t/io_uring.c

   1 #include <stdio.h>
   2 #include <errno.h>
   3 #include <assert.h>
   4 #include <stdlib.h>
   5 #include <stddef.h>
   6 #include <signal.h>
   7 #include <inttypes.h>
   8 #include <math.h>
   9
  10 #ifdef CONFIG_LIBAIO
  11 #include <libaio.h>
  12 #endif
  13
  14 #include <sys/types.h>
  15 #include <sys/stat.h>
  16 #include <sys/ioctl.h>
  17 #include <sys/syscall.h>
  18 #include <sys/resource.h>
  19 #include <sys/mman.h>
  20 #include <sys/uio.h>
  21 #include <linux/fs.h>
  22 #include <fcntl.h>
  23 #include <unistd.h>
  24 #include <string.h>
  25 #include <pthread.h>
  26 #include <sched.h>
  27
  28 #include "../arch/arch.h"
  29 #include "../lib/types.h"
  30 #include "../lib/roundup.h"
  31 #include "../lib/rand.h"
  32 #include "../minmax.h"
  33 #include "../os/linux/io_uring.h"
  34
  35 struct io_sq_ring {
  36         unsigned *head;
  37         unsigned *tail;
  38         unsigned *ring_mask;
  39         unsigned *ring_entries;
  40         unsigned *flags;
  41         unsigned *array;
  42 };
  43
  44 struct io_cq_ring {
  45         unsigned *head;
  46         unsigned *tail;
  47         unsigned *ring_mask;
  48         unsigned *ring_entries;
  49         struct io_uring_cqe *cqes;
  50 };
  51
  52 #define DEPTH                   128
  53 #define BATCH_SUBMIT            32
  54 #define BATCH_COMPLETE          32
  55 #define BS                      4096
  56
  57 #define MAX_FDS                 16
  58
  59 static unsigned sq_ring_mask, cq_ring_mask;
  60
  61 struct file {
  62         unsigned long max_blocks;
  63         unsigned long max_size;
  64         unsigned long cur_off;
  65         unsigned pending_ios;
  66         int real_fd;
  67         int fixed_fd;
  68         int fileno;
  69 };
  70
  71 #define PLAT_BITS               6
  72 #define PLAT_VAL                (1 << PLAT_BITS)
  73 #define PLAT_GROUP_NR           29
  74 #define PLAT_NR                 (PLAT_GROUP_NR * PLAT_VAL)
  75
  76 struct submitter {
  77         pthread_t thread;
  78         int ring_fd;
  79         int index;
  80         struct io_sq_ring sq_ring;
  81         struct io_uring_sqe *sqes;
  82         struct io_cq_ring cq_ring;
  83         int inflight;
  84         int tid;
  85         unsigned long reaps;
  86         unsigned long done;
  87         unsigned long calls;
  88         volatile int finish;
  89
  90         __s32 *fds;
  91
  92         struct taus258_state rand_state;
  93
  94         unsigned long *clock_batch;
  95         int clock_index;
  96         unsigned long *plat;
  97
  98 #ifdef CONFIG_LIBAIO
  99         io_context_t aio_ctx;
 100 #endif
 101
 102         struct file files[MAX_FDS];
 103         unsigned nr_files;
 104         unsigned cur_file;
 105         struct iovec iovecs[];
 106 };
 107
 108 static struct submitter *submitter;
 109 static volatile int finish;
 110 static int stats_running;
 111 static unsigned long max_iops;
 112
 113 static int depth = DEPTH;
 114 static int batch_submit = BATCH_SUBMIT;
 115 static int batch_complete = BATCH_COMPLETE;
 116 static int bs = BS;
 117 static int polled = 1;          /* use IO polling */
 118 static int fixedbufs = 1;       /* use fixed user buffers */
 119 static int dma_map;             /* pre-map DMA buffers */
 120 static int register_files = 1;  /* use fixed files */
 121 static int buffered = 0;        /* use buffered IO, not O_DIRECT */
 122 static int sq_thread_poll = 0;  /* use kernel submission/poller thread */
 123 static int sq_thread_cpu = -1;  /* pin above thread to this CPU */
 124 static int do_nop = 0;          /* no-op SQ ring commands */
 125 static int nthreads = 1;
 126 static int stats = 0;           /* generate IO stats */
 127 static int aio = 0;             /* use libaio */
 128 static int runtime = 0;         /* runtime */
 129 static int random_io = 1;       /* random or sequential IO */
 130
 131 static unsigned long tsc_rate;
 132
 133 #define TSC_RATE_FILE   "tsc-rate"
 134
 135 static int vectored = 1;
 136
 137 static float plist[] = { 1.0, 5.0, 10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0,
 138                         80.0, 90.0, 95.0, 99.0, 99.5, 99.9, 99.95, 99.99 };
 139 static int plist_len = 17;
 140
 141 #ifndef IORING_REGISTER_MAP_BUFFERS
 142 #define IORING_REGISTER_MAP_BUFFERS     22
 143 struct io_uring_map_buffers {
 144         __s32   fd;
 145         __u32   buf_start;
 146         __u32   buf_end;
 147         __u32   flags;
 148         __u64   rsvd[2];
 149 };
 150 #endif
 151
 152 static unsigned long cycles_to_nsec(unsigned long cycles)
 153 {
 154         uint64_t val;
 155
 156         if (!tsc_rate)
 157                 return cycles;
 158
 159         val = cycles * 1000000000ULL;
 160         return val / tsc_rate;
 161 }
 162
 163 static unsigned long plat_idx_to_val(unsigned int idx)
 164 {
 165         unsigned int error_bits;
 166         unsigned long k, base;
 167
 168         assert(idx < PLAT_NR);
 169
 170         /* MSB <= (PLAT_BITS-1), cannot be rounded off. Use
 171          * all bits of the sample as index */
 172         if (idx < (PLAT_VAL << 1))
 173                 return cycles_to_nsec(idx);
 174
 175         /* Find the group and compute the minimum value of that group */
 176         error_bits = (idx >> PLAT_BITS) - 1;
 177         base = ((unsigned long) 1) << (error_bits + PLAT_BITS);
 178
 179         /* Find its bucket number of the group */
 180         k = idx % PLAT_VAL;
 181
 182         /* Return the mean of the range of the bucket */
 183         return cycles_to_nsec(base + ((k + 0.5) * (1 << error_bits)));
 184 }
 185
 186 unsigned int calc_clat_percentiles(unsigned long *io_u_plat, unsigned long nr,
 187                                    unsigned long **output,
 188                                    unsigned long *maxv, unsigned long *minv)
 189 {
 190         unsigned long sum = 0;
 191         unsigned int len = plist_len, i, j = 0;
 192         unsigned long *ovals = NULL;
 193         bool is_last;
 194
 195         *minv = -1UL;
 196         *maxv = 0;
 197
 198         ovals = malloc(len * sizeof(*ovals));
 199         if (!ovals)
 200                 return 0;
 201
 202         /*
 203          * Calculate bucket values, note down max and min values
 204          */
 205         is_last = false;
 206         for (i = 0; i < PLAT_NR && !is_last; i++) {
 207                 sum += io_u_plat[i];
 208                 while (sum >= ((long double) plist[j] / 100.0 * nr)) {
 209                         assert(plist[j] <= 100.0);
 210
 211                         ovals[j] = plat_idx_to_val(i);
 212                         if (ovals[j] < *minv)
 213                                 *minv = ovals[j];
 214                         if (ovals[j] > *maxv)
 215                                 *maxv = ovals[j];
 216
 217                         is_last = (j == len - 1) != 0;
 218                         if (is_last)
 219                                 break;
 220
 221                         j++;
 222                 }
 223         }
 224
 225         if (!is_last)
 226                 fprintf(stderr, "error calculating latency percentiles\n");
 227
 228         *output = ovals;
 229         return len;
 230 }
 231
 232 static void show_clat_percentiles(unsigned long *io_u_plat, unsigned long nr,
 233                                   unsigned int precision)
 234 {
 235         unsigned int divisor, len, i, j = 0;
 236         unsigned long minv, maxv;
 237         unsigned long *ovals;
 238         int per_line, scale_down, time_width;
 239         bool is_last;
 240         char fmt[32];
 241
 242         len = calc_clat_percentiles(io_u_plat, nr, &ovals, &maxv, &minv);
 243         if (!len || !ovals)
 244                 goto out;
 245
 246         if (!tsc_rate) {
 247                 scale_down = 0;
 248                 divisor = 1;
 249                 printf("    percentiles (tsc ticks):\n     |");
 250         } else if (minv > 2000 && maxv > 99999) {
 251                 scale_down = 1;
 252                 divisor = 1000;
 253                 printf("    percentiles (usec):\n     |");
 254         } else {
 255                 scale_down = 0;
 256                 divisor = 1;
 257                 printf("    percentiles (nsec):\n     |");
 258         }
 259
 260         time_width = max(5, (int) (log10(maxv / divisor) + 1));
 261         snprintf(fmt, sizeof(fmt), " %%%u.%ufth=[%%%dllu]%%c", precision + 3,
 262                         precision, time_width);
 263         /* fmt will be something like " %5.2fth=[%4llu]%c" */
 264         per_line = (80 - 7) / (precision + 10 + time_width);
 265
 266         for (j = 0; j < len; j++) {
 267                 /* for formatting */
 268                 if (j != 0 && (j % per_line) == 0)
 269                         printf("     |");
 270
 271                 /* end of the list */
 272                 is_last = (j == len - 1) != 0;
 273
 274                 for (i = 0; i < scale_down; i++)
 275                         ovals[j] = (ovals[j] + 999) / 1000;
 276
 277                 printf(fmt, plist[j], ovals[j], is_last ? '\n' : ',');
 278
 279                 if (is_last)
 280                         break;
 281
 282                 if ((j % per_line) == per_line - 1)     /* for formatting */
 283                         printf("\n");
 284         }
 285
 286 out:
 287         free(ovals);
 288 }
 289
 290 #ifdef ARCH_HAVE_CPU_CLOCK
 291 static unsigned int plat_val_to_idx(unsigned long val)
 292 {
 293         unsigned int msb, error_bits, base, offset, idx;
 294
 295         /* Find MSB starting from bit 0 */
 296         if (val == 0)
 297                 msb = 0;
 298         else
 299                 msb = (sizeof(val)*8) - __builtin_clzll(val) - 1;
 300
 301         /*
 302          * MSB <= (PLAT_BITS-1), cannot be rounded off. Use
 303          * all bits of the sample as index
 304          */
 305         if (msb <= PLAT_BITS)
 306                 return val;
 307
 308         /* Compute the number of error bits to discard*/
 309         error_bits = msb - PLAT_BITS;
 310
 311         /* Compute the number of buckets before the group */
 312         base = (error_bits + 1) << PLAT_BITS;
 313
 314         /*
 315          * Discard the error bits and apply the mask to find the
 316          * index for the buckets in the group
 317          */
 318         offset = (PLAT_VAL - 1) & (val >> error_bits);
 319
 320         /* Make sure the index does not exceed (array size - 1) */
 321         idx = (base + offset) < (PLAT_NR - 1) ?
 322                 (base + offset) : (PLAT_NR - 1);
 323
 324         return idx;
 325 }
 326 #endif
 327
 328 static void add_stat(struct submitter *s, int clock_index, int nr)
 329 {
 330 #ifdef ARCH_HAVE_CPU_CLOCK
 331         unsigned long cycles;
 332         unsigned int pidx;
 333
 334         if (!s->finish && clock_index) {
 335                 cycles = get_cpu_clock();
 336                 cycles -= s->clock_batch[clock_index];
 337                 pidx = plat_val_to_idx(cycles);
 338                 s->plat[pidx] += nr;
 339         }
 340 #endif
 341 }
 342
 343 static int io_uring_map_buffers(struct submitter *s)
 344 {
 345         struct io_uring_map_buffers map = {
 346                 .fd             = s->files[0].real_fd,
 347                 .buf_end        = depth,
 348         };
 349
 350         if (do_nop)
 351                 return 0;
 352         if (s->nr_files > 1)
 353                 fprintf(stdout, "Mapping buffers may not work with multiple files\n");
 354
 355         return syscall(__NR_io_uring_register, s->ring_fd,
 356                         IORING_REGISTER_MAP_BUFFERS, &map, 1);
 357 }
 358
 359 static int io_uring_register_buffers(struct submitter *s)
 360 {
 361         if (do_nop)
 362                 return 0;
 363
 364         return syscall(__NR_io_uring_register, s->ring_fd,
 365                         IORING_REGISTER_BUFFERS, s->iovecs, roundup_pow2(depth));
 366 }
 367
 368 static int io_uring_register_files(struct submitter *s)
 369 {
 370         int i;
 371
 372         if (do_nop)
 373                 return 0;
 374
 375         s->fds = calloc(s->nr_files, sizeof(__s32));
 376         for (i = 0; i < s->nr_files; i++) {
 377                 s->fds[i] = s->files[i].real_fd;
 378                 s->files[i].fixed_fd = i;
 379         }
 380
 381         return syscall(__NR_io_uring_register, s->ring_fd,
 382                         IORING_REGISTER_FILES, s->fds, s->nr_files);
 383 }
 384
 385 static int io_uring_setup(unsigned entries, struct io_uring_params *p)
 386 {
 387         /*
 388          * Clamp CQ ring size at our SQ ring size, we don't need more entries
 389          * than that.
 390          */
 391         p->flags |= IORING_SETUP_CQSIZE;
 392         p->cq_entries = entries;
 393
 394         return syscall(__NR_io_uring_setup, entries, p);
 395 }
 396
 397 static void io_uring_probe(int fd)
 398 {
 399         struct io_uring_probe *p;
 400         int ret;
 401
 402         p = malloc(sizeof(*p) + 256 * sizeof(struct io_uring_probe_op));
 403         if (!p)
 404                 return;
 405
 406         memset(p, 0, sizeof(*p) + 256 * sizeof(struct io_uring_probe_op));
 407         ret = syscall(__NR_io_uring_register, fd, IORING_REGISTER_PROBE, p, 256);
 408         if (ret < 0)
 409                 goto out;
 410
 411         if (IORING_OP_READ > p->ops_len)
 412                 goto out;
 413
 414         if ((p->ops[IORING_OP_READ].flags & IO_URING_OP_SUPPORTED))
 415                 vectored = 0;
 416 out:
 417         free(p);
 418 }
 419
 420 static int io_uring_enter(struct submitter *s, unsigned int to_submit,
 421                           unsigned int min_complete, unsigned int flags)
 422 {
 423 #ifdef FIO_ARCH_HAS_SYSCALL
 424         return __do_syscall6(__NR_io_uring_enter, s->ring_fd, to_submit,
 425                                 min_complete, flags, NULL, 0);
 426 #else
 427         return syscall(__NR_io_uring_enter, s->ring_fd, to_submit, min_complete,
 428                         flags, NULL, 0);
 429 #endif
 430 }
 431
 432 #ifndef CONFIG_HAVE_GETTID
 433 static int gettid(void)
 434 {
 435         return syscall(__NR_gettid);
 436 }
 437 #endif
 438
 439 static unsigned file_depth(struct submitter *s)
 440 {
 441         return (depth + s->nr_files - 1) / s->nr_files;
 442 }
 443
 444 static void init_io(struct submitter *s, unsigned index)
 445 {
 446         struct io_uring_sqe *sqe = &s->sqes[index];
 447         unsigned long offset;
 448         struct file *f;
 449         long r;
 450
 451         if (do_nop) {
 452                 sqe->opcode = IORING_OP_NOP;
 453                 return;
 454         }
 455
 456         if (s->nr_files == 1) {
 457                 f = &s->files[0];
 458         } else {
 459                 f = &s->files[s->cur_file];
 460                 if (f->pending_ios >= file_depth(s)) {
 461                         s->cur_file++;
 462                         if (s->cur_file == s->nr_files)
 463                                 s->cur_file = 0;
 464                         f = &s->files[s->cur_file];
 465                 }
 466         }
 467         f->pending_ios++;
 468
 469         if (random_io) {
 470                 r = __rand64(&s->rand_state);
 471                 offset = (r % (f->max_blocks - 1)) * bs;
 472         } else {
 473                 offset = f->cur_off;
 474                 f->cur_off += bs;
 475                 if (f->cur_off + bs > f->max_size)
 476                         f->cur_off = 0;
 477         }
 478
 479         if (register_files) {
 480                 sqe->flags = IOSQE_FIXED_FILE;
 481                 sqe->fd = f->fixed_fd;
 482         } else {
 483                 sqe->flags = 0;
 484                 sqe->fd = f->real_fd;
 485         }
 486         if (fixedbufs) {
 487                 sqe->opcode = IORING_OP_READ_FIXED;
 488                 sqe->addr = (unsigned long) s->iovecs[index].iov_base;
 489                 sqe->len = bs;
 490                 sqe->buf_index = index;
 491         } else if (!vectored) {
 492                 sqe->opcode = IORING_OP_READ;
 493                 sqe->addr = (unsigned long) s->iovecs[index].iov_base;
 494                 sqe->len = bs;
 495                 sqe->buf_index = 0;
 496         } else {
 497                 sqe->opcode = IORING_OP_READV;
 498                 sqe->addr = (unsigned long) &s->iovecs[index];
 499                 sqe->len = 1;
 500                 sqe->buf_index = 0;
 501         }
 502         sqe->ioprio = 0;
 503         sqe->off = offset;
 504         sqe->user_data = (unsigned long) f->fileno;
 505         if (stats && stats_running)
 506                 sqe->user_data |= ((uint64_t)s->clock_index << 32);
 507 }
 508
 509 static int prep_more_ios_uring(struct submitter *s, int max_ios)
 510 {
 511         struct io_sq_ring *ring = &s->sq_ring;
 512         unsigned index, tail, next_tail, prepped = 0;
 513
 514         next_tail = tail = *ring->tail;
 515         do {
 516                 next_tail++;
 517                 if (next_tail == atomic_load_acquire(ring->head))
 518                         break;
 519
 520                 index = tail & sq_ring_mask;
 521                 init_io(s, index);
 522                 ring->array[index] = index;
 523                 prepped++;
 524                 tail = next_tail;
 525         } while (prepped < max_ios);
 526
 527         if (prepped)
 528                 atomic_store_release(ring->tail, tail);
 529         return prepped;
 530 }
 531
 532 static int get_file_size(struct file *f)
 533 {
 534         struct stat st;
 535
 536         if (fstat(f->real_fd, &st) < 0)
 537                 return -1;
 538         if (S_ISBLK(st.st_mode)) {
 539                 unsigned long long bytes;
 540
 541                 if (ioctl(f->real_fd, BLKGETSIZE64, &bytes) != 0)
 542                         return -1;
 543
 544                 f->max_blocks = bytes / bs;
 545                 f->max_size = bytes;
 546                 return 0;
 547         } else if (S_ISREG(st.st_mode)) {
 548                 f->max_blocks = st.st_size / bs;
 549                 f->max_size = st.st_size;
 550                 return 0;
 551         }
 552
 553         return -1;
 554 }
 555
 556 static int reap_events_uring(struct submitter *s)
 557 {
 558         struct io_cq_ring *ring = &s->cq_ring;
 559         struct io_uring_cqe *cqe;
 560         unsigned head, reaped = 0;
 561         int last_idx = -1, stat_nr = 0;
 562
 563         head = *ring->head;
 564         do {
 565                 struct file *f;
 566
 567                 read_barrier();
 568                 if (head == atomic_load_acquire(ring->tail))
 569                         break;
 570                 cqe = &ring->cqes[head & cq_ring_mask];
 571                 if (!do_nop) {
 572                         int fileno = cqe->user_data & 0xffffffff;
 573
 574                         f = &s->files[fileno];
 575                         f->pending_ios--;
 576                         if (cqe->res != bs) {
 577                                 printf("io: unexpected ret=%d\n", cqe->res);
 578                                 if (polled && cqe->res == -EOPNOTSUPP)
 579                                         printf("Your filesystem/driver/kernel doesn't support polled IO\n");
 580                                 return -1;
 581                         }
 582                 }
 583                 if (stats) {
 584                         int clock_index = cqe->user_data >> 32;
 585
 586                         if (last_idx != clock_index) {
 587                                 if (last_idx != -1) {
 588                                         add_stat(s, last_idx, stat_nr);
 589                                         stat_nr = 0;
 590                                 }
 591                                 last_idx = clock_index;
 592                         }
 593                         stat_nr++;
 594                 }
 595                 reaped++;
 596                 head++;
 597         } while (1);
 598
 599         if (stat_nr)
 600                 add_stat(s, last_idx, stat_nr);
 601
 602         if (reaped) {
 603                 s->inflight -= reaped;
 604                 atomic_store_release(ring->head, head);
 605         }
 606         return reaped;
 607 }
 608
 609 static int submitter_init(struct submitter *s)
 610 {
 611         int i, nr_batch;
 612
 613         s->tid = gettid();
 614         printf("submitter=%d, tid=%d\n", s->index, s->tid);
 615
 616         __init_rand64(&s->rand_state, pthread_self());
 617         srand48(pthread_self());
 618
 619         for (i = 0; i < MAX_FDS; i++)
 620                 s->files[i].fileno = i;
 621
 622         if (stats) {
 623                 nr_batch = roundup_pow2(depth / batch_submit);
 624                 if (nr_batch < 2)
 625                         nr_batch = 2;
 626                 s->clock_batch = calloc(nr_batch, sizeof(unsigned long));
 627                 s->clock_index = 1;
 628
 629                 s->plat = calloc(PLAT_NR, sizeof(unsigned long));
 630         } else {
 631                 s->clock_batch = NULL;
 632                 s->plat = NULL;
 633                 nr_batch = 0;
 634         }
 635
 636         return nr_batch;
 637 }
 638
 639 #ifdef CONFIG_LIBAIO
 640 static int prep_more_ios_aio(struct submitter *s, int max_ios, struct iocb *iocbs)
 641 {
 642         uint64_t data;
 643         long long offset;
 644         struct file *f;
 645         unsigned index;
 646         long r;
 647
 648         index = 0;
 649         while (index < max_ios) {
 650                 struct iocb *iocb = &iocbs[index];
 651
 652                 if (s->nr_files == 1) {
 653                         f = &s->files[0];
 654                 } else {
 655                         f = &s->files[s->cur_file];
 656                         if (f->pending_ios >= file_depth(s)) {
 657                                 s->cur_file++;
 658                                 if (s->cur_file == s->nr_files)
 659                                         s->cur_file = 0;
 660                                 f = &s->files[s->cur_file];
 661                         }
 662                 }
 663                 f->pending_ios++;
 664
 665                 r = lrand48();
 666                 offset = (r % (f->max_blocks - 1)) * bs;
 667                 io_prep_pread(iocb, f->real_fd, s->iovecs[index].iov_base,
 668                                 s->iovecs[index].iov_len, offset);
 669
 670                 data = f->fileno;
 671                 if (stats && stats_running)
 672                         data |= (((uint64_t) s->clock_index) << 32);
 673                 iocb->data = (void *) (uintptr_t) data;
 674                 index++;
 675         }
 676         return index;
 677 }
 678
 679 static int reap_events_aio(struct submitter *s, struct io_event *events, int evs)
 680 {
 681         int last_idx = -1, stat_nr = 0;
 682         int reaped = 0;
 683
 684         while (evs) {
 685                 uint64_t data = (uintptr_t) events[reaped].data;
 686                 struct file *f = &s->files[data & 0xffffffff];
 687
 688                 f->pending_ios--;
 689                 if (events[reaped].res != bs) {
 690                         printf("io: unexpected ret=%ld\n", events[reaped].res);
 691                         return -1;
 692                 }
 693                 if (stats) {
 694                         int clock_index = data >> 32;
 695
 696                         if (last_idx != clock_index) {
 697                                 if (last_idx != -1) {
 698                                         add_stat(s, last_idx, stat_nr);
 699                                         stat_nr = 0;
 700                                 }
 701                                 last_idx = clock_index;
 702                         }
 703                         stat_nr++;
 704                 }
 705                 reaped++;
 706                 evs--;
 707         }
 708
 709         if (stat_nr)
 710                 add_stat(s, last_idx, stat_nr);
 711
 712         s->inflight -= reaped;
 713         s->done += reaped;
 714         return reaped;
 715 }
 716
 717 static void *submitter_aio_fn(void *data)
 718 {
 719         struct submitter *s = data;
 720         int i, ret, prepped;
 721         struct iocb **iocbsptr;
 722         struct iocb *iocbs;
 723         struct io_event *events;
 724 #ifdef ARCH_HAVE_CPU_CLOCK
 725         int nr_batch = submitter_init(s);
 726 #else
 727         submitter_init(s);
 728 #endif
 729
 730         iocbsptr = calloc(depth, sizeof(struct iocb *));
 731         iocbs = calloc(depth, sizeof(struct iocb));
 732         events = calloc(depth, sizeof(struct io_event));
 733
 734         for (i = 0; i < depth; i++)
 735                 iocbsptr[i] = &iocbs[i];
 736
 737         prepped = 0;
 738         do {
 739                 int to_wait, to_submit, to_prep;
 740
 741                 if (!prepped && s->inflight < depth) {
 742                         to_prep = min(depth - s->inflight, batch_submit);
 743                         prepped = prep_more_ios_aio(s, to_prep, iocbs);
 744 #ifdef ARCH_HAVE_CPU_CLOCK
 745                         if (prepped && stats) {
 746                                 s->clock_batch[s->clock_index] = get_cpu_clock();
 747                                 s->clock_index = (s->clock_index + 1) & (nr_batch - 1);
 748                         }
 749 #endif
 750                 }
 751                 s->inflight += prepped;
 752                 to_submit = prepped;
 753
 754                 if (to_submit && (s->inflight + to_submit <= depth))
 755                         to_wait = 0;
 756                 else
 757                         to_wait = min(s->inflight + to_submit, batch_complete);
 758
 759                 ret = io_submit(s->aio_ctx, to_submit, iocbsptr);
 760                 s->calls++;
 761                 if (ret < 0) {
 762                         perror("io_submit");
 763                         break;
 764                 } else if (ret != to_submit) {
 765                         printf("submitted %d, wanted %d\n", ret, to_submit);
 766                         break;
 767                 }
 768                 prepped = 0;
 769
 770                 while (to_wait) {
 771                         int r;
 772
 773                         s->calls++;
 774                         r = io_getevents(s->aio_ctx, to_wait, to_wait, events, NULL);
 775                         if (r < 0) {
 776                                 perror("io_getevents");
 777                                 break;
 778                         } else if (r != to_wait) {
 779                                 printf("r=%d, wait=%d\n", r, to_wait);
 780                                 break;
 781                         }
 782                         r = reap_events_aio(s, events, r);
 783                         s->reaps += r;
 784                         to_wait -= r;
 785                 }
 786         } while (!s->finish);
 787
 788         free(iocbsptr);
 789         free(iocbs);
 790         free(events);
 791         finish = 1;
 792         return NULL;
 793 }
 794 #endif
 795
 796 static void *submitter_uring_fn(void *data)
 797 {
 798         struct submitter *s = data;
 799         struct io_sq_ring *ring = &s->sq_ring;
 800         int ret, prepped;
 801 #ifdef ARCH_HAVE_CPU_CLOCK
 802         int nr_batch = submitter_init(s);
 803 #else
 804         submitter_init(s);
 805 #endif
 806
 807         prepped = 0;
 808         do {
 809                 int to_wait, to_submit, this_reap, to_prep;
 810                 unsigned ring_flags = 0;
 811
 812                 if (!prepped && s->inflight < depth) {
 813                         to_prep = min(depth - s->inflight, batch_submit);
 814                         prepped = prep_more_ios_uring(s, to_prep);
 815 #ifdef ARCH_HAVE_CPU_CLOCK
 816                         if (prepped && stats) {
 817                                 s->clock_batch[s->clock_index] = get_cpu_clock();
 818                                 s->clock_index = (s->clock_index + 1) & (nr_batch - 1);
 819                         }
 820 #endif
 821                 }
 822                 s->inflight += prepped;
 823 submit_more:
 824                 to_submit = prepped;
 825 submit:
 826                 if (to_submit && (s->inflight + to_submit <= depth))
 827                         to_wait = 0;
 828                 else
 829                         to_wait = min(s->inflight + to_submit, batch_complete);
 830
 831                 /*
 832                  * Only need to call io_uring_enter if we're not using SQ thread
 833                  * poll, or if IORING_SQ_NEED_WAKEUP is set.
 834                  */
 835                 if (sq_thread_poll)
 836                         ring_flags = atomic_load_acquire(ring->flags);
 837                 if (!sq_thread_poll || ring_flags & IORING_SQ_NEED_WAKEUP) {
 838                         unsigned flags = 0;
 839
 840                         if (to_wait)
 841                                 flags = IORING_ENTER_GETEVENTS;
 842                         if (ring_flags & IORING_SQ_NEED_WAKEUP)
 843                                 flags |= IORING_ENTER_SQ_WAKEUP;
 844                         ret = io_uring_enter(s, to_submit, to_wait, flags);
 845                         s->calls++;
 846                 } else {
 847                         /* for SQPOLL, we submitted it all effectively */
 848                         ret = to_submit;
 849                 }
 850
 851                 /*
 852                  * For non SQ thread poll, we already got the events we needed
 853                  * through the io_uring_enter() above. For SQ thread poll, we
 854                  * need to loop here until we find enough events.
 855                  */
 856                 this_reap = 0;
 857                 do {
 858                         int r;
 859
 860                         r = reap_events_uring(s);
 861                         if (r == -1) {
 862                                 s->finish = 1;
 863                                 break;
 864                         } else if (r > 0)
 865                                 this_reap += r;
 866                 } while (sq_thread_poll && this_reap < to_wait);
 867                 s->reaps += this_reap;
 868
 869                 if (ret >= 0) {
 870                         if (!ret) {
 871                                 to_submit = 0;
 872                                 if (s->inflight)
 873                                         goto submit;
 874                                 continue;
 875                         } else if (ret < to_submit) {
 876                                 int diff = to_submit - ret;
 877
 878                                 s->done += ret;
 879                                 prepped -= diff;
 880                                 goto submit_more;
 881                         }
 882                         s->done += ret;
 883                         prepped = 0;
 884                         continue;
 885                 } else if (ret < 0) {
 886                         if (errno == EAGAIN) {
 887                                 if (s->finish)
 888                                         break;
 889                                 if (this_reap)
 890                                         goto submit;
 891                                 to_submit = 0;
 892                                 goto submit;
 893                         }
 894                         printf("io_submit: %s\n", strerror(errno));
 895                         break;
 896                 }
 897         } while (!s->finish);
 898
 899         finish = 1;
 900         return NULL;
 901 }
 902
 903 static struct submitter *get_submitter(int offset)
 904 {
 905         void *ret;
 906
 907         ret = submitter;
 908         if (offset)
 909                 ret += offset * (sizeof(*submitter) + depth * sizeof(struct iovec));
 910         return ret;
 911 }
 912
 913 static void do_finish(const char *reason)
 914 {
 915         int j;
 916         printf("Exiting on %s\n", reason);
 917         for (j = 0; j < nthreads; j++) {
 918                 struct submitter *s = get_submitter(j);
 919                 s->finish = 1;
 920         }
 921         if (max_iops > 100000)
 922                 printf("Maximum IOPS=%luK\n", max_iops / 1000);
 923         else if (max_iops)
 924                 printf("Maximum IOPS=%lu\n", max_iops);
 925         finish = 1;
 926 }
 927
 928 static void sig_int(int sig)
 929 {
 930         do_finish("signal");
 931 }
 932
 933 static void arm_sig_int(void)
 934 {
 935         struct sigaction act;
 936
 937         memset(&act, 0, sizeof(act));
 938         act.sa_handler = sig_int;
 939         act.sa_flags = SA_RESTART;
 940         sigaction(SIGINT, &act, NULL);
 941
 942         /* Windows uses SIGBREAK as a quit signal from other applications */
 943 #ifdef WIN32
 944         sigaction(SIGBREAK, &act, NULL);
 945 #endif
 946 }
 947
 948 static int setup_aio(struct submitter *s)
 949 {
 950 #ifdef CONFIG_LIBAIO
 951         if (polled) {
 952                 fprintf(stderr, "aio does not support polled IO\n");
 953                 polled = 0;
 954         }
 955         if (sq_thread_poll) {
 956                 fprintf(stderr, "aio does not support SQPOLL IO\n");
 957                 sq_thread_poll = 0;
 958         }
 959         if (do_nop) {
 960                 fprintf(stderr, "aio does not support polled IO\n");
 961                 do_nop = 0;
 962         }
 963         if (fixedbufs || register_files) {
 964                 fprintf(stderr, "aio does not support registered files or buffers\n");
 965                 fixedbufs = register_files = 0;
 966         }
 967
 968         return io_queue_init(roundup_pow2(depth), &s->aio_ctx);
 969 #else
 970         fprintf(stderr, "Legacy AIO not available on this system/build\n");
 971         errno = EINVAL;
 972         return -1;
 973 #endif
 974 }
 975
 976 static int setup_ring(struct submitter *s)
 977 {
 978         struct io_sq_ring *sring = &s->sq_ring;
 979         struct io_cq_ring *cring = &s->cq_ring;
 980         struct io_uring_params p;
 981         int ret, fd;
 982         void *ptr;
 983
 984         memset(&p, 0, sizeof(p));
 985
 986         if (polled && !do_nop)
 987                 p.flags |= IORING_SETUP_IOPOLL;
 988         if (sq_thread_poll) {
 989                 p.flags |= IORING_SETUP_SQPOLL;
 990                 if (sq_thread_cpu != -1) {
 991                         p.flags |= IORING_SETUP_SQ_AFF;
 992                         p.sq_thread_cpu = sq_thread_cpu;
 993                 }
 994         }
 995
 996         fd = io_uring_setup(depth, &p);
 997         if (fd < 0) {
 998                 perror("io_uring_setup");
 999                 return 1;
1000         }
1001         s->ring_fd = fd;
1002
1003         io_uring_probe(fd);
1004
1005         if (fixedbufs) {
1006                 struct rlimit rlim;
1007
1008                 rlim.rlim_cur = RLIM_INFINITY;
1009                 rlim.rlim_max = RLIM_INFINITY;
1010                 /* ignore potential error, not needed on newer kernels */
1011                 setrlimit(RLIMIT_MEMLOCK, &rlim);
1012
1013                 ret = io_uring_register_buffers(s);
1014                 if (ret < 0) {
1015                         perror("io_uring_register_buffers");
1016                         return 1;
1017                 }
1018
1019                 if (dma_map) {
1020                         ret = io_uring_map_buffers(s);
1021                         if (ret < 0) {
1022                                 perror("io_uring_map_buffers");
1023                                 return 1;
1024                         }
1025                 }
1026         }
1027
1028         if (register_files) {
1029                 ret = io_uring_register_files(s);
1030                 if (ret < 0) {
1031                         perror("io_uring_register_files");
1032                         return 1;
1033                 }
1034         }
1035
1036         ptr = mmap(0, p.sq_off.array + p.sq_entries * sizeof(__u32),
1037                         PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
1038                         IORING_OFF_SQ_RING);
1039         sring->head = ptr + p.sq_off.head;
1040         sring->tail = ptr + p.sq_off.tail;
1041         sring->ring_mask = ptr + p.sq_off.ring_mask;
1042         sring->ring_entries = ptr + p.sq_off.ring_entries;
1043         sring->flags = ptr + p.sq_off.flags;
1044         sring->array = ptr + p.sq_off.array;
1045         sq_ring_mask = *sring->ring_mask;
1046
1047         s->sqes = mmap(0, p.sq_entries * sizeof(struct io_uring_sqe),
1048                         PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
1049                         IORING_OFF_SQES);
1050
1051         ptr = mmap(0, p.cq_off.cqes + p.cq_entries * sizeof(struct io_uring_cqe),
1052                         PROT_READ | PROT_WRITE, MAP_SHARED | MAP_POPULATE, fd,
1053                         IORING_OFF_CQ_RING);
1054         cring->head = ptr + p.cq_off.head;
1055         cring->tail = ptr + p.cq_off.tail;
1056         cring->ring_mask = ptr + p.cq_off.ring_mask;
1057         cring->ring_entries = ptr + p.cq_off.ring_entries;
1058         cring->cqes = ptr + p.cq_off.cqes;
1059         cq_ring_mask = *cring->ring_mask;
1060         return 0;
1061 }
1062
1063 static void file_depths(char *buf)
1064 {
1065         bool prev = false;
1066         char *p;
1067         int i, j;
1068
1069         buf[0] = '\0';
1070         p = buf;
1071         for (j = 0; j < nthreads; j++) {
1072                 struct submitter *s = get_submitter(j);
1073
1074                 for (i = 0; i < s->nr_files; i++) {
1075                         struct file *f = &s->files[i];
1076
1077                         if (prev)
1078                                 p += sprintf(p, " %d", f->pending_ios);
1079                         else
1080                                 p += sprintf(p, "%d", f->pending_ios);
1081                         prev = true;
1082                 }
1083         }
1084 }
1085
1086 static void usage(char *argv, int status)
1087 {
1088         char runtime_str[16];
1089         snprintf(runtime_str, sizeof(runtime_str), "%d", runtime);
1090         printf("%s [options] -- [filenames]\n"
1091                 " -d <int>  : IO Depth, default %d\n"
1092                 " -s <int>  : Batch submit, default %d\n"
1093                 " -c <int>  : Batch complete, default %d\n"
1094                 " -b <int>  : Block size, default %d\n"
1095                 " -p <bool> : Polled IO, default %d\n"
1096                 " -B <bool> : Fixed buffers, default %d\n"
1097                 " -D <bool> : DMA map fixed buffers, default %d\n"
1098                 " -F <bool> : Register files, default %d\n"
1099                 " -n <int>  : Number of threads, default %d\n"
1100                 " -O <bool> : Use O_DIRECT, default %d\n"
1101                 " -N <bool> : Perform just no-op requests, default %d\n"
1102                 " -t <bool> : Track IO latencies, default %d\n"
1103                 " -T <int>  : TSC rate in HZ\n"
1104                 " -r <int>  : Runtime in seconds, default %s\n"
1105                 " -R <bool> : Use random IO, default %d\n"
1106                 " -a <bool> : Use legacy aio, default %d\n",
1107                 argv, DEPTH, BATCH_SUBMIT, BATCH_COMPLETE, BS, polled,
1108                 fixedbufs, dma_map, register_files, nthreads, !buffered, do_nop,
1109                 stats, runtime == 0 ? "unlimited" : runtime_str, random_io, aio);
1110         exit(status);
1111 }
1112
1113 static void read_tsc_rate(void)
1114 {
1115         char buffer[32];
1116         int fd, ret;
1117
1118         if (tsc_rate)
1119                 return;
1120
1121         fd = open(TSC_RATE_FILE, O_RDONLY);
1122         if (fd < 0)
1123                 return;
1124
1125         ret = read(fd, buffer, sizeof(buffer));
1126         if (ret < 0) {
1127                 close(fd);
1128                 return;
1129         }
1130
1131         tsc_rate = strtoul(buffer, NULL, 10);
1132         printf("Using TSC rate %luHz\n", tsc_rate);
1133         close(fd);
1134 }
1135
1136 static void write_tsc_rate(void)
1137 {
1138         char buffer[32];
1139         struct stat sb;
1140         int fd, ret;
1141
1142         if (!stat(TSC_RATE_FILE, &sb))
1143                 return;
1144
1145         fd = open(TSC_RATE_FILE, O_WRONLY | O_CREAT, 0644);
1146         if (fd < 0)
1147                 return;
1148
1149         memset(buffer, 0, sizeof(buffer));
1150         sprintf(buffer, "%lu", tsc_rate);
1151         ret = write(fd, buffer, strlen(buffer));
1152         if (ret < 0)
1153                 perror("write");
1154         close(fd);
1155 }
1156
1157 int main(int argc, char *argv[])
1158 {
1159         struct submitter *s;
1160         unsigned long done, calls, reap;
1161         int err, i, j, flags, fd, opt, threads_per_f, threads_rem = 0, nfiles;
1162         long page_size;
1163         struct file f;
1164         char *fdepths;
1165         void *ret;
1166
1167         if (!do_nop && argc < 2)
1168                 usage(argv[0], 1);
1169
1170         while ((opt = getopt(argc, argv, "d:s:c:b:p:B:F:n:N:O:t:T:a:r:D:R:h?")) != -1) {
1171                 switch (opt) {
1172                 case 'a':
1173                         aio = !!atoi(optarg);
1174                         break;
1175                 case 'd':
1176                         depth = atoi(optarg);
1177                         break;
1178                 case 's':
1179                         batch_submit = atoi(optarg);
1180                         if (!batch_submit)
1181                                 batch_submit = 1;
1182                         break;
1183                 case 'c':
1184                         batch_complete = atoi(optarg);
1185                         if (!batch_complete)
1186                                 batch_complete = 1;
1187                         break;
1188                 case 'b':
1189                         bs = atoi(optarg);
1190                         break;
1191                 case 'p':
1192                         polled = !!atoi(optarg);
1193                         break;
1194                 case 'B':
1195                         fixedbufs = !!atoi(optarg);
1196                         break;
1197                 case 'F':
1198                         register_files = !!atoi(optarg);
1199                         break;
1200                 case 'n':
1201                         nthreads = atoi(optarg);
1202                         if (!nthreads) {
1203                                 printf("Threads must be non-zero\n");
1204                                 usage(argv[0], 1);
1205                         }
1206                         break;
1207                 case 'N':
1208                         do_nop = !!atoi(optarg);
1209                         break;
1210                 case 'O':
1211                         buffered = !atoi(optarg);
1212                         break;
1213                 case 't':
1214 #ifndef ARCH_HAVE_CPU_CLOCK
1215                         fprintf(stderr, "Stats not supported on this CPU\n");
1216                         return 1;
1217 #endif
1218                         stats = !!atoi(optarg);
1219                         break;
1220                 case 'T':
1221 #ifndef ARCH_HAVE_CPU_CLOCK
1222                         fprintf(stderr, "Stats not supported on this CPU\n");
1223                         return 1;
1224 #endif
1225                         tsc_rate = strtoul(optarg, NULL, 10);
1226                         write_tsc_rate();
1227                         break;
1228                 case 'r':
1229                         runtime = atoi(optarg);
1230                         break;
1231                 case 'D':
1232                         dma_map = !!atoi(optarg);
1233                         break;
1234                 case 'R':
1235                         random_io = !!atoi(optarg);
1236                         break;
1237                 case 'h':
1238                 case '?':
1239                 default:
1240                         usage(argv[0], 0);
1241                         break;
1242                 }
1243         }
1244
1245         if (stats)
1246                 read_tsc_rate();
1247
1248         if (batch_complete > depth)
1249                 batch_complete = depth;
1250         if (batch_submit > depth)
1251                 batch_submit = depth;
1252         if (!fixedbufs && dma_map)
1253                 dma_map = 0;
1254
1255         submitter = calloc(nthreads, sizeof(*submitter) +
1256                                 roundup_pow2(depth) * sizeof(struct iovec));
1257         for (j = 0; j < nthreads; j++) {
1258                 s = get_submitter(j);
1259                 s->index = j;
1260                 s->done = s->calls = s->reaps = 0;
1261         }
1262
1263         flags = O_RDONLY | O_NOATIME;
1264         if (!buffered)
1265                 flags |= O_DIRECT;
1266
1267         j = 0;
1268         i = optind;
1269         nfiles = argc - i;
1270         if (!do_nop) {
1271                 if (!nfiles) {
1272                         printf("No files specified\n");
1273                         usage(argv[0], 1);
1274                 }
1275                 threads_per_f = nthreads / nfiles;
1276                 /* make sure each thread gets assigned files */
1277                 if (threads_per_f == 0) {
1278                         threads_per_f = 1;
1279                 } else {
1280                         threads_rem = nthreads - threads_per_f * nfiles;
1281                 }
1282         }
1283         while (!do_nop && i < argc) {
1284                 int k, limit;
1285
1286                 memset(&f, 0, sizeof(f));
1287
1288                 fd = open(argv[i], flags);
1289                 if (fd < 0) {
1290                         perror("open");
1291                         return 1;
1292                 }
1293                 f.real_fd = fd;
1294                 if (get_file_size(&f)) {
1295                         printf("failed getting size of device/file\n");
1296                         return 1;
1297                 }
1298                 if (f.max_blocks <= 1) {
1299                         printf("Zero file/device size?\n");
1300                         return 1;
1301                 }
1302                 f.max_blocks--;
1303
1304                 limit = threads_per_f;
1305                 limit += threads_rem > 0 ? 1 : 0;
1306                 for (k = 0; k < limit; k++) {
1307                         s = get_submitter((j + k) % nthreads);
1308
1309                         if (s->nr_files == MAX_FDS) {
1310                                 printf("Max number of files (%d) reached\n", MAX_FDS);
1311                                 break;
1312                         }
1313
1314                         memcpy(&s->files[s->nr_files], &f, sizeof(f));
1315
1316                         printf("Added file %s (submitter %d)\n", argv[i], s->index);
1317                         s->nr_files++;
1318                 }
1319                 threads_rem--;
1320                 i++;
1321                 j += limit;
1322         }
1323
1324         arm_sig_int();
1325
1326         page_size = sysconf(_SC_PAGESIZE);
1327         if (page_size < 0)
1328                 page_size = 4096;
1329
1330         for (j = 0; j < nthreads; j++) {
1331                 s = get_submitter(j);
1332                 for (i = 0; i < roundup_pow2(depth); i++) {
1333                         void *buf;
1334
1335                         if (posix_memalign(&buf, page_size, bs)) {
1336                                 printf("failed alloc\n");
1337                                 return 1;
1338                         }
1339                         s->iovecs[i].iov_base = buf;
1340                         s->iovecs[i].iov_len = bs;
1341                 }
1342         }
1343
1344         for (j = 0; j < nthreads; j++) {
1345                 s = get_submitter(j);
1346
1347                 if (!aio)
1348                         err = setup_ring(s);
1349                 else
1350                         err = setup_aio(s);
1351                 if (err) {
1352                         printf("ring setup failed: %s, %d\n", strerror(errno), err);
1353                         return 1;
1354                 }
1355         }
1356         s = get_submitter(0);
1357         printf("polled=%d, fixedbufs=%d/%d, register_files=%d, buffered=%d, QD=%d\n", polled, fixedbufs, dma_map, register_files, buffered, depth);
1358         if (!aio)
1359                 printf("Engine=io_uring, sq_ring=%d, cq_ring=%d\n", *s->sq_ring.ring_entries, *s->cq_ring.ring_entries);
1360         else
1361                 printf("Engine=aio\n");
1362
1363         for (j = 0; j < nthreads; j++) {
1364                 s = get_submitter(j);
1365                 if (!aio)
1366                         pthread_create(&s->thread, NULL, submitter_uring_fn, s);
1367 #ifdef CONFIG_LIBAIO
1368                 else
1369                         pthread_create(&s->thread, NULL, submitter_aio_fn, s);
1370 #endif
1371         }
1372
1373         fdepths = malloc(8 * s->nr_files * nthreads);
1374         reap = calls = done = 0;
1375         do {
1376                 unsigned long this_done = 0;
1377                 unsigned long this_reap = 0;
1378                 unsigned long this_call = 0;
1379                 unsigned long rpc = 0, ipc = 0;
1380                 unsigned long iops, bw;
1381
1382                 sleep(1);
1383                 if (runtime && !--runtime)
1384                         do_finish("timeout");
1385
1386                 /* don't print partial run, if interrupted by signal */
1387                 if (finish)
1388                         break;
1389
1390                 /* one second in to the run, enable stats */
1391                 if (stats)
1392                         stats_running = 1;
1393
1394                 for (j = 0; j < nthreads; j++) {
1395                         s = get_submitter(j);
1396                         this_done += s->done;
1397                         this_call += s->calls;
1398                         this_reap += s->reaps;
1399                 }
1400                 if (this_call - calls) {
1401                         rpc = (this_done - done) / (this_call - calls);
1402                         ipc = (this_reap - reap) / (this_call - calls);
1403                 } else
1404                         rpc = ipc = -1;
1405                 file_depths(fdepths);
1406                 iops = this_done - done;
1407                 if (bs > 1048576)
1408                         bw = iops * (bs / 1048576);
1409                 else
1410                         bw = iops / (1048576 / bs);
1411                 if (iops > 100000)
1412                         printf("IOPS=%luK, ", iops / 1000);
1413                 else
1414                         printf("IOPS=%lu, ", iops);
1415                 max_iops = max(max_iops, iops);
1416                 if (!do_nop)
1417                         printf("BW=%luMiB/s, ", bw);
1418                 printf("IOS/call=%ld/%ld, inflight=(%s)\n", rpc, ipc, fdepths);
1419                 done = this_done;
1420                 calls = this_call;
1421                 reap = this_reap;
1422         } while (!finish);
1423
1424         for (j = 0; j < nthreads; j++) {
1425                 s = get_submitter(j);
1426                 pthread_join(s->thread, &ret);
1427                 close(s->ring_fd);
1428
1429                 if (stats) {
1430                         unsigned long nr;
1431
1432                         printf("%d: Latency percentiles:\n", s->tid);
1433                         for (i = 0, nr = 0; i < PLAT_NR; i++)
1434                                 nr += s->plat[i];
1435                         show_clat_percentiles(s->plat, nr, 4);
1436                         free(s->clock_batch);
1437                         free(s->plat);
1438                 }
1439         }
1440
1441         free(fdepths);
1442         free(submitter);
1443         return 0;
1444 }