1 /* SPDX-License-Identifier: LGPL-2.1-only */
2 /*
3 * This code was taken from http://ccodearchive.net/info/hash.html
4 * The original file was modified to remove unwanted code
5 * and some changes to fit the current build environment
6 */
7 /*
8 -------------------------------------------------------------------------------
9 lookup3.c, by Bob Jenkins, May 2006, Public Domain.
10
11 These are functions for producing 32-bit hashes for hash table lookup.
12 hash_word(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
13 are externally useful functions. Routines to test the hash are included
14 if SELF_TEST is defined. You can use this free for any purpose. It's in
15 the public domain. It has no warranty.
16
17 You probably want to use hashlittle(). hashlittle() and hashbig()
18 hash byte arrays. hashlittle() is is faster than hashbig() on
19 little-endian machines. Intel and AMD are little-endian machines.
20 On second thought, you probably want hashlittle2(), which is identical to
21 hashlittle() except it returns two 32-bit hashes for the price of one.
22 You could implement hashbig2() if you wanted but I haven't bothered here.
23
24 If you want to find a hash of, say, exactly 7 integers, do
25 a = i1; b = i2; c = i3;
26 mix(a,b,c);
27 a += i4; b += i5; c += i6;
28 mix(a,b,c);
29 a += i7;
30 final(a,b,c);
31 then use c as the hash value. If you have a variable length array of
32 4-byte integers to hash, use hash_word(). If you have a byte array (like
33 a character string), use hashlittle(). If you have several byte arrays, or
34 a mix of things, see the comments above hashlittle().
35
36 Why is this so big? I read 12 bytes at a time into 3 4-byte integers,
37 then mix those integers. This is fast (you can do a lot more thorough
38 mixing with 12*3 instructions on 3 integers than you can with 3 instructions
39 on 1 byte), but shoehorning those bytes into integers efficiently is messy.
40 -------------------------------------------------------------------------------
41 */
42 #include "nl-default.h"
43
44 #include <netlink/hash.h>
45
46 #if HAVE_LITTLE_ENDIAN
47 #define HASH_LITTLE_ENDIAN 1
48 #define HASH_BIG_ENDIAN 0
49 #elif HAVE_BIG_ENDIAN
50 #define HASH_LITTLE_ENDIAN 0
51 #define HASH_BIG_ENDIAN 1
52 #else
53 #error Unknown endian
54 #endif
55
56 #define hashsize(n) ((uint32_t)1<<(n))
57 #define hashmask(n) (hashsize(n)-1)
58 #define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
59
60 /*
61 -------------------------------------------------------------------------------
62 mix -- mix 3 32-bit values reversibly.
63
64 This is reversible, so any information in (a,b,c) before mix() is
65 still in (a,b,c) after mix().
66
67 If four pairs of (a,b,c) inputs are run through mix(), or through
68 mix() in reverse, there are at least 32 bits of the output that
69 are sometimes the same for one pair and different for another pair.
70 This was tested for:
71 * pairs that differed by one bit, by two bits, in any combination
72 of top bits of (a,b,c), or in any combination of bottom bits of
73 (a,b,c).
74 * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
75 the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
76 is commonly produced by subtraction) look like a single 1-bit
77 difference.
78 * the base values were pseudorandom, all zero but one bit set, or
79 all zero plus a counter that starts at zero.
80
81 Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
82 satisfy this are
83 4 6 8 16 19 4
84 9 15 3 18 27 15
85 14 9 3 7 17 3
86 Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
87 for "differ" defined as + with a one-bit base and a two-bit delta. I
88 used http://burtleburtle.net/bob/hash/avalanche.html to choose
89 the operations, constants, and arrangements of the variables.
90
91 This does not achieve avalanche. There are input bits of (a,b,c)
92 that fail to affect some output bits of (a,b,c), especially of a. The
93 most thoroughly mixed value is c, but it doesn't really even achieve
94 avalanche in c.
95
96 This allows some parallelism. Read-after-writes are good at doubling
97 the number of bits affected, so the goal of mixing pulls in the opposite
98 direction as the goal of parallelism. I did what I could. Rotates
99 seem to cost as much as shifts on every machine I could lay my hands
100 on, and rotates are much kinder to the top and bottom bits, so I used
101 rotates.
102 -------------------------------------------------------------------------------
103 */
104 #define mix(a,b,c) \
105 { \
106 a -= c; a ^= rot(c, 4); c += b; \
107 b -= a; b ^= rot(a, 6); a += c; \
108 c -= b; c ^= rot(b, 8); b += a; \
109 a -= c; a ^= rot(c,16); c += b; \
110 b -= a; b ^= rot(a,19); a += c; \
111 c -= b; c ^= rot(b, 4); b += a; \
112 }
113
114 /*
115 -------------------------------------------------------------------------------
116 final -- final mixing of 3 32-bit values (a,b,c) into c
117
118 Pairs of (a,b,c) values differing in only a few bits will usually
119 produce values of c that look totally different. This was tested for
120 * pairs that differed by one bit, by two bits, in any combination
121 of top bits of (a,b,c), or in any combination of bottom bits of
122 (a,b,c).
123 * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
124 the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
125 is commonly produced by subtraction) look like a single 1-bit
126 difference.
127 * the base values were pseudorandom, all zero but one bit set, or
128 all zero plus a counter that starts at zero.
129
130 These constants passed:
131 14 11 25 16 4 14 24
132 12 14 25 16 4 14 24
133 and these came close:
134 4 8 15 26 3 22 24
135 10 8 15 26 3 22 24
136 11 8 15 26 3 22 24
137 -------------------------------------------------------------------------------
138 */
139 #define final(a,b,c) \
140 { \
141 c ^= b; c -= rot(b,14); \
142 a ^= c; a -= rot(c,11); \
143 b ^= a; b -= rot(a,25); \
144 c ^= b; c -= rot(b,16); \
145 a ^= c; a -= rot(c,4); \
146 b ^= a; b -= rot(a,14); \
147 c ^= b; c -= rot(b,24); \
148 }
149
150 /*
151 -------------------------------------------------------------------------------
152 hashlittle() -- hash a variable-length key into a 32-bit value
153 k : the key (the unaligned variable-length array of bytes)
154 length : the length of the key, counting by bytes
155 val2 : IN: can be any 4-byte value OUT: second 32 bit hash.
156 Returns a 32-bit value. Every bit of the key affects every bit of
157 the return value. Two keys differing by one or two bits will have
158 totally different hash values. Note that the return value is better
159 mixed than val2, so use that first.
160
161 The best hash table sizes are powers of 2. There is no need to do
162 mod a prime (mod is sooo slow!). If you need less than 32 bits,
163 use a bitmask. For example, if you need only 10 bits, do
164 h = (h & hashmask(10));
165 In which case, the hash table should have hashsize(10) elements.
166
167 If you are hashing n strings (uint8_t **)k, do it like this:
168 for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
169
170 By Bob Jenkins, 2006. [email protected]. You may use this
171 code any way you wish, private, educational, or commercial. It's free.
172
173 Use for hash table lookup, or anything where one collision in 2^^32 is
174 acceptable. Do NOT use for cryptographic purposes.
175 -------------------------------------------------------------------------------
176 */
177
hashlittle(const void * key,size_t length,uint32_t * val2)178 static uint32_t hashlittle( const void *key, size_t length, uint32_t *val2 )
179 {
180 uint32_t a,b,c; /* internal state */
181 union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */
182
183 /* Set up the internal state */
184 a = b = c = 0xdeadbeef + ((uint32_t)length) + *val2;
185
186 u.ptr = key;
187 if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
188 const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
189 const uint8_t *k8;
190
191 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
192 while (length > 12)
193 {
194 a += k[0];
195 b += k[1];
196 c += k[2];
197 mix(a,b,c);
198 length -= 12;
199 k += 3;
200 }
201
202 /*----------------------------- handle the last (probably partial) block */
203 /*
204 * "k[2]&0xffffff" actually reads beyond the end of the string, but
205 * then masks off the part it's not allowed to read. Because the
206 * string is aligned, the masked-off tail is in the same word as the
207 * rest of the string. Every machine with memory protection I've seen
208 * does it on word boundaries, so is OK with this. But VALGRIND will
209 * still catch it and complain. The masking trick does make the hash
210 * noticably faster for short strings (like English words).
211 *
212 * Not on my testing with gcc 4.5 on an intel i5 CPU, at least --RR.
213 */
214 #if 0
215 switch(length)
216 {
217 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
218 case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
219 case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
220 case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
221 case 8 : b+=k[1]; a+=k[0]; break;
222 case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
223 case 6 : b+=k[1]&0xffff; a+=k[0]; break;
224 case 5 : b+=k[1]&0xff; a+=k[0]; break;
225 case 4 : a+=k[0]; break;
226 case 3 : a+=k[0]&0xffffff; break;
227 case 2 : a+=k[0]&0xffff; break;
228 case 1 : a+=k[0]&0xff; break;
229 case 0 : return c; /* zero length strings require no mixing */
230 }
231
232 #else /* make valgrind happy */
233
234 k8 = (const uint8_t *)k;
235 switch(length)
236 {
237 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
238 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
239 case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
240 case 9 : c+=k8[8]; /* fall through */
241 case 8 : b+=k[1]; a+=k[0]; break;
242 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
243 case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
244 case 5 : b+=k8[4]; /* fall through */
245 case 4 : a+=k[0]; break;
246 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
247 case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
248 case 1 : a+=k8[0]; break;
249 case 0 : return c;
250 }
251
252 #endif /* !valgrind */
253
254 } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
255 const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
256 const uint8_t *k8;
257
258 /*--------------- all but last block: aligned reads and different mixing */
259 while (length > 12)
260 {
261 a += k[0] + (((uint32_t)k[1])<<16);
262 b += k[2] + (((uint32_t)k[3])<<16);
263 c += k[4] + (((uint32_t)k[5])<<16);
264 mix(a,b,c);
265 length -= 12;
266 k += 6;
267 }
268
269 /*----------------------------- handle the last (probably partial) block */
270 k8 = (const uint8_t *)k;
271 switch(length)
272 {
273 case 12: c+=k[4]+(((uint32_t)k[5])<<16);
274 b+=k[2]+(((uint32_t)k[3])<<16);
275 a+=k[0]+(((uint32_t)k[1])<<16);
276 break;
277 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
278 case 10: c+=k[4];
279 b+=k[2]+(((uint32_t)k[3])<<16);
280 a+=k[0]+(((uint32_t)k[1])<<16);
281 break;
282 case 9 : c+=k8[8]; /* fall through */
283 case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
284 a+=k[0]+(((uint32_t)k[1])<<16);
285 break;
286 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
287 case 6 : b+=k[2];
288 a+=k[0]+(((uint32_t)k[1])<<16);
289 break;
290 case 5 : b+=k8[4]; /* fall through */
291 case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
292 break;
293 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
294 case 2 : a+=k[0];
295 break;
296 case 1 : a+=k8[0];
297 break;
298 case 0 : return c; /* zero length requires no mixing */
299 }
300
301 } else { /* need to read the key one byte at a time */
302 const uint8_t *k = (const uint8_t *)key;
303
304 /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
305 while (length > 12)
306 {
307 a += k[0];
308 a += ((uint32_t)k[1])<<8;
309 a += ((uint32_t)k[2])<<16;
310 a += ((uint32_t)k[3])<<24;
311 b += k[4];
312 b += ((uint32_t)k[5])<<8;
313 b += ((uint32_t)k[6])<<16;
314 b += ((uint32_t)k[7])<<24;
315 c += k[8];
316 c += ((uint32_t)k[9])<<8;
317 c += ((uint32_t)k[10])<<16;
318 c += ((uint32_t)k[11])<<24;
319 mix(a,b,c);
320 length -= 12;
321 k += 12;
322 }
323
324 /*-------------------------------- last block: affect all 32 bits of (c) */
325 switch(length) /* all the case statements fall through */
326 {
327 case 12: c+=((uint32_t)k[11])<<24; /* fall through */
328 case 11: c+=((uint32_t)k[10])<<16; /* fall through */
329 case 10: c+=((uint32_t)k[9])<<8; /* fall through */
330 case 9 : c+=k[8]; /* fall through */
331 case 8 : b+=((uint32_t)k[7])<<24; /* fall through */
332 case 7 : b+=((uint32_t)k[6])<<16; /* fall through */
333 case 6 : b+=((uint32_t)k[5])<<8; /* fall through */
334 case 5 : b+=k[4]; /* fall through */
335 case 4 : a+=((uint32_t)k[3])<<24; /* fall through */
336 case 3 : a+=((uint32_t)k[2])<<16; /* fall through */
337 case 2 : a+=((uint32_t)k[1])<<8; /* fall through */
338 case 1 : a+=k[0];
339 break;
340 case 0 : return c;
341 }
342 }
343
344 final(a,b,c);
345 *val2 = b;
346 return c;
347 }
348
349 /*
350 * hashbig():
351 * This is the same as hash_word() on big-endian machines. It is different
352 * from hashlittle() on all machines. hashbig() takes advantage of
353 * big-endian byte ordering.
354 */
hashbig(const void * key,size_t length,uint32_t * val2)355 static uint32_t hashbig( const void *key, size_t length, uint32_t *val2)
356 {
357 uint32_t a,b,c;
358 union { const void *ptr; size_t i; } u; /* to cast key to (size_t) happily */
359
360 /* Set up the internal state */
361 a = b = c = 0xdeadbeef + ((uint32_t)length) + *val2;
362
363 u.ptr = key;
364 if (HASH_BIG_ENDIAN && ((u.i & 0x3) == 0)) {
365 const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
366 const uint8_t *k8;
367
368 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
369 while (length > 12)
370 {
371 a += k[0];
372 b += k[1];
373 c += k[2];
374 mix(a,b,c);
375 length -= 12;
376 k += 3;
377 }
378
379 /*----------------------------- handle the last (probably partial) block */
380 /*
381 * "k[2]<<8" actually reads beyond the end of the string, but
382 * then shifts out the part it's not allowed to read. Because the
383 * string is aligned, the illegal read is in the same word as the
384 * rest of the string. Every machine with memory protection I've seen
385 * does it on word boundaries, so is OK with this. But VALGRIND will
386 * still catch it and complain. The masking trick does make the hash
387 * noticably faster for short strings (like English words).
388 *
389 * Not on my testing with gcc 4.5 on an intel i5 CPU, at least --RR.
390 */
391 #if 0
392 switch(length)
393 {
394 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
395 case 11: c+=k[2]&0xffffff00; b+=k[1]; a+=k[0]; break;
396 case 10: c+=k[2]&0xffff0000; b+=k[1]; a+=k[0]; break;
397 case 9 : c+=k[2]&0xff000000; b+=k[1]; a+=k[0]; break;
398 case 8 : b+=k[1]; a+=k[0]; break;
399 case 7 : b+=k[1]&0xffffff00; a+=k[0]; break;
400 case 6 : b+=k[1]&0xffff0000; a+=k[0]; break;
401 case 5 : b+=k[1]&0xff000000; a+=k[0]; break;
402 case 4 : a+=k[0]; break;
403 case 3 : a+=k[0]&0xffffff00; break;
404 case 2 : a+=k[0]&0xffff0000; break;
405 case 1 : a+=k[0]&0xff000000; break;
406 case 0 : return c; /* zero length strings require no mixing */
407 }
408
409 #else /* make valgrind happy */
410
411 k8 = (const uint8_t *)k;
412 switch(length) /* all the case statements fall through */
413 {
414 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
415 case 11: c+=((uint32_t)k8[10])<<8; /* fall through */
416 case 10: c+=((uint32_t)k8[9])<<16; /* fall through */
417 case 9 : c+=((uint32_t)k8[8])<<24; /* fall through */
418 case 8 : b+=k[1]; a+=k[0]; break;
419 case 7 : b+=((uint32_t)k8[6])<<8; /* fall through */
420 case 6 : b+=((uint32_t)k8[5])<<16; /* fall through */
421 case 5 : b+=((uint32_t)k8[4])<<24; /* fall through */
422 case 4 : a+=k[0]; break;
423 case 3 : a+=((uint32_t)k8[2])<<8; /* fall through */
424 case 2 : a+=((uint32_t)k8[1])<<16; /* fall through */
425 case 1 : a+=((uint32_t)k8[0])<<24; break;
426 case 0 : return c;
427 }
428
429 #endif /* !VALGRIND */
430
431 } else { /* need to read the key one byte at a time */
432 const uint8_t *k = (const uint8_t *)key;
433
434 /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
435 while (length > 12)
436 {
437 a += ((uint32_t)k[0])<<24;
438 a += ((uint32_t)k[1])<<16;
439 a += ((uint32_t)k[2])<<8;
440 a += ((uint32_t)k[3]);
441 b += ((uint32_t)k[4])<<24;
442 b += ((uint32_t)k[5])<<16;
443 b += ((uint32_t)k[6])<<8;
444 b += ((uint32_t)k[7]);
445 c += ((uint32_t)k[8])<<24;
446 c += ((uint32_t)k[9])<<16;
447 c += ((uint32_t)k[10])<<8;
448 c += ((uint32_t)k[11]);
449 mix(a,b,c);
450 length -= 12;
451 k += 12;
452 }
453
454 /*-------------------------------- last block: affect all 32 bits of (c) */
455 switch(length) /* all the case statements fall through */
456 {
457 case 12: c+=k[11]; /* fall through */
458 case 11: c+=((uint32_t)k[10])<<8; /* fall through */
459 case 10: c+=((uint32_t)k[9])<<16; /* fall through */
460 case 9 : c+=((uint32_t)k[8])<<24; /* fall through */
461 case 8 : b+=k[7]; /* fall through */
462 case 7 : b+=((uint32_t)k[6])<<8; /* fall through */
463 case 6 : b+=((uint32_t)k[5])<<16; /* fall through */
464 case 5 : b+=((uint32_t)k[4])<<24; /* fall through */
465 case 4 : a+=k[3]; /* fall through */
466 case 3 : a+=((uint32_t)k[2])<<8; /* fall through */
467 case 2 : a+=((uint32_t)k[1])<<16; /* fall through */
468 case 1 : a+=((uint32_t)k[0])<<24; /* fall through */
469 break;
470 case 0 : return c;
471 }
472 }
473
474 final(a,b,c);
475 *val2 = b;
476 return c;
477 }
478
nl_hash_any(const void * key,size_t length,uint32_t base)479 uint32_t nl_hash_any(const void *key, size_t length, uint32_t base)
480 {
481 if (HASH_BIG_ENDIAN)
482 return hashbig(key, length, &base);
483 else
484 return hashlittle(key, length, &base);
485 }
486