git.ozlabs.org Git - ccan/blob - ccan/tdb/tools/replay_trace.c

   1 #include <ccan/tdb/tdb.h>
   2 #include <ccan/grab_file/grab_file.h>
   3 #include <ccan/hash/hash.h>
   4 #include <ccan/talloc/talloc.h>
   5 #include <ccan/str_talloc/str_talloc.h>
   6 #include <ccan/str/str.h>
   7 #include <ccan/list/list.h>
   8 #include <err.h>
   9 #include <ctype.h>
  10 #include <string.h>
  11 #include <unistd.h>
  12 #include <sys/types.h>
  13 #include <sys/wait.h>
  14 #include <sys/time.h>
  15 #include <errno.h>
  16 #include <signal.h>
  17 #include <assert.h>
  18
  19 #define STRINGIFY2(x) #x
  20 #define STRINGIFY(x) STRINGIFY2(x)
  21
  22 /* Avoid mod by zero */
  23 static unsigned int total_keys = 1;
  24
  25 /* #define DEBUG_DEPS 1 */
  26
  27 /* Traversals block transactions in the current implementation. */
  28 #define TRAVERSALS_TAKE_TRANSACTION_LOCK 1
  29
  30 struct pipe {
  31         int fd[2];
  32 };
  33 static struct pipe *pipes;
  34
  35 static void __attribute__((noreturn)) fail(const char *filename,
  36                                            unsigned int line,
  37                                            const char *fmt, ...)
  38 {
  39         va_list ap;
  40
  41         va_start(ap, fmt);
  42         fprintf(stderr, "%s:%u: FAIL: ", filename, line);
  43         vfprintf(stderr, fmt, ap);
  44         fprintf(stderr, "\n");
  45         va_end(ap);
  46         exit(1);
  47 }
  48
  49 /* Try or die. */
  50 #define try(expr, expect)                                               \
  51         do {                                                            \
  52                 int ret = (expr);                                       \
  53                 if (ret != (expect))                                    \
  54                         fail(filename[file], i+1,                       \
  55                              STRINGIFY(expr) "= %i", ret);              \
  56         } while (0)
  57
  58 /* Try or imitate results. */
  59 #define unreliable(expr, expect, force, undo)                           \
  60         do {                                                            \
  61                 int ret = expr;                                         \
  62                 if (ret != expect) {                                    \
  63                         fprintf(stderr, "%s:%u: %s gave %i not %i",     \
  64                                 filename[file], i+1, STRINGIFY(expr),   \
  65                                 ret, expect);                           \
  66                         if (expect == 0)                                \
  67                                 force;                                  \
  68                         else                                            \
  69                                 undo;                                   \
  70                 }                                                       \
  71         } while (0)
  72
  73 static bool key_eq(TDB_DATA a, TDB_DATA b)
  74 {
  75         if (a.dsize != b.dsize)
  76                 return false;
  77         return memcmp(a.dptr, b.dptr, a.dsize) == 0;
  78 }
  79
  80 /* This is based on the hash algorithm from gdbm */
  81 static unsigned int hash_key(TDB_DATA *key)
  82 {
  83         uint32_t value; /* Used to compute the hash value.  */
  84         uint32_t   i;   /* Used to cycle through random values. */
  85
  86         /* Set the initial value from the key size. */
  87         for (value = 0x238F13AF ^ key->dsize, i=0; i < key->dsize; i++)
  88                 value = (value + (key->dptr[i] << (i*5 % 24)));
  89
  90         return (1103515243 * value + 12345);
  91 }
  92
  93 enum op_type {
  94         OP_TDB_LOCKALL,
  95         OP_TDB_LOCKALL_MARK,
  96         OP_TDB_LOCKALL_UNMARK,
  97         OP_TDB_LOCKALL_NONBLOCK,
  98         OP_TDB_UNLOCKALL,
  99         OP_TDB_LOCKALL_READ,
 100         OP_TDB_LOCKALL_READ_NONBLOCK,
 101         OP_TDB_UNLOCKALL_READ,
 102         OP_TDB_CHAINLOCK,
 103         OP_TDB_CHAINLOCK_NONBLOCK,
 104         OP_TDB_CHAINLOCK_MARK,
 105         OP_TDB_CHAINLOCK_UNMARK,
 106         OP_TDB_CHAINUNLOCK,
 107         OP_TDB_CHAINLOCK_READ,
 108         OP_TDB_CHAINUNLOCK_READ,
 109         OP_TDB_PARSE_RECORD,
 110         OP_TDB_EXISTS,
 111         OP_TDB_STORE,
 112         OP_TDB_APPEND,
 113         OP_TDB_GET_SEQNUM,
 114         OP_TDB_WIPE_ALL,
 115         OP_TDB_TRANSACTION_START,
 116         OP_TDB_TRANSACTION_CANCEL,
 117         OP_TDB_TRANSACTION_COMMIT,
 118         OP_TDB_TRAVERSE_READ_START,
 119         OP_TDB_TRAVERSE_START,
 120         OP_TDB_TRAVERSE_END,
 121         OP_TDB_TRAVERSE,
 122         OP_TDB_FIRSTKEY,
 123         OP_TDB_NEXTKEY,
 124         OP_TDB_FETCH,
 125         OP_TDB_DELETE,
 126 };
 127
 128 struct op {
 129         unsigned int serial;
 130         enum op_type op;
 131         TDB_DATA key;
 132         TDB_DATA data;
 133         int ret;
 134
 135         /* Who is waiting for us? */
 136         struct list_head post;
 137         /* What are we waiting for? */
 138         struct list_head pre;
 139
 140         /* If I'm part of a group (traverse/transaction) where is
 141          * start?  (Otherwise, 0) */
 142         unsigned int group_start;
 143
 144         union {
 145                 int flag; /* open and store */
 146                 struct {  /* append */
 147                         TDB_DATA pre;
 148                         TDB_DATA post;
 149                 } append;
 150                 unsigned int group_len; /* transaction/traverse start */
 151         };
 152 };
 153
 154 static unsigned char hex_char(const char *filename, unsigned int line, char c)
 155 {
 156         c = toupper(c);
 157         if (c >= 'A' && c <= 'F')
 158                 return c - 'A' + 10;
 159         if (c >= '0' && c <= '9')
 160                 return c - '0';
 161         fail(filename, line, "invalid hex character '%c'", c);
 162 }
 163
 164 /* TDB data is <size>:<%02x>* */
 165 static TDB_DATA make_tdb_data(const void *ctx,
 166                               const char *filename, unsigned int line,
 167                               const char *word)
 168 {
 169         TDB_DATA data;
 170         unsigned int i;
 171         const char *p;
 172
 173         if (streq(word, "NULL"))
 174                 return tdb_null;
 175
 176         data.dsize = atoi(word);
 177         data.dptr = talloc_array(ctx, unsigned char, data.dsize);
 178         p = strchr(word, ':');
 179         if (!p)
 180                 fail(filename, line, "invalid tdb data '%s'", word);
 181         p++;
 182         for (i = 0; i < data.dsize; i++)
 183                 data.dptr[i] = hex_char(filename, line, p[i*2])*16
 184                         + hex_char(filename, line, p[i*2+1]);
 185
 186         return data;
 187 }
 188
 189 static void add_op(const char *filename, struct op **op, unsigned int i,
 190                    unsigned int serial, enum op_type type)
 191 {
 192         struct op *new;
 193         *op = talloc_realloc(NULL, *op, struct op, i+1);
 194         new = (*op) + i;
 195         new->op = type;
 196         new->serial = serial;
 197         new->ret = 0;
 198         new->group_start = 0;
 199 }
 200
 201 static void op_add_nothing(const char *filename,
 202                            struct op op[], unsigned int op_num, char *words[])
 203 {
 204         if (words[2])
 205                 fail(filename, op_num+1, "Expected no arguments");
 206         op[op_num].key = tdb_null;
 207 }
 208
 209 static void op_add_key(const char *filename,
 210                        struct op op[], unsigned int op_num, char *words[])
 211 {
 212         if (words[2] == NULL || words[3])
 213                 fail(filename, op_num+1, "Expected just a key");
 214
 215         op[op_num].key = make_tdb_data(op, filename, op_num+1, words[2]);
 216         if (op[op_num].op != OP_TDB_TRAVERSE)
 217                 total_keys++;
 218 }
 219
 220 static void op_add_key_ret(const char *filename,
 221                            struct op op[], unsigned int op_num, char *words[])
 222 {
 223         if (!words[2] || !words[3] || !words[4] || words[5]
 224             || !streq(words[3], "="))
 225                 fail(filename, op_num+1, "Expected <key> = <ret>");
 226         op[op_num].ret = atoi(words[4]);
 227         op[op_num].key = make_tdb_data(op, filename, op_num+1, words[2]);
 228         /* May only be a unique key if it fails */
 229         if (op[op_num].ret != 0)
 230                 total_keys++;
 231 }
 232
 233 static void op_add_key_data(const char *filename,
 234                             struct op op[], unsigned int op_num, char *words[])
 235 {
 236         if (!words[2] || !words[3] || !words[4] || words[5]
 237             || !streq(words[3], "="))
 238                 fail(filename, op_num+1, "Expected <key> = <data>");
 239         op[op_num].key = make_tdb_data(op, filename, op_num+1, words[2]);
 240         op[op_num].data = make_tdb_data(op, filename, op_num+1, words[4]);
 241         /* May only be a unique key if it fails */
 242         if (!op[op_num].data.dptr)
 243                 total_keys++;
 244 }
 245
 246 /* <serial> tdb_store <rec> <rec> <flag> = <ret> */
 247 static void op_add_store(const char *filename,
 248                          struct op op[], unsigned int op_num, char *words[])
 249 {
 250         if (!words[2] || !words[3] || !words[4] || !words[5] || !words[6]
 251             || words[7] || !streq(words[5], "="))
 252                 fail(filename, op_num+1, "Expect <key> <data> <flag> = <ret>");
 253
 254         op[op_num].flag = strtoul(words[4], NULL, 0);
 255         op[op_num].ret = atoi(words[6]);
 256         op[op_num].key = make_tdb_data(op, filename, op_num+1, words[2]);
 257         op[op_num].data = make_tdb_data(op, filename, op_num+1, words[3]);
 258         total_keys++;
 259 }
 260
 261 /* <serial> tdb_append <rec> <rec> = <rec> */
 262 static void op_add_append(const char *filename,
 263                           struct op op[], unsigned int op_num, char *words[])
 264 {
 265         if (!words[2] || !words[3] || !words[4] || !words[5] || words[6]
 266             || !streq(words[4], "="))
 267                 fail(filename, op_num+1, "Expect <key> <data> = <rec>");
 268
 269         op[op_num].key = make_tdb_data(op, filename, op_num+1, words[2]);
 270         op[op_num].data = make_tdb_data(op, filename, op_num+1, words[3]);
 271
 272         op[op_num].append.post
 273                 = make_tdb_data(op, filename, op_num+1, words[5]);
 274
 275         /* By subtraction, figure out what previous data was. */
 276         op[op_num].append.pre.dptr = op[op_num].append.post.dptr;
 277         op[op_num].append.pre.dsize
 278                 = op[op_num].append.post.dsize - op[op_num].data.dsize;
 279         total_keys++;
 280 }
 281
 282 /* <serial> tdb_get_seqnum = <ret> */
 283 static void op_add_seqnum(const char *filename,
 284                           struct op op[], unsigned int op_num, char *words[])
 285 {
 286         if (!words[2] || !words[3] || words[4] || !streq(words[2], "="))
 287                 fail(filename, op_num+1, "Expect = <ret>");
 288
 289         op[op_num].key = tdb_null;
 290         op[op_num].ret = atoi(words[3]);
 291 }
 292
 293 static void op_add_traverse(const char *filename,
 294                             struct op op[], unsigned int op_num, char *words[])
 295 {
 296         if (words[2])
 297                 fail(filename, op_num+1, "Expect no arguments");
 298
 299         op[op_num].key = tdb_null;
 300         op[op_num].group_len = 0;
 301 }
 302
 303 static void op_add_transaction(const char *filename, struct op op[],
 304                                unsigned int op_num, char *words[])
 305 {
 306         if (words[2])
 307                 fail(filename, op_num+1, "Expect no arguments");
 308
 309         op[op_num].key = tdb_null;
 310         op[op_num].group_len = 0;
 311 }
 312
 313 static void op_analyze_transaction(const char *filename,
 314                                    struct op op[], unsigned int op_num,
 315                                    char *words[])
 316 {
 317         int i, start;
 318
 319         op[op_num].key = tdb_null;
 320
 321         if (words[2])
 322                 fail(filename, op_num+1, "Expect no arguments");
 323
 324         for (i = op_num-1; i >= 0; i--) {
 325                 if (op[i].op == OP_TDB_TRANSACTION_START && !op[i].group_len)
 326                         break;
 327         }
 328
 329         if (i < 0)
 330                 fail(filename, op_num+1, "no transaction start found");
 331
 332         start = i;
 333         op[start].group_len = op_num - i;
 334
 335         /* This rolls in nested transactions.  I think that's right. */
 336         for (i++; i <= op_num; i++)
 337                 op[i].group_start = start;
 338 }
 339
 340 struct traverse_hash {
 341         TDB_DATA key;
 342         unsigned int index;
 343 };
 344
 345 static void op_analyze_traverse(const char *filename,
 346                                 struct op op[], unsigned int op_num,
 347                                 char *words[])
 348 {
 349         int i, start;
 350
 351         op[op_num].key = tdb_null;
 352
 353         /* = %u means traverse function terminated. */
 354         if (words[2]) {
 355                 if (!streq(words[2], "=") || !words[3] || words[4])
 356                         fail(filename, op_num+1, "expect = <num>");
 357                 op[op_num].ret = atoi(words[3]);
 358         } else
 359                 op[op_num].ret = 0;
 360
 361         for (i = op_num-1; i >= 0; i--) {
 362                 if (op[i].op != OP_TDB_TRAVERSE_READ_START
 363                     && op[i].op != OP_TDB_TRAVERSE_START)
 364                         continue;
 365                 if (op[i].group_len)
 366                         continue;
 367                 break;
 368         }
 369
 370         if (i < 0)
 371                 fail(filename, op_num+1, "no traversal start found");
 372
 373         start = i;
 374         op[start].group_len = op_num - start;
 375
 376         for (i = start; i <= op_num; i++)
 377                 op[i].group_start = start;
 378 }
 379
 380 /* Keep -Wmissing-declarations happy: */
 381 const struct op_table *
 382 find_keyword (register const char *str, register unsigned int len);
 383
 384 #include "keywords.c"
 385
 386 struct depend {
 387         /* We can have more than one */
 388         struct list_node pre_list;
 389         struct list_node post_list;
 390         unsigned int needs_file;
 391         unsigned int needs_opnum;
 392         unsigned int satisfies_file;
 393         unsigned int satisfies_opnum;
 394 };
 395
 396 static void check_deps(const char *filename, struct op op[], unsigned int num)
 397 {
 398 #ifdef DEBUG_DEPS
 399         unsigned int i;
 400
 401         for (i = 1; i < num; i++)
 402                 if (!list_empty(&op[i].pre))
 403                         fail(filename, i+1, "Still has dependencies");
 404 #endif
 405 }
 406
 407 static void dump_pre(char *filename[], unsigned int file,
 408                      struct op op[], unsigned int i)
 409 {
 410         struct depend *dep;
 411
 412         printf("%s:%u still waiting for:\n", filename[file], i+1);
 413         list_for_each(&op[i].pre, dep, pre_list)
 414                 printf("    %s:%u\n",
 415                        filename[dep->satisfies_file], dep->satisfies_opnum+1);
 416         check_deps(filename[file], op, i);
 417 }
 418
 419 /* We simply read/write pointers, since we all are children. */
 420 static void do_pre(char *filename[], unsigned int file, int pre_fd,
 421                    struct op op[], unsigned int i)
 422 {
 423         while (!list_empty(&op[i].pre)) {
 424                 struct depend *dep;
 425
 426 #if DEBUG_DEPS
 427                 printf("%s:%u:waiting for pre\n", filename[file], i+1);
 428                 fflush(stdout);
 429 #endif
 430                 alarm(10);
 431                 while (read(pre_fd, &dep, sizeof(dep)) != sizeof(dep)) {
 432                         if (errno == EINTR) {
 433                                 dump_pre(filename, file, op, i);
 434                                 exit(1);
 435                         } else
 436                                 errx(1, "Reading from pipe");
 437                 }
 438                 alarm(0);
 439
 440 #if DEBUG_DEPS
 441                 printf("%s:%u:got pre %u from %s:%u\n", filename[file], i+1,
 442                        dep->needs_opnum+1, filename[dep->satisfies_file],
 443                        dep->satisfies_opnum+1);
 444                 fflush(stdout);
 445 #endif
 446                 /* This could be any op, not just this one. */
 447                 talloc_free(dep);
 448         }
 449 }
 450
 451 static void do_post(char *filename[], unsigned int file,
 452                     const struct op op[], unsigned int i)
 453 {
 454         struct depend *dep;
 455
 456         list_for_each(&op[i].post, dep, post_list) {
 457 #if DEBUG_DEPS
 458                 printf("%s:%u:sending to file %s:%u\n", filename[file], i+1,
 459                        filename[dep->needs_file], dep->needs_opnum+1);
 460 #endif
 461                 if (write(pipes[dep->needs_file].fd[1], &dep, sizeof(dep))
 462                     != sizeof(dep))
 463                         err(1, "%s:%u failed to tell file %s",
 464                             filename[file], i+1, filename[dep->needs_file]);
 465         }
 466 }
 467
 468 static int get_len(TDB_DATA key, TDB_DATA data, void *private_data)
 469 {
 470         return data.dsize;
 471 }
 472
 473 static unsigned run_ops(struct tdb_context *tdb,
 474                         int pre_fd,
 475                         char *filename[],
 476                         unsigned int file,
 477                         struct op op[],
 478                         unsigned int start, unsigned int stop);
 479
 480 struct traverse_info {
 481         struct op *op;
 482         char **filename;
 483         unsigned file;
 484         int pre_fd;
 485         unsigned int start;
 486         unsigned int i;
 487 };
 488
 489 /* More complex.  Just do whatever's they did at the n'th entry. */
 490 static int nontrivial_traverse(struct tdb_context *tdb,
 491                                TDB_DATA key, TDB_DATA data,
 492                                void *_tinfo)
 493 {
 494         struct traverse_info *tinfo = _tinfo;
 495         unsigned int trav_len = tinfo->op[tinfo->start].group_len;
 496
 497         if (tinfo->i == tinfo->start + trav_len) {
 498                 /* This can happen if traverse expects to be empty. */
 499                 if (trav_len == 1)
 500                         return 1;
 501                 fail(tinfo->filename[tinfo->file], tinfo->start + 1,
 502                      "traverse did not terminate");
 503         }
 504
 505         if (tinfo->op[tinfo->i].op != OP_TDB_TRAVERSE)
 506                 fail(tinfo->filename[tinfo->file], tinfo->start + 1,
 507                      "%s:%u:traverse terminated early");
 508
 509         /* Run any normal ops. */
 510         tinfo->i = run_ops(tdb, tinfo->pre_fd, tinfo->filename, tinfo->file,
 511                            tinfo->op, tinfo->i+1, tinfo->start + trav_len);
 512
 513         if (tinfo->i == tinfo->start + trav_len)
 514                 return 1;
 515
 516         return 0;
 517 }
 518
 519 static unsigned op_traverse(struct tdb_context *tdb,
 520                             int pre_fd,
 521                             char *filename[],
 522                             unsigned int file,
 523                             int (*traversefn)(struct tdb_context *,
 524                                               tdb_traverse_func, void *),
 525                             struct op op[],
 526                             unsigned int start)
 527 {
 528         struct traverse_info tinfo = { op, filename, file, pre_fd,
 529                                        start, start+1 };
 530
 531         traversefn(tdb, nontrivial_traverse, &tinfo);
 532
 533         /* Traversing in wrong order can have strange effects: eg. if
 534          * original traverse went A (delete A), B, we might do B
 535          * (delete A).  So if we have ops left over, we do it now. */
 536         while (tinfo.i != start + op[start].group_len) {
 537                 if (op[tinfo.i].op == OP_TDB_TRAVERSE)
 538                         tinfo.i++;
 539                 else
 540                         tinfo.i = run_ops(tdb, pre_fd, filename, file, op,
 541                                           tinfo.i, start + op[start].group_len);
 542         }
 543
 544         return tinfo.i;
 545 }
 546
 547 static void break_out(int sig)
 548 {
 549 }
 550
 551 static __attribute__((noinline))
 552 unsigned run_ops(struct tdb_context *tdb,
 553                  int pre_fd,
 554                  char *filename[],
 555                  unsigned int file,
 556                  struct op op[], unsigned int start, unsigned int stop)
 557 {
 558         unsigned int i;
 559         struct sigaction sa;
 560
 561         sa.sa_handler = break_out;
 562         sa.sa_flags = 0;
 563
 564         sigaction(SIGALRM, &sa, NULL);
 565         for (i = start; i < stop; i++) {
 566                 do_pre(filename, file, pre_fd, op, i);
 567
 568                 switch (op[i].op) {
 569                 case OP_TDB_LOCKALL:
 570                         try(tdb_lockall(tdb), op[i].ret);
 571                         break;
 572                 case OP_TDB_LOCKALL_MARK:
 573                         try(tdb_lockall_mark(tdb), op[i].ret);
 574                         break;
 575                 case OP_TDB_LOCKALL_UNMARK:
 576                         try(tdb_lockall_unmark(tdb), op[i].ret);
 577                         break;
 578                 case OP_TDB_LOCKALL_NONBLOCK:
 579                         unreliable(tdb_lockall_nonblock(tdb), op[i].ret,
 580                                    tdb_lockall(tdb), tdb_unlockall(tdb));
 581                         break;
 582                 case OP_TDB_UNLOCKALL:
 583                         try(tdb_unlockall(tdb), op[i].ret);
 584                         break;
 585                 case OP_TDB_LOCKALL_READ:
 586                         try(tdb_lockall_read(tdb), op[i].ret);
 587                         break;
 588                 case OP_TDB_LOCKALL_READ_NONBLOCK:
 589                         unreliable(tdb_lockall_read_nonblock(tdb), op[i].ret,
 590                                    tdb_lockall_read(tdb),
 591                                    tdb_unlockall_read(tdb));
 592                         break;
 593                 case OP_TDB_UNLOCKALL_READ:
 594                         try(tdb_unlockall_read(tdb), op[i].ret);
 595                         break;
 596                 case OP_TDB_CHAINLOCK:
 597                         try(tdb_chainlock(tdb, op[i].key), op[i].ret);
 598                         break;
 599                 case OP_TDB_CHAINLOCK_NONBLOCK:
 600                         unreliable(tdb_chainlock_nonblock(tdb, op[i].key),
 601                                    op[i].ret,
 602                                    tdb_chainlock(tdb, op[i].key),
 603                                    tdb_chainunlock(tdb, op[i].key));
 604                         break;
 605                 case OP_TDB_CHAINLOCK_MARK:
 606                         try(tdb_chainlock_mark(tdb, op[i].key), op[i].ret);
 607                         break;
 608                 case OP_TDB_CHAINLOCK_UNMARK:
 609                         try(tdb_chainlock_unmark(tdb, op[i].key), op[i].ret);
 610                         break;
 611                 case OP_TDB_CHAINUNLOCK:
 612                         try(tdb_chainunlock(tdb, op[i].key), op[i].ret);
 613                         break;
 614                 case OP_TDB_CHAINLOCK_READ:
 615                         try(tdb_chainlock_read(tdb, op[i].key), op[i].ret);
 616                         break;
 617                 case OP_TDB_CHAINUNLOCK_READ:
 618                         try(tdb_chainunlock_read(tdb, op[i].key), op[i].ret);
 619                         break;
 620                 case OP_TDB_PARSE_RECORD:
 621                         try(tdb_parse_record(tdb, op[i].key, get_len, NULL),
 622                             op[i].ret);
 623                         break;
 624                 case OP_TDB_EXISTS:
 625                         try(tdb_exists(tdb, op[i].key), op[i].ret);
 626                         break;
 627                 case OP_TDB_STORE:
 628                         try(tdb_store(tdb, op[i].key, op[i].data, op[i].flag),
 629                             op[i].ret);
 630                         break;
 631                 case OP_TDB_APPEND:
 632                         try(tdb_append(tdb, op[i].key, op[i].data), op[i].ret);
 633                         break;
 634                 case OP_TDB_GET_SEQNUM:
 635                         try(tdb_get_seqnum(tdb), op[i].ret);
 636                         break;
 637                 case OP_TDB_WIPE_ALL:
 638                         try(tdb_wipe_all(tdb), op[i].ret);
 639                         break;
 640                 case OP_TDB_TRANSACTION_START:
 641                         try(tdb_transaction_start(tdb), op[i].ret);
 642                         break;
 643                 case OP_TDB_TRANSACTION_CANCEL:
 644                         try(tdb_transaction_cancel(tdb), op[i].ret);
 645                         break;
 646                 case OP_TDB_TRANSACTION_COMMIT:
 647                         try(tdb_transaction_commit(tdb), op[i].ret);
 648                         break;
 649                 case OP_TDB_TRAVERSE_READ_START:
 650                         i = op_traverse(tdb, pre_fd, filename, file,
 651                                         tdb_traverse_read, op, i);
 652                         break;
 653                 case OP_TDB_TRAVERSE_START:
 654                         i = op_traverse(tdb, pre_fd, filename, file,
 655                                         tdb_traverse, op, i);
 656                         break;
 657                 case OP_TDB_TRAVERSE:
 658                         /* Terminate: we're in a traverse, and we've
 659                          * done our ops. */
 660                         return i;
 661                 case OP_TDB_TRAVERSE_END:
 662                         fail(filename[file], i+1, "unexpected end traverse");
 663                 /* FIXME: These must be treated like traverse. */
 664                 case OP_TDB_FIRSTKEY:
 665                         if (!key_eq(tdb_firstkey(tdb), op[i].data))
 666                                 fail(filename[file], i+1, "bad firstkey");
 667                         break;
 668                 case OP_TDB_NEXTKEY:
 669                         if (!key_eq(tdb_nextkey(tdb, op[i].key), op[i].data))
 670                                 fail(filename[file], i+1, "bad nextkey");
 671                         break;
 672                 case OP_TDB_FETCH: {
 673                         TDB_DATA f = tdb_fetch(tdb, op[i].key);
 674                         if (!key_eq(f, op[i].data))
 675                                 fail(filename[file], i+1, "bad fetch %u",
 676                                      f.dsize);
 677                         break;
 678                 }
 679                 case OP_TDB_DELETE:
 680                         try(tdb_delete(tdb, op[i].key), op[i].ret);
 681                         break;
 682                 }
 683                 do_post(filename, file, op, i);
 684         }
 685         return i;
 686 }
 687
 688 static struct op *load_tracefile(const char *filename, unsigned int *num,
 689                                  unsigned int *hashsize,
 690                                  unsigned int *tdb_flags,
 691                                  unsigned int *open_flags)
 692 {
 693         unsigned int i;
 694         struct op *op = talloc_array(NULL, struct op, 1);
 695         char **words;
 696         char **lines;
 697         char *file;
 698
 699         file = grab_file(NULL, filename, NULL);
 700         if (!file)
 701                 err(1, "Reading %s", filename);
 702
 703         lines = strsplit(file, file, "\n", NULL);
 704         if (!lines[0])
 705                 errx(1, "%s is empty", filename);
 706
 707         words = strsplit(lines, lines[0], " ", NULL);
 708         if (!streq(words[1], "tdb_open"))
 709                 fail(filename, 1, "does not start with tdb_open");
 710
 711         *hashsize = atoi(words[2]);
 712         *tdb_flags = strtoul(words[3], NULL, 0);
 713         *open_flags = strtoul(words[4], NULL, 0);
 714
 715         for (i = 1; lines[i]; i++) {
 716                 const struct op_table *opt;
 717
 718                 words = strsplit(lines, lines[i], " ", NULL);
 719                 if (!words[0] || !words[1])
 720                         fail(filename, i+1, "Expected serial number and op");
 721
 722                 opt = find_keyword(words[1], strlen(words[1]));
 723                 if (!opt) {
 724                         if (streq(words[1], "tdb_close")) {
 725                                 if (lines[i+1])
 726                                         fail(filename, i+2,
 727                                              "lines after tdb_close");
 728                                 *num = i;
 729                                 talloc_free(lines);
 730                                 return op;
 731                         }
 732                         fail(filename, i+1, "Unknown operation '%s'", words[1]);
 733                 }
 734
 735                 add_op(filename, &op, i, atoi(words[0]), opt->type);
 736                 opt->enhance_op(filename, op, i, words);
 737         }
 738
 739         fprintf(stderr, "%s:%u:last operation is not tdb_close: incomplete?",
 740               filename, i);
 741         talloc_free(lines);
 742         *num = i - 1;
 743         return op;
 744 }
 745
 746 /* We remember all the keys we've ever seen, and who has them. */
 747 struct key_user {
 748         unsigned int file;
 749         unsigned int op_num;
 750 };
 751
 752 struct keyinfo {
 753         TDB_DATA key;
 754         unsigned int num_users;
 755         struct key_user *user;
 756 };
 757
 758 static const TDB_DATA must_not_exist;
 759 static const TDB_DATA must_exist;
 760 static const TDB_DATA not_exists_or_empty;
 761
 762 /* NULL means doesn't care if it exists or not, &must_exist means
 763  * it must exist but we don't care what, &must_not_exist means it must
 764  * not exist, otherwise the data it needs. */
 765 static const TDB_DATA *needs(const struct op *op)
 766 {
 767         switch (op->op) {
 768         /* FIXME: Pull forward deps, since we can deadlock */
 769         case OP_TDB_CHAINLOCK:
 770         case OP_TDB_CHAINLOCK_NONBLOCK:
 771         case OP_TDB_CHAINLOCK_MARK:
 772         case OP_TDB_CHAINLOCK_UNMARK:
 773         case OP_TDB_CHAINUNLOCK:
 774         case OP_TDB_CHAINLOCK_READ:
 775         case OP_TDB_CHAINUNLOCK_READ:
 776                 return NULL;
 777
 778         case OP_TDB_APPEND:
 779                 if (op->append.pre.dsize == 0)
 780                         return &not_exists_or_empty;
 781                 return &op->append.pre;
 782
 783         case OP_TDB_STORE:
 784                 if (op->flag == TDB_INSERT) {
 785                         if (op->ret < 0)
 786                                 return &must_exist;
 787                         else
 788                                 return &must_not_exist;
 789                 } else if (op->flag == TDB_MODIFY) {
 790                         if (op->ret < 0)
 791                                 return &must_not_exist;
 792                         else
 793                                 return &must_exist;
 794                 }
 795                 /* No flags?  Don't care */
 796                 return NULL;
 797
 798         case OP_TDB_EXISTS:
 799                 if (op->ret == 1)
 800                         return &must_exist;
 801                 else
 802                         return &must_not_exist;
 803
 804         case OP_TDB_PARSE_RECORD:
 805                 if (op->ret < 0)
 806                         return &must_not_exist;
 807                 return &must_exist;
 808
 809         /* FIXME: handle these. */
 810         case OP_TDB_WIPE_ALL:
 811         case OP_TDB_FIRSTKEY:
 812         case OP_TDB_NEXTKEY:
 813         case OP_TDB_GET_SEQNUM:
 814         case OP_TDB_TRAVERSE:
 815         case OP_TDB_TRANSACTION_COMMIT:
 816         case OP_TDB_TRANSACTION_CANCEL:
 817         case OP_TDB_TRANSACTION_START:
 818                 return NULL;
 819
 820         case OP_TDB_FETCH:
 821                 if (!op->data.dptr)
 822                         return &must_not_exist;
 823                 return &op->data;
 824
 825         case OP_TDB_DELETE:
 826                 if (op->ret < 0)
 827                         return &must_not_exist;
 828                 return &must_exist;
 829
 830         default:
 831                 errx(1, "Unexpected op %i", op->op);
 832         }
 833
 834 }
 835
 836 /* What's the data after this op?  pre if nothing changed. */
 837 static const TDB_DATA *gives(const struct op *op, const TDB_DATA *pre)
 838 {
 839         /* Failed ops don't change state of db. */
 840         if (op->ret < 0)
 841                 return pre;
 842
 843         if (op->op == OP_TDB_DELETE || op->op == OP_TDB_WIPE_ALL)
 844                 return &tdb_null;
 845
 846         if (op->op == OP_TDB_APPEND)
 847                 return &op->append.post;
 848
 849         if (op->op == OP_TDB_STORE)
 850                 return &op->data;
 851
 852         return pre;
 853 }
 854
 855 static struct keyinfo *hash_ops(struct op *op[], unsigned int num_ops[],
 856                                 unsigned int num)
 857 {
 858         unsigned int i, j, h;
 859         struct keyinfo *hash;
 860
 861         hash = talloc_zero_array(op[0], struct keyinfo, total_keys*2);
 862         for (i = 0; i < num; i++) {
 863                 for (j = 1; j < num_ops[i]; j++) {
 864                         /* We can't do this on allocation, due to realloc. */
 865                         list_head_init(&op[i][j].post);
 866                         list_head_init(&op[i][j].pre);
 867
 868                         if (!op[i][j].key.dptr)
 869                                 continue;
 870
 871                         /* We don't wait for traverse keys */
 872                         /* FIXME: We should, for trivial traversals. */
 873                         if (op[i][j].op == OP_TDB_TRAVERSE)
 874                                 continue;
 875
 876                         h = hash_key(&op[i][j].key) % (total_keys * 2);
 877                         while (!key_eq(hash[h].key, op[i][j].key)) {
 878                                 if (!hash[h].key.dptr) {
 879                                         hash[h].key = op[i][j].key;
 880                                         break;
 881                                 }
 882                                 h = (h + 1) % (total_keys * 2);
 883                         }
 884                         /* Might as well save some memory if we can. */
 885                         if (op[i][j].key.dptr != hash[h].key.dptr) {
 886                                 talloc_free(op[i][j].key.dptr);
 887                                 op[i][j].key.dptr = hash[h].key.dptr;
 888                         }
 889                         hash[h].user = talloc_realloc(hash, hash[h].user,
 890                                                      struct key_user,
 891                                                      hash[h].num_users+1);
 892                         hash[h].user[hash[h].num_users].op_num = j;
 893                         hash[h].user[hash[h].num_users].file = i;
 894                         hash[h].num_users++;
 895                 }
 896         }
 897
 898         return hash;
 899 }
 900
 901 static bool satisfies(const TDB_DATA *data, const TDB_DATA *need)
 902 {
 903         /* Don't need anything?  Cool. */
 904         if (!need)
 905                 return true;
 906
 907         /* This should be tdb_null or a real value. */
 908         assert(data != &must_exist);
 909         assert(data != &must_not_exist);
 910         assert(data != &not_exists_or_empty);
 911
 912         /* must_not_exist == must_not_exist, must_exist == must_exist, or
 913            not_exists_or_empty == not_exists_or_empty. */
 914         if (data->dsize == need->dsize && data->dptr == need->dptr)
 915                 return true;
 916
 917         /* Must not exist?  data must not exist. */
 918         if (need == &must_not_exist)
 919                 return data->dptr == NULL;
 920
 921         /* Must exist? */
 922         if (need == &must_exist)
 923                 return data->dptr != NULL;
 924
 925         /* Either noexist or empty. */
 926         if (need == &not_exists_or_empty)
 927                 return data->dsize == 0;
 928
 929         /* Needs something specific. */
 930         return key_eq(*data, *need);
 931 }
 932
 933 static void move_to_front(struct key_user res[], unsigned int elem)
 934 {
 935         if (elem != 0) {
 936                 struct key_user tmp = res[elem];
 937                 memmove(res + 1, res, elem*sizeof(res[0]));
 938                 res[0] = tmp;
 939         }
 940 }
 941
 942 static void restore_to_pos(struct key_user res[], unsigned int elem)
 943 {
 944         if (elem != 0) {
 945                 struct key_user tmp = res[0];
 946                 memmove(res, res + 1, elem*sizeof(res[0]));
 947                 res[elem] = tmp;
 948         }
 949 }
 950
 951 static bool sort_deps(char *filename[], struct op *op[],
 952                       struct key_user res[], unsigned num,
 953                       const TDB_DATA *data, unsigned num_files)
 954 {
 955         unsigned int i, files_done;
 956         struct op *this_op;
 957         bool done[num_files];
 958
 959         /* Nothing left?  We're sorted. */
 960         if (num == 0)
 961                 return true;
 962
 963         memset(done, 0, sizeof(done));
 964
 965         /* Since ops within a trace file are ordered, we just need to figure
 966          * out which file to try next.  Since we don't take into account
 967          * inter-key relationships (which exist by virtue of trace file order),
 968          * we minimize the chance of harm by trying to keep in serial order. */
 969         for (files_done = 0, i = 0; i < num && files_done < num_files; i++) {
 970                 if (done[res[i].file])
 971                         continue;
 972
 973                 this_op = &op[res[i].file][res[i].op_num];
 974                 /* Is what we have good enough for this op? */
 975                 if (satisfies(data, needs(this_op))) {
 976                         move_to_front(res, i);
 977                         if (sort_deps(filename, op, res+1, num-1,
 978                                       gives(this_op, data), num_files))
 979                                 return true;
 980                         restore_to_pos(res, i);
 981                 }
 982                 done[res[i].file] = true;
 983                 files_done++;
 984         }
 985
 986         /* No combination worked. */
 987         return false;
 988 }
 989
 990 static void check_dep_sorting(struct key_user user[], unsigned num_users,
 991                               unsigned num_files)
 992 {
 993 #if DEBUG_DEPS
 994         unsigned int i;
 995         unsigned minima[num_files];
 996
 997         memset(minima, 0, sizeof(minima));
 998         for (i = 0; i < num_users; i++) {
 999                 assert(minima[user[i].file] < user[i].op_num);
1000                 minima[user[i].file] = user[i].op_num;
1001         }
1002 #endif
1003 }
1004
1005 /* All these ops have the same serial number.  Which comes first?
1006  *
1007  * This can happen both because read ops or failed write ops don't
1008  * change serial number, and also due to race since we access the
1009  * number unlocked (the race can cause less detectable ordering problems,
1010  * in which case we'll deadlock and report: fix manually in that case).
1011  */
1012 static void figure_deps(char *filename[], struct op *op[],
1013                         struct key_user user[], unsigned num_users,
1014                         unsigned num_files)
1015 {
1016         /* We assume database starts empty. */
1017         const struct TDB_DATA *data = &tdb_null;
1018
1019         if (!sort_deps(filename, op, user, num_users, data, num_files))
1020                 fail(filename[user[0].file], user[0].op_num+1,
1021                      "Could not resolve inter-dependencies");
1022
1023         check_dep_sorting(user, num_users, num_files);
1024 }
1025
1026 static void sort_ops(struct keyinfo hash[], char *filename[], struct op *op[],
1027                      unsigned int num)
1028 {
1029         unsigned int h;
1030
1031         /* Gcc nexted function extension.  How cool is this? */
1032         int compare_serial(const void *_a, const void *_b)
1033         {
1034                 const struct key_user *a = _a, *b = _b;
1035
1036                 /* First, maintain order within any trace file. */
1037                 if (a->file == b->file)
1038                         return a->op_num - b->op_num;
1039
1040                 /* Otherwise, arrange by serial order. */
1041                 return op[a->file][a->op_num].serial
1042                         - op[b->file][b->op_num].serial;
1043         }
1044
1045         /* Now sort into serial order. */
1046         for (h = 0; h < total_keys * 2; h++) {
1047                 struct key_user *user = hash[h].user;
1048
1049                 qsort(user, hash[h].num_users, sizeof(user[0]), compare_serial);
1050                 figure_deps(filename, op, user, hash[h].num_users, num);
1051         }
1052 }
1053
1054 static int destroy_depend(struct depend *dep)
1055 {
1056         list_del(&dep->pre_list);
1057         list_del(&dep->post_list);
1058         return 0;
1059 }
1060
1061 static void add_dependency(void *ctx,
1062                            struct op *op[],
1063                            char *filename[],
1064                            unsigned int needs_file,
1065                            unsigned int needs_opnum,
1066                            unsigned int satisfies_file,
1067                            unsigned int satisfies_opnum)
1068 {
1069         struct depend *dep;
1070         unsigned int needs_start, sat_start;
1071
1072         /* We don't depend on ourselves. */
1073         if (needs_file == satisfies_file) {
1074                 assert(satisfies_opnum < needs_opnum);
1075                 return;
1076         }
1077
1078 #if DEBUG_DEPS
1079         printf("%s:%u: depends on %s:%u\n",
1080                filename[needs_file], needs_opnum+1,
1081                filename[satisfies_file], satisfies_opnum+1);
1082 #endif
1083
1084         needs_start = op[needs_file][needs_opnum].group_start;
1085         sat_start = op[satisfies_file][satisfies_opnum].group_start;
1086
1087         /* If needs is in a transaction, we need it before start. */
1088         if (needs_start) {
1089                 switch (op[needs_file][needs_start].op) {
1090                 case OP_TDB_TRANSACTION_START:
1091                         needs_opnum = needs_start;
1092 #ifdef DEBUG_DEPS
1093                         printf("  -> Back to %u\n", needs_start+1);
1094                         fflush(stdout);
1095 #endif
1096                         break;
1097                 default:
1098                         break;
1099                 }
1100         }
1101
1102         /* If satisfies is in a transaction, we wait until after commit. */
1103         /* FIXME: If transaction is cancelled, don't need dependency. */
1104         if (sat_start) {
1105                 if (op[satisfies_file][sat_start].op
1106                     == OP_TDB_TRANSACTION_START) {
1107                         satisfies_opnum = sat_start
1108                                 + op[satisfies_file][sat_start].group_len;
1109 #ifdef DEBUG_DEPS
1110                         printf("  -> Depends on %u\n", satisfies_opnum+1);
1111                         fflush(stdout);
1112 #endif
1113                 }
1114         }
1115
1116         assert(op[needs_file][needs_opnum].op != OP_TDB_TRAVERSE);
1117         assert(op[satisfies_file][satisfies_opnum].op != OP_TDB_TRAVERSE);
1118
1119         dep = talloc(ctx, struct depend);
1120         dep->needs_file = needs_file;
1121         dep->needs_opnum = needs_opnum;
1122         dep->satisfies_file = satisfies_file;
1123         dep->satisfies_opnum = satisfies_opnum;
1124         list_add(&op[satisfies_file][satisfies_opnum].post, &dep->post_list);
1125         list_add(&op[needs_file][needs_opnum].pre, &dep->pre_list);
1126         talloc_set_destructor(dep, destroy_depend);
1127 }
1128
1129 #if TRAVERSALS_TAKE_TRANSACTION_LOCK
1130 struct traverse_dep {
1131         unsigned int file;
1132         unsigned int op_num;
1133 };
1134
1135 /* Traversals can deadlock against each other.  Force order. */
1136 static void make_traverse_depends(char *filename[],
1137                                   struct op *op[], unsigned int num_ops[],
1138                                   unsigned int num)
1139 {
1140         unsigned int i, j, num_traversals = 0;
1141         struct traverse_dep *dep;
1142
1143         /* Sort by which one runs first. */
1144         int compare_traverse_dep(const void *_a, const void *_b)
1145         {
1146                 const struct traverse_dep *ta = _a, *tb = _b;
1147                 const struct op *a = &op[ta->file][ta->op_num],
1148                         *b = &op[tb->file][tb->op_num];
1149
1150                 if (a->serial != b->serial)
1151                         return a->serial - b->serial;
1152
1153                 /* If they have same serial, it means one didn't make any
1154                  * changes.  Thus sort by end in that case. */
1155                 return a[a->group_len].serial - b[b->group_len].serial;
1156         }
1157
1158         dep = talloc_array(NULL, struct traverse_dep, 1);
1159
1160         /* Count them. */
1161         for (i = 0; i < num; i++) {
1162                 for (j = 1; j < num_ops[i]; j++) {
1163                         /* Transaction on traverse start. */
1164                         if (op[i][j].group_start == j) {
1165                                 dep = talloc_realloc(NULL, dep,
1166                                                      struct traverse_dep,
1167                                                      num_traversals+1);
1168                                 dep[num_traversals].file = i;
1169                                 dep[num_traversals].op_num = j;
1170                                 num_traversals++;
1171                         }
1172                 }
1173         }
1174         qsort(dep, num_traversals, sizeof(dep[0]), compare_traverse_dep);
1175         for (i = 1; i < num_traversals; i++) {
1176                 /* i depends on end of traverse i-1. */
1177                 add_dependency(NULL, op, filename, dep[i].file, dep[i].op_num,
1178                                dep[i-1].file, dep[i-1].op_num
1179                                + op[dep[i-1].file][dep[i-1].op_num].group_len);
1180         }
1181         talloc_free(dep);
1182 }
1183 #endif /* TRAVERSALS_TAKE_TRANSACTION_LOCK */
1184
1185 static bool changes_db(const struct op *op)
1186 {
1187         return gives(op, NULL) != NULL;
1188 }
1189
1190 static void depend_on_previous(struct op *op[],
1191                                char *filename[],
1192                                unsigned int num,
1193                                struct key_user user[],
1194                                unsigned int i,
1195                                int prev)
1196 {
1197         bool deps[num];
1198         int j;
1199
1200         if (i == 0)
1201                 return;
1202
1203         if (prev == i - 1) {
1204                 /* Just depend on previous. */
1205                 add_dependency(NULL, op, filename,
1206                                user[i].file, user[i].op_num,
1207                                user[prev].file, user[prev].op_num);
1208                 return;
1209         }
1210
1211         /* We have to wait for the readers.  Find last one in *each* file. */
1212         memset(deps, 0, sizeof(deps));
1213         deps[user[i].file] = true;
1214         for (j = i - 1; j > prev; j--) {
1215                 if (!deps[user[j].file]) {
1216                         add_dependency(NULL, op, filename,
1217                                        user[i].file, user[i].op_num,
1218                                        user[j].file, user[j].op_num);
1219                         deps[user[j].file] = true;
1220                 }
1221         }
1222 }
1223
1224 /* This is simple, but not complete.  We don't take into account
1225  * indirect dependencies. */
1226 static void optimize_dependencies(struct op *op[], unsigned int num_ops[],
1227                                   unsigned int num)
1228 {
1229         unsigned int i, j;
1230
1231         /* There can only be one real dependency on each file */
1232         for (i = 0; i < num; i++) {
1233                 for (j = 1; j < num_ops[i]; j++) {
1234                         struct depend *dep, *next;
1235                         struct depend *prev[num];
1236
1237                         memset(prev, 0, sizeof(prev));
1238
1239                         list_for_each_safe(&op[i][j].pre, dep, next, pre_list) {
1240                                 if (!prev[dep->satisfies_file]) {
1241                                         prev[dep->satisfies_file] = dep;
1242                                         continue;
1243                                 }
1244                                 if (prev[dep->satisfies_file]->satisfies_opnum
1245                                     < dep->satisfies_opnum) {
1246                                         talloc_free(prev[dep->satisfies_file]);
1247                                         prev[dep->satisfies_file] = dep;
1248                                 } else
1249                                         talloc_free(dep);
1250                         }
1251                 }
1252         }
1253
1254         for (i = 0; i < num; i++) {
1255                 int deps[num];
1256
1257                 for (j = 0; j < num; j++)
1258                         deps[j] = -1;
1259
1260                 for (j = 1; j < num_ops[i]; j++) {
1261                         struct depend *dep, *next;
1262
1263                         list_for_each_safe(&op[i][j].pre, dep, next, pre_list) {
1264                                 if (deps[dep->satisfies_file]
1265                                     >= (int)dep->satisfies_opnum)
1266                                         talloc_free(dep);
1267                                 else
1268                                         deps[dep->satisfies_file]
1269                                                 = dep->satisfies_opnum;
1270                         }
1271                 }
1272         }
1273 }
1274
1275 static void derive_dependencies(char *filename[],
1276                                 struct op *op[], unsigned int num_ops[],
1277                                 unsigned int num)
1278 {
1279         struct keyinfo *hash;
1280         unsigned int h, i;
1281
1282         /* Create hash table for faster key lookup. */
1283         hash = hash_ops(op, num_ops, num);
1284
1285         /* Sort them by serial number. */
1286         sort_ops(hash, filename, op, num);
1287
1288         /* Create dependencies back to the last change, rather than
1289          * creating false dependencies by naively making each one
1290          * depend on the previous.  This has two purposes: it makes
1291          * later optimization simpler, and it also avoids deadlock with
1292          * same sequence number ops inside traversals (if one
1293          * traversal doesn't write anything, two ops can have the same
1294          * sequence number yet we can create a traversal dependency
1295          * the other way). */
1296         for (h = 0; h < total_keys * 2; h++) {
1297                 int prev = -1;
1298
1299                 if (hash[h].num_users < 2)
1300                         continue;
1301
1302                 for (i = 0; i < hash[h].num_users; i++) {
1303                         if (changes_db(&op[hash[h].user[i].file]
1304                                        [hash[h].user[i].op_num])) {
1305                                 depend_on_previous(op, filename, num,
1306                                                    hash[h].user, i, prev);
1307                                 prev = i;
1308                         } else if (prev >= 0)
1309                                 add_dependency(hash, op, filename,
1310                                                hash[h].user[i].file,
1311                                                hash[h].user[i].op_num,
1312                                                hash[h].user[prev].file,
1313                                                hash[h].user[prev].op_num);
1314                 }
1315         }
1316
1317 #if TRAVERSALS_TAKE_TRANSACTION_LOCK
1318         make_traverse_depends(filename, op, num_ops, num);
1319 #endif
1320
1321         optimize_dependencies(op, num_ops, num);
1322 }
1323
1324 int main(int argc, char *argv[])
1325 {
1326         struct timeval start, end;
1327         unsigned int i, num_ops[argc], hashsize[argc], tdb_flags[argc], open_flags[argc];
1328         struct op *op[argc];
1329         int fds[2];
1330         char c;
1331         bool ok = true;
1332
1333         if (argc < 3)
1334                 errx(1, "Usage: %s <tdbfile> <tracefile>...", argv[0]);
1335
1336         pipes = talloc_array(NULL, struct pipe, argc - 2);
1337         for (i = 0; i < argc - 2; i++) {
1338                 printf("Loading tracefile %s...", argv[2+i]);
1339                 fflush(stdout);
1340                 op[i] = load_tracefile(argv[2+i], &num_ops[i], &hashsize[i],
1341                                        &tdb_flags[i], &open_flags[i]);
1342                 if (pipe(pipes[i].fd) != 0)
1343                         err(1, "creating pipe");
1344                 printf("done\n");
1345         }
1346
1347         printf("Calculating inter-dependencies...");
1348         fflush(stdout);
1349         derive_dependencies(argv+2, op, num_ops, i);
1350         printf("done\n");
1351
1352         /* Don't fork for single arg case: simple debugging. */
1353         if (argc == 3) {
1354                 struct tdb_context *tdb;
1355                 tdb = tdb_open_ex(argv[1], hashsize[0], tdb_flags[0],
1356                                   open_flags[0], 0600,
1357                                   NULL, hash_key);
1358                 printf("Single threaded run...");
1359                 fflush(stdout);
1360
1361                 run_ops(tdb, pipes[0].fd[0], argv+2, 0, op[0], 1, num_ops[0]);
1362                 check_deps(argv[2], op[0], num_ops[0]);
1363
1364                 printf("done\n");
1365                 exit(0);
1366         }
1367
1368         if (pipe(fds) != 0)
1369                 err(1, "creating pipe");
1370
1371         for (i = 0; i < argc - 2; i++) {
1372                 struct tdb_context *tdb;
1373
1374                 switch (fork()) {
1375                 case -1:
1376                         err(1, "fork failed");
1377                 case 0:
1378                         close(fds[1]);
1379                         tdb = tdb_open_ex(argv[1], hashsize[i], tdb_flags[i],
1380                                           open_flags[i], 0600,
1381                                           NULL, hash_key);
1382                         if (!tdb)
1383                                 err(1, "Opening tdb %s", argv[1]);
1384
1385                         /* This catches parent exiting. */
1386                         if (read(fds[0], &c, 1) != 1)
1387                                 exit(1);
1388                         run_ops(tdb, pipes[i].fd[0], argv+2, i, op[i], 1,
1389                                 num_ops[i]);
1390                         check_deps(argv[2+i], op[i], num_ops[i]);
1391                         exit(0);
1392                 default:
1393                         break;
1394                 }
1395         }
1396
1397         /* Let everything settle. */
1398         sleep(1);
1399
1400         printf("Starting run...");
1401         fflush(stdout);
1402         gettimeofday(&start, NULL);
1403         /* Tell them all to go!  Any write of sufficient length will do. */
1404         if (write(fds[1], hashsize, i) != i)
1405                 err(1, "Writing to wakeup pipe");
1406
1407         for (i = 0; i < argc - 2; i++) {
1408                 int status;
1409                 wait(&status);
1410                 if (!WIFEXITED(status)) {
1411                         warnx("Child died with signal %i", WTERMSIG(status));
1412                         ok = false;
1413                 } else if (WEXITSTATUS(status) != 0)
1414                         /* Assume child spat out error. */
1415                         ok = false;
1416         }
1417         if (!ok)
1418                 exit(1);
1419
1420         gettimeofday(&end, NULL);
1421         printf("done\n");
1422
1423         end.tv_sec -= start.tv_sec;
1424         printf("Time replaying: %lu usec\n",
1425                end.tv_sec * 1000000UL + (end.tv_usec - start.tv_usec));
1426
1427         exit(0);
1428 }