blktrace.c

   1 /*
   2  * blktrace support code for fio
   3  */
   4 #include <stdio.h>
   5 #include <stdlib.h>
   6 #include <sys/stat.h>
   7 #include <sys/ioctl.h>
   8 #include <linux/fs.h>
   9 #include <dirent.h>
  10
  11 #include "flist.h"
  12 #include "fio.h"
  13 #include "blktrace_api.h"
  14 #include "oslib/linux-dev-lookup.h"
  15
  16 #define TRACE_FIFO_SIZE 8192
  17
  18 /*
  19  * fifo refill frontend, to avoid reading data in trace sized bites
  20  */
  21 static int refill_fifo(struct thread_data *td, struct fifo *fifo, int fd)
  22 {
  23         char buf[TRACE_FIFO_SIZE];
  24         unsigned int total;
  25         int ret;
  26
  27         total = sizeof(buf);
  28         if (total > fifo_room(fifo))
  29                 total = fifo_room(fifo);
  30
  31         ret = read(fd, buf, total);
  32         if (ret < 0) {
  33                 td_verror(td, errno, "read blktrace file");
  34                 return -1;
  35         }
  36
  37         if (ret > 0)
  38                 ret = fifo_put(fifo, buf, ret);
  39
  40         dprint(FD_BLKTRACE, "refill: filled %d bytes\n", ret);
  41         return ret;
  42 }
  43
  44 /*
  45  * Retrieve 'len' bytes from the fifo, refilling if necessary.
  46  */
  47 static int trace_fifo_get(struct thread_data *td, struct fifo *fifo, int fd,
  48                           void *buf, unsigned int len)
  49 {
  50         if (fifo_len(fifo) < len) {
  51                 int ret = refill_fifo(td, fifo, fd);
  52
  53                 if (ret < 0)
  54                         return ret;
  55         }
  56
  57         return fifo_get(fifo, buf, len);
  58 }
  59
  60 /*
  61  * Just discard the pdu by seeking past it.
  62  */
  63 static int discard_pdu(struct thread_data *td, struct fifo *fifo, int fd,
  64                        struct blk_io_trace *t)
  65 {
  66         if (t->pdu_len == 0)
  67                 return 0;
  68
  69         dprint(FD_BLKTRACE, "discard pdu len %u\n", t->pdu_len);
  70         return trace_fifo_get(td, fifo, fd, NULL, t->pdu_len);
  71 }
  72
  73 /*
  74  * Check if this is a blktrace binary data file. We read a single trace
  75  * into memory and check for the magic signature.
  76  */
  77 int is_blktrace(const char *filename, int *need_swap)
  78 {
  79         struct blk_io_trace t;
  80         int fd, ret;
  81
  82         fd = open(filename, O_RDONLY);
  83         if (fd < 0)
  84                 return 0;
  85
  86         ret = read(fd, &t, sizeof(t));
  87         close(fd);
  88
  89         if (ret < 0) {
  90                 perror("read blktrace");
  91                 return 0;
  92         } else if (ret != sizeof(t)) {
  93                 log_err("fio: short read on blktrace file\n");
  94                 return 0;
  95         }
  96
  97         if ((t.magic & 0xffffff00) == BLK_IO_TRACE_MAGIC) {
  98                 *need_swap = 0;
  99                 return 1;
 100         }
 101
 102         /*
 103          * Maybe it needs to be endian swapped...
 104          */
 105         t.magic = fio_swap32(t.magic);
 106         if ((t.magic & 0xffffff00) == BLK_IO_TRACE_MAGIC) {
 107                 *need_swap = 1;
 108                 return 1;
 109         }
 110
 111         return 0;
 112 }
 113
 114 #define FMINORBITS      20
 115 #define FMINORMASK      ((1U << FMINORBITS) - 1)
 116 #define FMAJOR(dev)     ((unsigned int) ((dev) >> FMINORBITS))
 117 #define FMINOR(dev)     ((unsigned int) ((dev) & FMINORMASK))
 118
 119 static void trace_add_open_close_event(struct thread_data *td, int fileno, enum file_log_act action)
 120 {
 121         struct io_piece *ipo;
 122
 123         ipo = calloc(1, sizeof(*ipo));
 124         init_ipo(ipo);
 125
 126         ipo->ddir = DDIR_INVAL;
 127         ipo->fileno = fileno;
 128         ipo->file_action = action;
 129         flist_add_tail(&ipo->list, &td->io_log_list);
 130 }
 131
 132 static int get_dev_blocksize(const char *dev, unsigned int *bs)
 133 {
 134         int fd;
 135
 136         fd = open(dev, O_RDONLY);
 137         if (fd < 0)
 138                 return 1;
 139
 140         if (ioctl(fd, BLKSSZGET, bs) < 0) {
 141                 close(fd);
 142                 return 1;
 143         }
 144
 145         close(fd);
 146         return 0;
 147 }
 148
 149 static int trace_add_file(struct thread_data *td, __u32 device,
 150                           unsigned int *bs)
 151 {
 152         static unsigned int last_maj, last_min, last_fileno, last_bs;
 153         unsigned int maj = FMAJOR(device);
 154         unsigned int min = FMINOR(device);
 155         struct fio_file *f;
 156         unsigned int i;
 157         char dev[256];
 158
 159         if (last_maj == maj && last_min == min) {
 160                 *bs = last_bs;
 161                 return last_fileno;
 162         }
 163
 164         last_maj = maj;
 165         last_min = min;
 166
 167         /*
 168          * check for this file in our list
 169          */
 170         for_each_file(td, f, i) {
 171                 if (f->major == maj && f->minor == min) {
 172                         last_fileno = f->fileno;
 173                         last_bs = f->bs;
 174                         goto out;
 175                 }
 176         }
 177
 178         strcpy(dev, "/dev");
 179         if (blktrace_lookup_device(td->o.replay_redirect, dev, maj, min)) {
 180                 unsigned int this_bs;
 181                 int fileno;
 182
 183                 if (td->o.replay_redirect)
 184                         dprint(FD_BLKTRACE, "device lookup: %d/%d\n overridden"
 185                                         " with: %s\n", maj, min,
 186                                         td->o.replay_redirect);
 187                 else
 188                         dprint(FD_BLKTRACE, "device lookup: %d/%d\n", maj, min);
 189
 190                 dprint(FD_BLKTRACE, "add devices %s\n", dev);
 191                 fileno = add_file_exclusive(td, dev);
 192
 193                 if (get_dev_blocksize(dev, &this_bs))
 194                         this_bs = 512;
 195
 196                 td->o.open_files++;
 197                 td->files[fileno]->major = maj;
 198                 td->files[fileno]->minor = min;
 199                 td->files[fileno]->bs = this_bs;
 200                 trace_add_open_close_event(td, fileno, FIO_LOG_OPEN_FILE);
 201
 202                 last_fileno = fileno;
 203                 last_bs = this_bs;
 204         }
 205
 206 out:
 207         *bs = last_bs;
 208         return last_fileno;
 209 }
 210
 211 static void t_bytes_align(struct thread_options *o, struct blk_io_trace *t)
 212 {
 213         if (!o->replay_align)
 214                 return;
 215
 216         t->bytes = (t->bytes + o->replay_align - 1) & ~(o->replay_align - 1);
 217 }
 218
 219 static void ipo_bytes_align(struct thread_options *o, struct io_piece *ipo)
 220 {
 221         if (!o->replay_align)
 222                 return;
 223
 224         ipo->offset &= ~(o->replay_align - 1);
 225 }
 226
 227
 228 /*
 229  * Store blk_io_trace data in an ipo for later retrieval.
 230  */
 231 static void store_ipo(struct thread_data *td, unsigned long long offset,
 232                       unsigned int bytes, int rw, unsigned long long ttime,
 233                       int fileno, unsigned int bs)
 234 {
 235         struct io_piece *ipo = malloc(sizeof(*ipo));
 236
 237         init_ipo(ipo);
 238
 239         ipo->offset = offset * bs;
 240         if (td->o.replay_scale)
 241                 ipo->offset = ipo->offset / td->o.replay_scale;
 242         ipo_bytes_align(&td->o, ipo);
 243         ipo->len = bytes;
 244         ipo->delay = ttime / 1000;
 245         if (rw)
 246                 ipo->ddir = DDIR_WRITE;
 247         else
 248                 ipo->ddir = DDIR_READ;
 249         ipo->fileno = fileno;
 250
 251         dprint(FD_BLKTRACE, "store ddir=%d, off=%llu, len=%lu, delay=%lu\n",
 252                                                         ipo->ddir, ipo->offset,
 253                                                         ipo->len, ipo->delay);
 254         queue_io_piece(td, ipo);
 255 }
 256
 257 static void handle_trace_notify(struct blk_io_trace *t)
 258 {
 259         switch (t->action) {
 260         case BLK_TN_PROCESS:
 261                 dprint(FD_BLKTRACE, "got process notify: %x, %d\n",
 262                                 t->action, t->pid);
 263                 break;
 264         case BLK_TN_TIMESTAMP:
 265                 dprint(FD_BLKTRACE, "got timestamp notify: %x, %d\n",
 266                                 t->action, t->pid);
 267                 break;
 268         case BLK_TN_MESSAGE:
 269                 break;
 270         default:
 271                 dprint(FD_BLKTRACE, "unknown trace act %x\n", t->action);
 272                 break;
 273         }
 274 }
 275
 276 static void handle_trace_discard(struct thread_data *td,
 277                                  struct blk_io_trace *t,
 278                                  unsigned long long ttime,
 279                                  unsigned long *ios, unsigned int *rw_bs)
 280 {
 281         struct io_piece *ipo = malloc(sizeof(*ipo));
 282         unsigned int bs;
 283         int fileno;
 284
 285         init_ipo(ipo);
 286         fileno = trace_add_file(td, t->device, &bs);
 287
 288         ios[DDIR_TRIM]++;
 289         if (t->bytes > rw_bs[DDIR_TRIM])
 290                 rw_bs[DDIR_TRIM] = t->bytes;
 291
 292         td->o.size += t->bytes;
 293
 294         memset(ipo, 0, sizeof(*ipo));
 295         INIT_FLIST_HEAD(&ipo->list);
 296
 297         ipo->offset = t->sector * bs;
 298         if (td->o.replay_scale)
 299                 ipo->offset = ipo->offset / td->o.replay_scale;
 300         ipo_bytes_align(&td->o, ipo);
 301         ipo->len = t->bytes;
 302         ipo->delay = ttime / 1000;
 303         ipo->ddir = DDIR_TRIM;
 304         ipo->fileno = fileno;
 305
 306         dprint(FD_BLKTRACE, "store discard, off=%llu, len=%lu, delay=%lu\n",
 307                                                         ipo->offset, ipo->len,
 308                                                         ipo->delay);
 309         queue_io_piece(td, ipo);
 310 }
 311
 312 static void handle_trace_fs(struct thread_data *td, struct blk_io_trace *t,
 313                             unsigned long long ttime, unsigned long *ios,
 314                             unsigned int *rw_bs)
 315 {
 316         unsigned int bs;
 317         int rw;
 318         int fileno;
 319
 320         fileno = trace_add_file(td, t->device, &bs);
 321
 322         rw = (t->action & BLK_TC_ACT(BLK_TC_WRITE)) != 0;
 323
 324         if (t->bytes > rw_bs[rw])
 325                 rw_bs[rw] = t->bytes;
 326
 327         ios[rw]++;
 328         td->o.size += t->bytes;
 329         store_ipo(td, t->sector, t->bytes, rw, ttime, fileno, bs);
 330 }
 331
 332 /*
 333  * We only care for queue traces, most of the others are side effects
 334  * due to internal workings of the block layer.
 335  */
 336 static void handle_trace(struct thread_data *td, struct blk_io_trace *t,
 337                          unsigned long *ios, unsigned int *bs)
 338 {
 339         static unsigned long long last_ttime;
 340         unsigned long long delay = 0;
 341
 342         if ((t->action & 0xffff) != __BLK_TA_QUEUE)
 343                 return;
 344
 345         if (!(t->action & BLK_TC_ACT(BLK_TC_NOTIFY))) {
 346                 if (!last_ttime || td->o.no_stall) {
 347                         last_ttime = t->time;
 348                         delay = 0;
 349                 } else {
 350                         delay = t->time - last_ttime;
 351                         last_ttime = t->time;
 352                 }
 353         }
 354
 355         t_bytes_align(&td->o, t);
 356
 357         if (t->action & BLK_TC_ACT(BLK_TC_NOTIFY))
 358                 handle_trace_notify(t);
 359         else if (t->action & BLK_TC_ACT(BLK_TC_DISCARD))
 360                 handle_trace_discard(td, t, delay, ios, bs);
 361         else
 362                 handle_trace_fs(td, t, delay, ios, bs);
 363 }
 364
 365 static void byteswap_trace(struct blk_io_trace *t)
 366 {
 367         t->magic = fio_swap32(t->magic);
 368         t->sequence = fio_swap32(t->sequence);
 369         t->time = fio_swap64(t->time);
 370         t->sector = fio_swap64(t->sector);
 371         t->bytes = fio_swap32(t->bytes);
 372         t->action = fio_swap32(t->action);
 373         t->pid = fio_swap32(t->pid);
 374         t->device = fio_swap32(t->device);
 375         t->cpu = fio_swap32(t->cpu);
 376         t->error = fio_swap16(t->error);
 377         t->pdu_len = fio_swap16(t->pdu_len);
 378 }
 379
 380 static int t_is_write(struct blk_io_trace *t)
 381 {
 382         return (t->action & BLK_TC_ACT(BLK_TC_WRITE | BLK_TC_DISCARD)) != 0;
 383 }
 384
 385 static enum fio_ddir t_get_ddir(struct blk_io_trace *t)
 386 {
 387         if (t->action & BLK_TC_ACT(BLK_TC_READ))
 388                 return DDIR_READ;
 389         else if (t->action & BLK_TC_ACT(BLK_TC_WRITE))
 390                 return DDIR_WRITE;
 391         else if (t->action & BLK_TC_ACT(BLK_TC_DISCARD))
 392                 return DDIR_TRIM;
 393
 394         return DDIR_INVAL;
 395 }
 396
 397 static void depth_inc(struct blk_io_trace *t, int *depth)
 398 {
 399         enum fio_ddir ddir;
 400
 401         ddir = t_get_ddir(t);
 402         if (ddir != DDIR_INVAL)
 403                 depth[ddir]++;
 404 }
 405
 406 static void depth_dec(struct blk_io_trace *t, int *depth)
 407 {
 408         enum fio_ddir ddir;
 409
 410         ddir = t_get_ddir(t);
 411         if (ddir != DDIR_INVAL)
 412                 depth[ddir]--;
 413 }
 414
 415 static void depth_end(struct blk_io_trace *t, int *this_depth, int *depth)
 416 {
 417         enum fio_ddir ddir = DDIR_INVAL;
 418
 419         ddir = t_get_ddir(t);
 420         if (ddir != DDIR_INVAL) {
 421                 depth[ddir] = max(depth[ddir], this_depth[ddir]);
 422                 this_depth[ddir] = 0;
 423         }
 424 }
 425
 426 /*
 427  * Load a blktrace file by reading all the blk_io_trace entries, and storing
 428  * them as io_pieces like the fio text version would do.
 429  */
 430 int load_blktrace(struct thread_data *td, const char *filename, int need_swap)
 431 {
 432         struct blk_io_trace t;
 433         unsigned long ios[DDIR_RWDIR_CNT], skipped_writes;
 434         unsigned int rw_bs[DDIR_RWDIR_CNT];
 435         struct fifo *fifo;
 436         int fd, i, old_state;
 437         struct fio_file *f;
 438         int this_depth[DDIR_RWDIR_CNT], depth[DDIR_RWDIR_CNT], max_depth;
 439
 440         fd = open(filename, O_RDONLY);
 441         if (fd < 0) {
 442                 td_verror(td, errno, "open blktrace file");
 443                 return 1;
 444         }
 445
 446         fifo = fifo_alloc(TRACE_FIFO_SIZE);
 447
 448         old_state = td_bump_runstate(td, TD_SETTING_UP);
 449
 450         td->o.size = 0;
 451
 452         for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 453                 ios[i] = 0;
 454                 rw_bs[i] = 0;
 455                 this_depth[i] = 0;
 456                 depth[i] = 0;
 457         }
 458
 459         skipped_writes = 0;
 460         do {
 461                 int ret = trace_fifo_get(td, fifo, fd, &t, sizeof(t));
 462
 463                 if (ret < 0)
 464                         goto err;
 465                 else if (!ret)
 466                         break;
 467                 else if (ret < (int) sizeof(t)) {
 468                         log_err("fio: short fifo get\n");
 469                         break;
 470                 }
 471
 472                 if (need_swap)
 473                         byteswap_trace(&t);
 474
 475                 if ((t.magic & 0xffffff00) != BLK_IO_TRACE_MAGIC) {
 476                         log_err("fio: bad magic in blktrace data: %x\n",
 477                                                                 t.magic);
 478                         goto err;
 479                 }
 480                 if ((t.magic & 0xff) != BLK_IO_TRACE_VERSION) {
 481                         log_err("fio: bad blktrace version %d\n",
 482                                                                 t.magic & 0xff);
 483                         goto err;
 484                 }
 485                 ret = discard_pdu(td, fifo, fd, &t);
 486                 if (ret < 0) {
 487                         td_verror(td, ret, "blktrace lseek");
 488                         goto err;
 489                 } else if (t.pdu_len != ret) {
 490                         log_err("fio: discarded %d of %d\n", ret, t.pdu_len);
 491                         goto err;
 492                 }
 493                 if ((t.action & BLK_TC_ACT(BLK_TC_NOTIFY)) == 0) {
 494                         if ((t.action & 0xffff) == __BLK_TA_QUEUE)
 495                                 depth_inc(&t, this_depth);
 496                         else if (((t.action & 0xffff) == __BLK_TA_BACKMERGE) ||
 497                                 ((t.action & 0xffff) == __BLK_TA_FRONTMERGE))
 498                                 depth_dec(&t, this_depth);
 499                         else if ((t.action & 0xffff) == __BLK_TA_COMPLETE)
 500                                 depth_end(&t, this_depth, depth);
 501
 502                         if (t_is_write(&t) && read_only) {
 503                                 skipped_writes++;
 504                                 continue;
 505                         }
 506                 }
 507
 508                 handle_trace(td, &t, ios, rw_bs);
 509         } while (1);
 510
 511         for (i = 0; i < td->files_index; i++) {
 512                 f = td->files[i];
 513                 trace_add_open_close_event(td, f->fileno, FIO_LOG_CLOSE_FILE);
 514         }
 515
 516         fifo_free(fifo);
 517         close(fd);
 518
 519         td_restore_runstate(td, old_state);
 520
 521         if (!td->files_index) {
 522                 log_err("fio: did not find replay device(s)\n");
 523                 return 1;
 524         }
 525
 526         /*
 527          * For stacked devices, we don't always get a COMPLETE event so
 528          * the depth grows to insane values. Limit it to something sane(r).
 529          */
 530         max_depth = 0;
 531         for (i = 0; i < DDIR_RWDIR_CNT; i++) {
 532                 if (depth[i] > 1024)
 533                         depth[i] = 1024;
 534                 else if (!depth[i] && ios[i])
 535                         depth[i] = 1;
 536                 max_depth = max(depth[i], max_depth);
 537         }
 538
 539         if (skipped_writes)
 540                 log_err("fio: %s skips replay of %lu writes due to read-only\n",
 541                                                 td->o.name, skipped_writes);
 542
 543         if (!ios[DDIR_READ] && !ios[DDIR_WRITE]) {
 544                 log_err("fio: found no ios in blktrace data\n");
 545                 return 1;
 546         } else if (ios[DDIR_READ] && !ios[DDIR_WRITE]) {
 547                 td->o.td_ddir = TD_DDIR_READ;
 548                 td->o.max_bs[DDIR_READ] = rw_bs[DDIR_READ];
 549         } else if (!ios[DDIR_READ] && ios[DDIR_WRITE]) {
 550                 td->o.td_ddir = TD_DDIR_WRITE;
 551                 td->o.max_bs[DDIR_WRITE] = rw_bs[DDIR_WRITE];
 552         } else {
 553                 td->o.td_ddir = TD_DDIR_RW;
 554                 td->o.max_bs[DDIR_READ] = rw_bs[DDIR_READ];
 555                 td->o.max_bs[DDIR_WRITE] = rw_bs[DDIR_WRITE];
 556                 td->o.max_bs[DDIR_TRIM] = rw_bs[DDIR_TRIM];
 557         }
 558
 559         /*
 560          * We need to do direct/raw ios to the device, to avoid getting
 561          * read-ahead in our way. But only do so if the minimum block size
 562          * is a multiple of 4k, otherwise we don't know if it's safe to do so.
 563          */
 564         if (!fio_option_is_set(&td->o, odirect) && !(td_min_bs(td) & 4095))
 565                 td->o.odirect = 1;
 566
 567         /*
 568          * If depth wasn't manually set, use probed depth
 569          */
 570         if (!fio_option_is_set(&td->o, iodepth))
 571                 td->o.iodepth = td->o.iodepth_low = max_depth;
 572
 573         return 0;
 574 err:
 575         close(fd);
 576         fifo_free(fifo);
 577         return 1;
 578 }