Wed, 31 Dec 2014 06:09:35 +0100
Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.
michael@0 | 1 | /*- |
michael@0 | 2 | * Copyright (c) 1990, 1993, 1994 |
michael@0 | 3 | * The Regents of the University of California. All rights reserved. |
michael@0 | 4 | * |
michael@0 | 5 | * This code is derived from software contributed to Berkeley by |
michael@0 | 6 | * Margo Seltzer. |
michael@0 | 7 | * |
michael@0 | 8 | * Redistribution and use in source and binary forms, with or without |
michael@0 | 9 | * modification, are permitted provided that the following conditions |
michael@0 | 10 | * are met: |
michael@0 | 11 | * 1. Redistributions of source code must retain the above copyright |
michael@0 | 12 | * notice, this list of conditions and the following disclaimer. |
michael@0 | 13 | * 2. Redistributions in binary form must reproduce the above copyright |
michael@0 | 14 | * notice, this list of conditions and the following disclaimer in the |
michael@0 | 15 | * documentation and/or other materials provided with the distribution. |
michael@0 | 16 | * 3. ***REMOVED*** - see |
michael@0 | 17 | * ftp://ftp.cs.berkeley.edu/pub/4bsd/README.Impt.License.Change |
michael@0 | 18 | * 4. Neither the name of the University nor the names of its contributors |
michael@0 | 19 | * may be used to endorse or promote products derived from this software |
michael@0 | 20 | * without specific prior written permission. |
michael@0 | 21 | * |
michael@0 | 22 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
michael@0 | 23 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
michael@0 | 24 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
michael@0 | 25 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
michael@0 | 26 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
michael@0 | 27 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
michael@0 | 28 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
michael@0 | 29 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
michael@0 | 30 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
michael@0 | 31 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
michael@0 | 32 | * SUCH DAMAGE. |
michael@0 | 33 | */ |
michael@0 | 34 | |
michael@0 | 35 | #if defined(unix) |
michael@0 | 36 | #define MY_LSEEK lseek |
michael@0 | 37 | #else |
michael@0 | 38 | #define MY_LSEEK new_lseek |
michael@0 | 39 | extern long new_lseek(int fd, long pos, int start); |
michael@0 | 40 | #endif |
michael@0 | 41 | |
michael@0 | 42 | #if defined(LIBC_SCCS) && !defined(lint) |
michael@0 | 43 | static char sccsid[] = "@(#)hash_page.c 8.7 (Berkeley) 8/16/94"; |
michael@0 | 44 | #endif /* LIBC_SCCS and not lint */ |
michael@0 | 45 | |
michael@0 | 46 | /* |
michael@0 | 47 | * PACKAGE: hashing |
michael@0 | 48 | * |
michael@0 | 49 | * DESCRIPTION: |
michael@0 | 50 | * Page manipulation for hashing package. |
michael@0 | 51 | * |
michael@0 | 52 | * ROUTINES: |
michael@0 | 53 | * |
michael@0 | 54 | * External |
michael@0 | 55 | * __get_page |
michael@0 | 56 | * __add_ovflpage |
michael@0 | 57 | * Internal |
michael@0 | 58 | * overflow_page |
michael@0 | 59 | * open_temp |
michael@0 | 60 | */ |
michael@0 | 61 | #ifndef macintosh |
michael@0 | 62 | #include <sys/types.h> |
michael@0 | 63 | #endif |
michael@0 | 64 | |
michael@0 | 65 | #if defined(macintosh) |
michael@0 | 66 | #include <unistd.h> |
michael@0 | 67 | #endif |
michael@0 | 68 | |
michael@0 | 69 | #include <errno.h> |
michael@0 | 70 | #include <fcntl.h> |
michael@0 | 71 | #if defined(_WIN32) || defined(_WINDOWS) |
michael@0 | 72 | #include <io.h> |
michael@0 | 73 | #endif |
michael@0 | 74 | #include <signal.h> |
michael@0 | 75 | #include <stdio.h> |
michael@0 | 76 | #include <stdlib.h> |
michael@0 | 77 | #include <string.h> |
michael@0 | 78 | |
michael@0 | 79 | #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh) |
michael@0 | 80 | #include <unistd.h> |
michael@0 | 81 | #endif |
michael@0 | 82 | |
michael@0 | 83 | #include <assert.h> |
michael@0 | 84 | |
michael@0 | 85 | #include "mcom_db.h" |
michael@0 | 86 | #include "hash.h" |
michael@0 | 87 | #include "page.h" |
michael@0 | 88 | /* #include "extern.h" */ |
michael@0 | 89 | |
michael@0 | 90 | extern int mkstempflags(char *path, int extraFlags); |
michael@0 | 91 | |
michael@0 | 92 | static uint32 *fetch_bitmap __P((HTAB *, uint32)); |
michael@0 | 93 | static uint32 first_free __P((uint32)); |
michael@0 | 94 | static int open_temp __P((HTAB *)); |
michael@0 | 95 | static uint16 overflow_page __P((HTAB *)); |
michael@0 | 96 | static void squeeze_key __P((uint16 *, const DBT *, const DBT *)); |
michael@0 | 97 | static int ugly_split |
michael@0 | 98 | __P((HTAB *, uint32, BUFHEAD *, BUFHEAD *, int, int)); |
michael@0 | 99 | |
michael@0 | 100 | #define PAGE_INIT(P) { \ |
michael@0 | 101 | ((uint16 *)(P))[0] = 0; \ |
michael@0 | 102 | ((uint16 *)(P))[1] = hashp->BSIZE - 3 * sizeof(uint16); \ |
michael@0 | 103 | ((uint16 *)(P))[2] = hashp->BSIZE; \ |
michael@0 | 104 | } |
michael@0 | 105 | |
michael@0 | 106 | /* implement a new lseek using lseek that |
michael@0 | 107 | * writes zero's when extending a file |
michael@0 | 108 | * beyond the end. |
michael@0 | 109 | */ |
michael@0 | 110 | long new_lseek(int fd, long offset, int origin) |
michael@0 | 111 | { |
michael@0 | 112 | long cur_pos=0; |
michael@0 | 113 | long end_pos=0; |
michael@0 | 114 | long seek_pos=0; |
michael@0 | 115 | |
michael@0 | 116 | if(origin == SEEK_CUR) |
michael@0 | 117 | { |
michael@0 | 118 | if(offset < 1) |
michael@0 | 119 | return(lseek(fd, offset, SEEK_CUR)); |
michael@0 | 120 | |
michael@0 | 121 | cur_pos = lseek(fd, 0, SEEK_CUR); |
michael@0 | 122 | |
michael@0 | 123 | if(cur_pos < 0) |
michael@0 | 124 | return(cur_pos); |
michael@0 | 125 | } |
michael@0 | 126 | |
michael@0 | 127 | end_pos = lseek(fd, 0, SEEK_END); |
michael@0 | 128 | if(end_pos < 0) |
michael@0 | 129 | return(end_pos); |
michael@0 | 130 | |
michael@0 | 131 | if(origin == SEEK_SET) |
michael@0 | 132 | seek_pos = offset; |
michael@0 | 133 | else if(origin == SEEK_CUR) |
michael@0 | 134 | seek_pos = cur_pos + offset; |
michael@0 | 135 | else if(origin == SEEK_END) |
michael@0 | 136 | seek_pos = end_pos + offset; |
michael@0 | 137 | else |
michael@0 | 138 | { |
michael@0 | 139 | assert(0); |
michael@0 | 140 | return(-1); |
michael@0 | 141 | } |
michael@0 | 142 | |
michael@0 | 143 | /* the seek position desired is before the |
michael@0 | 144 | * end of the file. We don't need |
michael@0 | 145 | * to do anything special except the seek. |
michael@0 | 146 | */ |
michael@0 | 147 | if(seek_pos <= end_pos) |
michael@0 | 148 | return(lseek(fd, seek_pos, SEEK_SET)); |
michael@0 | 149 | |
michael@0 | 150 | /* the seek position is beyond the end of the |
michael@0 | 151 | * file. Write zero's to the end. |
michael@0 | 152 | * |
michael@0 | 153 | * we are already at the end of the file so |
michael@0 | 154 | * we just need to "write()" zeros for the |
michael@0 | 155 | * difference between seek_pos-end_pos and |
michael@0 | 156 | * then seek to the position to finish |
michael@0 | 157 | * the call |
michael@0 | 158 | */ |
michael@0 | 159 | { |
michael@0 | 160 | char buffer[1024]; |
michael@0 | 161 | long len = seek_pos-end_pos; |
michael@0 | 162 | memset(&buffer, 0, 1024); |
michael@0 | 163 | while(len > 0) |
michael@0 | 164 | { |
michael@0 | 165 | write(fd, (char*)&buffer, (size_t)(1024 > len ? len : 1024)); |
michael@0 | 166 | len -= 1024; |
michael@0 | 167 | } |
michael@0 | 168 | return(lseek(fd, seek_pos, SEEK_SET)); |
michael@0 | 169 | } |
michael@0 | 170 | |
michael@0 | 171 | } |
michael@0 | 172 | |
michael@0 | 173 | /* |
michael@0 | 174 | * This is called AFTER we have verified that there is room on the page for |
michael@0 | 175 | * the pair (PAIRFITS has returned true) so we go right ahead and start moving |
michael@0 | 176 | * stuff on. |
michael@0 | 177 | */ |
michael@0 | 178 | static void |
michael@0 | 179 | putpair(char *p, const DBT *key, DBT * val) |
michael@0 | 180 | { |
michael@0 | 181 | register uint16 *bp, n, off; |
michael@0 | 182 | |
michael@0 | 183 | bp = (uint16 *)p; |
michael@0 | 184 | |
michael@0 | 185 | /* Enter the key first. */ |
michael@0 | 186 | n = bp[0]; |
michael@0 | 187 | |
michael@0 | 188 | off = OFFSET(bp) - key->size; |
michael@0 | 189 | memmove(p + off, key->data, key->size); |
michael@0 | 190 | bp[++n] = off; |
michael@0 | 191 | |
michael@0 | 192 | /* Now the data. */ |
michael@0 | 193 | off -= val->size; |
michael@0 | 194 | memmove(p + off, val->data, val->size); |
michael@0 | 195 | bp[++n] = off; |
michael@0 | 196 | |
michael@0 | 197 | /* Adjust page info. */ |
michael@0 | 198 | bp[0] = n; |
michael@0 | 199 | bp[n + 1] = off - ((n + 3) * sizeof(uint16)); |
michael@0 | 200 | bp[n + 2] = off; |
michael@0 | 201 | } |
michael@0 | 202 | |
michael@0 | 203 | /* |
michael@0 | 204 | * Returns: |
michael@0 | 205 | * 0 OK |
michael@0 | 206 | * -1 error |
michael@0 | 207 | */ |
michael@0 | 208 | extern int |
michael@0 | 209 | __delpair(HTAB *hashp, BUFHEAD *bufp, int ndx) |
michael@0 | 210 | { |
michael@0 | 211 | register uint16 *bp, newoff; |
michael@0 | 212 | register int n; |
michael@0 | 213 | uint16 pairlen; |
michael@0 | 214 | |
michael@0 | 215 | bp = (uint16 *)bufp->page; |
michael@0 | 216 | n = bp[0]; |
michael@0 | 217 | |
michael@0 | 218 | if (bp[ndx + 1] < REAL_KEY) |
michael@0 | 219 | return (__big_delete(hashp, bufp)); |
michael@0 | 220 | if (ndx != 1) |
michael@0 | 221 | newoff = bp[ndx - 1]; |
michael@0 | 222 | else |
michael@0 | 223 | newoff = hashp->BSIZE; |
michael@0 | 224 | pairlen = newoff - bp[ndx + 1]; |
michael@0 | 225 | |
michael@0 | 226 | if (ndx != (n - 1)) { |
michael@0 | 227 | /* Hard Case -- need to shuffle keys */ |
michael@0 | 228 | register int i; |
michael@0 | 229 | register char *src = bufp->page + (int)OFFSET(bp); |
michael@0 | 230 | uint32 dst_offset = (uint32)OFFSET(bp) + (uint32)pairlen; |
michael@0 | 231 | register char *dst = bufp->page + dst_offset; |
michael@0 | 232 | uint32 length = bp[ndx + 1] - OFFSET(bp); |
michael@0 | 233 | |
michael@0 | 234 | /* |
michael@0 | 235 | * +-----------+XXX+---------+XXX+---------+---------> +infinity |
michael@0 | 236 | * | | | | |
michael@0 | 237 | * 0 src_offset dst_offset BSIZE |
michael@0 | 238 | * |
michael@0 | 239 | * Dst_offset is > src_offset, so if src_offset were bad, dst_offset |
michael@0 | 240 | * would be too, therefore we check only dst_offset. |
michael@0 | 241 | * |
michael@0 | 242 | * If dst_offset is >= BSIZE, either OFFSET(bp), or pairlen, or both |
michael@0 | 243 | * is corrupted. |
michael@0 | 244 | * |
michael@0 | 245 | * Once we know dst_offset is < BSIZE, we can subtract it from BSIZE |
michael@0 | 246 | * to get an upper bound on length. |
michael@0 | 247 | */ |
michael@0 | 248 | if(dst_offset > (uint32)hashp->BSIZE) |
michael@0 | 249 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 250 | |
michael@0 | 251 | if(length > (uint32)(hashp->BSIZE - dst_offset)) |
michael@0 | 252 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 253 | |
michael@0 | 254 | memmove(dst, src, length); |
michael@0 | 255 | |
michael@0 | 256 | /* Now adjust the pointers */ |
michael@0 | 257 | for (i = ndx + 2; i <= n; i += 2) { |
michael@0 | 258 | if (bp[i + 1] == OVFLPAGE) { |
michael@0 | 259 | bp[i - 2] = bp[i]; |
michael@0 | 260 | bp[i - 1] = bp[i + 1]; |
michael@0 | 261 | } else { |
michael@0 | 262 | bp[i - 2] = bp[i] + pairlen; |
michael@0 | 263 | bp[i - 1] = bp[i + 1] + pairlen; |
michael@0 | 264 | } |
michael@0 | 265 | } |
michael@0 | 266 | } |
michael@0 | 267 | /* Finally adjust the page data */ |
michael@0 | 268 | bp[n] = OFFSET(bp) + pairlen; |
michael@0 | 269 | bp[n - 1] = bp[n + 1] + pairlen + 2 * sizeof(uint16); |
michael@0 | 270 | bp[0] = n - 2; |
michael@0 | 271 | hashp->NKEYS--; |
michael@0 | 272 | |
michael@0 | 273 | bufp->flags |= BUF_MOD; |
michael@0 | 274 | return (0); |
michael@0 | 275 | } |
michael@0 | 276 | /* |
michael@0 | 277 | * Returns: |
michael@0 | 278 | * 0 ==> OK |
michael@0 | 279 | * -1 ==> Error |
michael@0 | 280 | */ |
michael@0 | 281 | extern int |
michael@0 | 282 | __split_page(HTAB *hashp, uint32 obucket, uint32 nbucket) |
michael@0 | 283 | { |
michael@0 | 284 | register BUFHEAD *new_bufp, *old_bufp; |
michael@0 | 285 | register uint16 *ino; |
michael@0 | 286 | register uint16 *tmp_uint16_array; |
michael@0 | 287 | register char *np; |
michael@0 | 288 | DBT key, val; |
michael@0 | 289 | uint16 n, ndx; |
michael@0 | 290 | int retval; |
michael@0 | 291 | uint16 copyto, diff, moved; |
michael@0 | 292 | size_t off; |
michael@0 | 293 | char *op; |
michael@0 | 294 | |
michael@0 | 295 | copyto = (uint16)hashp->BSIZE; |
michael@0 | 296 | off = (uint16)hashp->BSIZE; |
michael@0 | 297 | old_bufp = __get_buf(hashp, obucket, NULL, 0); |
michael@0 | 298 | if (old_bufp == NULL) |
michael@0 | 299 | return (-1); |
michael@0 | 300 | new_bufp = __get_buf(hashp, nbucket, NULL, 0); |
michael@0 | 301 | if (new_bufp == NULL) |
michael@0 | 302 | return (-1); |
michael@0 | 303 | |
michael@0 | 304 | old_bufp->flags |= (BUF_MOD | BUF_PIN); |
michael@0 | 305 | new_bufp->flags |= (BUF_MOD | BUF_PIN); |
michael@0 | 306 | |
michael@0 | 307 | ino = (uint16 *)(op = old_bufp->page); |
michael@0 | 308 | np = new_bufp->page; |
michael@0 | 309 | |
michael@0 | 310 | moved = 0; |
michael@0 | 311 | |
michael@0 | 312 | for (n = 1, ndx = 1; n < ino[0]; n += 2) { |
michael@0 | 313 | if (ino[n + 1] < REAL_KEY) { |
michael@0 | 314 | retval = ugly_split(hashp, obucket, old_bufp, new_bufp, |
michael@0 | 315 | (int)copyto, (int)moved); |
michael@0 | 316 | old_bufp->flags &= ~BUF_PIN; |
michael@0 | 317 | new_bufp->flags &= ~BUF_PIN; |
michael@0 | 318 | return (retval); |
michael@0 | 319 | |
michael@0 | 320 | } |
michael@0 | 321 | key.data = (uint8 *)op + ino[n]; |
michael@0 | 322 | |
michael@0 | 323 | /* check here for ino[n] being greater than |
michael@0 | 324 | * off. If it is then the database has |
michael@0 | 325 | * been corrupted. |
michael@0 | 326 | */ |
michael@0 | 327 | if(ino[n] > off) |
michael@0 | 328 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 329 | |
michael@0 | 330 | key.size = off - ino[n]; |
michael@0 | 331 | |
michael@0 | 332 | #ifdef DEBUG |
michael@0 | 333 | /* make sure the size is positive */ |
michael@0 | 334 | assert(((int)key.size) > -1); |
michael@0 | 335 | #endif |
michael@0 | 336 | |
michael@0 | 337 | if (__call_hash(hashp, (char *)key.data, key.size) == obucket) { |
michael@0 | 338 | /* Don't switch page */ |
michael@0 | 339 | diff = copyto - off; |
michael@0 | 340 | if (diff) { |
michael@0 | 341 | copyto = ino[n + 1] + diff; |
michael@0 | 342 | memmove(op + copyto, op + ino[n + 1], |
michael@0 | 343 | off - ino[n + 1]); |
michael@0 | 344 | ino[ndx] = copyto + ino[n] - ino[n + 1]; |
michael@0 | 345 | ino[ndx + 1] = copyto; |
michael@0 | 346 | } else |
michael@0 | 347 | copyto = ino[n + 1]; |
michael@0 | 348 | ndx += 2; |
michael@0 | 349 | } else { |
michael@0 | 350 | /* Switch page */ |
michael@0 | 351 | val.data = (uint8 *)op + ino[n + 1]; |
michael@0 | 352 | val.size = ino[n] - ino[n + 1]; |
michael@0 | 353 | |
michael@0 | 354 | /* if the pair doesn't fit something is horribly |
michael@0 | 355 | * wrong. LJM |
michael@0 | 356 | */ |
michael@0 | 357 | tmp_uint16_array = (uint16*)np; |
michael@0 | 358 | if(!PAIRFITS(tmp_uint16_array, &key, &val)) |
michael@0 | 359 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 360 | |
michael@0 | 361 | putpair(np, &key, &val); |
michael@0 | 362 | moved += 2; |
michael@0 | 363 | } |
michael@0 | 364 | |
michael@0 | 365 | off = ino[n + 1]; |
michael@0 | 366 | } |
michael@0 | 367 | |
michael@0 | 368 | /* Now clean up the page */ |
michael@0 | 369 | ino[0] -= moved; |
michael@0 | 370 | FREESPACE(ino) = copyto - sizeof(uint16) * (ino[0] + 3); |
michael@0 | 371 | OFFSET(ino) = copyto; |
michael@0 | 372 | |
michael@0 | 373 | #ifdef DEBUG3 |
michael@0 | 374 | (void)fprintf(stderr, "split %d/%d\n", |
michael@0 | 375 | ((uint16 *)np)[0] / 2, |
michael@0 | 376 | ((uint16 *)op)[0] / 2); |
michael@0 | 377 | #endif |
michael@0 | 378 | /* unpin both pages */ |
michael@0 | 379 | old_bufp->flags &= ~BUF_PIN; |
michael@0 | 380 | new_bufp->flags &= ~BUF_PIN; |
michael@0 | 381 | return (0); |
michael@0 | 382 | } |
michael@0 | 383 | |
michael@0 | 384 | /* |
michael@0 | 385 | * Called when we encounter an overflow or big key/data page during split |
michael@0 | 386 | * handling. This is special cased since we have to begin checking whether |
michael@0 | 387 | * the key/data pairs fit on their respective pages and because we may need |
michael@0 | 388 | * overflow pages for both the old and new pages. |
michael@0 | 389 | * |
michael@0 | 390 | * The first page might be a page with regular key/data pairs in which case |
michael@0 | 391 | * we have a regular overflow condition and just need to go on to the next |
michael@0 | 392 | * page or it might be a big key/data pair in which case we need to fix the |
michael@0 | 393 | * big key/data pair. |
michael@0 | 394 | * |
michael@0 | 395 | * Returns: |
michael@0 | 396 | * 0 ==> success |
michael@0 | 397 | * -1 ==> failure |
michael@0 | 398 | */ |
michael@0 | 399 | |
michael@0 | 400 | /* the maximum number of loops we will allow UGLY split to chew |
michael@0 | 401 | * on before we assume the database is corrupted and throw it |
michael@0 | 402 | * away. |
michael@0 | 403 | */ |
michael@0 | 404 | #define MAX_UGLY_SPLIT_LOOPS 10000 |
michael@0 | 405 | |
michael@0 | 406 | static int |
michael@0 | 407 | ugly_split(HTAB *hashp, uint32 obucket, BUFHEAD *old_bufp, |
michael@0 | 408 | BUFHEAD *new_bufp,/* Same as __split_page. */ int copyto, int moved) |
michael@0 | 409 | /* int copyto; First byte on page which contains key/data values. */ |
michael@0 | 410 | /* int moved; Number of pairs moved to new page. */ |
michael@0 | 411 | { |
michael@0 | 412 | register BUFHEAD *bufp; /* Buffer header for ino */ |
michael@0 | 413 | register uint16 *ino; /* Page keys come off of */ |
michael@0 | 414 | register uint16 *np; /* New page */ |
michael@0 | 415 | register uint16 *op; /* Page keys go on to if they aren't moving */ |
michael@0 | 416 | uint32 loop_detection=0; |
michael@0 | 417 | |
michael@0 | 418 | BUFHEAD *last_bfp; /* Last buf header OVFL needing to be freed */ |
michael@0 | 419 | DBT key, val; |
michael@0 | 420 | SPLIT_RETURN ret; |
michael@0 | 421 | uint16 n, off, ov_addr, scopyto; |
michael@0 | 422 | char *cino; /* Character value of ino */ |
michael@0 | 423 | int status; |
michael@0 | 424 | |
michael@0 | 425 | bufp = old_bufp; |
michael@0 | 426 | ino = (uint16 *)old_bufp->page; |
michael@0 | 427 | np = (uint16 *)new_bufp->page; |
michael@0 | 428 | op = (uint16 *)old_bufp->page; |
michael@0 | 429 | last_bfp = NULL; |
michael@0 | 430 | scopyto = (uint16)copyto; /* ANSI */ |
michael@0 | 431 | |
michael@0 | 432 | n = ino[0] - 1; |
michael@0 | 433 | while (n < ino[0]) { |
michael@0 | 434 | |
michael@0 | 435 | |
michael@0 | 436 | /* this function goes nuts sometimes and never returns. |
michael@0 | 437 | * I havent found the problem yet but I need a solution |
michael@0 | 438 | * so if we loop too often we assume a database curruption error |
michael@0 | 439 | * :LJM |
michael@0 | 440 | */ |
michael@0 | 441 | loop_detection++; |
michael@0 | 442 | |
michael@0 | 443 | if(loop_detection > MAX_UGLY_SPLIT_LOOPS) |
michael@0 | 444 | return DATABASE_CORRUPTED_ERROR; |
michael@0 | 445 | |
michael@0 | 446 | if (ino[2] < REAL_KEY && ino[2] != OVFLPAGE) { |
michael@0 | 447 | if ((status = __big_split(hashp, old_bufp, |
michael@0 | 448 | new_bufp, bufp, bufp->addr, obucket, &ret))) |
michael@0 | 449 | return (status); |
michael@0 | 450 | old_bufp = ret.oldp; |
michael@0 | 451 | if (!old_bufp) |
michael@0 | 452 | return (-1); |
michael@0 | 453 | op = (uint16 *)old_bufp->page; |
michael@0 | 454 | new_bufp = ret.newp; |
michael@0 | 455 | if (!new_bufp) |
michael@0 | 456 | return (-1); |
michael@0 | 457 | np = (uint16 *)new_bufp->page; |
michael@0 | 458 | bufp = ret.nextp; |
michael@0 | 459 | if (!bufp) |
michael@0 | 460 | return (0); |
michael@0 | 461 | cino = (char *)bufp->page; |
michael@0 | 462 | ino = (uint16 *)cino; |
michael@0 | 463 | last_bfp = ret.nextp; |
michael@0 | 464 | } else if (ino[n + 1] == OVFLPAGE) { |
michael@0 | 465 | ov_addr = ino[n]; |
michael@0 | 466 | /* |
michael@0 | 467 | * Fix up the old page -- the extra 2 are the fields |
michael@0 | 468 | * which contained the overflow information. |
michael@0 | 469 | */ |
michael@0 | 470 | ino[0] -= (moved + 2); |
michael@0 | 471 | FREESPACE(ino) = |
michael@0 | 472 | scopyto - sizeof(uint16) * (ino[0] + 3); |
michael@0 | 473 | OFFSET(ino) = scopyto; |
michael@0 | 474 | |
michael@0 | 475 | bufp = __get_buf(hashp, ov_addr, bufp, 0); |
michael@0 | 476 | if (!bufp) |
michael@0 | 477 | return (-1); |
michael@0 | 478 | |
michael@0 | 479 | ino = (uint16 *)bufp->page; |
michael@0 | 480 | n = 1; |
michael@0 | 481 | scopyto = hashp->BSIZE; |
michael@0 | 482 | moved = 0; |
michael@0 | 483 | |
michael@0 | 484 | if (last_bfp) |
michael@0 | 485 | __free_ovflpage(hashp, last_bfp); |
michael@0 | 486 | last_bfp = bufp; |
michael@0 | 487 | } |
michael@0 | 488 | /* Move regular sized pairs of there are any */ |
michael@0 | 489 | off = hashp->BSIZE; |
michael@0 | 490 | for (n = 1; (n < ino[0]) && (ino[n + 1] >= REAL_KEY); n += 2) { |
michael@0 | 491 | cino = (char *)ino; |
michael@0 | 492 | key.data = (uint8 *)cino + ino[n]; |
michael@0 | 493 | key.size = off - ino[n]; |
michael@0 | 494 | val.data = (uint8 *)cino + ino[n + 1]; |
michael@0 | 495 | val.size = ino[n] - ino[n + 1]; |
michael@0 | 496 | off = ino[n + 1]; |
michael@0 | 497 | |
michael@0 | 498 | if (__call_hash(hashp, (char*)key.data, key.size) == obucket) { |
michael@0 | 499 | /* Keep on old page */ |
michael@0 | 500 | if (PAIRFITS(op, (&key), (&val))) |
michael@0 | 501 | putpair((char *)op, &key, &val); |
michael@0 | 502 | else { |
michael@0 | 503 | old_bufp = |
michael@0 | 504 | __add_ovflpage(hashp, old_bufp); |
michael@0 | 505 | if (!old_bufp) |
michael@0 | 506 | return (-1); |
michael@0 | 507 | op = (uint16 *)old_bufp->page; |
michael@0 | 508 | putpair((char *)op, &key, &val); |
michael@0 | 509 | } |
michael@0 | 510 | old_bufp->flags |= BUF_MOD; |
michael@0 | 511 | } else { |
michael@0 | 512 | /* Move to new page */ |
michael@0 | 513 | if (PAIRFITS(np, (&key), (&val))) |
michael@0 | 514 | putpair((char *)np, &key, &val); |
michael@0 | 515 | else { |
michael@0 | 516 | new_bufp = |
michael@0 | 517 | __add_ovflpage(hashp, new_bufp); |
michael@0 | 518 | if (!new_bufp) |
michael@0 | 519 | return (-1); |
michael@0 | 520 | np = (uint16 *)new_bufp->page; |
michael@0 | 521 | putpair((char *)np, &key, &val); |
michael@0 | 522 | } |
michael@0 | 523 | new_bufp->flags |= BUF_MOD; |
michael@0 | 524 | } |
michael@0 | 525 | } |
michael@0 | 526 | } |
michael@0 | 527 | if (last_bfp) |
michael@0 | 528 | __free_ovflpage(hashp, last_bfp); |
michael@0 | 529 | return (0); |
michael@0 | 530 | } |
michael@0 | 531 | |
michael@0 | 532 | /* |
michael@0 | 533 | * Add the given pair to the page |
michael@0 | 534 | * |
michael@0 | 535 | * Returns: |
michael@0 | 536 | * 0 ==> OK |
michael@0 | 537 | * 1 ==> failure |
michael@0 | 538 | */ |
michael@0 | 539 | extern int |
michael@0 | 540 | __addel(HTAB *hashp, BUFHEAD *bufp, const DBT *key, const DBT * val) |
michael@0 | 541 | { |
michael@0 | 542 | register uint16 *bp, *sop; |
michael@0 | 543 | int do_expand; |
michael@0 | 544 | |
michael@0 | 545 | bp = (uint16 *)bufp->page; |
michael@0 | 546 | do_expand = 0; |
michael@0 | 547 | while (bp[0] && (bp[2] < REAL_KEY || bp[bp[0]] < REAL_KEY)) |
michael@0 | 548 | /* Exception case */ |
michael@0 | 549 | if (bp[2] == FULL_KEY_DATA && bp[0] == 2) |
michael@0 | 550 | /* This is the last page of a big key/data pair |
michael@0 | 551 | and we need to add another page */ |
michael@0 | 552 | break; |
michael@0 | 553 | else if (bp[2] < REAL_KEY && bp[bp[0]] != OVFLPAGE) { |
michael@0 | 554 | bufp = __get_buf(hashp, bp[bp[0] - 1], bufp, 0); |
michael@0 | 555 | if (!bufp) |
michael@0 | 556 | { |
michael@0 | 557 | #ifdef DEBUG |
michael@0 | 558 | assert(0); |
michael@0 | 559 | #endif |
michael@0 | 560 | return (-1); |
michael@0 | 561 | } |
michael@0 | 562 | bp = (uint16 *)bufp->page; |
michael@0 | 563 | } else |
michael@0 | 564 | /* Try to squeeze key on this page */ |
michael@0 | 565 | if (FREESPACE(bp) > PAIRSIZE(key, val)) { |
michael@0 | 566 | { |
michael@0 | 567 | squeeze_key(bp, key, val); |
michael@0 | 568 | |
michael@0 | 569 | /* LJM: I added this because I think it was |
michael@0 | 570 | * left out on accident. |
michael@0 | 571 | * if this isn't incremented nkeys will not |
michael@0 | 572 | * be the actual number of keys in the db. |
michael@0 | 573 | */ |
michael@0 | 574 | hashp->NKEYS++; |
michael@0 | 575 | return (0); |
michael@0 | 576 | } |
michael@0 | 577 | } else { |
michael@0 | 578 | bufp = __get_buf(hashp, bp[bp[0] - 1], bufp, 0); |
michael@0 | 579 | if (!bufp) |
michael@0 | 580 | { |
michael@0 | 581 | #ifdef DEBUG |
michael@0 | 582 | assert(0); |
michael@0 | 583 | #endif |
michael@0 | 584 | return (-1); |
michael@0 | 585 | } |
michael@0 | 586 | bp = (uint16 *)bufp->page; |
michael@0 | 587 | } |
michael@0 | 588 | |
michael@0 | 589 | if (PAIRFITS(bp, key, val)) |
michael@0 | 590 | putpair(bufp->page, key, (DBT *)val); |
michael@0 | 591 | else { |
michael@0 | 592 | do_expand = 1; |
michael@0 | 593 | bufp = __add_ovflpage(hashp, bufp); |
michael@0 | 594 | if (!bufp) |
michael@0 | 595 | { |
michael@0 | 596 | #ifdef DEBUG |
michael@0 | 597 | assert(0); |
michael@0 | 598 | #endif |
michael@0 | 599 | return (-1); |
michael@0 | 600 | } |
michael@0 | 601 | sop = (uint16 *)bufp->page; |
michael@0 | 602 | |
michael@0 | 603 | if (PAIRFITS(sop, key, val)) |
michael@0 | 604 | putpair((char *)sop, key, (DBT *)val); |
michael@0 | 605 | else |
michael@0 | 606 | if (__big_insert(hashp, bufp, key, val)) |
michael@0 | 607 | { |
michael@0 | 608 | #ifdef DEBUG |
michael@0 | 609 | assert(0); |
michael@0 | 610 | #endif |
michael@0 | 611 | return (-1); |
michael@0 | 612 | } |
michael@0 | 613 | } |
michael@0 | 614 | bufp->flags |= BUF_MOD; |
michael@0 | 615 | /* |
michael@0 | 616 | * If the average number of keys per bucket exceeds the fill factor, |
michael@0 | 617 | * expand the table. |
michael@0 | 618 | */ |
michael@0 | 619 | hashp->NKEYS++; |
michael@0 | 620 | if (do_expand || |
michael@0 | 621 | (hashp->NKEYS / (hashp->MAX_BUCKET + 1) > hashp->FFACTOR)) |
michael@0 | 622 | return (__expand_table(hashp)); |
michael@0 | 623 | return (0); |
michael@0 | 624 | } |
michael@0 | 625 | |
michael@0 | 626 | /* |
michael@0 | 627 | * |
michael@0 | 628 | * Returns: |
michael@0 | 629 | * pointer on success |
michael@0 | 630 | * NULL on error |
michael@0 | 631 | */ |
michael@0 | 632 | extern BUFHEAD * |
michael@0 | 633 | __add_ovflpage(HTAB *hashp, BUFHEAD *bufp) |
michael@0 | 634 | { |
michael@0 | 635 | register uint16 *sp; |
michael@0 | 636 | uint16 ndx, ovfl_num; |
michael@0 | 637 | #ifdef DEBUG1 |
michael@0 | 638 | int tmp1, tmp2; |
michael@0 | 639 | #endif |
michael@0 | 640 | sp = (uint16 *)bufp->page; |
michael@0 | 641 | |
michael@0 | 642 | /* Check if we are dynamically determining the fill factor */ |
michael@0 | 643 | if (hashp->FFACTOR == DEF_FFACTOR) { |
michael@0 | 644 | hashp->FFACTOR = sp[0] >> 1; |
michael@0 | 645 | if (hashp->FFACTOR < MIN_FFACTOR) |
michael@0 | 646 | hashp->FFACTOR = MIN_FFACTOR; |
michael@0 | 647 | } |
michael@0 | 648 | bufp->flags |= BUF_MOD; |
michael@0 | 649 | ovfl_num = overflow_page(hashp); |
michael@0 | 650 | #ifdef DEBUG1 |
michael@0 | 651 | tmp1 = bufp->addr; |
michael@0 | 652 | tmp2 = bufp->ovfl ? bufp->ovfl->addr : 0; |
michael@0 | 653 | #endif |
michael@0 | 654 | if (!ovfl_num || !(bufp->ovfl = __get_buf(hashp, ovfl_num, bufp, 1))) |
michael@0 | 655 | return (NULL); |
michael@0 | 656 | bufp->ovfl->flags |= BUF_MOD; |
michael@0 | 657 | #ifdef DEBUG1 |
michael@0 | 658 | (void)fprintf(stderr, "ADDOVFLPAGE: %d->ovfl was %d is now %d\n", |
michael@0 | 659 | tmp1, tmp2, bufp->ovfl->addr); |
michael@0 | 660 | #endif |
michael@0 | 661 | ndx = sp[0]; |
michael@0 | 662 | /* |
michael@0 | 663 | * Since a pair is allocated on a page only if there's room to add |
michael@0 | 664 | * an overflow page, we know that the OVFL information will fit on |
michael@0 | 665 | * the page. |
michael@0 | 666 | */ |
michael@0 | 667 | sp[ndx + 4] = OFFSET(sp); |
michael@0 | 668 | sp[ndx + 3] = FREESPACE(sp) - OVFLSIZE; |
michael@0 | 669 | sp[ndx + 1] = ovfl_num; |
michael@0 | 670 | sp[ndx + 2] = OVFLPAGE; |
michael@0 | 671 | sp[0] = ndx + 2; |
michael@0 | 672 | #ifdef HASH_STATISTICS |
michael@0 | 673 | hash_overflows++; |
michael@0 | 674 | #endif |
michael@0 | 675 | return (bufp->ovfl); |
michael@0 | 676 | } |
michael@0 | 677 | |
michael@0 | 678 | /* |
michael@0 | 679 | * Returns: |
michael@0 | 680 | * 0 indicates SUCCESS |
michael@0 | 681 | * -1 indicates FAILURE |
michael@0 | 682 | */ |
michael@0 | 683 | extern int |
michael@0 | 684 | __get_page(HTAB *hashp, |
michael@0 | 685 | char * p, |
michael@0 | 686 | uint32 bucket, |
michael@0 | 687 | int is_bucket, |
michael@0 | 688 | int is_disk, |
michael@0 | 689 | int is_bitmap) |
michael@0 | 690 | { |
michael@0 | 691 | register int fd, page; |
michael@0 | 692 | size_t size; |
michael@0 | 693 | int rsize; |
michael@0 | 694 | uint16 *bp; |
michael@0 | 695 | |
michael@0 | 696 | fd = hashp->fp; |
michael@0 | 697 | size = hashp->BSIZE; |
michael@0 | 698 | |
michael@0 | 699 | if ((fd == -1) || !is_disk) { |
michael@0 | 700 | PAGE_INIT(p); |
michael@0 | 701 | return (0); |
michael@0 | 702 | } |
michael@0 | 703 | if (is_bucket) |
michael@0 | 704 | page = BUCKET_TO_PAGE(bucket); |
michael@0 | 705 | else |
michael@0 | 706 | page = OADDR_TO_PAGE(bucket); |
michael@0 | 707 | if ((MY_LSEEK(fd, (off_t)page << hashp->BSHIFT, SEEK_SET) == -1) || |
michael@0 | 708 | ((rsize = read(fd, p, size)) == -1)) |
michael@0 | 709 | return (-1); |
michael@0 | 710 | |
michael@0 | 711 | bp = (uint16 *)p; |
michael@0 | 712 | if (!rsize) |
michael@0 | 713 | bp[0] = 0; /* We hit the EOF, so initialize a new page */ |
michael@0 | 714 | else |
michael@0 | 715 | if ((unsigned)rsize != size) { |
michael@0 | 716 | errno = EFTYPE; |
michael@0 | 717 | return (-1); |
michael@0 | 718 | } |
michael@0 | 719 | |
michael@0 | 720 | if (!is_bitmap && !bp[0]) { |
michael@0 | 721 | PAGE_INIT(p); |
michael@0 | 722 | } else { |
michael@0 | 723 | |
michael@0 | 724 | #ifdef DEBUG |
michael@0 | 725 | if(BYTE_ORDER == LITTLE_ENDIAN) |
michael@0 | 726 | { |
michael@0 | 727 | int is_little_endian; |
michael@0 | 728 | is_little_endian = BYTE_ORDER; |
michael@0 | 729 | } |
michael@0 | 730 | else if(BYTE_ORDER == BIG_ENDIAN) |
michael@0 | 731 | { |
michael@0 | 732 | int is_big_endian; |
michael@0 | 733 | is_big_endian = BYTE_ORDER; |
michael@0 | 734 | } |
michael@0 | 735 | else |
michael@0 | 736 | { |
michael@0 | 737 | assert(0); |
michael@0 | 738 | } |
michael@0 | 739 | #endif |
michael@0 | 740 | |
michael@0 | 741 | if (hashp->LORDER != BYTE_ORDER) { |
michael@0 | 742 | register int i, max; |
michael@0 | 743 | |
michael@0 | 744 | if (is_bitmap) { |
michael@0 | 745 | max = hashp->BSIZE >> 2; /* divide by 4 */ |
michael@0 | 746 | for (i = 0; i < max; i++) |
michael@0 | 747 | M_32_SWAP(((int *)p)[i]); |
michael@0 | 748 | } else { |
michael@0 | 749 | M_16_SWAP(bp[0]); |
michael@0 | 750 | max = bp[0] + 2; |
michael@0 | 751 | |
michael@0 | 752 | /* bound the size of max by |
michael@0 | 753 | * the maximum number of entries |
michael@0 | 754 | * in the array |
michael@0 | 755 | */ |
michael@0 | 756 | if((unsigned)max > (size / sizeof(uint16))) |
michael@0 | 757 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 758 | |
michael@0 | 759 | /* do the byte order swap |
michael@0 | 760 | */ |
michael@0 | 761 | for (i = 1; i <= max; i++) |
michael@0 | 762 | M_16_SWAP(bp[i]); |
michael@0 | 763 | } |
michael@0 | 764 | } |
michael@0 | 765 | |
michael@0 | 766 | /* check the validity of the page here |
michael@0 | 767 | * (after doing byte order swaping if necessary) |
michael@0 | 768 | */ |
michael@0 | 769 | if(!is_bitmap && bp[0] != 0) |
michael@0 | 770 | { |
michael@0 | 771 | uint16 num_keys = bp[0]; |
michael@0 | 772 | uint16 offset; |
michael@0 | 773 | uint16 i; |
michael@0 | 774 | |
michael@0 | 775 | /* bp[0] is supposed to be the number of |
michael@0 | 776 | * entries currently in the page. If |
michael@0 | 777 | * bp[0] is too large (larger than the whole |
michael@0 | 778 | * page) then the page is corrupted |
michael@0 | 779 | */ |
michael@0 | 780 | if(bp[0] > (size / sizeof(uint16))) |
michael@0 | 781 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 782 | |
michael@0 | 783 | /* bound free space */ |
michael@0 | 784 | if(FREESPACE(bp) > size) |
michael@0 | 785 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 786 | |
michael@0 | 787 | /* check each key and data offset to make |
michael@0 | 788 | * sure they are all within bounds they |
michael@0 | 789 | * should all be less than the previous |
michael@0 | 790 | * offset as well. |
michael@0 | 791 | */ |
michael@0 | 792 | offset = size; |
michael@0 | 793 | for(i=1 ; i <= num_keys; i+=2) |
michael@0 | 794 | { |
michael@0 | 795 | /* ignore overflow pages etc. */ |
michael@0 | 796 | if(bp[i+1] >= REAL_KEY) |
michael@0 | 797 | { |
michael@0 | 798 | |
michael@0 | 799 | if(bp[i] > offset || bp[i+1] > bp[i]) |
michael@0 | 800 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 801 | |
michael@0 | 802 | offset = bp[i+1]; |
michael@0 | 803 | } |
michael@0 | 804 | else |
michael@0 | 805 | { |
michael@0 | 806 | /* there are no other valid keys after |
michael@0 | 807 | * seeing a non REAL_KEY |
michael@0 | 808 | */ |
michael@0 | 809 | break; |
michael@0 | 810 | } |
michael@0 | 811 | } |
michael@0 | 812 | } |
michael@0 | 813 | } |
michael@0 | 814 | return (0); |
michael@0 | 815 | } |
michael@0 | 816 | |
michael@0 | 817 | /* |
michael@0 | 818 | * Write page p to disk |
michael@0 | 819 | * |
michael@0 | 820 | * Returns: |
michael@0 | 821 | * 0 ==> OK |
michael@0 | 822 | * -1 ==>failure |
michael@0 | 823 | */ |
michael@0 | 824 | extern int |
michael@0 | 825 | __put_page(HTAB *hashp, char *p, uint32 bucket, int is_bucket, int is_bitmap) |
michael@0 | 826 | { |
michael@0 | 827 | register int fd, page; |
michael@0 | 828 | size_t size; |
michael@0 | 829 | int wsize; |
michael@0 | 830 | off_t offset; |
michael@0 | 831 | |
michael@0 | 832 | size = hashp->BSIZE; |
michael@0 | 833 | if ((hashp->fp == -1) && open_temp(hashp)) |
michael@0 | 834 | return (-1); |
michael@0 | 835 | fd = hashp->fp; |
michael@0 | 836 | |
michael@0 | 837 | if (hashp->LORDER != BYTE_ORDER) { |
michael@0 | 838 | register int i; |
michael@0 | 839 | register int max; |
michael@0 | 840 | |
michael@0 | 841 | if (is_bitmap) { |
michael@0 | 842 | max = hashp->BSIZE >> 2; /* divide by 4 */ |
michael@0 | 843 | for (i = 0; i < max; i++) |
michael@0 | 844 | M_32_SWAP(((int *)p)[i]); |
michael@0 | 845 | } else { |
michael@0 | 846 | max = ((uint16 *)p)[0] + 2; |
michael@0 | 847 | |
michael@0 | 848 | /* bound the size of max by |
michael@0 | 849 | * the maximum number of entries |
michael@0 | 850 | * in the array |
michael@0 | 851 | */ |
michael@0 | 852 | if((unsigned)max > (size / sizeof(uint16))) |
michael@0 | 853 | return(DATABASE_CORRUPTED_ERROR); |
michael@0 | 854 | |
michael@0 | 855 | for (i = 0; i <= max; i++) |
michael@0 | 856 | M_16_SWAP(((uint16 *)p)[i]); |
michael@0 | 857 | |
michael@0 | 858 | } |
michael@0 | 859 | } |
michael@0 | 860 | |
michael@0 | 861 | if (is_bucket) |
michael@0 | 862 | page = BUCKET_TO_PAGE(bucket); |
michael@0 | 863 | else |
michael@0 | 864 | page = OADDR_TO_PAGE(bucket); |
michael@0 | 865 | offset = (off_t)page << hashp->BSHIFT; |
michael@0 | 866 | if ((MY_LSEEK(fd, offset, SEEK_SET) == -1) || |
michael@0 | 867 | ((wsize = write(fd, p, size)) == -1)) |
michael@0 | 868 | /* Errno is set */ |
michael@0 | 869 | return (-1); |
michael@0 | 870 | if ((unsigned)wsize != size) { |
michael@0 | 871 | errno = EFTYPE; |
michael@0 | 872 | return (-1); |
michael@0 | 873 | } |
michael@0 | 874 | #if defined(_WIN32) || defined(_WINDOWS) |
michael@0 | 875 | if (offset + size > hashp->file_size) { |
michael@0 | 876 | hashp->updateEOF = 1; |
michael@0 | 877 | } |
michael@0 | 878 | #endif |
michael@0 | 879 | /* put the page back the way it was so that it isn't byteswapped |
michael@0 | 880 | * if it remains in memory - LJM |
michael@0 | 881 | */ |
michael@0 | 882 | if (hashp->LORDER != BYTE_ORDER) { |
michael@0 | 883 | register int i; |
michael@0 | 884 | register int max; |
michael@0 | 885 | |
michael@0 | 886 | if (is_bitmap) { |
michael@0 | 887 | max = hashp->BSIZE >> 2; /* divide by 4 */ |
michael@0 | 888 | for (i = 0; i < max; i++) |
michael@0 | 889 | M_32_SWAP(((int *)p)[i]); |
michael@0 | 890 | } else { |
michael@0 | 891 | uint16 *bp = (uint16 *)p; |
michael@0 | 892 | |
michael@0 | 893 | M_16_SWAP(bp[0]); |
michael@0 | 894 | max = bp[0] + 2; |
michael@0 | 895 | |
michael@0 | 896 | /* no need to bound the size if max again |
michael@0 | 897 | * since it was done already above |
michael@0 | 898 | */ |
michael@0 | 899 | |
michael@0 | 900 | /* do the byte order re-swap |
michael@0 | 901 | */ |
michael@0 | 902 | for (i = 1; i <= max; i++) |
michael@0 | 903 | M_16_SWAP(bp[i]); |
michael@0 | 904 | } |
michael@0 | 905 | } |
michael@0 | 906 | |
michael@0 | 907 | return (0); |
michael@0 | 908 | } |
michael@0 | 909 | |
michael@0 | 910 | #define BYTE_MASK ((1 << INT_BYTE_SHIFT) -1) |
michael@0 | 911 | /* |
michael@0 | 912 | * Initialize a new bitmap page. Bitmap pages are left in memory |
michael@0 | 913 | * once they are read in. |
michael@0 | 914 | */ |
michael@0 | 915 | extern int |
michael@0 | 916 | __ibitmap(HTAB *hashp, int pnum, int nbits, int ndx) |
michael@0 | 917 | { |
michael@0 | 918 | uint32 *ip; |
michael@0 | 919 | size_t clearbytes, clearints; |
michael@0 | 920 | |
michael@0 | 921 | if ((ip = (uint32 *)malloc((size_t)hashp->BSIZE)) == NULL) |
michael@0 | 922 | return (1); |
michael@0 | 923 | hashp->nmaps++; |
michael@0 | 924 | clearints = ((nbits - 1) >> INT_BYTE_SHIFT) + 1; |
michael@0 | 925 | clearbytes = clearints << INT_TO_BYTE; |
michael@0 | 926 | (void)memset((char *)ip, 0, clearbytes); |
michael@0 | 927 | (void)memset(((char *)ip) + clearbytes, 0xFF, |
michael@0 | 928 | hashp->BSIZE - clearbytes); |
michael@0 | 929 | ip[clearints - 1] = ALL_SET << (nbits & BYTE_MASK); |
michael@0 | 930 | SETBIT(ip, 0); |
michael@0 | 931 | hashp->BITMAPS[ndx] = (uint16)pnum; |
michael@0 | 932 | hashp->mapp[ndx] = ip; |
michael@0 | 933 | return (0); |
michael@0 | 934 | } |
michael@0 | 935 | |
michael@0 | 936 | static uint32 |
michael@0 | 937 | first_free(uint32 map) |
michael@0 | 938 | { |
michael@0 | 939 | register uint32 i, mask; |
michael@0 | 940 | |
michael@0 | 941 | mask = 0x1; |
michael@0 | 942 | for (i = 0; i < BITS_PER_MAP; i++) { |
michael@0 | 943 | if (!(mask & map)) |
michael@0 | 944 | return (i); |
michael@0 | 945 | mask = mask << 1; |
michael@0 | 946 | } |
michael@0 | 947 | return (i); |
michael@0 | 948 | } |
michael@0 | 949 | |
michael@0 | 950 | static uint16 |
michael@0 | 951 | overflow_page(HTAB *hashp) |
michael@0 | 952 | { |
michael@0 | 953 | register uint32 *freep=NULL; |
michael@0 | 954 | register int max_free, offset, splitnum; |
michael@0 | 955 | uint16 addr; |
michael@0 | 956 | uint32 i; |
michael@0 | 957 | int bit, first_page, free_bit, free_page, in_use_bits, j; |
michael@0 | 958 | #ifdef DEBUG2 |
michael@0 | 959 | int tmp1, tmp2; |
michael@0 | 960 | #endif |
michael@0 | 961 | splitnum = hashp->OVFL_POINT; |
michael@0 | 962 | max_free = hashp->SPARES[splitnum]; |
michael@0 | 963 | |
michael@0 | 964 | free_page = (max_free - 1) >> (hashp->BSHIFT + BYTE_SHIFT); |
michael@0 | 965 | free_bit = (max_free - 1) & ((hashp->BSIZE << BYTE_SHIFT) - 1); |
michael@0 | 966 | |
michael@0 | 967 | /* Look through all the free maps to find the first free block */ |
michael@0 | 968 | first_page = hashp->LAST_FREED >>(hashp->BSHIFT + BYTE_SHIFT); |
michael@0 | 969 | for ( i = first_page; i <= (unsigned)free_page; i++ ) { |
michael@0 | 970 | if (!(freep = (uint32 *)hashp->mapp[i]) && |
michael@0 | 971 | !(freep = fetch_bitmap(hashp, i))) |
michael@0 | 972 | return (0); |
michael@0 | 973 | if (i == (unsigned)free_page) |
michael@0 | 974 | in_use_bits = free_bit; |
michael@0 | 975 | else |
michael@0 | 976 | in_use_bits = (hashp->BSIZE << BYTE_SHIFT) - 1; |
michael@0 | 977 | |
michael@0 | 978 | if (i == (unsigned)first_page) { |
michael@0 | 979 | bit = hashp->LAST_FREED & |
michael@0 | 980 | ((hashp->BSIZE << BYTE_SHIFT) - 1); |
michael@0 | 981 | j = bit / BITS_PER_MAP; |
michael@0 | 982 | bit = bit & ~(BITS_PER_MAP - 1); |
michael@0 | 983 | } else { |
michael@0 | 984 | bit = 0; |
michael@0 | 985 | j = 0; |
michael@0 | 986 | } |
michael@0 | 987 | for (; bit <= in_use_bits; j++, bit += BITS_PER_MAP) |
michael@0 | 988 | if (freep[j] != ALL_SET) |
michael@0 | 989 | goto found; |
michael@0 | 990 | } |
michael@0 | 991 | |
michael@0 | 992 | /* No Free Page Found */ |
michael@0 | 993 | hashp->LAST_FREED = hashp->SPARES[splitnum]; |
michael@0 | 994 | hashp->SPARES[splitnum]++; |
michael@0 | 995 | offset = hashp->SPARES[splitnum] - |
michael@0 | 996 | (splitnum ? hashp->SPARES[splitnum - 1] : 0); |
michael@0 | 997 | |
michael@0 | 998 | #define OVMSG "HASH: Out of overflow pages. Increase page size\n" |
michael@0 | 999 | if (offset > SPLITMASK) { |
michael@0 | 1000 | if (++splitnum >= NCACHED) { |
michael@0 | 1001 | #ifndef macintosh |
michael@0 | 1002 | (void)write(STDERR_FILENO, OVMSG, sizeof(OVMSG) - 1); |
michael@0 | 1003 | #endif |
michael@0 | 1004 | return (0); |
michael@0 | 1005 | } |
michael@0 | 1006 | hashp->OVFL_POINT = splitnum; |
michael@0 | 1007 | hashp->SPARES[splitnum] = hashp->SPARES[splitnum-1]; |
michael@0 | 1008 | hashp->SPARES[splitnum-1]--; |
michael@0 | 1009 | offset = 1; |
michael@0 | 1010 | } |
michael@0 | 1011 | |
michael@0 | 1012 | /* Check if we need to allocate a new bitmap page */ |
michael@0 | 1013 | if (free_bit == (hashp->BSIZE << BYTE_SHIFT) - 1) { |
michael@0 | 1014 | free_page++; |
michael@0 | 1015 | if (free_page >= NCACHED) { |
michael@0 | 1016 | #ifndef macintosh |
michael@0 | 1017 | (void)write(STDERR_FILENO, OVMSG, sizeof(OVMSG) - 1); |
michael@0 | 1018 | #endif |
michael@0 | 1019 | return (0); |
michael@0 | 1020 | } |
michael@0 | 1021 | /* |
michael@0 | 1022 | * This is tricky. The 1 indicates that you want the new page |
michael@0 | 1023 | * allocated with 1 clear bit. Actually, you are going to |
michael@0 | 1024 | * allocate 2 pages from this map. The first is going to be |
michael@0 | 1025 | * the map page, the second is the overflow page we were |
michael@0 | 1026 | * looking for. The init_bitmap routine automatically, sets |
michael@0 | 1027 | * the first bit of itself to indicate that the bitmap itself |
michael@0 | 1028 | * is in use. We would explicitly set the second bit, but |
michael@0 | 1029 | * don't have to if we tell init_bitmap not to leave it clear |
michael@0 | 1030 | * in the first place. |
michael@0 | 1031 | */ |
michael@0 | 1032 | if (__ibitmap(hashp, |
michael@0 | 1033 | (int)OADDR_OF(splitnum, offset), 1, free_page)) |
michael@0 | 1034 | return (0); |
michael@0 | 1035 | hashp->SPARES[splitnum]++; |
michael@0 | 1036 | #ifdef DEBUG2 |
michael@0 | 1037 | free_bit = 2; |
michael@0 | 1038 | #endif |
michael@0 | 1039 | offset++; |
michael@0 | 1040 | if (offset > SPLITMASK) { |
michael@0 | 1041 | if (++splitnum >= NCACHED) { |
michael@0 | 1042 | #ifndef macintosh |
michael@0 | 1043 | (void)write(STDERR_FILENO, OVMSG, |
michael@0 | 1044 | sizeof(OVMSG) - 1); |
michael@0 | 1045 | #endif |
michael@0 | 1046 | return (0); |
michael@0 | 1047 | } |
michael@0 | 1048 | hashp->OVFL_POINT = splitnum; |
michael@0 | 1049 | hashp->SPARES[splitnum] = hashp->SPARES[splitnum-1]; |
michael@0 | 1050 | hashp->SPARES[splitnum-1]--; |
michael@0 | 1051 | offset = 0; |
michael@0 | 1052 | } |
michael@0 | 1053 | } else { |
michael@0 | 1054 | /* |
michael@0 | 1055 | * Free_bit addresses the last used bit. Bump it to address |
michael@0 | 1056 | * the first available bit. |
michael@0 | 1057 | */ |
michael@0 | 1058 | free_bit++; |
michael@0 | 1059 | SETBIT(freep, free_bit); |
michael@0 | 1060 | } |
michael@0 | 1061 | |
michael@0 | 1062 | /* Calculate address of the new overflow page */ |
michael@0 | 1063 | addr = OADDR_OF(splitnum, offset); |
michael@0 | 1064 | #ifdef DEBUG2 |
michael@0 | 1065 | (void)fprintf(stderr, "OVERFLOW_PAGE: ADDR: %d BIT: %d PAGE %d\n", |
michael@0 | 1066 | addr, free_bit, free_page); |
michael@0 | 1067 | #endif |
michael@0 | 1068 | return (addr); |
michael@0 | 1069 | |
michael@0 | 1070 | found: |
michael@0 | 1071 | bit = bit + first_free(freep[j]); |
michael@0 | 1072 | SETBIT(freep, bit); |
michael@0 | 1073 | #ifdef DEBUG2 |
michael@0 | 1074 | tmp1 = bit; |
michael@0 | 1075 | tmp2 = i; |
michael@0 | 1076 | #endif |
michael@0 | 1077 | /* |
michael@0 | 1078 | * Bits are addressed starting with 0, but overflow pages are addressed |
michael@0 | 1079 | * beginning at 1. Bit is a bit addressnumber, so we need to increment |
michael@0 | 1080 | * it to convert it to a page number. |
michael@0 | 1081 | */ |
michael@0 | 1082 | bit = 1 + bit + (i * (hashp->BSIZE << BYTE_SHIFT)); |
michael@0 | 1083 | if (bit >= hashp->LAST_FREED) |
michael@0 | 1084 | hashp->LAST_FREED = bit - 1; |
michael@0 | 1085 | |
michael@0 | 1086 | /* Calculate the split number for this page */ |
michael@0 | 1087 | for (i = 0; (i < (unsigned)splitnum) && (bit > hashp->SPARES[i]); i++) {} |
michael@0 | 1088 | offset = (i ? bit - hashp->SPARES[i - 1] : bit); |
michael@0 | 1089 | if (offset >= SPLITMASK) |
michael@0 | 1090 | return (0); /* Out of overflow pages */ |
michael@0 | 1091 | addr = OADDR_OF(i, offset); |
michael@0 | 1092 | #ifdef DEBUG2 |
michael@0 | 1093 | (void)fprintf(stderr, "OVERFLOW_PAGE: ADDR: %d BIT: %d PAGE %d\n", |
michael@0 | 1094 | addr, tmp1, tmp2); |
michael@0 | 1095 | #endif |
michael@0 | 1096 | |
michael@0 | 1097 | /* Allocate and return the overflow page */ |
michael@0 | 1098 | return (addr); |
michael@0 | 1099 | } |
michael@0 | 1100 | |
michael@0 | 1101 | /* |
michael@0 | 1102 | * Mark this overflow page as free. |
michael@0 | 1103 | */ |
michael@0 | 1104 | extern void |
michael@0 | 1105 | __free_ovflpage(HTAB *hashp, BUFHEAD *obufp) |
michael@0 | 1106 | { |
michael@0 | 1107 | uint16 addr; |
michael@0 | 1108 | uint32 *freep; |
michael@0 | 1109 | uint32 bit_address, free_page, free_bit; |
michael@0 | 1110 | uint16 ndx; |
michael@0 | 1111 | |
michael@0 | 1112 | if(!obufp || !obufp->addr) |
michael@0 | 1113 | return; |
michael@0 | 1114 | |
michael@0 | 1115 | addr = obufp->addr; |
michael@0 | 1116 | #ifdef DEBUG1 |
michael@0 | 1117 | (void)fprintf(stderr, "Freeing %d\n", addr); |
michael@0 | 1118 | #endif |
michael@0 | 1119 | ndx = (((uint16)addr) >> SPLITSHIFT); |
michael@0 | 1120 | bit_address = |
michael@0 | 1121 | (ndx ? hashp->SPARES[ndx - 1] : 0) + (addr & SPLITMASK) - 1; |
michael@0 | 1122 | if (bit_address < (uint32)hashp->LAST_FREED) |
michael@0 | 1123 | hashp->LAST_FREED = bit_address; |
michael@0 | 1124 | free_page = (bit_address >> (hashp->BSHIFT + BYTE_SHIFT)); |
michael@0 | 1125 | free_bit = bit_address & ((hashp->BSIZE << BYTE_SHIFT) - 1); |
michael@0 | 1126 | |
michael@0 | 1127 | if (!(freep = hashp->mapp[free_page])) |
michael@0 | 1128 | freep = fetch_bitmap(hashp, free_page); |
michael@0 | 1129 | |
michael@0 | 1130 | #ifdef DEBUG |
michael@0 | 1131 | /* |
michael@0 | 1132 | * This had better never happen. It means we tried to read a bitmap |
michael@0 | 1133 | * that has already had overflow pages allocated off it, and we |
michael@0 | 1134 | * failed to read it from the file. |
michael@0 | 1135 | */ |
michael@0 | 1136 | if (!freep) |
michael@0 | 1137 | { |
michael@0 | 1138 | assert(0); |
michael@0 | 1139 | return; |
michael@0 | 1140 | } |
michael@0 | 1141 | #endif |
michael@0 | 1142 | CLRBIT(freep, free_bit); |
michael@0 | 1143 | #ifdef DEBUG2 |
michael@0 | 1144 | (void)fprintf(stderr, "FREE_OVFLPAGE: ADDR: %d BIT: %d PAGE %d\n", |
michael@0 | 1145 | obufp->addr, free_bit, free_page); |
michael@0 | 1146 | #endif |
michael@0 | 1147 | __reclaim_buf(hashp, obufp); |
michael@0 | 1148 | } |
michael@0 | 1149 | |
michael@0 | 1150 | /* |
michael@0 | 1151 | * Returns: |
michael@0 | 1152 | * 0 success |
michael@0 | 1153 | * -1 failure |
michael@0 | 1154 | */ |
michael@0 | 1155 | static int |
michael@0 | 1156 | open_temp(HTAB *hashp) |
michael@0 | 1157 | { |
michael@0 | 1158 | #ifdef XP_OS2 |
michael@0 | 1159 | hashp->fp = mkstemp(NULL); |
michael@0 | 1160 | #else |
michael@0 | 1161 | #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh) |
michael@0 | 1162 | sigset_t set, oset; |
michael@0 | 1163 | #endif |
michael@0 | 1164 | #if !defined(macintosh) |
michael@0 | 1165 | char * tmpdir; |
michael@0 | 1166 | size_t len; |
michael@0 | 1167 | char last; |
michael@0 | 1168 | #endif |
michael@0 | 1169 | static const char namestr[] = "/_hashXXXXXX"; |
michael@0 | 1170 | char filename[1024]; |
michael@0 | 1171 | |
michael@0 | 1172 | #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh) |
michael@0 | 1173 | /* Block signals; make sure file goes away at process exit. */ |
michael@0 | 1174 | (void)sigfillset(&set); |
michael@0 | 1175 | (void)sigprocmask(SIG_BLOCK, &set, &oset); |
michael@0 | 1176 | #endif |
michael@0 | 1177 | |
michael@0 | 1178 | filename[0] = 0; |
michael@0 | 1179 | #if defined(macintosh) |
michael@0 | 1180 | strcat(filename, namestr + 1); |
michael@0 | 1181 | #else |
michael@0 | 1182 | tmpdir = getenv("TMP"); |
michael@0 | 1183 | if (!tmpdir) |
michael@0 | 1184 | tmpdir = getenv("TMPDIR"); |
michael@0 | 1185 | if (!tmpdir) |
michael@0 | 1186 | tmpdir = getenv("TEMP"); |
michael@0 | 1187 | if (!tmpdir) |
michael@0 | 1188 | tmpdir = "."; |
michael@0 | 1189 | len = strlen(tmpdir); |
michael@0 | 1190 | if (len && len < (sizeof filename - sizeof namestr)) { |
michael@0 | 1191 | strcpy(filename, tmpdir); |
michael@0 | 1192 | } |
michael@0 | 1193 | len = strlen(filename); |
michael@0 | 1194 | last = tmpdir[len - 1]; |
michael@0 | 1195 | strcat(filename, (last == '/' || last == '\\') ? namestr + 1 : namestr); |
michael@0 | 1196 | #endif |
michael@0 | 1197 | |
michael@0 | 1198 | #if defined(_WIN32) || defined(_WINDOWS) |
michael@0 | 1199 | if ((hashp->fp = mkstempflags(filename, _O_BINARY|_O_TEMPORARY)) != -1) { |
michael@0 | 1200 | if (hashp->filename) { |
michael@0 | 1201 | free(hashp->filename); |
michael@0 | 1202 | } |
michael@0 | 1203 | hashp->filename = strdup(filename); |
michael@0 | 1204 | hashp->is_temp = 1; |
michael@0 | 1205 | } |
michael@0 | 1206 | #else |
michael@0 | 1207 | if ((hashp->fp = mkstemp(filename)) != -1) { |
michael@0 | 1208 | (void)unlink(filename); |
michael@0 | 1209 | #if !defined(macintosh) |
michael@0 | 1210 | (void)fcntl(hashp->fp, F_SETFD, 1); |
michael@0 | 1211 | #endif |
michael@0 | 1212 | } |
michael@0 | 1213 | #endif |
michael@0 | 1214 | |
michael@0 | 1215 | #if !defined(_WIN32) && !defined(_WINDOWS) && !defined(macintosh) |
michael@0 | 1216 | (void)sigprocmask(SIG_SETMASK, &oset, (sigset_t *)NULL); |
michael@0 | 1217 | #endif |
michael@0 | 1218 | #endif /* !OS2 */ |
michael@0 | 1219 | return (hashp->fp != -1 ? 0 : -1); |
michael@0 | 1220 | } |
michael@0 | 1221 | |
michael@0 | 1222 | /* |
michael@0 | 1223 | * We have to know that the key will fit, but the last entry on the page is |
michael@0 | 1224 | * an overflow pair, so we need to shift things. |
michael@0 | 1225 | */ |
michael@0 | 1226 | static void |
michael@0 | 1227 | squeeze_key(uint16 *sp, const DBT * key, const DBT * val) |
michael@0 | 1228 | { |
michael@0 | 1229 | register char *p; |
michael@0 | 1230 | uint16 free_space, n, off, pageno; |
michael@0 | 1231 | |
michael@0 | 1232 | p = (char *)sp; |
michael@0 | 1233 | n = sp[0]; |
michael@0 | 1234 | free_space = FREESPACE(sp); |
michael@0 | 1235 | off = OFFSET(sp); |
michael@0 | 1236 | |
michael@0 | 1237 | pageno = sp[n - 1]; |
michael@0 | 1238 | off -= key->size; |
michael@0 | 1239 | sp[n - 1] = off; |
michael@0 | 1240 | memmove(p + off, key->data, key->size); |
michael@0 | 1241 | off -= val->size; |
michael@0 | 1242 | sp[n] = off; |
michael@0 | 1243 | memmove(p + off, val->data, val->size); |
michael@0 | 1244 | sp[0] = n + 2; |
michael@0 | 1245 | sp[n + 1] = pageno; |
michael@0 | 1246 | sp[n + 2] = OVFLPAGE; |
michael@0 | 1247 | FREESPACE(sp) = free_space - PAIRSIZE(key, val); |
michael@0 | 1248 | OFFSET(sp) = off; |
michael@0 | 1249 | } |
michael@0 | 1250 | |
michael@0 | 1251 | static uint32 * |
michael@0 | 1252 | fetch_bitmap(HTAB *hashp, uint32 ndx) |
michael@0 | 1253 | { |
michael@0 | 1254 | if (ndx >= (unsigned)hashp->nmaps) |
michael@0 | 1255 | return (NULL); |
michael@0 | 1256 | if ((hashp->mapp[ndx] = (uint32 *)malloc((size_t)hashp->BSIZE)) == NULL) |
michael@0 | 1257 | return (NULL); |
michael@0 | 1258 | if (__get_page(hashp, |
michael@0 | 1259 | (char *)hashp->mapp[ndx], hashp->BITMAPS[ndx], 0, 1, 1)) { |
michael@0 | 1260 | free(hashp->mapp[ndx]); |
michael@0 | 1261 | hashp->mapp[ndx] = NULL; /* NEW: 9-11-95 */ |
michael@0 | 1262 | return (NULL); |
michael@0 | 1263 | } |
michael@0 | 1264 | return (hashp->mapp[ndx]); |
michael@0 | 1265 | } |
michael@0 | 1266 | |
michael@0 | 1267 | #ifdef DEBUG4 |
michael@0 | 1268 | int |
michael@0 | 1269 | print_chain(int addr) |
michael@0 | 1270 | { |
michael@0 | 1271 | BUFHEAD *bufp; |
michael@0 | 1272 | short *bp, oaddr; |
michael@0 | 1273 | |
michael@0 | 1274 | (void)fprintf(stderr, "%d ", addr); |
michael@0 | 1275 | bufp = __get_buf(hashp, addr, NULL, 0); |
michael@0 | 1276 | bp = (short *)bufp->page; |
michael@0 | 1277 | while (bp[0] && ((bp[bp[0]] == OVFLPAGE) || |
michael@0 | 1278 | ((bp[0] > 2) && bp[2] < REAL_KEY))) { |
michael@0 | 1279 | oaddr = bp[bp[0] - 1]; |
michael@0 | 1280 | (void)fprintf(stderr, "%d ", (int)oaddr); |
michael@0 | 1281 | bufp = __get_buf(hashp, (int)oaddr, bufp, 0); |
michael@0 | 1282 | bp = (short *)bufp->page; |
michael@0 | 1283 | } |
michael@0 | 1284 | (void)fprintf(stderr, "\n"); |
michael@0 | 1285 | } |
michael@0 | 1286 | #endif |