the-tree/lib/xxhash.c

0001 /*
0002  * xxHash - Extremely Fast Hash algorithm
0003  * Copyright (C) 2012-2016, Yann Collet.
0004  *
0005  * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
0006  *
0007  * Redistribution and use in source and binary forms, with or without
0008  * modification, are permitted provided that the following conditions are
0009  * met:
0010  *
0011  *   * Redistributions of source code must retain the above copyright
0012  *     notice, this list of conditions and the following disclaimer.
0013  *   * Redistributions in binary form must reproduce the above
0014  *     copyright notice, this list of conditions and the following disclaimer
0015  *     in the documentation and/or other materials provided with the
0016  *     distribution.
0017  *
0018  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
0019  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
0020  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
0021  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
0022  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
0023  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
0024  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
0025  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
0026  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
0027  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
0028  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
0029  *
0030  * This program is free software; you can redistribute it and/or modify it under
0031  * the terms of the GNU General Public License version 2 as published by the
0032  * Free Software Foundation. This program is dual-licensed; you may select
0033  * either version 2 of the GNU General Public License ("GPL") or BSD license
0034  * ("BSD").
0035  *
0036  * You can contact the author at:
0037  * - xxHash homepage: https://cyan4973.github.io/xxHash/
0038  * - xxHash source repository: https://github.com/Cyan4973/xxHash
0039  */
0040
0041 #include <asm/unaligned.h>
0042 #include <linux/errno.h>
0043 #include <linux/compiler.h>
0044 #include <linux/kernel.h>
0045 #include <linux/module.h>
0046 #include <linux/string.h>
0047 #include <linux/xxhash.h>
0048
0049 /*-*************************************
0050  * Macros
0051  **************************************/
0052 #define xxh_rotl32(x, r) ((x << r) | (x >> (32 - r)))
0053 #define xxh_rotl64(x, r) ((x << r) | (x >> (64 - r)))
0054
0055 #ifdef __LITTLE_ENDIAN
0056 # define XXH_CPU_LITTLE_ENDIAN 1
0057 #else
0058 # define XXH_CPU_LITTLE_ENDIAN 0
0059 #endif
0060
0061 /*-*************************************
0062  * Constants
0063  **************************************/
0064 static const uint32_t PRIME32_1 = 2654435761U;
0065 static const uint32_t PRIME32_2 = 2246822519U;
0066 static const uint32_t PRIME32_3 = 3266489917U;
0067 static const uint32_t PRIME32_4 =  668265263U;
0068 static const uint32_t PRIME32_5 =  374761393U;
0069
0070 static const uint64_t PRIME64_1 = 11400714785074694791ULL;
0071 static const uint64_t PRIME64_2 = 14029467366897019727ULL;
0072 static const uint64_t PRIME64_3 =  1609587929392839161ULL;
0073 static const uint64_t PRIME64_4 =  9650029242287828579ULL;
0074 static const uint64_t PRIME64_5 =  2870177450012600261ULL;
0075
0076 /*-**************************
0077  *  Utils
0078  ***************************/
0079 void xxh32_copy_state(struct xxh32_state *dst, const struct xxh32_state *src)
0080 {
0081     memcpy(dst, src, sizeof(*dst));
0082 }
0083 EXPORT_SYMBOL(xxh32_copy_state);
0084
0085 void xxh64_copy_state(struct xxh64_state *dst, const struct xxh64_state *src)
0086 {
0087     memcpy(dst, src, sizeof(*dst));
0088 }
0089 EXPORT_SYMBOL(xxh64_copy_state);
0090
0091 /*-***************************
0092  * Simple Hash Functions
0093  ****************************/
0094 static uint32_t xxh32_round(uint32_t seed, const uint32_t input)
0095 {
0096     seed += input * PRIME32_2;
0097     seed = xxh_rotl32(seed, 13);
0098     seed *= PRIME32_1;
0099     return seed;
0100 }
0101
0102 uint32_t xxh32(const void *input, const size_t len, const uint32_t seed)
0103 {
0104     const uint8_t *p = (const uint8_t *)input;
0105     const uint8_t *b_end = p + len;
0106     uint32_t h32;
0107
0108     if (len >= 16) {
0109         const uint8_t *const limit = b_end - 16;
0110         uint32_t v1 = seed + PRIME32_1 + PRIME32_2;
0111         uint32_t v2 = seed + PRIME32_2;
0112         uint32_t v3 = seed + 0;
0113         uint32_t v4 = seed - PRIME32_1;
0114
0115         do {
0116             v1 = xxh32_round(v1, get_unaligned_le32(p));
0117             p += 4;
0118             v2 = xxh32_round(v2, get_unaligned_le32(p));
0119             p += 4;
0120             v3 = xxh32_round(v3, get_unaligned_le32(p));
0121             p += 4;
0122             v4 = xxh32_round(v4, get_unaligned_le32(p));
0123             p += 4;
0124         } while (p <= limit);
0125
0126         h32 = xxh_rotl32(v1, 1) + xxh_rotl32(v2, 7) +
0127             xxh_rotl32(v3, 12) + xxh_rotl32(v4, 18);
0128     } else {
0129         h32 = seed + PRIME32_5;
0130     }
0131
0132     h32 += (uint32_t)len;
0133
0134     while (p + 4 <= b_end) {
0135         h32 += get_unaligned_le32(p) * PRIME32_3;
0136         h32 = xxh_rotl32(h32, 17) * PRIME32_4;
0137         p += 4;
0138     }
0139
0140     while (p < b_end) {
0141         h32 += (*p) * PRIME32_5;
0142         h32 = xxh_rotl32(h32, 11) * PRIME32_1;
0143         p++;
0144     }
0145
0146     h32 ^= h32 >> 15;
0147     h32 *= PRIME32_2;
0148     h32 ^= h32 >> 13;
0149     h32 *= PRIME32_3;
0150     h32 ^= h32 >> 16;
0151
0152     return h32;
0153 }
0154 EXPORT_SYMBOL(xxh32);
0155
0156 static uint64_t xxh64_round(uint64_t acc, const uint64_t input)
0157 {
0158     acc += input * PRIME64_2;
0159     acc = xxh_rotl64(acc, 31);
0160     acc *= PRIME64_1;
0161     return acc;
0162 }
0163
0164 static uint64_t xxh64_merge_round(uint64_t acc, uint64_t val)
0165 {
0166     val = xxh64_round(0, val);
0167     acc ^= val;
0168     acc = acc * PRIME64_1 + PRIME64_4;
0169     return acc;
0170 }
0171
0172 uint64_t xxh64(const void *input, const size_t len, const uint64_t seed)
0173 {
0174     const uint8_t *p = (const uint8_t *)input;
0175     const uint8_t *const b_end = p + len;
0176     uint64_t h64;
0177
0178     if (len >= 32) {
0179         const uint8_t *const limit = b_end - 32;
0180         uint64_t v1 = seed + PRIME64_1 + PRIME64_2;
0181         uint64_t v2 = seed + PRIME64_2;
0182         uint64_t v3 = seed + 0;
0183         uint64_t v4 = seed - PRIME64_1;
0184
0185         do {
0186             v1 = xxh64_round(v1, get_unaligned_le64(p));
0187             p += 8;
0188             v2 = xxh64_round(v2, get_unaligned_le64(p));
0189             p += 8;
0190             v3 = xxh64_round(v3, get_unaligned_le64(p));
0191             p += 8;
0192             v4 = xxh64_round(v4, get_unaligned_le64(p));
0193             p += 8;
0194         } while (p <= limit);
0195
0196         h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
0197             xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
0198         h64 = xxh64_merge_round(h64, v1);
0199         h64 = xxh64_merge_round(h64, v2);
0200         h64 = xxh64_merge_round(h64, v3);
0201         h64 = xxh64_merge_round(h64, v4);
0202
0203     } else {
0204         h64  = seed + PRIME64_5;
0205     }
0206
0207     h64 += (uint64_t)len;
0208
0209     while (p + 8 <= b_end) {
0210         const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
0211
0212         h64 ^= k1;
0213         h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
0214         p += 8;
0215     }
0216
0217     if (p + 4 <= b_end) {
0218         h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
0219         h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
0220         p += 4;
0221     }
0222
0223     while (p < b_end) {
0224         h64 ^= (*p) * PRIME64_5;
0225         h64 = xxh_rotl64(h64, 11) * PRIME64_1;
0226         p++;
0227     }
0228
0229     h64 ^= h64 >> 33;
0230     h64 *= PRIME64_2;
0231     h64 ^= h64 >> 29;
0232     h64 *= PRIME64_3;
0233     h64 ^= h64 >> 32;
0234
0235     return h64;
0236 }
0237 EXPORT_SYMBOL(xxh64);
0238
0239 /*-**************************************************
0240  * Advanced Hash Functions
0241  ***************************************************/
0242 void xxh32_reset(struct xxh32_state *statePtr, const uint32_t seed)
0243 {
0244     /* use a local state for memcpy() to avoid strict-aliasing warnings */
0245     struct xxh32_state state;
0246
0247     memset(&state, 0, sizeof(state));
0248     state.v1 = seed + PRIME32_1 + PRIME32_2;
0249     state.v2 = seed + PRIME32_2;
0250     state.v3 = seed + 0;
0251     state.v4 = seed - PRIME32_1;
0252     memcpy(statePtr, &state, sizeof(state));
0253 }
0254 EXPORT_SYMBOL(xxh32_reset);
0255
0256 void xxh64_reset(struct xxh64_state *statePtr, const uint64_t seed)
0257 {
0258     /* use a local state for memcpy() to avoid strict-aliasing warnings */
0259     struct xxh64_state state;
0260
0261     memset(&state, 0, sizeof(state));
0262     state.v1 = seed + PRIME64_1 + PRIME64_2;
0263     state.v2 = seed + PRIME64_2;
0264     state.v3 = seed + 0;
0265     state.v4 = seed - PRIME64_1;
0266     memcpy(statePtr, &state, sizeof(state));
0267 }
0268 EXPORT_SYMBOL(xxh64_reset);
0269
0270 int xxh32_update(struct xxh32_state *state, const void *input, const size_t len)
0271 {
0272     const uint8_t *p = (const uint8_t *)input;
0273     const uint8_t *const b_end = p + len;
0274
0275     if (input == NULL)
0276         return -EINVAL;
0277
0278     state->total_len_32 += (uint32_t)len;
0279     state->large_len |= (len >= 16) | (state->total_len_32 >= 16);
0280
0281     if (state->memsize + len < 16) { /* fill in tmp buffer */
0282         memcpy((uint8_t *)(state->mem32) + state->memsize, input, len);
0283         state->memsize += (uint32_t)len;
0284         return 0;
0285     }
0286
0287     if (state->memsize) { /* some data left from previous update */
0288         const uint32_t *p32 = state->mem32;
0289
0290         memcpy((uint8_t *)(state->mem32) + state->memsize, input,
0291             16 - state->memsize);
0292
0293         state->v1 = xxh32_round(state->v1, get_unaligned_le32(p32));
0294         p32++;
0295         state->v2 = xxh32_round(state->v2, get_unaligned_le32(p32));
0296         p32++;
0297         state->v3 = xxh32_round(state->v3, get_unaligned_le32(p32));
0298         p32++;
0299         state->v4 = xxh32_round(state->v4, get_unaligned_le32(p32));
0300         p32++;
0301
0302         p += 16-state->memsize;
0303         state->memsize = 0;
0304     }
0305
0306     if (p <= b_end - 16) {
0307         const uint8_t *const limit = b_end - 16;
0308         uint32_t v1 = state->v1;
0309         uint32_t v2 = state->v2;
0310         uint32_t v3 = state->v3;
0311         uint32_t v4 = state->v4;
0312
0313         do {
0314             v1 = xxh32_round(v1, get_unaligned_le32(p));
0315             p += 4;
0316             v2 = xxh32_round(v2, get_unaligned_le32(p));
0317             p += 4;
0318             v3 = xxh32_round(v3, get_unaligned_le32(p));
0319             p += 4;
0320             v4 = xxh32_round(v4, get_unaligned_le32(p));
0321             p += 4;
0322         } while (p <= limit);
0323
0324         state->v1 = v1;
0325         state->v2 = v2;
0326         state->v3 = v3;
0327         state->v4 = v4;
0328     }
0329
0330     if (p < b_end) {
0331         memcpy(state->mem32, p, (size_t)(b_end-p));
0332         state->memsize = (uint32_t)(b_end-p);
0333     }
0334
0335     return 0;
0336 }
0337 EXPORT_SYMBOL(xxh32_update);
0338
0339 uint32_t xxh32_digest(const struct xxh32_state *state)
0340 {
0341     const uint8_t *p = (const uint8_t *)state->mem32;
0342     const uint8_t *const b_end = (const uint8_t *)(state->mem32) +
0343         state->memsize;
0344     uint32_t h32;
0345
0346     if (state->large_len) {
0347         h32 = xxh_rotl32(state->v1, 1) + xxh_rotl32(state->v2, 7) +
0348             xxh_rotl32(state->v3, 12) + xxh_rotl32(state->v4, 18);
0349     } else {
0350         h32 = state->v3 /* == seed */ + PRIME32_5;
0351     }
0352
0353     h32 += state->total_len_32;
0354
0355     while (p + 4 <= b_end) {
0356         h32 += get_unaligned_le32(p) * PRIME32_3;
0357         h32 = xxh_rotl32(h32, 17) * PRIME32_4;
0358         p += 4;
0359     }
0360
0361     while (p < b_end) {
0362         h32 += (*p) * PRIME32_5;
0363         h32 = xxh_rotl32(h32, 11) * PRIME32_1;
0364         p++;
0365     }
0366
0367     h32 ^= h32 >> 15;
0368     h32 *= PRIME32_2;
0369     h32 ^= h32 >> 13;
0370     h32 *= PRIME32_3;
0371     h32 ^= h32 >> 16;
0372
0373     return h32;
0374 }
0375 EXPORT_SYMBOL(xxh32_digest);
0376
0377 int xxh64_update(struct xxh64_state *state, const void *input, const size_t len)
0378 {
0379     const uint8_t *p = (const uint8_t *)input;
0380     const uint8_t *const b_end = p + len;
0381
0382     if (input == NULL)
0383         return -EINVAL;
0384
0385     state->total_len += len;
0386
0387     if (state->memsize + len < 32) { /* fill in tmp buffer */
0388         memcpy(((uint8_t *)state->mem64) + state->memsize, input, len);
0389         state->memsize += (uint32_t)len;
0390         return 0;
0391     }
0392
0393     if (state->memsize) { /* tmp buffer is full */
0394         uint64_t *p64 = state->mem64;
0395
0396         memcpy(((uint8_t *)p64) + state->memsize, input,
0397             32 - state->memsize);
0398
0399         state->v1 = xxh64_round(state->v1, get_unaligned_le64(p64));
0400         p64++;
0401         state->v2 = xxh64_round(state->v2, get_unaligned_le64(p64));
0402         p64++;
0403         state->v3 = xxh64_round(state->v3, get_unaligned_le64(p64));
0404         p64++;
0405         state->v4 = xxh64_round(state->v4, get_unaligned_le64(p64));
0406
0407         p += 32 - state->memsize;
0408         state->memsize = 0;
0409     }
0410
0411     if (p + 32 <= b_end) {
0412         const uint8_t *const limit = b_end - 32;
0413         uint64_t v1 = state->v1;
0414         uint64_t v2 = state->v2;
0415         uint64_t v3 = state->v3;
0416         uint64_t v4 = state->v4;
0417
0418         do {
0419             v1 = xxh64_round(v1, get_unaligned_le64(p));
0420             p += 8;
0421             v2 = xxh64_round(v2, get_unaligned_le64(p));
0422             p += 8;
0423             v3 = xxh64_round(v3, get_unaligned_le64(p));
0424             p += 8;
0425             v4 = xxh64_round(v4, get_unaligned_le64(p));
0426             p += 8;
0427         } while (p <= limit);
0428
0429         state->v1 = v1;
0430         state->v2 = v2;
0431         state->v3 = v3;
0432         state->v4 = v4;
0433     }
0434
0435     if (p < b_end) {
0436         memcpy(state->mem64, p, (size_t)(b_end-p));
0437         state->memsize = (uint32_t)(b_end - p);
0438     }
0439
0440     return 0;
0441 }
0442 EXPORT_SYMBOL(xxh64_update);
0443
0444 uint64_t xxh64_digest(const struct xxh64_state *state)
0445 {
0446     const uint8_t *p = (const uint8_t *)state->mem64;
0447     const uint8_t *const b_end = (const uint8_t *)state->mem64 +
0448         state->memsize;
0449     uint64_t h64;
0450
0451     if (state->total_len >= 32) {
0452         const uint64_t v1 = state->v1;
0453         const uint64_t v2 = state->v2;
0454         const uint64_t v3 = state->v3;
0455         const uint64_t v4 = state->v4;
0456
0457         h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
0458             xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
0459         h64 = xxh64_merge_round(h64, v1);
0460         h64 = xxh64_merge_round(h64, v2);
0461         h64 = xxh64_merge_round(h64, v3);
0462         h64 = xxh64_merge_round(h64, v4);
0463     } else {
0464         h64  = state->v3 + PRIME64_5;
0465     }
0466
0467     h64 += (uint64_t)state->total_len;
0468
0469     while (p + 8 <= b_end) {
0470         const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
0471
0472         h64 ^= k1;
0473         h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
0474         p += 8;
0475     }
0476
0477     if (p + 4 <= b_end) {
0478         h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
0479         h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
0480         p += 4;
0481     }
0482
0483     while (p < b_end) {
0484         h64 ^= (*p) * PRIME64_5;
0485         h64 = xxh_rotl64(h64, 11) * PRIME64_1;
0486         p++;
0487     }
0488
0489     h64 ^= h64 >> 33;
0490     h64 *= PRIME64_2;
0491     h64 ^= h64 >> 29;
0492     h64 *= PRIME64_3;
0493     h64 ^= h64 >> 32;
0494
0495     return h64;
0496 }
0497 EXPORT_SYMBOL(xxh64_digest);
0498
0499 MODULE_LICENSE("Dual BSD/GPL");
0500 MODULE_DESCRIPTION("xxHash");