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 /*
42  * Notice extracted from xxHash homepage:
43  *
44  * xxHash is an extremely fast Hash algorithm, running at RAM speed limits.
45  * It also successfully passes all tests from the SMHasher suite.
46  *
47  * Comparison (single thread, Windows Seven 32 bits, using SMHasher on a Core 2
48  * Duo @3GHz)
49  *
50  * Name            Speed       Q.Score   Author
51  * xxHash          5.4 GB/s     10
52  * CrapWow         3.2 GB/s      2       Andrew
53  * MumurHash 3a    2.7 GB/s     10       Austin Appleby
54  * SpookyHash      2.0 GB/s     10       Bob Jenkins
55  * SBox            1.4 GB/s      9       Bret Mulvey
56  * Lookup3         1.2 GB/s      9       Bob Jenkins
57  * SuperFastHash   1.2 GB/s      1       Paul Hsieh
58  * CityHash64      1.05 GB/s    10       Pike & Alakuijala
59  * FNV             0.55 GB/s     5       Fowler, Noll, Vo
60  * CRC32           0.43 GB/s     9
61  * MD5-32          0.33 GB/s    10       Ronald L. Rivest
62  * SHA1-32         0.28 GB/s    10
63  *
64  * Q.Score is a measure of quality of the hash function.
65  * It depends on successfully passing SMHasher test set.
66  * 10 is a perfect score.
67  *
68  * A 64-bits version, named xxh64 offers much better speed,
69  * but for 64-bits applications only.
70  * Name     Speed on 64 bits    Speed on 32 bits
71  * xxh64       13.8 GB/s            1.9 GB/s
72  * xxh32        6.8 GB/s            6.0 GB/s
73  */
74 
75 #ifndef XXHASH_H
76 #define XXHASH_H
77 
78 #include <linux/types.h>
79 
80 #define XXH_API static inline __attribute__((unused))
81 /*-****************************
82  * Simple Hash Functions
83  *****************************/
84 
85 /**
86  * xxh32() - calculate the 32-bit hash of the input with a given seed.
87  *
88  * @input:  The data to hash.
89  * @length: The length of the data to hash.
90  * @seed:   The seed can be used to alter the result predictably.
91  *
92  * Speed on Core 2 Duo @ 3 GHz (single thread, SMHasher benchmark) : 5.4 GB/s
93  *
94  * Return:  The 32-bit hash of the data.
95  */
96 XXH_API uint32_t xxh32(const void *input, size_t length, uint32_t seed);
97 
98 /**
99  * xxh64() - calculate the 64-bit hash of the input with a given seed.
100  *
101  * @input:  The data to hash.
102  * @length: The length of the data to hash.
103  * @seed:   The seed can be used to alter the result predictably.
104  *
105  * This function runs 2x faster on 64-bit systems, but slower on 32-bit systems.
106  *
107  * Return:  The 64-bit hash of the data.
108  */
109 XXH_API uint64_t xxh64(const void *input, size_t length, uint64_t seed);
110 
111 /**
112  * xxhash() - calculate wordsize hash of the input with a given seed
113  * @input:  The data to hash.
114  * @length: The length of the data to hash.
115  * @seed:   The seed can be used to alter the result predictably.
116  *
117  * If the hash does not need to be comparable between machines with
118  * different word sizes, this function will call whichever of xxh32()
119  * or xxh64() is faster.
120  *
121  * Return:  wordsize hash of the data.
122  */
123 
xxhash(const void * input,size_t length,uint64_t seed)124 static inline unsigned long xxhash(const void *input, size_t length,
125 				   uint64_t seed)
126 {
127 #if BITS_PER_LONG == 64
128        return xxh64(input, length, seed);
129 #else
130        return xxh32(input, length, seed);
131 #endif
132 }
133 
134 /*-****************************
135  * Streaming Hash Functions
136  *****************************/
137 
138 /*
139  * These definitions are only meant to allow allocation of XXH state
140  * statically, on stack, or in a struct for example.
141  * Do not use members directly.
142  */
143 
144 /**
145  * struct xxh32_state - private xxh32 state, do not use members directly
146  */
147 struct xxh32_state {
148 	uint32_t total_len_32;
149 	uint32_t large_len;
150 	uint32_t v1;
151 	uint32_t v2;
152 	uint32_t v3;
153 	uint32_t v4;
154 	uint32_t mem32[4];
155 	uint32_t memsize;
156 };
157 
158 /**
159  * struct xxh32_state - private xxh64 state, do not use members directly
160  */
161 struct xxh64_state {
162 	uint64_t total_len;
163 	uint64_t v1;
164 	uint64_t v2;
165 	uint64_t v3;
166 	uint64_t v4;
167 	uint64_t mem64[4];
168 	uint32_t memsize;
169 };
170 
171 /**
172  * xxh32_reset() - reset the xxh32 state to start a new hashing operation
173  *
174  * @state: The xxh32 state to reset.
175  * @seed:  Initialize the hash state with this seed.
176  *
177  * Call this function on any xxh32_state to prepare for a new hashing operation.
178  */
179 XXH_API void xxh32_reset(struct xxh32_state *state, uint32_t seed);
180 
181 /**
182  * xxh32_update() - hash the data given and update the xxh32 state
183  *
184  * @state:  The xxh32 state to update.
185  * @input:  The data to hash.
186  * @length: The length of the data to hash.
187  *
188  * After calling xxh32_reset() call xxh32_update() as many times as necessary.
189  *
190  * Return:  Zero on success, otherwise an error code.
191  */
192 XXH_API int xxh32_update(struct xxh32_state *state, const void *input, size_t length);
193 
194 /**
195  * xxh32_digest() - produce the current xxh32 hash
196  *
197  * @state: Produce the current xxh32 hash of this state.
198  *
199  * A hash value can be produced at any time. It is still possible to continue
200  * inserting input into the hash state after a call to xxh32_digest(), and
201  * generate new hashes later on, by calling xxh32_digest() again.
202  *
203  * Return: The xxh32 hash stored in the state.
204  */
205 XXH_API uint32_t xxh32_digest(const struct xxh32_state *state);
206 
207 /**
208  * xxh64_reset() - reset the xxh64 state to start a new hashing operation
209  *
210  * @state: The xxh64 state to reset.
211  * @seed:  Initialize the hash state with this seed.
212  */
213 XXH_API void xxh64_reset(struct xxh64_state *state, uint64_t seed);
214 
215 /**
216  * xxh64_update() - hash the data given and update the xxh64 state
217  * @state:  The xxh64 state to update.
218  * @input:  The data to hash.
219  * @length: The length of the data to hash.
220  *
221  * After calling xxh64_reset() call xxh64_update() as many times as necessary.
222  *
223  * Return:  Zero on success, otherwise an error code.
224  */
225 XXH_API int xxh64_update(struct xxh64_state *state, const void *input, size_t length);
226 
227 /**
228  * xxh64_digest() - produce the current xxh64 hash
229  *
230  * @state: Produce the current xxh64 hash of this state.
231  *
232  * A hash value can be produced at any time. It is still possible to continue
233  * inserting input into the hash state after a call to xxh64_digest(), and
234  * generate new hashes later on, by calling xxh64_digest() again.
235  *
236  * Return: The xxh64 hash stored in the state.
237  */
238 XXH_API uint64_t xxh64_digest(const struct xxh64_state *state);
239 
240 /*-**************************
241  * Utils
242  ***************************/
243 
244 /**
245  * xxh32_copy_state() - copy the source state into the destination state
246  *
247  * @src: The source xxh32 state.
248  * @dst: The destination xxh32 state.
249  */
250 XXH_API void xxh32_copy_state(struct xxh32_state *dst, const struct xxh32_state *src);
251 
252 /**
253  * xxh64_copy_state() - copy the source state into the destination state
254  *
255  * @src: The source xxh64 state.
256  * @dst: The destination xxh64 state.
257  */
258 XXH_API void xxh64_copy_state(struct xxh64_state *dst, const struct xxh64_state *src);
259 
260 /*
261  * xxHash - Extremely Fast Hash algorithm
262  * Copyright (C) 2012-2016, Yann Collet.
263  *
264  * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
265  *
266  * Redistribution and use in source and binary forms, with or without
267  * modification, are permitted provided that the following conditions are
268  * met:
269  *
270  *   * Redistributions of source code must retain the above copyright
271  *     notice, this list of conditions and the following disclaimer.
272  *   * Redistributions in binary form must reproduce the above
273  *     copyright notice, this list of conditions and the following disclaimer
274  *     in the documentation and/or other materials provided with the
275  *     distribution.
276  *
277  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
278  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
279  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
280  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
281  * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
282  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
283  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
284  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
285  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
286  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
287  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
288  *
289  * This program is free software; you can redistribute it and/or modify it under
290  * the terms of the GNU General Public License version 2 as published by the
291  * Free Software Foundation. This program is dual-licensed; you may select
292  * either version 2 of the GNU General Public License ("GPL") or BSD license
293  * ("BSD").
294  *
295  * You can contact the author at:
296  * - xxHash homepage: https://cyan4973.github.io/xxHash/
297  * - xxHash source repository: https://github.com/Cyan4973/xxHash
298  */
299 
300 #include <asm/unaligned.h>
301 #include <linux/errno.h>
302 #include <linux/kernel.h>
303 #include <linux/module.h>
304 #include <linux/xxhash.h>
305 
306 /*-*************************************
307  * Macros
308  **************************************/
309 #define xxh_rotl32(x, r) ((x << r) | (x >> (32 - r)))
310 #define xxh_rotl64(x, r) ((x << r) | (x >> (64 - r)))
311 
312 #ifdef __LITTLE_ENDIAN
313 # define XXH_CPU_LITTLE_ENDIAN 1
314 #else
315 # define XXH_CPU_LITTLE_ENDIAN 0
316 #endif
317 
318 /*-*************************************
319  * Constants
320  **************************************/
321 static const uint32_t PRIME32_1 = 2654435761U;
322 static const uint32_t PRIME32_2 = 2246822519U;
323 static const uint32_t PRIME32_3 = 3266489917U;
324 static const uint32_t PRIME32_4 =  668265263U;
325 static const uint32_t PRIME32_5 =  374761393U;
326 
327 static const uint64_t PRIME64_1 = 11400714785074694791ULL;
328 static const uint64_t PRIME64_2 = 14029467366897019727ULL;
329 static const uint64_t PRIME64_3 =  1609587929392839161ULL;
330 static const uint64_t PRIME64_4 =  9650029242287828579ULL;
331 static const uint64_t PRIME64_5 =  2870177450012600261ULL;
332 
333 /*-**************************
334  *  Utils
335  ***************************/
xxh32_copy_state(struct xxh32_state * dst,const struct xxh32_state * src)336 XXH_API void xxh32_copy_state(struct xxh32_state *dst, const struct xxh32_state *src)
337 {
338 	__builtin_memcpy(dst, src, sizeof(*dst));
339 }
340 
xxh64_copy_state(struct xxh64_state * dst,const struct xxh64_state * src)341 XXH_API void xxh64_copy_state(struct xxh64_state *dst, const struct xxh64_state *src)
342 {
343 	__builtin_memcpy(dst, src, sizeof(*dst));
344 }
345 
346 /*-***************************
347  * Simple Hash Functions
348  ****************************/
xxh32_round(uint32_t seed,const uint32_t input)349 static uint32_t xxh32_round(uint32_t seed, const uint32_t input)
350 {
351 	seed += input * PRIME32_2;
352 	seed = xxh_rotl32(seed, 13);
353 	seed *= PRIME32_1;
354 	return seed;
355 }
356 
xxh32(const void * input,const size_t len,const uint32_t seed)357 XXH_API uint32_t xxh32(const void *input, const size_t len, const uint32_t seed)
358 {
359 	const uint8_t *p = (const uint8_t *)input;
360 	const uint8_t *b_end = p + len;
361 	uint32_t h32;
362 
363 	if (len >= 16) {
364 		const uint8_t *const limit = b_end - 16;
365 		uint32_t v1 = seed + PRIME32_1 + PRIME32_2;
366 		uint32_t v2 = seed + PRIME32_2;
367 		uint32_t v3 = seed + 0;
368 		uint32_t v4 = seed - PRIME32_1;
369 
370 		do {
371 			v1 = xxh32_round(v1, get_unaligned_le32(p));
372 			p += 4;
373 			v2 = xxh32_round(v2, get_unaligned_le32(p));
374 			p += 4;
375 			v3 = xxh32_round(v3, get_unaligned_le32(p));
376 			p += 4;
377 			v4 = xxh32_round(v4, get_unaligned_le32(p));
378 			p += 4;
379 		} while (p <= limit);
380 
381 		h32 = xxh_rotl32(v1, 1) + xxh_rotl32(v2, 7) +
382 			xxh_rotl32(v3, 12) + xxh_rotl32(v4, 18);
383 	} else {
384 		h32 = seed + PRIME32_5;
385 	}
386 
387 	h32 += (uint32_t)len;
388 
389 	while (p + 4 <= b_end) {
390 		h32 += get_unaligned_le32(p) * PRIME32_3;
391 		h32 = xxh_rotl32(h32, 17) * PRIME32_4;
392 		p += 4;
393 	}
394 
395 	while (p < b_end) {
396 		h32 += (*p) * PRIME32_5;
397 		h32 = xxh_rotl32(h32, 11) * PRIME32_1;
398 		p++;
399 	}
400 
401 	h32 ^= h32 >> 15;
402 	h32 *= PRIME32_2;
403 	h32 ^= h32 >> 13;
404 	h32 *= PRIME32_3;
405 	h32 ^= h32 >> 16;
406 
407 	return h32;
408 }
409 
xxh64_round(uint64_t acc,const uint64_t input)410 static uint64_t xxh64_round(uint64_t acc, const uint64_t input)
411 {
412 	acc += input * PRIME64_2;
413 	acc = xxh_rotl64(acc, 31);
414 	acc *= PRIME64_1;
415 	return acc;
416 }
417 
xxh64_merge_round(uint64_t acc,uint64_t val)418 static uint64_t xxh64_merge_round(uint64_t acc, uint64_t val)
419 {
420 	val = xxh64_round(0, val);
421 	acc ^= val;
422 	acc = acc * PRIME64_1 + PRIME64_4;
423 	return acc;
424 }
425 
xxh64(const void * input,const size_t len,const uint64_t seed)426 XXH_API uint64_t xxh64(const void *input, const size_t len, const uint64_t seed)
427 {
428 	const uint8_t *p = (const uint8_t *)input;
429 	const uint8_t *const b_end = p + len;
430 	uint64_t h64;
431 
432 	if (len >= 32) {
433 		const uint8_t *const limit = b_end - 32;
434 		uint64_t v1 = seed + PRIME64_1 + PRIME64_2;
435 		uint64_t v2 = seed + PRIME64_2;
436 		uint64_t v3 = seed + 0;
437 		uint64_t v4 = seed - PRIME64_1;
438 
439 		do {
440 			v1 = xxh64_round(v1, get_unaligned_le64(p));
441 			p += 8;
442 			v2 = xxh64_round(v2, get_unaligned_le64(p));
443 			p += 8;
444 			v3 = xxh64_round(v3, get_unaligned_le64(p));
445 			p += 8;
446 			v4 = xxh64_round(v4, get_unaligned_le64(p));
447 			p += 8;
448 		} while (p <= limit);
449 
450 		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
451 			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
452 		h64 = xxh64_merge_round(h64, v1);
453 		h64 = xxh64_merge_round(h64, v2);
454 		h64 = xxh64_merge_round(h64, v3);
455 		h64 = xxh64_merge_round(h64, v4);
456 
457 	} else {
458 		h64  = seed + PRIME64_5;
459 	}
460 
461 	h64 += (uint64_t)len;
462 
463 	while (p + 8 <= b_end) {
464 		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
465 
466 		h64 ^= k1;
467 		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
468 		p += 8;
469 	}
470 
471 	if (p + 4 <= b_end) {
472 		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
473 		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
474 		p += 4;
475 	}
476 
477 	while (p < b_end) {
478 		h64 ^= (*p) * PRIME64_5;
479 		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
480 		p++;
481 	}
482 
483 	h64 ^= h64 >> 33;
484 	h64 *= PRIME64_2;
485 	h64 ^= h64 >> 29;
486 	h64 *= PRIME64_3;
487 	h64 ^= h64 >> 32;
488 
489 	return h64;
490 }
491 
492 /*-**************************************************
493  * Advanced Hash Functions
494  ***************************************************/
xxh32_reset(struct xxh32_state * statePtr,const uint32_t seed)495 XXH_API void xxh32_reset(struct xxh32_state *statePtr, const uint32_t seed)
496 {
497 	/* use a local state for memcpy() to avoid strict-aliasing warnings */
498 	struct xxh32_state state;
499 
500 	__builtin_memset(&state, 0, sizeof(state));
501 	state.v1 = seed + PRIME32_1 + PRIME32_2;
502 	state.v2 = seed + PRIME32_2;
503 	state.v3 = seed + 0;
504 	state.v4 = seed - PRIME32_1;
505 	__builtin_memcpy(statePtr, &state, sizeof(state));
506 }
507 
xxh64_reset(struct xxh64_state * statePtr,const uint64_t seed)508 XXH_API void xxh64_reset(struct xxh64_state *statePtr, const uint64_t seed)
509 {
510 	/* use a local state for memcpy() to avoid strict-aliasing warnings */
511 	struct xxh64_state state;
512 
513 	__builtin_memset(&state, 0, sizeof(state));
514 	state.v1 = seed + PRIME64_1 + PRIME64_2;
515 	state.v2 = seed + PRIME64_2;
516 	state.v3 = seed + 0;
517 	state.v4 = seed - PRIME64_1;
518 	__builtin_memcpy(statePtr, &state, sizeof(state));
519 }
520 
xxh32_update(struct xxh32_state * state,const void * input,const size_t len)521 XXH_API int xxh32_update(struct xxh32_state *state, const void *input, const size_t len)
522 {
523 	const uint8_t *p = (const uint8_t *)input;
524 	const uint8_t *const b_end = p + len;
525 
526 	if (input == NULL)
527 		return -EINVAL;
528 
529 	state->total_len_32 += (uint32_t)len;
530 	state->large_len |= (len >= 16) | (state->total_len_32 >= 16);
531 
532 	if (state->memsize + len < 16) { /* fill in tmp buffer */
533 		__builtin_memcpy((uint8_t *)(state->mem32) + state->memsize, input, len);
534 		state->memsize += (uint32_t)len;
535 		return 0;
536 	}
537 
538 	if (state->memsize) { /* some data left from previous update */
539 		const uint32_t *p32 = state->mem32;
540 
541 		__builtin_memcpy((uint8_t *)(state->mem32) + state->memsize, input,
542 			16 - state->memsize);
543 
544 		state->v1 = xxh32_round(state->v1, get_unaligned_le32(p32));
545 		p32++;
546 		state->v2 = xxh32_round(state->v2, get_unaligned_le32(p32));
547 		p32++;
548 		state->v3 = xxh32_round(state->v3, get_unaligned_le32(p32));
549 		p32++;
550 		state->v4 = xxh32_round(state->v4, get_unaligned_le32(p32));
551 		p32++;
552 
553 		p += 16-state->memsize;
554 		state->memsize = 0;
555 	}
556 
557 	if (p <= b_end - 16) {
558 		const uint8_t *const limit = b_end - 16;
559 		uint32_t v1 = state->v1;
560 		uint32_t v2 = state->v2;
561 		uint32_t v3 = state->v3;
562 		uint32_t v4 = state->v4;
563 
564 		do {
565 			v1 = xxh32_round(v1, get_unaligned_le32(p));
566 			p += 4;
567 			v2 = xxh32_round(v2, get_unaligned_le32(p));
568 			p += 4;
569 			v3 = xxh32_round(v3, get_unaligned_le32(p));
570 			p += 4;
571 			v4 = xxh32_round(v4, get_unaligned_le32(p));
572 			p += 4;
573 		} while (p <= limit);
574 
575 		state->v1 = v1;
576 		state->v2 = v2;
577 		state->v3 = v3;
578 		state->v4 = v4;
579 	}
580 
581 	if (p < b_end) {
582 		__builtin_memcpy(state->mem32, p, (size_t)(b_end-p));
583 		state->memsize = (uint32_t)(b_end-p);
584 	}
585 
586 	return 0;
587 }
588 
xxh32_digest(const struct xxh32_state * state)589 XXH_API uint32_t xxh32_digest(const struct xxh32_state *state)
590 {
591 	const uint8_t *p = (const uint8_t *)state->mem32;
592 	const uint8_t *const b_end = (const uint8_t *)(state->mem32) +
593 		state->memsize;
594 	uint32_t h32;
595 
596 	if (state->large_len) {
597 		h32 = xxh_rotl32(state->v1, 1) + xxh_rotl32(state->v2, 7) +
598 			xxh_rotl32(state->v3, 12) + xxh_rotl32(state->v4, 18);
599 	} else {
600 		h32 = state->v3 /* == seed */ + PRIME32_5;
601 	}
602 
603 	h32 += state->total_len_32;
604 
605 	while (p + 4 <= b_end) {
606 		h32 += get_unaligned_le32(p) * PRIME32_3;
607 		h32 = xxh_rotl32(h32, 17) * PRIME32_4;
608 		p += 4;
609 	}
610 
611 	while (p < b_end) {
612 		h32 += (*p) * PRIME32_5;
613 		h32 = xxh_rotl32(h32, 11) * PRIME32_1;
614 		p++;
615 	}
616 
617 	h32 ^= h32 >> 15;
618 	h32 *= PRIME32_2;
619 	h32 ^= h32 >> 13;
620 	h32 *= PRIME32_3;
621 	h32 ^= h32 >> 16;
622 
623 	return h32;
624 }
625 
xxh64_update(struct xxh64_state * state,const void * input,const size_t len)626 XXH_API int xxh64_update(struct xxh64_state *state, const void *input, const size_t len)
627 {
628 	const uint8_t *p = (const uint8_t *)input;
629 	const uint8_t *const b_end = p + len;
630 
631 	if (input == NULL)
632 		return -EINVAL;
633 
634 	state->total_len += len;
635 
636 	if (state->memsize + len < 32) { /* fill in tmp buffer */
637 		__builtin_memcpy(((uint8_t *)state->mem64) + state->memsize, input, len);
638 		state->memsize += (uint32_t)len;
639 		return 0;
640 	}
641 
642 	if (state->memsize) { /* tmp buffer is full */
643 		uint64_t *p64 = state->mem64;
644 
645 		__builtin_memcpy(((uint8_t *)p64) + state->memsize, input,
646 			32 - state->memsize);
647 
648 		state->v1 = xxh64_round(state->v1, get_unaligned_le64(p64));
649 		p64++;
650 		state->v2 = xxh64_round(state->v2, get_unaligned_le64(p64));
651 		p64++;
652 		state->v3 = xxh64_round(state->v3, get_unaligned_le64(p64));
653 		p64++;
654 		state->v4 = xxh64_round(state->v4, get_unaligned_le64(p64));
655 
656 		p += 32 - state->memsize;
657 		state->memsize = 0;
658 	}
659 
660 	if (p + 32 <= b_end) {
661 		const uint8_t *const limit = b_end - 32;
662 		uint64_t v1 = state->v1;
663 		uint64_t v2 = state->v2;
664 		uint64_t v3 = state->v3;
665 		uint64_t v4 = state->v4;
666 
667 		do {
668 			v1 = xxh64_round(v1, get_unaligned_le64(p));
669 			p += 8;
670 			v2 = xxh64_round(v2, get_unaligned_le64(p));
671 			p += 8;
672 			v3 = xxh64_round(v3, get_unaligned_le64(p));
673 			p += 8;
674 			v4 = xxh64_round(v4, get_unaligned_le64(p));
675 			p += 8;
676 		} while (p <= limit);
677 
678 		state->v1 = v1;
679 		state->v2 = v2;
680 		state->v3 = v3;
681 		state->v4 = v4;
682 	}
683 
684 	if (p < b_end) {
685 		__builtin_memcpy(state->mem64, p, (size_t)(b_end-p));
686 		state->memsize = (uint32_t)(b_end - p);
687 	}
688 
689 	return 0;
690 }
691 
xxh64_digest(const struct xxh64_state * state)692 XXH_API uint64_t xxh64_digest(const struct xxh64_state *state)
693 {
694 	const uint8_t *p = (const uint8_t *)state->mem64;
695 	const uint8_t *const b_end = (const uint8_t *)state->mem64 +
696 		state->memsize;
697 	uint64_t h64;
698 
699 	if (state->total_len >= 32) {
700 		const uint64_t v1 = state->v1;
701 		const uint64_t v2 = state->v2;
702 		const uint64_t v3 = state->v3;
703 		const uint64_t v4 = state->v4;
704 
705 		h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) +
706 			xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18);
707 		h64 = xxh64_merge_round(h64, v1);
708 		h64 = xxh64_merge_round(h64, v2);
709 		h64 = xxh64_merge_round(h64, v3);
710 		h64 = xxh64_merge_round(h64, v4);
711 	} else {
712 		h64  = state->v3 + PRIME64_5;
713 	}
714 
715 	h64 += (uint64_t)state->total_len;
716 
717 	while (p + 8 <= b_end) {
718 		const uint64_t k1 = xxh64_round(0, get_unaligned_le64(p));
719 
720 		h64 ^= k1;
721 		h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4;
722 		p += 8;
723 	}
724 
725 	if (p + 4 <= b_end) {
726 		h64 ^= (uint64_t)(get_unaligned_le32(p)) * PRIME64_1;
727 		h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
728 		p += 4;
729 	}
730 
731 	while (p < b_end) {
732 		h64 ^= (*p) * PRIME64_5;
733 		h64 = xxh_rotl64(h64, 11) * PRIME64_1;
734 		p++;
735 	}
736 
737 	h64 ^= h64 >> 33;
738 	h64 *= PRIME64_2;
739 	h64 ^= h64 >> 29;
740 	h64 *= PRIME64_3;
741 	h64 ^= h64 >> 32;
742 
743 	return h64;
744 }
745 
746 #endif /* XXHASH_H */
747