lib/minitasn1/hash.c

   1 /*
   2  * Copyright (C) 2011-2012 Free Software Foundation, Inc.
   3  *
   4  * Author: Nikos Mavrogiannopoulos
   5  *
   6  * This file is part of GnuTLS.
   7  *
   8  * The GnuTLS is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public License
  10  * as published by the Free Software Foundation; either version 3 of
  11  * the License, or (at your option) any later version.
  12  *
  13  * This library is distributed in the hope that it will be useful, but
  14  * WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public License
  19  * along with this program.  If not, see <http://www.gnu.org/licenses/>
  20  *
  21  */
  22
  23 #include <int.h>
  24
  25 #define INIT_VAL 0x33a1
  26 /*
  27 -------------------------------------------------------------------------------
  28 lookup3.c, by Bob Jenkins, May 2006, Public Domain.
  29
  30 These are functions for producing 32-bit hashes for hash table lookup.
  31 hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
  32 are externally useful functions.  Routines to test the hash are included
  33 if SELF_TEST is defined.  You can use this free for any purpose.  It's in
  34 the public domain.  It has no warranty.
  35
  36 You probably want to use hashlittle().  hashlittle() and hashbig()
  37 hash byte arrays.  hashlittle() is faster than hashbig() on
  38 little-endian machines.  Intel and AMD are little-endian machines.
  39 On second thought, you probably want hashlittle2(), which is identical to
  40 hashlittle() except it returns two 32-bit hashes for the price of one.
  41 You could implement hashbig2() if you wanted but I haven't bothered here.
  42
  43 If you want to find a hash of, say, exactly 7 integers, do
  44   a = i1;  b = i2;  c = i3;
  45   mix(a,b,c);
  46   a += i4; b += i5; c += i6;
  47   mix(a,b,c);
  48   a += i7;
  49   final(a,b,c);
  50 then use c as the hash value.  If you have a variable length array of
  51 4-byte integers to hash, use hashword().  If you have a byte array (like
  52 a character string), use hashlittle().  If you have several byte arrays, or
  53 a mix of things, see the comments above hashlittle().
  54
  55 Why is this so big?  I read 12 bytes at a time into 3 4-byte integers,
  56 then mix those integers.  This is fast (you can do a lot more thorough
  57 mixing with 12*3 instructions on 3 integers than you can with 3 instructions
  58 on 1 byte), but shoehorning those bytes into integers efficiently is messy.
  59 -------------------------------------------------------------------------------
  60 */
  61 #define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
  62
  63 /*
  64 -------------------------------------------------------------------------------
  65 mix -- mix 3 32-bit values reversibly.
  66
  67 This is reversible, so any information in (a,b,c) before mix() is
  68 still in (a,b,c) after mix().
  69
  70 If four pairs of (a,b,c) inputs are run through mix(), or through
  71 mix() in reverse, there are at least 32 bits of the output that
  72 are sometimes the same for one pair and different for another pair.
  73 This was tested for:
  74 * pairs that differed by one bit, by two bits, in any combination
  75   of top bits of (a,b,c), or in any combination of bottom bits of
  76   (a,b,c).
  77 * "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
  78   the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
  79   is commonly produced by subtraction) look like a single 1-bit
  80   difference.
  81 * the base values were pseudorandom, all zero but one bit set, or
  82   all zero plus a counter that starts at zero.
  83
  84 Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
  85 satisfy this are
  86     4  6  8 16 19  4
  87     9 15  3 18 27 15
  88    14  9  3  7 17  3
  89 Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
  90 for "differ" defined as + with a one-bit base and a two-bit delta.  I
  91 used http://burtleburtle.net/bob/hash/avalanche.html to choose
  92 the operations, constants, and arrangements of the variables.
  93
  94 This does not achieve avalanche.  There are input bits of (a,b,c)
  95 that fail to affect some output bits of (a,b,c), especially of a.  The
  96 most thoroughly mixed value is c, but it doesn't really even achieve
  97 avalanche in c.
  98
  99 This allows some parallelism.  Read-after-writes are good at doubling
 100 the number of bits affected, so the goal of mixing pulls in the opposite
 101 direction as the goal of parallelism.  I did what I could.  Rotates
 102 seem to cost as much as shifts on every machine I could lay my hands
 103 on, and rotates are much kinder to the top and bottom bits, so I used
 104 rotates.
 105 -------------------------------------------------------------------------------
 106 */
 107 #define mix(a,b,c) \
 108 { \
 109   a -= c;  a ^= rot(c, 4);  c += b; \
 110   b -= a;  b ^= rot(a, 6);  a += c; \
 111   c -= b;  c ^= rot(b, 8);  b += a; \
 112   a -= c;  a ^= rot(c,16);  c += b; \
 113   b -= a;  b ^= rot(a,19);  a += c; \
 114   c -= b;  c ^= rot(b, 4);  b += a; \
 115 }
 116
 117 /*
 118 -------------------------------------------------------------------------------
 119 final -- final mixing of 3 32-bit values (a,b,c) into c
 120
 121 Pairs of (a,b,c) values differing in only a few bits will usually
 122 produce values of c that look totally different.  This was tested for
 123 * pairs that differed by one bit, by two bits, in any combination
 124   of top bits of (a,b,c), or in any combination of bottom bits of
 125   (a,b,c).
 126 * "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
 127   the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
 128   is commonly produced by subtraction) look like a single 1-bit
 129   difference.
 130 * the base values were pseudorandom, all zero but one bit set, or
 131   all zero plus a counter that starts at zero.
 132
 133 These constants passed:
 134  14 11 25 16 4 14 24
 135  12 14 25 16 4 14 24
 136 and these came close:
 137   4  8 15 26 3 22 24
 138  10  8 15 26 3 22 24
 139  11  8 15 26 3 22 24
 140 -------------------------------------------------------------------------------
 141 */
 142 #define final(a,b,c) \
 143 { \
 144   c ^= b; c -= rot(b,14); \
 145   a ^= c; a -= rot(c,11); \
 146   b ^= a; b -= rot(a,25); \
 147   c ^= b; c -= rot(b,16); \
 148   a ^= c; a -= rot(c,4);  \
 149   b ^= a; b -= rot(a,14); \
 150   c ^= b; c -= rot(b,24); \
 151 }
 152
 153
 154 /*
 155 -------------------------------------------------------------------------------
 156 hashlittle() -- hash a variable-length key into a 32-bit value
 157   k       : the key (the unaligned variable-length array of bytes)
 158   length  : the length of the key, counting by bytes
 159   initval : can be any 4-byte value
 160 Returns a 32-bit value.  Every bit of the key affects every bit of
 161 the return value.  Two keys differing by one or two bits will have
 162 totally different hash values.
 163
 164 The best hash table sizes are powers of 2.  There is no need to do
 165 mod a prime (mod is sooo slow!).  If you need less than 32 bits,
 166 use a bitmask.  For example, if you need only 10 bits, do
 167   h = (h & hashmask(10));
 168 In which case, the hash table should have hashsize(10) elements.
 169
 170 If you are hashing n strings (uint8_t **)k, do it like this:
 171   for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
 172
 173 By Bob Jenkins, 2006.  bob_jenkins@burtleburtle.net.  You may use this
 174 code any way you wish, private, educational, or commercial.  It's free.
 175
 176 Use for hash table lookup, or anything where one collision in 2^^32 is
 177 acceptable.  Do NOT use for cryptographic purposes.
 178 -------------------------------------------------------------------------------
 179 */
 180
 181 uint32_t _asn1_bhash( const void *key, size_t length)
 182 {
 183   uint32_t a,b,c;                                          /* internal state */
 184   const uint8_t *k;
 185
 186   /* Set up the internal state */
 187   a = b = c = 0xdeadbeef + ((uint32_t)length) + INIT_VAL;
 188
 189   k = (const uint8_t *)key;
 190
 191   /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
 192   while (length > 12)
 193   {
 194     a += k[0];
 195     a += ((uint32_t)k[1])<<8;
 196     a += ((uint32_t)k[2])<<16;
 197     a += ((uint32_t)k[3])<<24;
 198     b += k[4];
 199     b += ((uint32_t)k[5])<<8;
 200     b += ((uint32_t)k[6])<<16;
 201     b += ((uint32_t)k[7])<<24;
 202     c += k[8];
 203     c += ((uint32_t)k[9])<<8;
 204     c += ((uint32_t)k[10])<<16;
 205     c += ((uint32_t)k[11])<<24;
 206     mix(a,b,c);
 207     length -= 12;
 208     k += 12;
 209   }
 210
 211   /*-------------------------------- last block: affect all 32 bits of (c) */
 212   switch(length)                   /* all the case statements fall through */
 213   {
 214     case 12: c+=((uint32_t)k[11])<<24;
 215     case 11: c+=((uint32_t)k[10])<<16;
 216     case 10: c+=((uint32_t)k[9])<<8;
 217     case 9 : c+=k[8];
 218     case 8 : b+=((uint32_t)k[7])<<24;
 219     case 7 : b+=((uint32_t)k[6])<<16;
 220     case 6 : b+=((uint32_t)k[5])<<8;
 221     case 5 : b+=k[4];
 222     case 4 : a+=((uint32_t)k[3])<<24;
 223     case 3 : a+=((uint32_t)k[2])<<16;
 224     case 2 : a+=((uint32_t)k[1])<<8;
 225     case 1 : a+=k[0];
 226              break;
 227     case 0 : return c;
 228   }
 229
 230   final(a,b,c);
 231   return c;
 232 }
 233
 234
 235