 | */ |
| 1023 | total = ic.len; |
| 1024 | do { |
| 1025 | size_t iret; |
| 1026 | |
| 1027 | iret = inflate_chunk(&ic, 1, stdout, &stream, &iter); |
| 1028 | total -= iret; |
| 1029 | if (!total) |
| 1030 | break; |
| 1031 | if (iter.err) |
| 1032 | break; |
| 1033 | |
| 1034 | ic.seq++; |
| 1035 | ic.len -= iret; |
| 1036 | ic.buf += iret; |
| 1037 | } while (1); |
| 1038 | |
| 1039 | if (iter.seq) { |
| 1040 | finish_chunk(&stream, stdout, &iter); |
| 1041 | free(iter.buf); |
| 1042 | } |
| 1043 | |
| 1044 | free(buf); |
| 1045 | return iter.err; |
| 1046 | } |
| 1047 | |
| 1048 | #else |
| 1049 | |
| 1050 | static int inflate_gz_chunks(struct io_log *log, FILE *f) |
| 1051 | { |
| 1052 | return 0; |
| 1053 | } |
| 1054 | |
| 1055 | int iolog_file_inflate(const char *file) |
| 1056 | { |
| 1057 | log_err("fio: log inflation not possible without zlib\n"); |
| 1058 | return 1; |
| 1059 | } |
| 1060 | |
| 1061 | #endif |
| 1062 | |
| 1063 | void flush_log(struct io_log *log, bool do_append) |
| 1064 | { |
| 1065 | void *buf; |
| 1066 | FILE *f; |
| 1067 | |
| 1068 | if (!do_append) |
| 1069 | f = fopen(log->filename, "w"); |
| 1070 | else |
| 1071 | f = fopen(log->filename, "a"); |
| 1072 | if (!f) { |
| 1073 | perror("fopen log"); |
| 1074 | return; |
| 1075 | } |
| 1076 | |
| 1077 | buf = set_file_buffer(f); |
| 1078 | |
| 1079 | inflate_gz_chunks(log, f); |
| 1080 | |
| 1081 | while (!flist_empty(&log->io_logs)) { |
| 1082 | struct io_logs *cur_log; |
| 1083 | |
| 1084 | cur_log = flist_first_entry(&log->io_logs, struct io_logs, list); |
| 1085 | flist_del_init(&cur_log->list); |
| 1086 | |
| 1087 | if (log->td && log == log->td->clat_hist_log) |
| 1088 | flush_hist_samples(f, log->hist_coarseness, cur_log->log, |
| 1089 | log_sample_sz(log, cur_log)); |
| 1090 | else |
| 1091 | flush_samples(f, cur_log->log, log_sample_sz(log, cur_log)); |
| 1092 | |
| 1093 | sfree(cur_log); |
| 1094 | } |
| 1095 | |
| 1096 | fclose(f); |
| 1097 | clear_file_buffer(buf); |
| 1098 | } |
| 1099 | |
| 1100 | static int finish_log(struct thread_data *td, struct io_log *log, int trylock) |
| 1101 | { |
| 1102 | if (td->flags & TD_F_COMPRESS_LOG) |
| 1103 | iolog_flush(log); |
| 1104 | |
| 1105 | if (trylock) { |
| 1106 | if (fio_trylock_file(log->filename)) |
| 1107 | return 1; |
| 1108 | } else |
| 1109 | fio_lock_file(log->filename); |
| 1110 | |
| 1111 | if (td->client_type == FIO_CLIENT_TYPE_GUI || is_backend) |
| 1112 | fio_send_iolog(td, log, log->filename); |
| 1113 | else |
| 1114 | flush_log(log, !td->o.per_job_logs); |
| 1115 | |
| 1116 | fio_unlock_file(log->filename); |
| 1117 | free_log(log); |
| 1118 | return 0; |
| 1119 | } |
| 1120 | |
| 1121 | size_t log_chunk_sizes(struct io_log *log) |
| 1122 | { |
| 1123 | struct flist_head *entry; |
| 1124 | size_t ret; |
| 1125 | |
| 1126 | if (flist_empty(&log->chunk_list)) |
| 1127 | return 0; |
| 1128 | |
| 1129 | ret = 0; |
| 1130 | pthread_mutex_lock(&log->chunk_lock); |
| 1131 | flist_for_each(entry, &log->chunk_list) { |
| 1132 | struct iolog_compress *c; |
| 1133 | |
| 1134 | c = flist_entry(entry, struct iolog_compress, list); |
| 1135 | ret += c->len; |
| 1136 | } |
| 1137 | pthread_mutex_unlock(&log->chunk_lock); |
| 1138 | return ret; |
| 1139 | } |
| 1140 | |
| 1141 | #ifdef CONFIG_ZLIB |
| 1142 | |
| 1143 | static void iolog_put_deferred(struct io_log *log, void *ptr) |
| 1144 | { |
| 1145 | if (!ptr) |
| 1146 | return; |
| 1147 | |
| 1148 | pthread_mutex_lock(&log->deferred_free_lock); |
| 1149 | if (log->deferred < IOLOG_MAX_DEFER) { |
| 1150 | log->deferred_items[log->deferred] = ptr; |
| 1151 | log->deferred++; |
| 1152 | } else if (!fio_did_warn(FIO_WARN_IOLOG_DROP)) |
| 1153 | log_err("fio: had to drop log entry free\n"); |
| 1154 | pthread_mutex_unlock(&log->deferred_free_lock); |
| 1155 | } |
| 1156 | |
| 1157 | static void iolog_free_deferred(struct io_log *log) |
| 1158 | { |
| 1159 | int i; |
| 1160 | |
| 1161 | if (!log->deferred) |
| 1162 | return; |
| 1163 | |
| 1164 | pthread_mutex_lock(&log->deferred_free_lock); |
| 1165 | |
| 1166 | for (i = 0; i < log->deferred; i++) { |
| 1167 | free(log->deferred_items[i]); |
| 1168 | log->deferred_items[i] = NULL; |
| 1169 | } |
| 1170 | |
| 1171 | log->deferred = 0; |
| 1172 | pthread_mutex_unlock(&log->deferred_free_lock); |
| 1173 | } |
| 1174 | |
| 1175 | static int gz_work(struct iolog_flush_data *data) |
| 1176 | { |
| 1177 | struct iolog_compress *c = NULL; |
| 1178 | struct flist_head list; |
| 1179 | unsigned int seq; |
| 1180 | z_stream stream; |
| 1181 | size_t total = 0; |
| 1182 | int ret; |
| 1183 | |
| 1184 | INIT_FLIST_HEAD(&list); |
| 1185 | |
| 1186 | memset(&stream, 0, sizeof(stream)); |
| 1187 | stream.zalloc = Z_NULL; |
| 1188 | stream.zfree = Z_NULL; |
| 1189 | stream.opaque = Z_NULL; |
| 1190 | |
| 1191 | ret = deflateInit(&stream, Z_DEFAULT_COMPRESSION); |
| 1192 | if (ret != Z_OK) { |
| 1193 | log_err("fio: failed to init gz stream\n"); |
| 1194 | goto err; |
| 1195 | } |
| 1196 | |
| 1197 | seq = ++data->log->chunk_seq; |
| 1198 | |
| 1199 | stream.next_in = (void *) data->samples; |
| 1200 | stream.avail_in = data->nr_samples * log_entry_sz(data->log); |
| 1201 | |
| 1202 | dprint(FD_COMPRESS, "deflate input size=%lu, seq=%u, log=%s\n", |
| 1203 | (unsigned long) stream.avail_in, seq, |
| 1204 | data->log->filename); |
| 1205 | do { |
| 1206 | if (c) |
| 1207 | dprint(FD_COMPRESS, "seq=%d, chunk=%lu\n", seq, |
| 1208 | (unsigned long) c->len); |
| 1209 | c = get_new_chunk(seq); |
| 1210 | stream.avail_out = GZ_CHUNK; |
| 1211 | stream.next_out = c->buf; |
| 1212 | ret = deflate(&stream, Z_NO_FLUSH); |
| 1213 | if (ret < 0) { |
| 1214 | log_err("fio: deflate log (%d)\n", ret); |
| 1215 | free_chunk(c); |
| 1216 | goto err; |
| 1217 | } |
| 1218 | |
| 1219 | c->len = GZ_CHUNK - stream.avail_out; |
| 1220 | flist_add_tail(&c->list, &list); |
| 1221 | total += c->len; |
| 1222 | } while (stream.avail_in); |
| 1223 | |
| 1224 | stream.next_out = c->buf + c->len; |
| 1225 | stream.avail_out = GZ_CHUNK - c->len; |
| 1226 | |
| 1227 | ret = deflate(&stream, Z_FINISH); |
| 1228 | if (ret < 0) { |
| 1229 | /* |
| 1230 | * Z_BUF_ERROR is special, it just means we need more |
| 1231 | * output space. We'll handle that below. Treat any other |
| 1232 | * error as fatal. |
| 1233 | */ |
| 1234 | if (ret != Z_BUF_ERROR) { |
| 1235 | log_err("fio: deflate log (%d)\n", ret); |
| 1236 | flist_del(&c->list); |
| 1237 | free_chunk(c); |
| 1238 | goto err; |
| 1239 | } |
| 1240 | } |
| 1241 | |
| 1242 | total -= c->len; |
| 1243 | c->len = GZ_CHUNK - stream.avail_out; |
| 1244 | total += c->len; |
| 1245 | dprint(FD_COMPRESS, "seq=%d, chunk=%lu\n", seq, (unsigned long) c->len); |
| 1246 | |
| 1247 | if (ret != Z_STREAM_END) { |
| 1248 | do { |
| 1249 | c = get_new_chunk(seq); |
| 1250 | stream.avail_out = GZ_CHUNK; |
| 1251 | stream.next_out = c->buf; |
| 1252 | ret = deflate(&stream, Z_FINISH); |
| 1253 | c->len = GZ_CHUNK - stream.avail_out; |
| 1254 | total += c->len; |
| 1255 | flist_add_tail(&c->list, &list); |
| 1256 | dprint(FD_COMPRESS, "seq=%d, chunk=%lu\n", seq, |
| 1257 | (unsigned long) c->len); |
| 1258 | } while (ret != Z_STREAM_END); |
| 1259 | } |
| 1260 | |
| 1261 | dprint(FD_COMPRESS, "deflated to size=%lu\n", (unsigned long) total); |
| 1262 | |
| 1263 | ret = deflateEnd(&stream); |
| 1264 | if (ret != Z_OK) |
| 1265 | log_err("fio: deflateEnd %d\n", ret); |
| 1266 | |
| 1267 | iolog_put_deferred(data->log, data->samples); |
| 1268 | |
| 1269 | if (!flist_empty(&list)) { |
| 1270 | pthread_mutex_lock(&data->log->chunk_lock); |
| 1271 | flist_splice_tail(&list, &data->log->chunk_list); |
| 1272 | pthread_mutex_unlock(&data->log->chunk_lock); |
| 1273 | } |
| 1274 | |
| 1275 | ret = 0; |
| 1276 | done: |
| 1277 | if (data->free) |
| 1278 | sfree(data); |
| 1279 | return ret; |
| 1280 | err: |
| 1281 | while (!flist_empty(&list)) { |
| 1282 | c = flist_first_entry(list.next, struct iolog_compress, list); |
| 1283 | flist_del(&c->list); |
| 1284 | free_chunk(c); |
| 1285 | } |
| 1286 | ret = 1; |
| 1287 | goto done; |
| 1288 | } |
| 1289 | |
| 1290 | /* |
| 1291 | * Invoked from our compress helper thread, when logging would have exceeded |
| 1292 | * the specified memory limitation. Compresses the previously stored |
| 1293 | * entries. |
| 1294 | */ |
| 1295 | static int gz_work_async(struct submit_worker *sw, struct workqueue_work *work) |
| 1296 | { |
| 1297 | return gz_work(container_of(work, struct iolog_flush_data, work)); |
| 1298 | } |
| 1299 | |
| 1300 | static int gz_init_worker(struct submit_worker *sw) |
| 1301 | { |
| 1302 | struct thread_data *td = sw->wq->td; |
| 1303 | |
| 1304 | if (!fio_option_is_set(&td->o, log_gz_cpumask)) |
| 1305 | return 0; |
| 1306 | |
| 1307 | if (fio_setaffinity(gettid(), td->o.log_gz_cpumask) == -1) { |
| 1308 | log_err("gz: failed to set CPU affinity\n"); |
| 1309 | return 1; |
| 1310 | } |
| 1311 | |
| 1312 | return 0; |
| 1313 | } |
| 1314 | |
| 1315 | static struct workqueue_ops log_compress_wq_ops = { |
| 1316 | .fn = gz_work_async, |
| 1317 | .init_worker_fn = gz_init_worker, |
| 1318 | .nice = 1, |
| 1319 | }; |
| 1320 | |
| 1321 | int iolog_compress_init(struct thread_data *td, struct sk_out *sk_out) |
| 1322 | { |
| 1323 | if (!(td->flags & TD_F_COMPRESS_LOG)) |
| 1324 | return 0; |
| 1325 | |
| 1326 | workqueue_init(td, &td->log_compress_wq, &log_compress_wq_ops, 1, sk_out); |
| 1327 | return 0; |
| 1328 | } |
| 1329 | |
| 1330 | void iolog_compress_exit(struct thread_data *td) |
| 1331 | { |
| 1332 | if (!(td->flags & TD_F_COMPRESS_LOG)) |
| 1333 | return; |
| 1334 | |
| 1335 | workqueue_exit(&td->log_compress_wq); |
| 1336 | } |
| 1337 | |
| 1338 | /* |
| 1339 | * Queue work item to compress the existing log entries. We reset the |
| 1340 | * current log to a small size, and reference the existing log in the |
| 1341 | * data that we queue for compression. Once compression has been done, |
| 1342 | * this old log is freed. If called with finish == true, will not return |
| 1343 | * until the log compression has completed, and will flush all previous |
| 1344 | * logs too |
| 1345 | */ |
| 1346 | static int iolog_flush(struct io_log *log) |
| 1347 | { |
| 1348 | struct iolog_flush_data *data; |
| 1349 | |
| 1350 | data = malloc(sizeof(*data)); |
| 1351 | if (!data) |
| 1352 | return 1; |
| 1353 | |
| 1354 | data->log = log; |
| 1355 | data->free = false; |
| 1356 | |
| 1357 | while (!flist_empty(&log->io_logs)) { |
| 1358 | struct io_logs *cur_log; |
| 1359 | |
| 1360 | cur_log = flist_first_entry(&log->io_logs, struct io_logs, list); |
| 1361 | flist_del_init(&cur_log->list); |
| 1362 | |
| 1363 | data->samples = cur_log->log; |
| 1364 | data->nr_samples = cur_log->nr_samples; |
| 1365 | |
| 1366 | sfree(cur_log); |
| 1367 | |
| 1368 | gz_work(data); |
| 1369 | } |
| 1370 | |
| 1371 | free(data); |
| 1372 | return 0; |
| 1373 | } |
| 1374 | |
| 1375 | int iolog_cur_flush(struct io_log *log, struct io_logs *cur_log) |
| 1376 | { |
| 1377 | struct iolog_flush_data *data; |
| 1378 | |
| 1379 | data = smalloc(sizeof(*data)); |
| 1380 | if (!data) |
| 1381 | return 1; |
| 1382 | |
| 1383 | data->log = log; |
| 1384 | |
| 1385 | data->samples = cur_log->log; |
| 1386 | data->nr_samples = cur_log->nr_samples; |
| 1387 | data->free = true; |
| 1388 | |
| 1389 | cur_log->nr_samples = cur_log->max_samples = 0; |
| 1390 | cur_log->log = NULL; |
| 1391 | |
| 1392 | workqueue_enqueue(&log->td->log_compress_wq, &data->work); |
| 1393 | |
| 1394 | iolog_free_deferred(log); |
| 1395 | |
| 1396 | return 0; |
| 1397 | } |
| 1398 | #else |
| 1399 | |
| 1400 | static int iolog_flush(struct io_log *log) |
| 1401 | { |
| 1402 | return 1; |
| 1403 | } |
| 1404 | |
| 1405 | int iolog_cur_flush(struct io_log *log, struct io_logs *cur_log) |
| 1406 | { |
| 1407 | return 1; |
| 1408 | } |
| 1409 | |
| 1410 | int iolog_compress_init(struct thread_data *td, struct sk_out *sk_out) |
| 1411 | { |
| 1412 | return 0; |
| 1413 | } |
| 1414 | |
| 1415 | void iolog_compress_exit(struct thread_data *td) |
| 1416 | { |
| 1417 | } |
| 1418 | |
| 1419 | #endif |
| 1420 | |
| 1421 | struct io_logs *iolog_cur_log(struct io_log *log) |
| 1422 | { |
| 1423 | if (flist_empty(&log->io_logs)) |
| 1424 | return NULL; |
| 1425 | |
| 1426 | return flist_last_entry(&log->io_logs, struct io_logs, list); |
| 1427 | } |
| 1428 | |
| 1429 | uint64_t iolog_nr_samples(struct io_log *iolog) |
| 1430 | { |
| 1431 | struct flist_head *entry; |
| 1432 | uint64_t ret = 0; |
| 1433 | |
| 1434 | flist_for_each(entry, &iolog->io_logs) { |
| 1435 | struct io_logs *cur_log; |
| 1436 | |
| 1437 | cur_log = flist_entry(entry, struct io_logs, list); |
| 1438 | ret += cur_log->nr_samples; |
| 1439 | } |
| 1440 | |
| 1441 | return ret; |
| 1442 | } |
| 1443 | |
| 1444 | static int __write_log(struct thread_data *td, struct io_log *log, int try) |
| 1445 | { |
| 1446 | if (log) |
| 1447 | return finish_log(td, log, try); |
| 1448 | |
| 1449 | return 0; |
| 1450 | } |
| 1451 | |
| 1452 | static int write_iops_log(struct thread_data *td, int try, bool unit_log) |
| 1453 | { |
| 1454 | int ret; |
| 1455 | |
| 1456 | if (per_unit_log(td->iops_log) != unit_log) |
| 1457 | return 0; |
| 1458 | |
| 1459 | ret = __write_log(td, td->iops_log, try); |
| 1460 | if (!ret) |
| 1461 | td->iops_log = NULL; |
| 1462 | |
| 1463 | return ret; |
| 1464 | } |
| 1465 | |
| 1466 | static int write_slat_log(struct thread_data *td, int try, bool unit_log) |
| 1467 | { |
| 1468 | int ret; |
| 1469 | |
| 1470 | if (!unit_log) |
| 1471 | return 0; |
| 1472 | |
| 1473 | ret = __write_log(td, td->slat_log, try); |
| 1474 | if (!ret) |
| 1475 | td->slat_log = NULL; |
| 1476 | |
| 1477 | return ret; |
| 1478 | } |
| 1479 | |
| 1480 | static int write_clat_log(struct thread_data *td, int try, bool unit_log) |
| 1481 | { |
| 1482 | int ret; |
| 1483 | |
| 1484 | if (!unit_log) |
| 1485 | return 0; |
| 1486 | |
| 1487 | ret = __write_log(td, td->clat_log, try); |
| 1488 | if (!ret) |
| 1489 | td->clat_log = NULL; |
| 1490 | |
| 1491 | return ret; |
| 1492 | } |
| 1493 | |
| 1494 | static int write_clat_hist_log(struct thread_data *td, int try, bool unit_log) |
| 1495 | { |
| 1496 | int ret; |
| 1497 | |
| 1498 | if (!unit_log) |
| 1499 | return 0; |
| 1500 | |
| 1501 | ret = __write_log(td, td->clat_hist_log, try); |
| 1502 | if (!ret) |
| 1503 | td->clat_hist_log = NULL; |
| 1504 | |
| 1505 | return ret; |
| 1506 | } |
| 1507 | |
| 1508 | static int write_lat_log(struct thread_data *td, int try, bool unit_log) |
| 1509 | { |
| 1510 | int ret; |
| 1511 | |
| 1512 | if (!unit_log) |
| 1513 | return 0; |
| 1514 | |
| 1515 | ret = __write_log(td, td->lat_log, try); |
| 1516 | if (!ret) |
| 1517 | td->lat_log = NULL; |
| 1518 | |
| 1519 | return ret; |
| 1520 | } |
| 1521 | |
| 1522 | static int write_bandw_log(struct thread_data *td, int try, bool unit_log) |
| 1523 | { |
| 1524 | int ret; |
| 1525 | |
| 1526 | if (per_unit_log(td->bw_log) != unit_log) |
| 1527 | return 0; |
| 1528 | |
| 1529 | ret = __write_log(td, td->bw_log, try); |
| 1530 | if (!ret) |
| 1531 | td->bw_log = NULL; |
| 1532 | |
| 1533 | return ret; |
| 1534 | } |
| 1535 | |
| 1536 | enum { |
| 1537 | BW_LOG_MASK = 1, |
| 1538 | LAT_LOG_MASK = 2, |
| 1539 | SLAT_LOG_MASK = 4, |
| 1540 | CLAT_LOG_MASK = 8, |
| 1541 | IOPS_LOG_MASK = 16, |
| 1542 | CLAT_HIST_LOG_MASK = 32, |
| 1543 | |
| 1544 | ALL_LOG_NR = 6, |
| 1545 | }; |
| 1546 | |
| 1547 | struct log_type { |
| 1548 | unsigned int mask; |
| 1549 | int (*fn)(struct thread_data *, int, bool); |
| 1550 | }; |
| 1551 | |
| 1552 | static struct log_type log_types[] = { |
| 1553 | { |
| 1554 | .mask = BW_LOG_MASK, |
| 1555 | .fn = write_bandw_log, |
| 1556 | }, |
| 1557 | { |
| 1558 | .mask = LAT_LOG_MASK, |
| 1559 | .fn = write_lat_log, |
| 1560 | }, |
| 1561 | { |
| 1562 | .mask = SLAT_LOG_MASK, |
| 1563 | .fn = write_slat_log, |
| 1564 | }, |
| 1565 | { |
| 1566 | .mask = CLAT_LOG_MASK, |
| 1567 | .fn = write_clat_log, |
| 1568 | }, |
| 1569 | { |
| 1570 | .mask = IOPS_LOG_MASK, |
| 1571 | .fn = write_iops_log, |
| 1572 | }, |
| 1573 | { |
| 1574 | .mask = CLAT_HIST_LOG_MASK, |
| 1575 | .fn = write_clat_hist_log, |
| 1576 | } |
| 1577 | }; |
| 1578 | |
| 1579 | void td_writeout_logs(struct thread_data *td, bool unit_logs) |
| 1580 | { |
| 1581 | unsigned int log_mask = 0; |
| 1582 | unsigned int log_left = ALL_LOG_NR; |
| 1583 | int old_state, i; |
| 1584 | |
| 1585 | old_state = td_bump_runstate(td, TD_FINISHING); |
| 1586 | |
| 1587 | finalize_logs(td, unit_logs); |
| 1588 | |
| 1589 | while (log_left) { |
| 1590 | int prev_log_left = log_left; |
| 1591 | |
| 1592 | for (i = 0; i < ALL_LOG_NR && log_left; i++) { |
| 1593 | struct log_type *lt = &log_types[i]; |
| 1594 | int ret; |
| 1595 | |
| 1596 | if (!(log_mask & lt->mask)) { |
| 1597 | ret = lt->fn(td, log_left != 1, unit_logs); |
| 1598 | if (!ret) { |
| 1599 | log_left--; |
| 1600 | log_mask |= lt->mask; |
| 1601 | } |
| 1602 | } |
| 1603 | } |
| 1604 | |
| 1605 | if (prev_log_left == log_left) |
| 1606 | usleep(5000); |
| 1607 | } |
| 1608 | |
| 1609 | td_restore_runstate(td, old_state); |
| 1610 | } |
| 1611 | |
| 1612 | void fio_writeout_logs(bool unit_logs) |
| 1613 | { |
| 1614 | struct thread_data *td; |
| 1615 | int i; |
| 1616 | |
| 1617 | for_each_td(td, i) |
| 1618 | td_writeout_logs(td, unit_logs); |
| 1619 | } |