crypto/lhash/lhash.c

   1 /*
   2  * Copyright 1995-2017 The OpenSSL Project Authors. All Rights Reserved.
   3  *
   4  * Licensed under the OpenSSL license (the "License").  You may not use
   5  * this file except in compliance with the License.  You can obtain a copy
   6  * in the file LICENSE in the source distribution or at
   7  * https://www.openssl.org/source/license.html
   8  */
   9
  10 #include <stdio.h>
  11 #include <string.h>
  12 #include <stdlib.h>
  13 #include <openssl/crypto.h>
  14 #include <openssl/lhash.h>
  15 #include <openssl/err.h>
  16 #include "lhash_lcl.h"
  17
  18 /*
  19  * A hashing implementation that appears to be based on the linear hashing
  20  * alogrithm:
  21  * https://en.wikipedia.org/wiki/Linear_hashing
  22  *
  23  * Litwin, Witold (1980), "Linear hashing: A new tool for file and table
  24  * addressing", Proc. 6th Conference on Very Large Databases: 212-223
  25  * http://hackthology.com/pdfs/Litwin-1980-Linear_Hashing.pdf
  26  *
  27  * From the wikipedia article "Linear hashing is used in the BDB Berkeley
  28  * database system, which in turn is used by many software systems such as
  29  * OpenLDAP, using a C implementation derived from the CACM article and first
  30  * published on the Usenet in 1988 by Esmond Pitt."
  31  *
  32  * The CACM paper is available here:
  33  * https://pdfs.semanticscholar.org/ff4d/1c5deca6269cc316bfd952172284dbf610ee.pdf
  34  */
  35
  36 #undef MIN_NODES
  37 #define MIN_NODES       16
  38 #define UP_LOAD         (2*LH_LOAD_MULT) /* load times 256 (default 2) */
  39 #define DOWN_LOAD       (LH_LOAD_MULT) /* load times 256 (default 1) */
  40
  41 static int expand(OPENSSL_LHASH *lh);
  42 static void contract(OPENSSL_LHASH *lh);
  43 static OPENSSL_LH_NODE **getrn(OPENSSL_LHASH *lh, const void *data, unsigned long *rhash);
  44
  45 OPENSSL_LHASH *OPENSSL_LH_new(OPENSSL_LH_HASHFUNC h, OPENSSL_LH_COMPFUNC c)
  46 {
  47     OPENSSL_LHASH *ret;
  48
  49     if ((ret = OPENSSL_zalloc(sizeof(*ret))) == NULL) {
  50         /*
  51          * Do not set the error code, because the ERR code uses LHASH
  52          * and we want to avoid possible endless error loop.
  53          * CRYPTOerr(CRYPTO_F_OPENSSL_LH_NEW, ERR_R_MALLOC_FAILURE);
  54          */
  55         return NULL;
  56     }
  57     if ((ret->b = OPENSSL_zalloc(sizeof(*ret->b) * MIN_NODES)) == NULL)
  58         goto err;
  59     ret->comp = ((c == NULL) ? (OPENSSL_LH_COMPFUNC)strcmp : c);
  60     ret->hash = ((h == NULL) ? (OPENSSL_LH_HASHFUNC)OPENSSL_LH_strhash : h);
  61     ret->num_nodes = MIN_NODES / 2;
  62     ret->num_alloc_nodes = MIN_NODES;
  63     ret->pmax = MIN_NODES / 2;
  64     ret->up_load = UP_LOAD;
  65     ret->down_load = DOWN_LOAD;
  66     return ret;
  67
  68 err:
  69     OPENSSL_free(ret->b);
  70     OPENSSL_free(ret);
  71     return NULL;
  72 }
  73
  74 void OPENSSL_LH_free(OPENSSL_LHASH *lh)
  75 {
  76     unsigned int i;
  77     OPENSSL_LH_NODE *n, *nn;
  78
  79     if (lh == NULL)
  80         return;
  81
  82     for (i = 0; i < lh->num_nodes; i++) {
  83         n = lh->b[i];
  84         while (n != NULL) {
  85             nn = n->next;
  86             OPENSSL_free(n);
  87             n = nn;
  88         }
  89     }
  90     OPENSSL_free(lh->b);
  91     OPENSSL_free(lh);
  92 }
  93
  94 void *OPENSSL_LH_insert(OPENSSL_LHASH *lh, void *data)
  95 {
  96     unsigned long hash;
  97     OPENSSL_LH_NODE *nn, **rn;
  98     void *ret;
  99
 100     lh->error = 0;
 101     if ((lh->up_load <= (lh->num_items * LH_LOAD_MULT / lh->num_nodes)) && !expand(lh))
 102         return NULL;        /* 'lh->error++' already done in 'expand' */
 103
 104     rn = getrn(lh, data, &hash);
 105
 106     if (*rn == NULL) {
 107         if ((nn = OPENSSL_malloc(sizeof(*nn))) == NULL) {
 108             lh->error++;
 109             return NULL;
 110         }
 111         nn->data = data;
 112         nn->next = NULL;
 113         nn->hash = hash;
 114         *rn = nn;
 115         ret = NULL;
 116         lh->num_insert++;
 117         lh->num_items++;
 118     } else {                    /* replace same key */
 119         ret = (*rn)->data;
 120         (*rn)->data = data;
 121         lh->num_replace++;
 122     }
 123     return ret;
 124 }
 125
 126 void *OPENSSL_LH_delete(OPENSSL_LHASH *lh, const void *data)
 127 {
 128     unsigned long hash;
 129     OPENSSL_LH_NODE *nn, **rn;
 130     void *ret;
 131
 132     lh->error = 0;
 133     rn = getrn(lh, data, &hash);
 134
 135     if (*rn == NULL) {
 136         lh->num_no_delete++;
 137         return NULL;
 138     } else {
 139         nn = *rn;
 140         *rn = nn->next;
 141         ret = nn->data;
 142         OPENSSL_free(nn);
 143         lh->num_delete++;
 144     }
 145
 146     lh->num_items--;
 147     if ((lh->num_nodes > MIN_NODES) &&
 148         (lh->down_load >= (lh->num_items * LH_LOAD_MULT / lh->num_nodes)))
 149         contract(lh);
 150
 151     return ret;
 152 }
 153
 154 void *OPENSSL_LH_retrieve(OPENSSL_LHASH *lh, const void *data)
 155 {
 156     unsigned long hash;
 157     OPENSSL_LH_NODE **rn;
 158     void *ret;
 159
 160     lh->error = 0;
 161     rn = getrn(lh, data, &hash);
 162
 163     if (*rn == NULL) {
 164         lh->num_retrieve_miss++;
 165         return NULL;
 166     } else {
 167         ret = (*rn)->data;
 168         lh->num_retrieve++;
 169     }
 170     return ret;
 171 }
 172
 173 static void doall_util_fn(OPENSSL_LHASH *lh, int use_arg,
 174                           OPENSSL_LH_DOALL_FUNC func,
 175                           OPENSSL_LH_DOALL_FUNCARG func_arg, void *arg)
 176 {
 177     int i;
 178     OPENSSL_LH_NODE *a, *n;
 179
 180     if (lh == NULL)
 181         return;
 182
 183     /*
 184      * reverse the order so we search from 'top to bottom' We were having
 185      * memory leaks otherwise
 186      */
 187     for (i = lh->num_nodes - 1; i >= 0; i--) {
 188         a = lh->b[i];
 189         while (a != NULL) {
 190             n = a->next;
 191             if (use_arg)
 192                 func_arg(a->data, arg);
 193             else
 194                 func(a->data);
 195             a = n;
 196         }
 197     }
 198 }
 199
 200 void OPENSSL_LH_doall(OPENSSL_LHASH *lh, OPENSSL_LH_DOALL_FUNC func)
 201 {
 202     doall_util_fn(lh, 0, func, (OPENSSL_LH_DOALL_FUNCARG)0, NULL);
 203 }
 204
 205 void OPENSSL_LH_doall_arg(OPENSSL_LHASH *lh, OPENSSL_LH_DOALL_FUNCARG func, void *arg)
 206 {
 207     doall_util_fn(lh, 1, (OPENSSL_LH_DOALL_FUNC)0, func, arg);
 208 }
 209
 210 static int expand(OPENSSL_LHASH *lh)
 211 {
 212     OPENSSL_LH_NODE **n, **n1, **n2, *np;
 213     unsigned int p, pmax, nni, j;
 214     unsigned long hash;
 215
 216     nni = lh->num_alloc_nodes;
 217     p = lh->p;
 218     pmax = lh->pmax;
 219     if (p + 1 >= pmax) {
 220         j = nni * 2;
 221         n = OPENSSL_realloc(lh->b, sizeof(OPENSSL_LH_NODE *) * j);
 222         if (n == NULL) {
 223             lh->error++;
 224             return 0;
 225         }
 226         lh->b = n;
 227         memset(n + nni, 0, sizeof(*n) * (j - nni));
 228         lh->pmax = nni;
 229         lh->num_alloc_nodes = j;
 230         lh->num_expand_reallocs++;
 231         lh->p = 0;
 232     } else {
 233         lh->p++;
 234     }
 235
 236     lh->num_nodes++;
 237     lh->num_expands++;
 238     n1 = &(lh->b[p]);
 239     n2 = &(lh->b[p + pmax]);
 240     *n2 = NULL;
 241
 242     for (np = *n1; np != NULL;) {
 243         hash = np->hash;
 244         if ((hash % nni) != p) { /* move it */
 245             *n1 = (*n1)->next;
 246             np->next = *n2;
 247             *n2 = np;
 248         } else
 249             n1 = &((*n1)->next);
 250         np = *n1;
 251     }
 252
 253     return 1;
 254 }
 255
 256 static void contract(OPENSSL_LHASH *lh)
 257 {
 258     OPENSSL_LH_NODE **n, *n1, *np;
 259
 260     np = lh->b[lh->p + lh->pmax - 1];
 261     lh->b[lh->p + lh->pmax - 1] = NULL; /* 24/07-92 - eay - weird but :-( */
 262     if (lh->p == 0) {
 263         n = OPENSSL_realloc(lh->b,
 264                             (unsigned int)(sizeof(OPENSSL_LH_NODE *) * lh->pmax));
 265         if (n == NULL) {
 266             /* fputs("realloc error in lhash",stderr); */
 267             lh->error++;
 268             return;
 269         }
 270         lh->num_contract_reallocs++;
 271         lh->num_alloc_nodes /= 2;
 272         lh->pmax /= 2;
 273         lh->p = lh->pmax - 1;
 274         lh->b = n;
 275     } else
 276         lh->p--;
 277
 278     lh->num_nodes--;
 279     lh->num_contracts++;
 280
 281     n1 = lh->b[(int)lh->p];
 282     if (n1 == NULL)
 283         lh->b[(int)lh->p] = np;
 284     else {
 285         while (n1->next != NULL)
 286             n1 = n1->next;
 287         n1->next = np;
 288     }
 289 }
 290
 291 static OPENSSL_LH_NODE **getrn(OPENSSL_LHASH *lh,
 292                                const void *data, unsigned long *rhash)
 293 {
 294     OPENSSL_LH_NODE **ret, *n1;
 295     unsigned long hash, nn;
 296     OPENSSL_LH_COMPFUNC cf;
 297
 298     hash = (*(lh->hash)) (data);
 299     lh->num_hash_calls++;
 300     *rhash = hash;
 301
 302     nn = hash % lh->pmax;
 303     if (nn < lh->p)
 304         nn = hash % lh->num_alloc_nodes;
 305
 306     cf = lh->comp;
 307     ret = &(lh->b[(int)nn]);
 308     for (n1 = *ret; n1 != NULL; n1 = n1->next) {
 309         lh->num_hash_comps++;
 310         if (n1->hash != hash) {
 311             ret = &(n1->next);
 312             continue;
 313         }
 314         lh->num_comp_calls++;
 315         if (cf(n1->data, data) == 0)
 316             break;
 317         ret = &(n1->next);
 318     }
 319     return ret;
 320 }
 321
 322 /*
 323  * The following hash seems to work very well on normal text strings no
 324  * collisions on /usr/dict/words and it distributes on %2^n quite well, not
 325  * as good as MD5, but still good.
 326  */
 327 unsigned long OPENSSL_LH_strhash(const char *c)
 328 {
 329     unsigned long ret = 0;
 330     long n;
 331     unsigned long v;
 332     int r;
 333
 334     if ((c == NULL) || (*c == '\0'))
 335         return ret;
 336
 337     n = 0x100;
 338     while (*c) {
 339         v = n | (*c);
 340         n += 0x100;
 341         r = (int)((v >> 2) ^ v) & 0x0f;
 342         ret = (ret << r) | (ret >> (32 - r));
 343         ret &= 0xFFFFFFFFL;
 344         ret ^= v * v;
 345         c++;
 346     }
 347     return (ret >> 16) ^ ret;
 348 }
 349
 350 unsigned long OPENSSL_LH_num_items(const OPENSSL_LHASH *lh)
 351 {
 352     return lh ? lh->num_items : 0;
 353 }
 354
 355 unsigned long OPENSSL_LH_get_down_load(const OPENSSL_LHASH *lh)
 356 {
 357     return lh->down_load;
 358 }
 359
 360 void OPENSSL_LH_set_down_load(OPENSSL_LHASH *lh, unsigned long down_load)
 361 {
 362     lh->down_load = down_load;
 363 }
 364
 365 int OPENSSL_LH_error(OPENSSL_LHASH *lh)
 366 {
 367     return lh->error;
 368 }