dbm/src/hash.c

   1 /*-
   2  * Copyright (c) 1990, 1993, 1994
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Margo Seltzer.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. ***REMOVED*** - see
  17  *    ftp://ftp.cs.berkeley.edu/pub/4bsd/README.Impt.License.Change
  18  * 4. Neither the name of the University nor the names of its contributors
  19  *    may be used to endorse or promote products derived from this software
  20  *    without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  */
  34
  35 #if defined(LIBC_SCCS) && !defined(lint)
  36 static char sccsid[] = "@(#)hash.c      8.9 (Berkeley) 6/16/94";
  37 #endif /* LIBC_SCCS and not lint */
  38
  39 #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh)
  40 #include <sys/param.h>
  41 #endif
  42
  43 #if !defined(macintosh)
  44 #ifdef XP_OS2
  45 #include <sys/types.h>
  46 #endif
  47 #include <sys/stat.h>
  48 #endif
  49
  50 #if defined(macintosh)
  51 #include <unix.h>
  52 #include <unistd.h>
  53 #endif
  54
  55 #include <errno.h>
  56 #include <fcntl.h>
  57 #include <stdio.h>
  58 #include <stdlib.h>
  59 #include <string.h>
  60
  61 #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh)
  62 #include <unistd.h>
  63 #endif
  64 #if defined(_WIN32) || defined(_WINDOWS)
  65 #include <windows.h>
  66 #endif
  67
  68 #include <assert.h>
  69
  70 #include "mcom_db.h"
  71 #include "hash.h"
  72 #include "page.h"
  73
  74 /*
  75 #include "extern.h"
  76 */
  77 static int   alloc_segs __P((HTAB *, int));
  78 static int   flush_meta __P((HTAB *));
  79 static int   hash_access __P((HTAB *, ACTION, DBT *, DBT *));
  80 static int   hash_close __P((DB *));
  81 static int   hash_delete __P((const DB *, const DBT *, uint));
  82 static int   hash_fd __P((const DB *));
  83 static int   hash_get __P((const DB *, const DBT *, DBT *, uint));
  84 static int   hash_put __P((const DB *, DBT *, const DBT *, uint));
  85 static void *hash_realloc __P((SEGMENT **, size_t, size_t));
  86 static int   hash_seq __P((const DB *, DBT *, DBT *, uint));
  87 static int   hash_sync __P((const DB *, uint));
  88 static int   hdestroy __P((HTAB *));
  89 static HTAB *init_hash __P((HTAB *, const char *, HASHINFO *));
  90 static int   init_htab __P((HTAB *, int));
  91 #if BYTE_ORDER == LITTLE_ENDIAN
  92 static void  swap_header __P((HTAB *));
  93 static void  swap_header_copy __P((HASHHDR *, HASHHDR *));
  94 #endif
  95
  96 /* Fast arithmetic, relying on powers of 2, */
  97 #define MOD(x, y)               ((x) & ((y) - 1))
  98
  99 #define RETURN_ERROR(ERR, LOC)  { save_errno = ERR; goto LOC; }
 100
 101 /* Return values */
 102 #define SUCCESS  (0)
 103 #define DBM_ERROR       (-1)
 104 #define ABNORMAL (1)
 105
 106 #ifdef HASH_STATISTICS
 107 int hash_accesses, hash_collisions, hash_expansions, hash_overflows;
 108 #endif
 109
 110 /* A new Lou (montulli@mozilla.com) routine.
 111  *
 112  * The database is screwed.
 113  *
 114  * This closes the file, flushing buffers as appropriate.
 115  */
 116 static void
 117 __remove_database(DB *dbp)
 118 {
 119         HTAB *hashp = (HTAB *)dbp->internal;
 120
 121         assert(0);
 122
 123         if (!hashp)
 124                 return;
 125         hdestroy(hashp);
 126         dbp->internal = NULL;
 127 }
 128
 129 /************************** INTERFACE ROUTINES ***************************/
 130 /* OPEN/CLOSE */
 131
 132
 133 extern DB *
 134 __hash_open(const char *file, int flags, int mode, const HASHINFO *info, int dflags)
 135 {
 136         HTAB *hashp=NULL;
 137         struct stat statbuf;
 138         DB *dbp;
 139         int bpages, hdrsize, new_table, nsegs, save_errno;
 140
 141         if ((flags & O_ACCMODE) == O_WRONLY) {
 142                 errno = EINVAL;
 143                 return NULL;
 144         }
 145
 146         /* zero the statbuffer so that
 147          * we can check it for a non-zero
 148          * date to see if stat succeeded
 149          */
 150         memset(&statbuf, 0, sizeof(struct stat));
 151
 152         if (!(hashp = (HTAB *)calloc(1, sizeof(HTAB)))) {
 153                 errno = ENOMEM;
 154                 return NULL;
 155         }
 156         hashp->fp = NO_FILE;
 157         if(file)
 158                 hashp->filename = strdup(file);
 159
 160         /*
 161          * Even if user wants write only, we need to be able to read
 162          * the actual file, so we need to open it read/write. But, the
 163          * field in the hashp structure needs to be accurate so that
 164          * we can check accesses.
 165          */
 166         hashp->flags = flags;
 167
 168         new_table = 0;
 169         if (!file || (flags & O_TRUNC)  || (stat(file, &statbuf)  && (errno == ENOENT)))
 170         {
 171                 if (errno == ENOENT)
 172                         errno = 0; /* Just in case someone looks at errno */
 173                 new_table = 1;
 174         }
 175         else if(statbuf.st_mtime && statbuf.st_size == 0)
 176         {
 177                 /* check for a zero length file and delete it
 178                  * if it exists
 179                  */
 180                 new_table = 1;
 181         }
 182         hashp->file_size = statbuf.st_size;
 183
 184         if (file) {
 185 #if defined(_WIN32) || defined(_WINDOWS) || defined (macintosh)  || defined(XP_OS2)
 186                 if ((hashp->fp = DBFILE_OPEN(file, flags | O_BINARY, mode)) == -1)
 187                         RETURN_ERROR(errno, error1);
 188 #else
 189                 if ((hashp->fp = open(file, flags, mode)) == -1)
 190                         RETURN_ERROR(errno, error1);
 191                 (void)fcntl(hashp->fp, F_SETFD, 1);
 192 #endif
 193         }
 194         if (new_table) {
 195                 if (!init_hash(hashp, file, (HASHINFO *)info))
 196                         RETURN_ERROR(errno, error1);
 197         } else {
 198                 /* Table already exists */
 199                 if (info && info->hash)
 200                         hashp->hash = info->hash;
 201                 else
 202                         hashp->hash = __default_hash;
 203
 204                 hdrsize = read(hashp->fp, (char *)&hashp->hdr, sizeof(HASHHDR));
 205                 if (hdrsize == -1)
 206                         RETURN_ERROR(errno, error1);
 207                 if (hdrsize != sizeof(HASHHDR))
 208                         RETURN_ERROR(EFTYPE, error1);
 209 #if BYTE_ORDER == LITTLE_ENDIAN
 210                 swap_header(hashp);
 211 #endif
 212                 /* Verify file type, versions and hash function */
 213                 if (hashp->MAGIC != HASHMAGIC)
 214                         RETURN_ERROR(EFTYPE, error1);
 215 #define OLDHASHVERSION  1
 216                 if (hashp->VERSION != HASHVERSION &&
 217                     hashp->VERSION != OLDHASHVERSION)
 218                         RETURN_ERROR(EFTYPE, error1);
 219                 if (hashp->hash(CHARKEY, sizeof(CHARKEY)) != hashp->H_CHARKEY)
 220                         RETURN_ERROR(EFTYPE, error1);
 221                 if (hashp->NKEYS < 0) /* Old bad database. */
 222                         RETURN_ERROR(EFTYPE, error1);
 223
 224                 /*
 225                  * Figure out how many segments we need.  Max_Bucket is the
 226                  * maximum bucket number, so the number of buckets is
 227                  * max_bucket + 1.
 228                  */
 229                 nsegs = (hashp->MAX_BUCKET + 1 + hashp->SGSIZE - 1) /
 230                          hashp->SGSIZE;
 231                 hashp->nsegs = 0;
 232                 if (alloc_segs(hashp, nsegs))
 233                         /* If alloc_segs fails, errno will have been set.  */
 234                         RETURN_ERROR(errno, error1);
 235                 /* Read in bitmaps */
 236                 bpages = (hashp->SPARES[hashp->OVFL_POINT] +
 237                     (hashp->BSIZE << BYTE_SHIFT) - 1) >>
 238                     (hashp->BSHIFT + BYTE_SHIFT);
 239
 240                 hashp->nmaps = bpages;
 241                 (void)memset(&hashp->mapp[0], 0, bpages * sizeof(uint32 *));
 242         }
 243
 244         /* Initialize Buffer Manager */
 245         if (info && info->cachesize)
 246                 __buf_init(hashp, (int32) info->cachesize);
 247         else
 248                 __buf_init(hashp, DEF_BUFSIZE);
 249
 250         hashp->new_file = new_table;
 251 #ifdef macintosh
 252         hashp->save_file = file && !(hashp->flags & O_RDONLY);
 253 #else
 254         hashp->save_file = file && (hashp->flags & O_RDWR);
 255 #endif
 256         hashp->cbucket = -1;
 257         if (!(dbp = (DB *)malloc(sizeof(DB)))) {
 258                 RETURN_ERROR(ENOMEM, error1);
 259         }
 260         dbp->internal = hashp;
 261         dbp->close = hash_close;
 262         dbp->del = hash_delete;
 263         dbp->fd = hash_fd;
 264         dbp->get = hash_get;
 265         dbp->put = hash_put;
 266         dbp->seq = hash_seq;
 267         dbp->sync = hash_sync;
 268         dbp->type = DB_HASH;
 269
 270 #ifdef HASH_STATISTICS
 271         hash_overflows = hash_accesses = hash_collisions = hash_expansions = 0;
 272 #endif
 273         return (dbp);
 274
 275 error1:
 276         hdestroy(hashp);
 277         errno = save_errno;
 278         return (NULL);
 279 }
 280
 281 static int
 282 hash_close(DB *dbp)
 283 {
 284         HTAB *hashp;
 285         int retval;
 286
 287         if (!dbp)
 288                 return (DBM_ERROR);
 289
 290         hashp = (HTAB *)dbp->internal;
 291         if(!hashp)
 292                 return (DBM_ERROR);
 293
 294         retval = hdestroy(hashp);
 295         free(dbp);
 296         return (retval);
 297 }
 298
 299 static int hash_fd(const DB *dbp)
 300 {
 301         HTAB *hashp;
 302
 303         if (!dbp)
 304                 return (DBM_ERROR);
 305
 306         hashp = (HTAB *)dbp->internal;
 307         if(!hashp)
 308                 return (DBM_ERROR);
 309
 310         if (hashp->fp == -1) {
 311                 errno = ENOENT;
 312                 return (-1);
 313         }
 314         return (hashp->fp);
 315 }
 316
 317 /************************** LOCAL CREATION ROUTINES **********************/
 318 static HTAB *
 319 init_hash(HTAB *hashp, const char *file, HASHINFO *info)
 320 {
 321         struct stat statbuf;
 322         int nelem;
 323
 324         nelem = 1;
 325         hashp->NKEYS = 0;
 326         hashp->LORDER = BYTE_ORDER;
 327         hashp->BSIZE = DEF_BUCKET_SIZE;
 328         hashp->BSHIFT = DEF_BUCKET_SHIFT;
 329         hashp->SGSIZE = DEF_SEGSIZE;
 330         hashp->SSHIFT = DEF_SEGSIZE_SHIFT;
 331         hashp->DSIZE = DEF_DIRSIZE;
 332         hashp->FFACTOR = DEF_FFACTOR;
 333         hashp->hash = __default_hash;
 334         memset(hashp->SPARES, 0, sizeof(hashp->SPARES));
 335         memset(hashp->BITMAPS, 0, sizeof (hashp->BITMAPS));
 336
 337         /* Fix bucket size to be optimal for file system */
 338         if (file != NULL) {
 339                 if (stat(file, &statbuf))
 340                         return (NULL);
 341
 342 #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh) && !defined(XP_OS2)
 343 #if defined(__QNX__) && !defined(__QNXNTO__)
 344                 hashp->BSIZE = 512; /* preferred blk size on qnx4 */
 345 #else
 346                 hashp->BSIZE = statbuf.st_blksize;
 347 #endif
 348
 349                 /* new code added by Lou to reduce block
 350                  * size down below MAX_BSIZE
 351                  */
 352                 if (hashp->BSIZE > MAX_BSIZE)
 353                         hashp->BSIZE = MAX_BSIZE;
 354 #endif
 355                 hashp->BSHIFT = __log2((uint32)hashp->BSIZE);
 356         }
 357
 358         if (info) {
 359                 if (info->bsize) {
 360                         /* Round pagesize up to power of 2 */
 361                         hashp->BSHIFT = __log2(info->bsize);
 362                         hashp->BSIZE = 1 << hashp->BSHIFT;
 363                         if (hashp->BSIZE > MAX_BSIZE) {
 364                                 errno = EINVAL;
 365                                 return (NULL);
 366                         }
 367                 }
 368                 if (info->ffactor)
 369                         hashp->FFACTOR = info->ffactor;
 370                 if (info->hash)
 371                         hashp->hash = info->hash;
 372                 if (info->nelem)
 373                         nelem = info->nelem;
 374                 if (info->lorder) {
 375                         if (info->lorder != BIG_ENDIAN &&
 376                             info->lorder != LITTLE_ENDIAN) {
 377                                 errno = EINVAL;
 378                                 return (NULL);
 379                         }
 380                         hashp->LORDER = info->lorder;
 381                 }
 382         }
 383         /* init_htab sets errno if it fails */
 384         if (init_htab(hashp, nelem))
 385                 return (NULL);
 386         else
 387                 return (hashp);
 388 }
 389 /*
 390  * This calls alloc_segs which may run out of memory.  Alloc_segs will
 391  * set errno, so we just pass the error information along.
 392  *
 393  * Returns 0 on No Error
 394  */
 395 static int
 396 init_htab(HTAB *hashp, int nelem)
 397 {
 398         register int nbuckets, nsegs;
 399         int l2;
 400
 401         /*
 402          * Divide number of elements by the fill factor and determine a
 403          * desired number of buckets.  Allocate space for the next greater
 404          * power of two number of buckets.
 405          */
 406         nelem = (nelem - 1) / hashp->FFACTOR + 1;
 407
 408         l2 = __log2((uint32)PR_MAX(nelem, 2));
 409         nbuckets = 1 << l2;
 410
 411         hashp->SPARES[l2] = l2 + 1;
 412         hashp->SPARES[l2 + 1] = l2 + 1;
 413         hashp->OVFL_POINT = l2;
 414         hashp->LAST_FREED = 2;
 415
 416         /* First bitmap page is at: splitpoint l2 page offset 1 */
 417         if (__ibitmap(hashp, (int)OADDR_OF(l2, 1), l2 + 1, 0))
 418                 return (-1);
 419
 420         hashp->MAX_BUCKET = hashp->LOW_MASK = nbuckets - 1;
 421         hashp->HIGH_MASK = (nbuckets << 1) - 1;
 422         hashp->HDRPAGES = ((PR_MAX(sizeof(HASHHDR), MINHDRSIZE) - 1) >>
 423             hashp->BSHIFT) + 1;
 424
 425         nsegs = (nbuckets - 1) / hashp->SGSIZE + 1;
 426         nsegs = 1 << __log2((uint32)nsegs);
 427
 428         if (nsegs > hashp->DSIZE)
 429                 hashp->DSIZE = nsegs;
 430         return (alloc_segs(hashp, nsegs));
 431 }
 432
 433 /********************** DESTROY/CLOSE ROUTINES ************************/
 434
 435 /*
 436  * Flushes any changes to the file if necessary and destroys the hashp
 437  * structure, freeing all allocated space.
 438  */
 439 static int
 440 hdestroy(HTAB *hashp)
 441 {
 442         int i, save_errno;
 443
 444         save_errno = 0;
 445
 446 #ifdef HASH_STATISTICS
 447         (void)fprintf(stderr, "hdestroy: accesses %ld collisions %ld\n",
 448             hash_accesses, hash_collisions);
 449         (void)fprintf(stderr, "hdestroy: expansions %ld\n",
 450             hash_expansions);
 451         (void)fprintf(stderr, "hdestroy: overflows %ld\n",
 452             hash_overflows);
 453         (void)fprintf(stderr, "keys %ld maxp %d segmentcount %d\n",
 454             hashp->NKEYS, hashp->MAX_BUCKET, hashp->nsegs);
 455
 456         for (i = 0; i < NCACHED; i++)
 457                 (void)fprintf(stderr,
 458                     "spares[%d] = %d\n", i, hashp->SPARES[i]);
 459 #endif
 460         /*
 461          * Call on buffer manager to free buffers, and if required,
 462          * write them to disk.
 463          */
 464         if (__buf_free(hashp, 1, hashp->save_file))
 465                 save_errno = errno;
 466         if (hashp->dir) {
 467                 free(*hashp->dir);      /* Free initial segments */
 468                 /* Free extra segments */
 469                 while (hashp->exsegs--)
 470                         free(hashp->dir[--hashp->nsegs]);
 471                 free(hashp->dir);
 472         }
 473         if (flush_meta(hashp) && !save_errno)
 474                 save_errno = errno;
 475         /* Free Bigmaps */
 476         for (i = 0; i < hashp->nmaps; i++)
 477                 if (hashp->mapp[i])
 478                         free(hashp->mapp[i]);
 479
 480         if (hashp->fp != -1)
 481                 (void)close(hashp->fp);
 482
 483         if(hashp->filename) {
 484 #if defined(_WIN32) || defined(_WINDOWS) || defined(XP_OS2)
 485                 if (hashp->is_temp)
 486                         (void)unlink(hashp->filename);
 487 #endif
 488                 free(hashp->filename);
 489         }
 490         if (hashp->tmp_buf)
 491             free(hashp->tmp_buf);
 492         if (hashp->tmp_key)
 493             free(hashp->tmp_key);
 494         free(hashp);
 495         if (save_errno) {
 496                 errno = save_errno;
 497                 return (DBM_ERROR);
 498         }
 499         return (SUCCESS);
 500 }
 501
 502 #if defined(_WIN32) || defined(_WINDOWS)
 503 /*
 504  * Close and reopen file to force file length update on windows.
 505  *
 506  * Returns:
 507  *       0 == OK
 508  *      -1 DBM_ERROR
 509  */
 510 static int
 511 update_EOF(HTAB *hashp)
 512 {
 513 #if defined(DBM_REOPEN_ON_FLUSH)
 514         char *      file       = hashp->filename;
 515         off_t       file_size;
 516         int         flags;
 517         int         mode       = -1;
 518         struct stat statbuf;
 519
 520         memset(&statbuf, 0, sizeof statbuf);
 521
 522         /* make sure we won't lose the file by closing it. */
 523         if (!file || (stat(file, &statbuf)  && (errno == ENOENT)))  {
 524                 /* pretend we did it. */
 525                 return 0;
 526         }
 527
 528         (void)close(hashp->fp);
 529
 530         flags = hashp->flags & ~(O_TRUNC | O_CREAT | O_EXCL);
 531
 532         if ((hashp->fp = DBFILE_OPEN(file, flags | O_BINARY, mode)) == -1)
 533                 return -1;
 534         file_size = lseek(hashp->fp, (off_t)0, SEEK_END);
 535         if (file_size == -1)
 536                 return -1;
 537         hashp->file_size = file_size;
 538         return 0;
 539 #else
 540         int    fd        = hashp->fp;
 541         off_t  file_size = lseek(fd, (off_t)0, SEEK_END);
 542         HANDLE handle    = (HANDLE)_get_osfhandle(fd);
 543         BOOL   cool      = FlushFileBuffers(handle);
 544 #ifdef DEBUG3
 545         if (!cool) {
 546                 DWORD err = GetLastError();
 547                 (void)fprintf(stderr,
 548                         "FlushFileBuffers failed, last error = %d, 0x%08x\n",
 549                         err, err);
 550         }
 551 #endif
 552         if (file_size == -1)
 553                 return -1;
 554         hashp->file_size = file_size;
 555         return cool ? 0 : -1;
 556 #endif
 557 }
 558 #endif
 559
 560 /*
 561  * Write modified pages to disk
 562  *
 563  * Returns:
 564  *       0 == OK
 565  *      -1 DBM_ERROR
 566  */
 567 static int
 568 hash_sync(const DB *dbp, uint flags)
 569 {
 570         HTAB *hashp;
 571
 572         if (flags != 0) {
 573                 errno = EINVAL;
 574                 return (DBM_ERROR);
 575         }
 576
 577         if (!dbp)
 578                 return (DBM_ERROR);
 579
 580         hashp = (HTAB *)dbp->internal;
 581         if(!hashp)
 582                 return (DBM_ERROR);
 583
 584         if (!hashp->save_file)
 585                 return (0);
 586         if (__buf_free(hashp, 0, 1) || flush_meta(hashp))
 587                 return (DBM_ERROR);
 588 #if defined(_WIN32) || defined(_WINDOWS)
 589         if (hashp->updateEOF && hashp->filename && !hashp->is_temp) {
 590                 int status = update_EOF(hashp);
 591                 hashp->updateEOF = 0;
 592                 if (status)
 593                         return status;
 594         }
 595 #endif
 596         hashp->new_file = 0;
 597         return (0);
 598 }
 599
 600 /*
 601  * Returns:
 602  *       0 == OK
 603  *      -1 indicates that errno should be set
 604  */
 605 static int
 606 flush_meta(HTAB *hashp)
 607 {
 608         HASHHDR *whdrp;
 609 #if BYTE_ORDER == LITTLE_ENDIAN
 610         HASHHDR whdr;
 611 #endif
 612         int fp, i, wsize;
 613
 614         if (!hashp->save_file)
 615                 return (0);
 616         hashp->MAGIC = HASHMAGIC;
 617         hashp->VERSION = HASHVERSION;
 618         hashp->H_CHARKEY = hashp->hash(CHARKEY, sizeof(CHARKEY));
 619
 620         fp = hashp->fp;
 621         whdrp = &hashp->hdr;
 622 #if BYTE_ORDER == LITTLE_ENDIAN
 623         whdrp = &whdr;
 624         swap_header_copy(&hashp->hdr, whdrp);
 625 #endif
 626         if ((lseek(fp, (off_t)0, SEEK_SET) == -1) ||
 627             ((wsize = write(fp, (char*)whdrp, sizeof(HASHHDR))) == -1))
 628                 return (-1);
 629         else
 630                 if (wsize != sizeof(HASHHDR)) {
 631                         errno = EFTYPE;
 632                         hashp->dbmerrno = errno;
 633                         return (-1);
 634                 }
 635         for (i = 0; i < NCACHED; i++)
 636                 if (hashp->mapp[i])
 637                         if (__put_page(hashp, (char *)hashp->mapp[i],
 638                                 hashp->BITMAPS[i], 0, 1))
 639                                 return (-1);
 640         return (0);
 641 }
 642
 643 /*******************************SEARCH ROUTINES *****************************/
 644 /*
 645  * All the access routines return
 646  *
 647  * Returns:
 648  *       0 on SUCCESS
 649  *       1 to indicate an external DBM_ERROR (i.e. key not found, etc)
 650  *      -1 to indicate an internal DBM_ERROR (i.e. out of memory, etc)
 651  */
 652 static int
 653 hash_get(
 654         const DB *dbp,
 655         const DBT *key,
 656         DBT *data,
 657         uint flag)
 658 {
 659         HTAB *hashp;
 660         int rv;
 661
 662         hashp = (HTAB *)dbp->internal;
 663         if (!hashp)
 664                 return (DBM_ERROR);
 665
 666         if (flag) {
 667                 hashp->dbmerrno = errno = EINVAL;
 668                 return (DBM_ERROR);
 669         }
 670
 671         rv = hash_access(hashp, HASH_GET, (DBT *)key, data);
 672
 673         if(rv == DATABASE_CORRUPTED_ERROR)
 674           {
 675 #if defined(unix) && defined(DEBUG)
 676                 printf("\n\nDBM Database has been corrupted, tell Lou...\n\n");
 677 #endif
 678                 __remove_database((DB *)dbp);
 679           }
 680
 681         return(rv);
 682 }
 683
 684 static int
 685 hash_put(
 686         const DB *dbp,
 687         DBT *key,
 688         const DBT *data,
 689         uint flag)
 690 {
 691         HTAB *hashp;
 692         int rv;
 693
 694         hashp = (HTAB *)dbp->internal;
 695         if (!hashp)
 696                 return (DBM_ERROR);
 697
 698         if (flag && flag != R_NOOVERWRITE) {
 699                 hashp->dbmerrno = errno = EINVAL;
 700                 return (DBM_ERROR);
 701         }
 702         if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
 703                 hashp->dbmerrno = errno = EPERM;
 704                 return (DBM_ERROR);
 705         }
 706
 707         rv =  hash_access(hashp, flag == R_NOOVERWRITE ?
 708             HASH_PUTNEW : HASH_PUT, (DBT *)key, (DBT *)data);
 709
 710         if(rv == DATABASE_CORRUPTED_ERROR)
 711           {
 712 #if defined(unix) && defined(DEBUG)
 713                 printf("\n\nDBM Database has been corrupted, tell Lou...\n\n");
 714 #endif
 715                 __remove_database((DB *)dbp);
 716           }
 717
 718         return(rv);
 719 }
 720
 721 static int
 722 hash_delete(
 723         const DB *dbp,
 724         const DBT *key,
 725         uint flag)              /* Ignored */
 726 {
 727         HTAB *hashp;
 728         int rv;
 729
 730         hashp = (HTAB *)dbp->internal;
 731         if (!hashp)
 732                 return (DBM_ERROR);
 733
 734         if (flag && flag != R_CURSOR) {
 735                 hashp->dbmerrno = errno = EINVAL;
 736                 return (DBM_ERROR);
 737         }
 738         if ((hashp->flags & O_ACCMODE) == O_RDONLY) {
 739                 hashp->dbmerrno = errno = EPERM;
 740                 return (DBM_ERROR);
 741         }
 742         rv = hash_access(hashp, HASH_DELETE, (DBT *)key, NULL);
 743
 744         if(rv == DATABASE_CORRUPTED_ERROR)
 745           {
 746 #if defined(unix) && defined(DEBUG)
 747                 printf("\n\nDBM Database has been corrupted, tell Lou...\n\n");
 748 #endif
 749                 __remove_database((DB *)dbp);
 750           }
 751
 752         return(rv);
 753 }
 754
 755 #define MAX_OVERFLOW_HASH_ACCESS_LOOPS 2000
 756 /*
 757  * Assume that hashp has been set in wrapper routine.
 758  */
 759 static int
 760 hash_access(
 761         HTAB *hashp,
 762         ACTION action,
 763         DBT *key, DBT *val)
 764 {
 765         register BUFHEAD *rbufp;
 766         BUFHEAD *bufp, *save_bufp;
 767         register uint16 *bp;
 768         register long n, ndx, off;
 769         register size_t size;
 770         register char *kp;
 771         uint16 pageno;
 772         uint32 ovfl_loop_count=0;
 773     int32 last_overflow_page_no = -1;
 774
 775 #ifdef HASH_STATISTICS
 776         hash_accesses++;
 777 #endif
 778
 779         off = hashp->BSIZE;
 780         size = key->size;
 781         kp = (char *)key->data;
 782         rbufp = __get_buf(hashp, __call_hash(hashp, kp, size), NULL, 0);
 783         if (!rbufp)
 784                 return (DATABASE_CORRUPTED_ERROR);
 785         save_bufp = rbufp;
 786
 787         /* Pin the bucket chain */
 788         rbufp->flags |= BUF_PIN;
 789         for (bp = (uint16 *)rbufp->page, n = *bp++, ndx = 1; ndx < n;)
 790         {
 791
 792                 if (bp[1] >= REAL_KEY) {
 793                         /* Real key/data pair */
 794                         if (size == (unsigned long)(off - *bp) &&
 795                             memcmp(kp, rbufp->page + *bp, size) == 0)
 796                                 goto found;
 797                         off = bp[1];
 798 #ifdef HASH_STATISTICS
 799                         hash_collisions++;
 800 #endif
 801                         bp += 2;
 802                         ndx += 2;
 803                 } else if (bp[1] == OVFLPAGE) {
 804
 805             /* database corruption: overflow loop detection */
 806             if(last_overflow_page_no == (int32)*bp)
 807                         return (DATABASE_CORRUPTED_ERROR);
 808
 809             last_overflow_page_no = *bp;
 810
 811                         rbufp = __get_buf(hashp, *bp, rbufp, 0);
 812                         if (!rbufp) {
 813                                 save_bufp->flags &= ~BUF_PIN;
 814                                 return (DBM_ERROR);
 815                         }
 816
 817             ovfl_loop_count++;
 818             if(ovfl_loop_count > MAX_OVERFLOW_HASH_ACCESS_LOOPS)
 819                         return (DATABASE_CORRUPTED_ERROR);
 820
 821                         /* FOR LOOP INIT */
 822                         bp = (uint16 *)rbufp->page;
 823                         n = *bp++;
 824                         ndx = 1;
 825                         off = hashp->BSIZE;
 826                 } else if (bp[1] < REAL_KEY) {
 827                         if ((ndx =
 828                             __find_bigpair(hashp, rbufp, ndx, kp, (int)size)) > 0)
 829                                 goto found;
 830                         if (ndx == -2) {
 831                                 bufp = rbufp;
 832                                 if (!(pageno =
 833                                     __find_last_page(hashp, &bufp))) {
 834                                         ndx = 0;
 835                                         rbufp = bufp;
 836                                         break;  /* FOR */
 837                                 }
 838                                 rbufp = __get_buf(hashp, pageno, bufp, 0);
 839                                 if (!rbufp) {
 840                                         save_bufp->flags &= ~BUF_PIN;
 841                                         return (DBM_ERROR);
 842                                 }
 843                                 /* FOR LOOP INIT */
 844                                 bp = (uint16 *)rbufp->page;
 845                                 n = *bp++;
 846                                 ndx = 1;
 847                                 off = hashp->BSIZE;
 848                         } else {
 849                                 save_bufp->flags &= ~BUF_PIN;
 850                                 return (DBM_ERROR);
 851
 852                         }
 853                 }
 854         }
 855
 856         /* Not found */
 857         switch (action) {
 858         case HASH_PUT:
 859         case HASH_PUTNEW:
 860                 if (__addel(hashp, rbufp, key, val)) {
 861                         save_bufp->flags &= ~BUF_PIN;
 862                         return (DBM_ERROR);
 863                 } else {
 864                         save_bufp->flags &= ~BUF_PIN;
 865                         return (SUCCESS);
 866                 }
 867         case HASH_GET:
 868         case HASH_DELETE:
 869         default:
 870                 save_bufp->flags &= ~BUF_PIN;
 871                 return (ABNORMAL);
 872         }
 873
 874 found:
 875         switch (action) {
 876         case HASH_PUTNEW:
 877                 save_bufp->flags &= ~BUF_PIN;
 878                 return (ABNORMAL);
 879         case HASH_GET:
 880                 bp = (uint16 *)rbufp->page;
 881                 if (bp[ndx + 1] < REAL_KEY) {
 882                         if (__big_return(hashp, rbufp, ndx, val, 0))
 883                                 return (DBM_ERROR);
 884                 } else {
 885                         val->data = (uint8 *)rbufp->page + (int)bp[ndx + 1];
 886                         val->size = bp[ndx] - bp[ndx + 1];
 887                 }
 888                 break;
 889         case HASH_PUT:
 890                 if ((__delpair(hashp, rbufp, ndx)) ||
 891                     (__addel(hashp, rbufp, key, val))) {
 892                         save_bufp->flags &= ~BUF_PIN;
 893                         return (DBM_ERROR);
 894                 }
 895                 break;
 896         case HASH_DELETE:
 897                 if (__delpair(hashp, rbufp, ndx))
 898                         return (DBM_ERROR);
 899                 break;
 900         default:
 901                 abort();
 902         }
 903         save_bufp->flags &= ~BUF_PIN;
 904         return (SUCCESS);
 905 }
 906
 907 static int
 908 hash_seq(
 909         const DB *dbp,
 910         DBT *key, DBT *data,
 911         uint flag)
 912 {
 913         register uint32 bucket;
 914         register BUFHEAD *bufp;
 915         HTAB *hashp;
 916         uint16 *bp, ndx;
 917
 918         hashp = (HTAB *)dbp->internal;
 919         if (!hashp)
 920                 return (DBM_ERROR);
 921
 922         if (flag && flag != R_FIRST && flag != R_NEXT) {
 923                 hashp->dbmerrno = errno = EINVAL;
 924                 return (DBM_ERROR);
 925         }
 926 #ifdef HASH_STATISTICS
 927         hash_accesses++;
 928 #endif
 929         if ((hashp->cbucket < 0) || (flag == R_FIRST)) {
 930                 hashp->cbucket = 0;
 931                 hashp->cndx = 1;
 932                 hashp->cpage = NULL;
 933         }
 934
 935         for (bp = NULL; !bp || !bp[0]; ) {
 936                 if (!(bufp = hashp->cpage)) {
 937                         for (bucket = hashp->cbucket;
 938                             bucket <= (uint32)hashp->MAX_BUCKET;
 939                             bucket++, hashp->cndx = 1) {
 940                                 bufp = __get_buf(hashp, bucket, NULL, 0);
 941                                 if (!bufp)
 942                                         return (DBM_ERROR);
 943                                 hashp->cpage = bufp;
 944                                 bp = (uint16 *)bufp->page;
 945                                 if (bp[0])
 946                                         break;
 947                         }
 948                         hashp->cbucket = bucket;
 949                         if (hashp->cbucket > hashp->MAX_BUCKET) {
 950                                 hashp->cbucket = -1;
 951                                 return (ABNORMAL);
 952                         }
 953                 } else
 954                         bp = (uint16 *)hashp->cpage->page;
 955
 956 #ifdef DEBUG
 957                 assert(bp);
 958                 assert(bufp);
 959 #endif
 960                 while (bp[hashp->cndx + 1] == OVFLPAGE) {
 961                         bufp = hashp->cpage =
 962                             __get_buf(hashp, bp[hashp->cndx], bufp, 0);
 963                         if (!bufp)
 964                                 return (DBM_ERROR);
 965                         bp = (uint16 *)(bufp->page);
 966                         hashp->cndx = 1;
 967                 }
 968                 if (!bp[0]) {
 969                         hashp->cpage = NULL;
 970                         ++hashp->cbucket;
 971                 }
 972         }
 973         ndx = hashp->cndx;
 974         if (bp[ndx + 1] < REAL_KEY) {
 975                 if (__big_keydata(hashp, bufp, key, data, 1))
 976                         return (DBM_ERROR);
 977         } else {
 978                 key->data = (uint8 *)hashp->cpage->page + bp[ndx];
 979                 key->size = (ndx > 1 ? bp[ndx - 1] : hashp->BSIZE) - bp[ndx];
 980                 data->data = (uint8 *)hashp->cpage->page + bp[ndx + 1];
 981                 data->size = bp[ndx] - bp[ndx + 1];
 982                 ndx += 2;
 983                 if (ndx > bp[0]) {
 984                         hashp->cpage = NULL;
 985                         hashp->cbucket++;
 986                         hashp->cndx = 1;
 987                 } else
 988                         hashp->cndx = ndx;
 989         }
 990         return (SUCCESS);
 991 }
 992
 993 /********************************* UTILITIES ************************/
 994
 995 /*
 996  * Returns:
 997  *       0 ==> OK
 998  *      -1 ==> Error
 999  */
1000 extern int
1001 __expand_table(HTAB *hashp)
1002 {
1003         uint32 old_bucket, new_bucket;
1004         int new_segnum, spare_ndx;
1005         size_t dirsize;
1006
1007 #ifdef HASH_STATISTICS
1008         hash_expansions++;
1009 #endif
1010         new_bucket = ++hashp->MAX_BUCKET;
1011         old_bucket = (hashp->MAX_BUCKET & hashp->LOW_MASK);
1012
1013         new_segnum = new_bucket >> hashp->SSHIFT;
1014
1015         /* Check if we need a new segment */
1016         if (new_segnum >= hashp->nsegs) {
1017                 /* Check if we need to expand directory */
1018                 if (new_segnum >= hashp->DSIZE) {
1019                         /* Reallocate directory */
1020                         dirsize = hashp->DSIZE * sizeof(SEGMENT *);
1021                         if (!hash_realloc(&hashp->dir, dirsize, dirsize << 1))
1022                                 return (-1);
1023                         hashp->DSIZE = dirsize << 1;
1024                 }
1025                 if ((hashp->dir[new_segnum] =
1026                     (SEGMENT)calloc((size_t)hashp->SGSIZE, sizeof(SEGMENT))) == NULL)
1027                         return (-1);
1028                 hashp->exsegs++;
1029                 hashp->nsegs++;
1030         }
1031         /*
1032          * If the split point is increasing (MAX_BUCKET's log base 2
1033          * * increases), we need to copy the current contents of the spare
1034          * split bucket to the next bucket.
1035          */
1036         spare_ndx = __log2((uint32)(hashp->MAX_BUCKET + 1));
1037         if (spare_ndx > hashp->OVFL_POINT) {
1038                 hashp->SPARES[spare_ndx] = hashp->SPARES[hashp->OVFL_POINT];
1039                 hashp->OVFL_POINT = spare_ndx;
1040         }
1041
1042         if (new_bucket > (uint32)hashp->HIGH_MASK) {
1043                 /* Starting a new doubling */
1044                 hashp->LOW_MASK = hashp->HIGH_MASK;
1045                 hashp->HIGH_MASK = new_bucket | hashp->LOW_MASK;
1046         }
1047         /* Relocate records to the new bucket */
1048         return (__split_page(hashp, old_bucket, new_bucket));
1049 }
1050
1051 /*
1052  * If realloc guarantees that the pointer is not destroyed if the realloc
1053  * fails, then this routine can go away.
1054  */
1055 static void *
1056 hash_realloc(
1057         SEGMENT **p_ptr,
1058         size_t oldsize, size_t newsize)
1059 {
1060         register void *p;
1061
1062         if ((p = malloc(newsize))) {
1063                 memmove(p, *p_ptr, oldsize);
1064                 memset((char *)p + oldsize, 0, newsize - oldsize);
1065                 free(*p_ptr);
1066                 *p_ptr = (SEGMENT *)p;
1067         }
1068         return (p);
1069 }
1070
1071 extern uint32
1072 __call_hash(HTAB *hashp, char *k, size_t len)
1073 {
1074         uint32 n, bucket;
1075
1076         n = hashp->hash(k, len);
1077         bucket = n & hashp->HIGH_MASK;
1078         if (bucket > (uint32)hashp->MAX_BUCKET)
1079                 bucket = bucket & hashp->LOW_MASK;
1080         return (bucket);
1081 }
1082
1083 /*
1084  * Allocate segment table.  On error, set errno.
1085  *
1086  * Returns 0 on success
1087  */
1088 static int
1089 alloc_segs(
1090         HTAB *hashp,
1091         int nsegs)
1092 {
1093         register int i;
1094         register SEGMENT store;
1095
1096         if ((hashp->dir =
1097             (SEGMENT *)calloc((size_t)hashp->DSIZE, sizeof(SEGMENT *))) == NULL) {
1098                 errno = ENOMEM;
1099                 return (-1);
1100         }
1101         /* Allocate segments */
1102         if ((store =
1103             (SEGMENT)calloc((size_t)nsegs << hashp->SSHIFT, sizeof(SEGMENT))) == NULL) {
1104                 errno = ENOMEM;
1105                 return (-1);
1106         }
1107         for (i = 0; i < nsegs; i++, hashp->nsegs++)
1108                 hashp->dir[i] = &store[i << hashp->SSHIFT];
1109         return (0);
1110 }
1111
1112 #if BYTE_ORDER == LITTLE_ENDIAN
1113 /*
1114  * Hashp->hdr needs to be byteswapped.
1115  */
1116 static void
1117 swap_header_copy(
1118         HASHHDR *srcp, HASHHDR *destp)
1119 {
1120         int i;
1121
1122         P_32_COPY(srcp->magic, destp->magic);
1123         P_32_COPY(srcp->version, destp->version);
1124         P_32_COPY(srcp->lorder, destp->lorder);
1125         P_32_COPY(srcp->bsize, destp->bsize);
1126         P_32_COPY(srcp->bshift, destp->bshift);
1127         P_32_COPY(srcp->dsize, destp->dsize);
1128         P_32_COPY(srcp->ssize, destp->ssize);
1129         P_32_COPY(srcp->sshift, destp->sshift);
1130         P_32_COPY(srcp->ovfl_point, destp->ovfl_point);
1131         P_32_COPY(srcp->last_freed, destp->last_freed);
1132         P_32_COPY(srcp->max_bucket, destp->max_bucket);
1133         P_32_COPY(srcp->high_mask, destp->high_mask);
1134         P_32_COPY(srcp->low_mask, destp->low_mask);
1135         P_32_COPY(srcp->ffactor, destp->ffactor);
1136         P_32_COPY(srcp->nkeys, destp->nkeys);
1137         P_32_COPY(srcp->hdrpages, destp->hdrpages);
1138         P_32_COPY(srcp->h_charkey, destp->h_charkey);
1139         for (i = 0; i < NCACHED; i++) {
1140                 P_32_COPY(srcp->spares[i], destp->spares[i]);
1141                 P_16_COPY(srcp->bitmaps[i], destp->bitmaps[i]);
1142         }
1143 }
1144
1145 static void
1146 swap_header(HTAB *hashp)
1147 {
1148         HASHHDR *hdrp;
1149         int i;
1150
1151         hdrp = &hashp->hdr;
1152
1153         M_32_SWAP(hdrp->magic);
1154         M_32_SWAP(hdrp->version);
1155         M_32_SWAP(hdrp->lorder);
1156         M_32_SWAP(hdrp->bsize);
1157         M_32_SWAP(hdrp->bshift);
1158         M_32_SWAP(hdrp->dsize);
1159         M_32_SWAP(hdrp->ssize);
1160         M_32_SWAP(hdrp->sshift);
1161         M_32_SWAP(hdrp->ovfl_point);
1162         M_32_SWAP(hdrp->last_freed);
1163         M_32_SWAP(hdrp->max_bucket);
1164         M_32_SWAP(hdrp->high_mask);
1165         M_32_SWAP(hdrp->low_mask);
1166         M_32_SWAP(hdrp->ffactor);
1167         M_32_SWAP(hdrp->nkeys);
1168         M_32_SWAP(hdrp->hdrpages);
1169         M_32_SWAP(hdrp->h_charkey);
1170         for (i = 0; i < NCACHED; i++) {
1171                 M_32_SWAP(hdrp->spares[i]);
1172                 M_16_SWAP(hdrp->bitmaps[i]);
1173         }
1174 }
1175 #endif