xref: /linux/lib/xxhash.c (revision 0526b56cbc3c489642bd6a5fe4b718dea7ef0ee8)
1 /*
2  * xxHash - Extremely Fast Hash algorithm
3  * Copyright (C) 2012-2016, Yann Collet.
4  *
5  * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions are
9  * met:
10  *
11  *   * Redistributions of source code must retain the above copyright
12  *     notice, this list of conditions and the following disclaimer.
13  *   * Redistributions in binary form must reproduce the above
14  *     copyright notice, this list of conditions and the following disclaimer
15  *     in the documentation and/or other materials provided with the
16  *     distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29  *
30  * This program is free software; you can redistribute it and/or modify it under
31  * the terms of the GNU General Public License version 2 as published by the
32  * Free Software Foundation. This program is dual-licensed; you may select
33  * either version 2 of the GNU General Public License ("GPL") or BSD license
34  * ("BSD").
35  *
36  * You can contact the author at:
37  * - xxHash homepage: https://cyan4973.github.io/xxHash/
38  * - xxHash source repository: https://github.com/Cyan4973/xxHash
39  */
40 
41 #include <asm/unaligned.h>
42 #include <linux/errno.h>
43 #include <linux/compiler.h>
44 #include <linux/kernel.h>
45 #include <linux/module.h>
46 #include <linux/string.h>
47 #include <linux/xxhash.h>
48 
49 /*-*************************************
50  * Macros
51  **************************************/
52 #define xxh_rotl32(x, r) ((x << r) | (x >> (32 - r)))
53 #define xxh_rotl64(x, r) ((x << r) | (x >> (64 - r)))
54 
55 #ifdef __LITTLE_ENDIAN
56 # define XXH_CPU_LITTLE_ENDIAN 1
57 #else
58 # define XXH_CPU_LITTLE_ENDIAN 0
59 #endif
60 
61 /*-*************************************
62  * Constants
63  **************************************/
64 static const uint32_t PRIME32_1 = 2654435761U;
65 static const uint32_t PRIME32_2 = 2246822519U;
66 static const uint32_t PRIME32_3 = 3266489917U;
67 static const uint32_t PRIME32_4 =  668265263U;
68 static const uint32_t PRIME32_5 =  374761393U;
69 
70 static const uint64_t PRIME64_1 = 11400714785074694791ULL;
71 static const uint64_t PRIME64_2 = 14029467366897019727ULL;
72 static const uint64_t PRIME64_3 =  1609587929392839161ULL;
73 static const uint64_t PRIME64_4 =  9650029242287828579ULL;
74 static const uint64_t PRIME64_5 =  2870177450012600261ULL;
75 
76 /*-**************************
77  *  Utils
78  ***************************/
79 void xxh32_copy_state(struct xxh32_state *dst, const struct xxh32_state *src)
80 {
81 	memcpy(dst, src, sizeof(*dst));
82 }
83 EXPORT_SYMBOL(xxh32_copy_state);
84 
85 void xxh64_copy_state(struct xxh64_state *dst, const struct xxh64_state *src)
86 {
87 	memcpy(dst, src, sizeof(*dst));
88 }
89 EXPORT_SYMBOL(xxh64_copy_state);
90 
91 /*-***************************
92  * Simple Hash Functions
93  ****************************/
94 static uint32_t xxh32_round(uint32_t seed, const uint32_t input)
95 {
96 	seed += input * PRIME32_2;
97 	seed = xxh_rotl32(seed, 13);
98 	seed *= PRIME32_1;
99 	return seed;
100 }
101 
102 uint32_t xxh32(const void *input, const size_t len, const uint32_t seed)
103 {
104 	const uint8_t *p = (const uint8_t *)input;
105 	const uint8_t *b_end = p + len;
106 	uint32_t h32;
107 
108 	if (len >= 16) {
109 		const uint8_t *const limit = b_end - 16;
110 		uint32_t v1 = seed + PRIME32_1 + PRIME32_2;
111 		uint32_t v2 = seed + PRIME32_2;
112 		uint32_t v3 = seed + 0;
113 		uint32_t v4 = seed - PRIME32_1;
114 
115 		do {
116 			v1 = xxh32_round(v1, get_unaligned_le32(p));
117 			p += 4;
118 			v2 = xxh32_round(v2, get_unaligned_le32(p));
119 			p += 4;
120 			v3 = xxh32_round(v3, get_unaligned_le32(p));
121 			p += 4;
122 			v4 = xxh32_round(v4, get_unaligned_le32(p));
123 			p += 4;
124 		} while (p <= limit);
125 
126 		h32 = xxh_rotl32(v1, 1) + xxh_rotl32(v2, 7) +
127 			xxh_rotl32(v3, 12) + xxh_rotl32(v4, 18);
128 	} else {
129 		h32 = seed + PRIME32_5;
130 	}
131 
132 	h32 += (uint32_t)len;
133 
134 	while (p + 4 <= b_end) {
135 		h32 += get_unaligned_le32(p) * PRIME32_3;
136 		h32 = xxh_rotl32(h32, 17) * PRIME32_4;
137 		p += 4;
138 	}
139 
140 	while (p < b_end) {
141 		h32 += (*p) * PRIME32_5;
142 		h32 = xxh_rotl32(h32, 11) * PRIME32_1;
143 		p++;
144 	}
145 
146 	h32 ^= h32 >> 15;
147 	h32 *= PRIME32_2;
148 	h32 ^= h32 >> 13;
149 	h32 *= PRIME32_3;
150 	h32 ^= h32 >> 16;
151 
152 	return h32;
153 }
154 EXPORT_SYMBOL(xxh32);
155 
156 static uint64_t xxh64_round(uint64_t acc, const uint64_t input)
157 {
158 	acc += input * PRIME64_2;
159 	acc = xxh_rotl64(acc, 31);
160 	acc *= PRIME64_1;
161 	return acc;
162 }
163 
164 static uint64_t xxh64_merge_round(uint64_t acc, uint64_t val)
165 {
166 	val = xxh64_round(0, val);
167 	acc ^= val;
168 	acc = acc * PRIME64_1 + PRIME64_4;
169 	return acc;
170 }
171 
172 uint64_t xxh64(const void *input, const size_t len, const uint64_t seed)
173 {
174 	const uint8_t *p = (const uint8_t *)input;
175 	const uint8_t *const b_end = p + len;
176 	uint64_t h64;
177 
178 	if (len >= 32) {
179 		const uint8_t *const limit = b_end - 32;
180 		uint64_t v1 = seed + PRIME64_1 + PRIME64_2;
181 		uint64_t v2 = seed + PRIME64_2;
182 		uint64_t v3 = seed + 0;
183 		uint64_t v4 = seed - PRIME64_1;
184 
185 		do {
186 			v1 = xxh64_round(v1, get_unaligned_le64(p));
187 			p += 8;
188 			v2 = xxh64_round(v2, get_unaligned_le64(p));
189 			p += 8;
190 			v3 = xxh64_round(v3, get_unaligned_le64(p));
191 			p += 8;
192 			v4 = xxh64_round(v4, get_unaligned_le64(p));
193 			p += 8;
194 		} while (p <= limit);
195 
196 		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
197 			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
198 		h64 = xxh64_merge_round(h64, v1);
199 		h64 = xxh64_merge_round(h64, v2);
200 		h64 = xxh64_merge_round(h64, v3);
201 		h64 = xxh64_merge_round(h64, v4);
202 
203 	} else {
204 		h64  = seed + PRIME64_5;
205 	}
206 
207 	h64 += (uint64_t)len;
208 
209 	while (p + 8 <= b_end) {
210 		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
211 
212 		h64 ^= k1;
213 		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
214 		p += 8;
215 	}
216 
217 	if (p + 4 <= b_end) {
218 		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
219 		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
220 		p += 4;
221 	}
222 
223 	while (p < b_end) {
224 		h64 ^= (*p) * PRIME64_5;
225 		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
226 		p++;
227 	}
228 
229 	h64 ^= h64 >> 33;
230 	h64 *= PRIME64_2;
231 	h64 ^= h64 >> 29;
232 	h64 *= PRIME64_3;
233 	h64 ^= h64 >> 32;
234 
235 	return h64;
236 }
237 EXPORT_SYMBOL(xxh64);
238 
239 /*-**************************************************
240  * Advanced Hash Functions
241  ***************************************************/
242 void xxh32_reset(struct xxh32_state *statePtr, const uint32_t seed)
243 {
244 	/* use a local state for memcpy() to avoid strict-aliasing warnings */
245 	struct xxh32_state state;
246 
247 	memset(&state, 0, sizeof(state));
248 	state.v1 = seed + PRIME32_1 + PRIME32_2;
249 	state.v2 = seed + PRIME32_2;
250 	state.v3 = seed + 0;
251 	state.v4 = seed - PRIME32_1;
252 	memcpy(statePtr, &state, sizeof(state));
253 }
254 EXPORT_SYMBOL(xxh32_reset);
255 
256 void xxh64_reset(struct xxh64_state *statePtr, const uint64_t seed)
257 {
258 	/* use a local state for memcpy() to avoid strict-aliasing warnings */
259 	struct xxh64_state state;
260 
261 	memset(&state, 0, sizeof(state));
262 	state.v1 = seed + PRIME64_1 + PRIME64_2;
263 	state.v2 = seed + PRIME64_2;
264 	state.v3 = seed + 0;
265 	state.v4 = seed - PRIME64_1;
266 	memcpy(statePtr, &state, sizeof(state));
267 }
268 EXPORT_SYMBOL(xxh64_reset);
269 
270 int xxh32_update(struct xxh32_state *state, const void *input, const size_t len)
271 {
272 	const uint8_t *p = (const uint8_t *)input;
273 	const uint8_t *const b_end = p + len;
274 
275 	if (input == NULL)
276 		return -EINVAL;
277 
278 	state->total_len_32 += (uint32_t)len;
279 	state->large_len |= (len >= 16) | (state->total_len_32 >= 16);
280 
281 	if (state->memsize + len < 16) { /* fill in tmp buffer */
282 		memcpy((uint8_t *)(state->mem32) + state->memsize, input, len);
283 		state->memsize += (uint32_t)len;
284 		return 0;
285 	}
286 
287 	if (state->memsize) { /* some data left from previous update */
288 		const uint32_t *p32 = state->mem32;
289 
290 		memcpy((uint8_t *)(state->mem32) + state->memsize, input,
291 			16 - state->memsize);
292 
293 		state->v1 = xxh32_round(state->v1, get_unaligned_le32(p32));
294 		p32++;
295 		state->v2 = xxh32_round(state->v2, get_unaligned_le32(p32));
296 		p32++;
297 		state->v3 = xxh32_round(state->v3, get_unaligned_le32(p32));
298 		p32++;
299 		state->v4 = xxh32_round(state->v4, get_unaligned_le32(p32));
300 		p32++;
301 
302 		p += 16-state->memsize;
303 		state->memsize = 0;
304 	}
305 
306 	if (p <= b_end - 16) {
307 		const uint8_t *const limit = b_end - 16;
308 		uint32_t v1 = state->v1;
309 		uint32_t v2 = state->v2;
310 		uint32_t v3 = state->v3;
311 		uint32_t v4 = state->v4;
312 
313 		do {
314 			v1 = xxh32_round(v1, get_unaligned_le32(p));
315 			p += 4;
316 			v2 = xxh32_round(v2, get_unaligned_le32(p));
317 			p += 4;
318 			v3 = xxh32_round(v3, get_unaligned_le32(p));
319 			p += 4;
320 			v4 = xxh32_round(v4, get_unaligned_le32(p));
321 			p += 4;
322 		} while (p <= limit);
323 
324 		state->v1 = v1;
325 		state->v2 = v2;
326 		state->v3 = v3;
327 		state->v4 = v4;
328 	}
329 
330 	if (p < b_end) {
331 		memcpy(state->mem32, p, (size_t)(b_end-p));
332 		state->memsize = (uint32_t)(b_end-p);
333 	}
334 
335 	return 0;
336 }
337 EXPORT_SYMBOL(xxh32_update);
338 
339 uint32_t xxh32_digest(const struct xxh32_state *state)
340 {
341 	const uint8_t *p = (const uint8_t *)state->mem32;
342 	const uint8_t *const b_end = (const uint8_t *)(state->mem32) +
343 		state->memsize;
344 	uint32_t h32;
345 
346 	if (state->large_len) {
347 		h32 = xxh_rotl32(state->v1, 1) + xxh_rotl32(state->v2, 7) +
348 			xxh_rotl32(state->v3, 12) + xxh_rotl32(state->v4, 18);
349 	} else {
350 		h32 = state->v3 /* == seed */ + PRIME32_5;
351 	}
352 
353 	h32 += state->total_len_32;
354 
355 	while (p + 4 <= b_end) {
356 		h32 += get_unaligned_le32(p) * PRIME32_3;
357 		h32 = xxh_rotl32(h32, 17) * PRIME32_4;
358 		p += 4;
359 	}
360 
361 	while (p < b_end) {
362 		h32 += (*p) * PRIME32_5;
363 		h32 = xxh_rotl32(h32, 11) * PRIME32_1;
364 		p++;
365 	}
366 
367 	h32 ^= h32 >> 15;
368 	h32 *= PRIME32_2;
369 	h32 ^= h32 >> 13;
370 	h32 *= PRIME32_3;
371 	h32 ^= h32 >> 16;
372 
373 	return h32;
374 }
375 EXPORT_SYMBOL(xxh32_digest);
376 
377 int xxh64_update(struct xxh64_state *state, const void *input, const size_t len)
378 {
379 	const uint8_t *p = (const uint8_t *)input;
380 	const uint8_t *const b_end = p + len;
381 
382 	if (input == NULL)
383 		return -EINVAL;
384 
385 	state->total_len += len;
386 
387 	if (state->memsize + len < 32) { /* fill in tmp buffer */
388 		memcpy(((uint8_t *)state->mem64) + state->memsize, input, len);
389 		state->memsize += (uint32_t)len;
390 		return 0;
391 	}
392 
393 	if (state->memsize) { /* tmp buffer is full */
394 		uint64_t *p64 = state->mem64;
395 
396 		memcpy(((uint8_t *)p64) + state->memsize, input,
397 			32 - state->memsize);
398 
399 		state->v1 = xxh64_round(state->v1, get_unaligned_le64(p64));
400 		p64++;
401 		state->v2 = xxh64_round(state->v2, get_unaligned_le64(p64));
402 		p64++;
403 		state->v3 = xxh64_round(state->v3, get_unaligned_le64(p64));
404 		p64++;
405 		state->v4 = xxh64_round(state->v4, get_unaligned_le64(p64));
406 
407 		p += 32 - state->memsize;
408 		state->memsize = 0;
409 	}
410 
411 	if (p + 32 <= b_end) {
412 		const uint8_t *const limit = b_end - 32;
413 		uint64_t v1 = state->v1;
414 		uint64_t v2 = state->v2;
415 		uint64_t v3 = state->v3;
416 		uint64_t v4 = state->v4;
417 
418 		do {
419 			v1 = xxh64_round(v1, get_unaligned_le64(p));
420 			p += 8;
421 			v2 = xxh64_round(v2, get_unaligned_le64(p));
422 			p += 8;
423 			v3 = xxh64_round(v3, get_unaligned_le64(p));
424 			p += 8;
425 			v4 = xxh64_round(v4, get_unaligned_le64(p));
426 			p += 8;
427 		} while (p <= limit);
428 
429 		state->v1 = v1;
430 		state->v2 = v2;
431 		state->v3 = v3;
432 		state->v4 = v4;
433 	}
434 
435 	if (p < b_end) {
436 		memcpy(state->mem64, p, (size_t)(b_end-p));
437 		state->memsize = (uint32_t)(b_end - p);
438 	}
439 
440 	return 0;
441 }
442 EXPORT_SYMBOL(xxh64_update);
443 
444 uint64_t xxh64_digest(const struct xxh64_state *state)
445 {
446 	const uint8_t *p = (const uint8_t *)state->mem64;
447 	const uint8_t *const b_end = (const uint8_t *)state->mem64 +
448 		state->memsize;
449 	uint64_t h64;
450 
451 	if (state->total_len >= 32) {
452 		const uint64_t v1 = state->v1;
453 		const uint64_t v2 = state->v2;
454 		const uint64_t v3 = state->v3;
455 		const uint64_t v4 = state->v4;
456 
457 		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
458 			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
459 		h64 = xxh64_merge_round(h64, v1);
460 		h64 = xxh64_merge_round(h64, v2);
461 		h64 = xxh64_merge_round(h64, v3);
462 		h64 = xxh64_merge_round(h64, v4);
463 	} else {
464 		h64  = state->v3 + PRIME64_5;
465 	}
466 
467 	h64 += (uint64_t)state->total_len;
468 
469 	while (p + 8 <= b_end) {
470 		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
471 
472 		h64 ^= k1;
473 		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
474 		p += 8;
475 	}
476 
477 	if (p + 4 <= b_end) {
478 		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
479 		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
480 		p += 4;
481 	}
482 
483 	while (p < b_end) {
484 		h64 ^= (*p) * PRIME64_5;
485 		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
486 		p++;
487 	}
488 
489 	h64 ^= h64 >> 33;
490 	h64 *= PRIME64_2;
491 	h64 ^= h64 >> 29;
492 	h64 *= PRIME64_3;
493 	h64 ^= h64 >> 32;
494 
495 	return h64;
496 }
497 EXPORT_SYMBOL(xxh64_digest);
498 
499 MODULE_LICENSE("Dual BSD/GPL");
500 MODULE_DESCRIPTION("xxHash");
501