iolog.c

   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 <libgen.h>
   8 #include <assert.h>
   9 #include "flist.h"
  10 #include "fio.h"
  11 #include "verify.h"
  12 #include "trim.h"
  13
  14 static const char iolog_ver2[] = "fio version 2 iolog";
  15
  16 void queue_io_piece(struct thread_data *td, struct io_piece *ipo)
  17 {
  18         flist_add_tail(&ipo->list, &td->io_log_list);
  19         td->total_io_size += ipo->len;
  20 }
  21
  22 void log_io_u(struct thread_data *td, struct io_u *io_u)
  23 {
  24         const char *act[] = { "read", "write", "sync", "datasync",
  25                                 "sync_file_range", "wait", "trim" };
  26
  27         assert(io_u->ddir <= 6);
  28
  29         if (!td->o.write_iolog_file)
  30                 return;
  31
  32         fprintf(td->iolog_f, "%s %s %llu %lu\n", io_u->file->file_name,
  33                                                 act[io_u->ddir], io_u->offset,
  34                                                 io_u->buflen);
  35 }
  36
  37 void log_file(struct thread_data *td, struct fio_file *f,
  38               enum file_log_act what)
  39 {
  40         const char *act[] = { "add", "open", "close" };
  41
  42         assert(what < 3);
  43
  44         if (!td->o.write_iolog_file)
  45                 return;
  46
  47
  48         /*
  49          * this happens on the pre-open/close done before the job starts
  50          */
  51         if (!td->iolog_f)
  52                 return;
  53
  54         fprintf(td->iolog_f, "%s %s\n", f->file_name, act[what]);
  55 }
  56
  57 static void iolog_delay(struct thread_data *td, unsigned long delay)
  58 {
  59         unsigned long usec = utime_since_now(&td->last_issue);
  60
  61         if (delay < usec)
  62                 return;
  63
  64         delay -= usec;
  65
  66         /*
  67          * less than 100 usec delay, just regard it as noise
  68          */
  69         if (delay < 100)
  70                 return;
  71
  72         usec_sleep(td, delay);
  73 }
  74
  75 static int ipo_special(struct thread_data *td, struct io_piece *ipo)
  76 {
  77         struct fio_file *f;
  78         int ret;
  79
  80         /*
  81          * Not a special ipo
  82          */
  83         if (ipo->ddir != DDIR_INVAL)
  84                 return 0;
  85
  86         f = td->files[ipo->fileno];
  87
  88         switch (ipo->file_action) {
  89         case FIO_LOG_OPEN_FILE:
  90                 ret = td_io_open_file(td, f);
  91                 if (!ret)
  92                         break;
  93                 td_verror(td, ret, "iolog open file");
  94                 return -1;
  95         case FIO_LOG_CLOSE_FILE:
  96                 td_io_close_file(td, f);
  97                 break;
  98         case FIO_LOG_UNLINK_FILE:
  99                 unlink(f->file_name);
 100                 break;
 101         default:
 102                 log_err("fio: bad file action %d\n", ipo->file_action);
 103                 break;
 104         }
 105
 106         return 1;
 107 }
 108
 109 int read_iolog_get(struct thread_data *td, struct io_u *io_u)
 110 {
 111         struct io_piece *ipo;
 112         unsigned long elapsed;
 113
 114         while (!flist_empty(&td->io_log_list)) {
 115                 int ret;
 116
 117                 ipo = flist_entry(td->io_log_list.next, struct io_piece, list);
 118                 flist_del(&ipo->list);
 119                 remove_trim_entry(td, ipo);
 120
 121                 ret = ipo_special(td, ipo);
 122                 if (ret < 0) {
 123                         free(ipo);
 124                         break;
 125                 } else if (ret > 0) {
 126                         free(ipo);
 127                         continue;
 128                 }
 129
 130                 io_u->ddir = ipo->ddir;
 131                 if (ipo->ddir != DDIR_WAIT) {
 132                         io_u->offset = ipo->offset;
 133                         io_u->buflen = ipo->len;
 134                         io_u->file = td->files[ipo->fileno];
 135                         get_file(io_u->file);
 136                         dprint(FD_IO, "iolog: get %llu/%lu/%s\n", io_u->offset,
 137                                                 io_u->buflen, io_u->file->file_name);
 138                         if (ipo->delay)
 139                                 iolog_delay(td, ipo->delay);
 140                 } else {
 141                         elapsed = mtime_since_genesis();
 142                         if (ipo->delay > elapsed)
 143                                 usec_sleep(td, (ipo->delay - elapsed) * 1000);
 144
 145                 }
 146
 147                 free(ipo);
 148
 149                 if (io_u->ddir != DDIR_WAIT)
 150                         return 0;
 151         }
 152
 153         td->done = 1;
 154         return 1;
 155 }
 156
 157 void prune_io_piece_log(struct thread_data *td)
 158 {
 159         struct io_piece *ipo;
 160         struct rb_node *n;
 161
 162         while ((n = rb_first(&td->io_hist_tree)) != NULL) {
 163                 ipo = rb_entry(n, struct io_piece, rb_node);
 164                 rb_erase(n, &td->io_hist_tree);
 165                 remove_trim_entry(td, ipo);
 166                 td->io_hist_len--;
 167                 free(ipo);
 168         }
 169
 170         while (!flist_empty(&td->io_hist_list)) {
 171                 ipo = flist_entry(td->io_hist_list.next, struct io_piece, list);
 172                 flist_del(&ipo->list);
 173                 remove_trim_entry(td, ipo);
 174                 td->io_hist_len--;
 175                 free(ipo);
 176         }
 177 }
 178
 179 /*
 180  * log a successful write, so we can unwind the log for verify
 181  */
 182 void log_io_piece(struct thread_data *td, struct io_u *io_u)
 183 {
 184         struct rb_node **p, *parent;
 185         struct io_piece *ipo, *__ipo;
 186
 187         ipo = malloc(sizeof(struct io_piece));
 188         init_ipo(ipo);
 189         ipo->file = io_u->file;
 190         ipo->offset = io_u->offset;
 191         ipo->len = io_u->buflen;
 192
 193         if (io_u_should_trim(td, io_u)) {
 194                 flist_add_tail(&ipo->trim_list, &td->trim_list);
 195                 td->trim_entries++;
 196         }
 197
 198         /*
 199          * We don't need to sort the entries, if:
 200          *
 201          *      Sequential writes, or
 202          *      Random writes that lay out the file as it goes along
 203          *
 204          * For both these cases, just reading back data in the order we
 205          * wrote it out is the fastest.
 206          *
 207          * One exception is if we don't have a random map AND we are doing
 208          * verifies, in that case we need to check for duplicate blocks and
 209          * drop the old one, which we rely on the rb insert/lookup for
 210          * handling.
 211          */
 212         if ((!td_random(td) || !td->o.overwrite) &&
 213               (file_randommap(td, ipo->file) || td->o.verify == VERIFY_NONE)) {
 214                 INIT_FLIST_HEAD(&ipo->list);
 215                 flist_add_tail(&ipo->list, &td->io_hist_list);
 216                 ipo->flags |= IP_F_ONLIST;
 217                 td->io_hist_len++;
 218                 return;
 219         }
 220
 221         RB_CLEAR_NODE(&ipo->rb_node);
 222
 223         /*
 224          * Sort the entry into the verification list
 225          */
 226 restart:
 227         p = &td->io_hist_tree.rb_node;
 228         parent = NULL;
 229         while (*p) {
 230                 parent = *p;
 231
 232                 __ipo = rb_entry(parent, struct io_piece, rb_node);
 233                 if (ipo->file < __ipo->file)
 234                         p = &(*p)->rb_left;
 235                 else if (ipo->file > __ipo->file)
 236                         p = &(*p)->rb_right;
 237                 else if (ipo->offset < __ipo->offset)
 238                         p = &(*p)->rb_left;
 239                 else if (ipo->offset > __ipo->offset)
 240                         p = &(*p)->rb_right;
 241                 else {
 242                         assert(ipo->len == __ipo->len);
 243                         td->io_hist_len--;
 244                         rb_erase(parent, &td->io_hist_tree);
 245                         remove_trim_entry(td, __ipo);
 246                         free(__ipo);
 247                         goto restart;
 248                 }
 249         }
 250
 251         rb_link_node(&ipo->rb_node, parent, p);
 252         rb_insert_color(&ipo->rb_node, &td->io_hist_tree);
 253         ipo->flags |= IP_F_ONRB;
 254         td->io_hist_len++;
 255 }
 256
 257 void write_iolog_close(struct thread_data *td)
 258 {
 259         fflush(td->iolog_f);
 260         fclose(td->iolog_f);
 261         free(td->iolog_buf);
 262         td->iolog_f = NULL;
 263         td->iolog_buf = NULL;
 264 }
 265
 266 /*
 267  * Read version 2 iolog data. It is enhanced to include per-file logging,
 268  * syncs, etc.
 269  */
 270 static int read_iolog2(struct thread_data *td, FILE *f)
 271 {
 272         unsigned long long offset;
 273         unsigned int bytes;
 274         int reads, writes, waits, fileno = 0, file_action = 0; /* stupid gcc */
 275         char *fname, *act;
 276         char *str, *p;
 277         enum fio_ddir rw;
 278
 279         free_release_files(td);
 280
 281         /*
 282          * Read in the read iolog and store it, reuse the infrastructure
 283          * for doing verifications.
 284          */
 285         str = malloc(4096);
 286         fname = malloc(256+16);
 287         act = malloc(256+16);
 288
 289         reads = writes = waits = 0;
 290         while ((p = fgets(str, 4096, f)) != NULL) {
 291                 struct io_piece *ipo;
 292                 int r;
 293
 294                 r = sscanf(p, "%256s %256s %llu %u", fname, act, &offset,
 295                                                                         &bytes);
 296                 if (r == 4) {
 297                         /*
 298                          * Check action first
 299                          */
 300                         if (!strcmp(act, "wait"))
 301                                 rw = DDIR_WAIT;
 302                         else if (!strcmp(act, "read"))
 303                                 rw = DDIR_READ;
 304                         else if (!strcmp(act, "write"))
 305                                 rw = DDIR_WRITE;
 306                         else if (!strcmp(act, "sync"))
 307                                 rw = DDIR_SYNC;
 308                         else if (!strcmp(act, "datasync"))
 309                                 rw = DDIR_DATASYNC;
 310                         else if (!strcmp(act, "trim"))
 311                                 rw = DDIR_TRIM;
 312                         else {
 313                                 log_err("fio: bad iolog file action: %s\n",
 314                                                                         act);
 315                                 continue;
 316                         }
 317                 } else if (r == 2) {
 318                         rw = DDIR_INVAL;
 319                         if (!strcmp(act, "add")) {
 320                                 td->o.nr_files++;
 321                                 fileno = add_file(td, fname);
 322                                 file_action = FIO_LOG_ADD_FILE;
 323                                 continue;
 324                         } else if (!strcmp(act, "open")) {
 325                                 fileno = get_fileno(td, fname);
 326                                 file_action = FIO_LOG_OPEN_FILE;
 327                         } else if (!strcmp(act, "close")) {
 328                                 fileno = get_fileno(td, fname);
 329                                 file_action = FIO_LOG_CLOSE_FILE;
 330                         } else {
 331                                 log_err("fio: bad iolog file action: %s\n",
 332                                                                         act);
 333                                 continue;
 334                         }
 335                 } else {
 336                         log_err("bad iolog2: %s", p);
 337                         continue;
 338                 }
 339
 340                 if (rw == DDIR_READ)
 341                         reads++;
 342                 else if (rw == DDIR_WRITE) {
 343                         /*
 344                          * Don't add a write for ro mode
 345                          */
 346                         if (read_only)
 347                                 continue;
 348                         writes++;
 349                 } else if (rw == DDIR_WAIT) {
 350                         waits++;
 351                 } else if (rw == DDIR_INVAL) {
 352                 } else if (!ddir_sync(rw)) {
 353                         log_err("bad ddir: %d\n", rw);
 354                         continue;
 355                 }
 356
 357                 /*
 358                  * Make note of file
 359                  */
 360                 ipo = malloc(sizeof(*ipo));
 361                 init_ipo(ipo);
 362                 ipo->ddir = rw;
 363                 if (rw == DDIR_WAIT) {
 364                         ipo->delay = offset;
 365                 } else {
 366                         ipo->offset = offset;
 367                         ipo->len = bytes;
 368                         if (bytes > td->o.max_bs[rw])
 369                                 td->o.max_bs[rw] = bytes;
 370                         ipo->fileno = fileno;
 371                         ipo->file_action = file_action;
 372                 }
 373
 374                 queue_io_piece(td, ipo);
 375         }
 376
 377         free(str);
 378         free(act);
 379         free(fname);
 380
 381         if (writes && read_only) {
 382                 log_err("fio: <%s> skips replay of %d writes due to"
 383                         " read-only\n", td->o.name, writes);
 384                 writes = 0;
 385         }
 386
 387         if (!reads && !writes && !waits)
 388                 return 1;
 389         else if (reads && !writes)
 390                 td->o.td_ddir = TD_DDIR_READ;
 391         else if (!reads && writes)
 392                 td->o.td_ddir = TD_DDIR_WRITE;
 393         else
 394                 td->o.td_ddir = TD_DDIR_RW;
 395
 396         return 0;
 397 }
 398
 399 /*
 400  * open iolog, check version, and call appropriate parser
 401  */
 402 static int init_iolog_read(struct thread_data *td)
 403 {
 404         char buffer[256], *p;
 405         FILE *f;
 406         int ret;
 407
 408         f = fopen(td->o.read_iolog_file, "r");
 409         if (!f) {
 410                 perror("fopen read iolog");
 411                 return 1;
 412         }
 413
 414         p = fgets(buffer, sizeof(buffer), f);
 415         if (!p) {
 416                 td_verror(td, errno, "iolog read");
 417                 log_err("fio: unable to read iolog\n");
 418                 fclose(f);
 419                 return 1;
 420         }
 421
 422         /*
 423          * version 2 of the iolog stores a specific string as the
 424          * first line, check for that
 425          */
 426         if (!strncmp(iolog_ver2, buffer, strlen(iolog_ver2)))
 427                 ret = read_iolog2(td, f);
 428         else {
 429                 log_err("fio: iolog version 1 is no longer supported\n");
 430                 ret = 1;
 431         }
 432
 433         fclose(f);
 434         return ret;
 435 }
 436
 437 /*
 438  * Set up a log for storing io patterns.
 439  */
 440 static int init_iolog_write(struct thread_data *td)
 441 {
 442         struct fio_file *ff;
 443         FILE *f;
 444         unsigned int i;
 445
 446         f = fopen(td->o.write_iolog_file, "a");
 447         if (!f) {
 448                 perror("fopen write iolog");
 449                 return 1;
 450         }
 451
 452         /*
 453          * That's it for writing, setup a log buffer and we're done.
 454           */
 455         td->iolog_f = f;
 456         td->iolog_buf = malloc(8192);
 457         setvbuf(f, td->iolog_buf, _IOFBF, 8192);
 458
 459         /*
 460          * write our version line
 461          */
 462         if (fprintf(f, "%s\n", iolog_ver2) < 0) {
 463                 perror("iolog init\n");
 464                 return 1;
 465         }
 466
 467         /*
 468          * add all known files
 469          */
 470         for_each_file(td, ff, i)
 471                 log_file(td, ff, FIO_LOG_ADD_FILE);
 472
 473         return 0;
 474 }
 475
 476 int init_iolog(struct thread_data *td)
 477 {
 478         int ret = 0;
 479
 480         if (td->o.read_iolog_file) {
 481                 /*
 482                  * Check if it's a blktrace file and load that if possible.
 483                  * Otherwise assume it's a normal log file and load that.
 484                  */
 485                 if (is_blktrace(td->o.read_iolog_file))
 486                         ret = load_blktrace(td, td->o.read_iolog_file);
 487                 else
 488                         ret = init_iolog_read(td);
 489         } else if (td->o.write_iolog_file)
 490                 ret = init_iolog_write(td);
 491
 492         return ret;
 493 }
 494
 495 void setup_log(struct io_log **log)
 496 {
 497         struct io_log *l = malloc(sizeof(*l));
 498
 499         l->nr_samples = 0;
 500         l->max_samples = 1024;
 501         l->log = malloc(l->max_samples * sizeof(struct io_sample));
 502         *log = l;
 503 }
 504
 505 void __finish_log(struct io_log *log, const char *name)
 506 {
 507         unsigned int i;
 508         FILE *f;
 509
 510         f = fopen(name, "a");
 511         if (!f) {
 512                 perror("fopen log");
 513                 return;
 514         }
 515
 516         for (i = 0; i < log->nr_samples; i++) {
 517                 fprintf(f, "%lu, %lu, %u, %u\n", log->log[i].time,
 518                                                 log->log[i].val,
 519                                                 log->log[i].ddir,
 520                                                 log->log[i].bs);
 521         }
 522
 523         fclose(f);
 524         free(log->log);
 525         free(log);
 526 }
 527
 528 void finish_log_named(struct thread_data *td, struct io_log *log,
 529                        const char *prefix, const char *postfix)
 530 {
 531         char file_name[256], *p;
 532
 533         snprintf(file_name, 200, "%s_%s.log", prefix, postfix);
 534         p = basename(file_name);
 535         __finish_log(log, p);
 536 }
 537
 538 void finish_log(struct thread_data *td, struct io_log *log, const char *name)
 539 {
 540         finish_log_named(td, log, td->o.name, name);
 541 }