xxhash.c

  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 <linux/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 **************************************/
 64static const uint32_t PRIME32_1 = 2654435761U;
 65static const uint32_t PRIME32_2 = 2246822519U;
 66static const uint32_t PRIME32_3 = 3266489917U;
 67static const uint32_t PRIME32_4 =  668265263U;
 68static const uint32_t PRIME32_5 =  374761393U;
 69
 70static const uint64_t PRIME64_1 = 11400714785074694791ULL;
 71static const uint64_t PRIME64_2 = 14029467366897019727ULL;
 72static const uint64_t PRIME64_3 =  1609587929392839161ULL;
 73static const uint64_t PRIME64_4 =  9650029242287828579ULL;
 74static const uint64_t PRIME64_5 =  2870177450012600261ULL;
 75
 76/*-***************************
 77 * Simple Hash Functions
 78 ****************************/
 79static uint32_t xxh32_round(uint32_t seed, const uint32_t input)
 80{
 81	seed += input * PRIME32_2;
 82	seed = xxh_rotl32(seed, 13);
 83	seed *= PRIME32_1;
 84	return seed;
 85}
 86
 87uint32_t xxh32(const void *input, const size_t len, const uint32_t seed)
 88{
 89	const uint8_t *p = (const uint8_t *)input;
 90	const uint8_t *b_end = p + len;
 91	uint32_t h32;
 92
 93	if (len >= 16) {
 94		const uint8_t *const limit = b_end - 16;
 95		uint32_t v1 = seed + PRIME32_1 + PRIME32_2;
 96		uint32_t v2 = seed + PRIME32_2;
 97		uint32_t v3 = seed + 0;
 98		uint32_t v4 = seed - PRIME32_1;
 99
100		do {
101			v1 = xxh32_round(v1, get_unaligned_le32(p));
102			p += 4;
103			v2 = xxh32_round(v2, get_unaligned_le32(p));
104			p += 4;
105			v3 = xxh32_round(v3, get_unaligned_le32(p));
106			p += 4;
107			v4 = xxh32_round(v4, get_unaligned_le32(p));
108			p += 4;
109		} while (p <= limit);
110
111		h32 = xxh_rotl32(v1, 1) + xxh_rotl32(v2, 7) +
112			xxh_rotl32(v3, 12) + xxh_rotl32(v4, 18);
113	} else {
114		h32 = seed + PRIME32_5;
115	}
116
117	h32 += (uint32_t)len;
118
119	while (p + 4 <= b_end) {
120		h32 += get_unaligned_le32(p) * PRIME32_3;
121		h32 = xxh_rotl32(h32, 17) * PRIME32_4;
122		p += 4;
123	}
124
125	while (p < b_end) {
126		h32 += (*p) * PRIME32_5;
127		h32 = xxh_rotl32(h32, 11) * PRIME32_1;
128		p++;
129	}
130
131	h32 ^= h32 >> 15;
132	h32 *= PRIME32_2;
133	h32 ^= h32 >> 13;
134	h32 *= PRIME32_3;
135	h32 ^= h32 >> 16;
136
137	return h32;
138}
139EXPORT_SYMBOL(xxh32);
140
141static uint64_t xxh64_round(uint64_t acc, const uint64_t input)
142{
143	acc += input * PRIME64_2;
144	acc = xxh_rotl64(acc, 31);
145	acc *= PRIME64_1;
146	return acc;
147}
148
149static uint64_t xxh64_merge_round(uint64_t acc, uint64_t val)
150{
151	val = xxh64_round(0, val);
152	acc ^= val;
153	acc = acc * PRIME64_1 + PRIME64_4;
154	return acc;
155}
156
157uint64_t xxh64(const void *input, const size_t len, const uint64_t seed)
158{
159	const uint8_t *p = (const uint8_t *)input;
160	const uint8_t *const b_end = p + len;
161	uint64_t h64;
162
163	if (len >= 32) {
164		const uint8_t *const limit = b_end - 32;
165		uint64_t v1 = seed + PRIME64_1 + PRIME64_2;
166		uint64_t v2 = seed + PRIME64_2;
167		uint64_t v3 = seed + 0;
168		uint64_t v4 = seed - PRIME64_1;
169
170		do {
171			v1 = xxh64_round(v1, get_unaligned_le64(p));
172			p += 8;
173			v2 = xxh64_round(v2, get_unaligned_le64(p));
174			p += 8;
175			v3 = xxh64_round(v3, get_unaligned_le64(p));
176			p += 8;
177			v4 = xxh64_round(v4, get_unaligned_le64(p));
178			p += 8;
179		} while (p <= limit);
180
181		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
182			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
183		h64 = xxh64_merge_round(h64, v1);
184		h64 = xxh64_merge_round(h64, v2);
185		h64 = xxh64_merge_round(h64, v3);
186		h64 = xxh64_merge_round(h64, v4);
187
188	} else {
189		h64  = seed + PRIME64_5;
190	}
191
192	h64 += (uint64_t)len;
193
194	while (p + 8 <= b_end) {
195		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
196
197		h64 ^= k1;
198		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
199		p += 8;
200	}
201
202	if (p + 4 <= b_end) {
203		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
204		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
205		p += 4;
206	}
207
208	while (p < b_end) {
209		h64 ^= (*p) * PRIME64_5;
210		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
211		p++;
212	}
213
214	h64 ^= h64 >> 33;
215	h64 *= PRIME64_2;
216	h64 ^= h64 >> 29;
217	h64 *= PRIME64_3;
218	h64 ^= h64 >> 32;
219
220	return h64;
221}
222EXPORT_SYMBOL(xxh64);
223
224/*-**************************************************
225 * Advanced Hash Functions
226 ***************************************************/
227void xxh64_reset(struct xxh64_state *statePtr, const uint64_t seed)
228{
229	/* use a local state for memcpy() to avoid strict-aliasing warnings */
230	struct xxh64_state state;
231
232	memset(&state, 0, sizeof(state));
233	state.v1 = seed + PRIME64_1 + PRIME64_2;
234	state.v2 = seed + PRIME64_2;
235	state.v3 = seed + 0;
236	state.v4 = seed - PRIME64_1;
237	memcpy(statePtr, &state, sizeof(state));
238}
239EXPORT_SYMBOL(xxh64_reset);
240
241int xxh64_update(struct xxh64_state *state, const void *input, const size_t len)
242{
243	const uint8_t *p = (const uint8_t *)input;
244	const uint8_t *const b_end = p + len;
245
246	if (input == NULL)
247		return -EINVAL;
248
249	state->total_len += len;
250
251	if (state->memsize + len < 32) { /* fill in tmp buffer */
252		memcpy(((uint8_t *)state->mem64) + state->memsize, input, len);
253		state->memsize += (uint32_t)len;
254		return 0;
255	}
256
257	if (state->memsize) { /* tmp buffer is full */
258		uint64_t *p64 = state->mem64;
259
260		memcpy(((uint8_t *)p64) + state->memsize, input,
261			32 - state->memsize);
262
263		state->v1 = xxh64_round(state->v1, get_unaligned_le64(p64));
264		p64++;
265		state->v2 = xxh64_round(state->v2, get_unaligned_le64(p64));
266		p64++;
267		state->v3 = xxh64_round(state->v3, get_unaligned_le64(p64));
268		p64++;
269		state->v4 = xxh64_round(state->v4, get_unaligned_le64(p64));
270
271		p += 32 - state->memsize;
272		state->memsize = 0;
273	}
274
275	if (p + 32 <= b_end) {
276		const uint8_t *const limit = b_end - 32;
277		uint64_t v1 = state->v1;
278		uint64_t v2 = state->v2;
279		uint64_t v3 = state->v3;
280		uint64_t v4 = state->v4;
281
282		do {
283			v1 = xxh64_round(v1, get_unaligned_le64(p));
284			p += 8;
285			v2 = xxh64_round(v2, get_unaligned_le64(p));
286			p += 8;
287			v3 = xxh64_round(v3, get_unaligned_le64(p));
288			p += 8;
289			v4 = xxh64_round(v4, get_unaligned_le64(p));
290			p += 8;
291		} while (p <= limit);
292
293		state->v1 = v1;
294		state->v2 = v2;
295		state->v3 = v3;
296		state->v4 = v4;
297	}
298
299	if (p < b_end) {
300		memcpy(state->mem64, p, (size_t)(b_end-p));
301		state->memsize = (uint32_t)(b_end - p);
302	}
303
304	return 0;
305}
306EXPORT_SYMBOL(xxh64_update);
307
308uint64_t xxh64_digest(const struct xxh64_state *state)
309{
310	const uint8_t *p = (const uint8_t *)state->mem64;
311	const uint8_t *const b_end = (const uint8_t *)state->mem64 +
312		state->memsize;
313	uint64_t h64;
314
315	if (state->total_len >= 32) {
316		const uint64_t v1 = state->v1;
317		const uint64_t v2 = state->v2;
318		const uint64_t v3 = state->v3;
319		const uint64_t v4 = state->v4;
320
321		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
322			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
323		h64 = xxh64_merge_round(h64, v1);
324		h64 = xxh64_merge_round(h64, v2);
325		h64 = xxh64_merge_round(h64, v3);
326		h64 = xxh64_merge_round(h64, v4);
327	} else {
328		h64  = state->v3 + PRIME64_5;
329	}
330
331	h64 += (uint64_t)state->total_len;
332
333	while (p + 8 <= b_end) {
334		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
335
336		h64 ^= k1;
337		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
338		p += 8;
339	}
340
341	if (p + 4 <= b_end) {
342		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
343		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
344		p += 4;
345	}
346
347	while (p < b_end) {
348		h64 ^= (*p) * PRIME64_5;
349		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
350		p++;
351	}
352
353	h64 ^= h64 >> 33;
354	h64 *= PRIME64_2;
355	h64 ^= h64 >> 29;
356	h64 *= PRIME64_3;
357	h64 ^= h64 >> 32;
358
359	return h64;
360}
361EXPORT_SYMBOL(xxh64_digest);
362
363MODULE_LICENSE("Dual BSD/GPL");
364MODULE_DESCRIPTION("xxHash");