Update ABD stats for linear page Linux
[zfs.git] / module / icp / algs / blake3 / blake3_generic.c
blobfbe184969672d2ec36c4eb7ebe2b9c84f16514e1
1 /*
2 * CDDL HEADER START
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or https://opensource.org/licenses/CDDL-1.0.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
19 * CDDL HEADER END
23 * Based on BLAKE3 v1.3.1, https://github.com/BLAKE3-team/BLAKE3
24 * Copyright (c) 2019-2020 Samuel Neves and Jack O'Connor
25 * Copyright (c) 2021-2022 Tino Reichardt <milky-zfs@mcmilk.de>
28 #include <sys/simd.h>
29 #include <sys/zfs_context.h>
30 #include "blake3_impl.h"
32 #define rotr32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
33 static inline void g(uint32_t *state, size_t a, size_t b, size_t c, size_t d,
34 uint32_t x, uint32_t y)
36 state[a] = state[a] + state[b] + x;
37 state[d] = rotr32(state[d] ^ state[a], 16);
38 state[c] = state[c] + state[d];
39 state[b] = rotr32(state[b] ^ state[c], 12);
40 state[a] = state[a] + state[b] + y;
41 state[d] = rotr32(state[d] ^ state[a], 8);
42 state[c] = state[c] + state[d];
43 state[b] = rotr32(state[b] ^ state[c], 7);
46 static inline void round_fn(uint32_t state[16], const uint32_t *msg,
47 size_t round)
49 /* Select the message schedule based on the round. */
50 const uint8_t *schedule = BLAKE3_MSG_SCHEDULE[round];
52 /* Mix the columns. */
53 g(state, 0, 4, 8, 12, msg[schedule[0]], msg[schedule[1]]);
54 g(state, 1, 5, 9, 13, msg[schedule[2]], msg[schedule[3]]);
55 g(state, 2, 6, 10, 14, msg[schedule[4]], msg[schedule[5]]);
56 g(state, 3, 7, 11, 15, msg[schedule[6]], msg[schedule[7]]);
58 /* Mix the rows. */
59 g(state, 0, 5, 10, 15, msg[schedule[8]], msg[schedule[9]]);
60 g(state, 1, 6, 11, 12, msg[schedule[10]], msg[schedule[11]]);
61 g(state, 2, 7, 8, 13, msg[schedule[12]], msg[schedule[13]]);
62 g(state, 3, 4, 9, 14, msg[schedule[14]], msg[schedule[15]]);
65 static inline void compress_pre(uint32_t state[16], const uint32_t cv[8],
66 const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
67 uint64_t counter, uint8_t flags)
69 uint32_t block_words[16];
70 block_words[0] = load32(block + 4 * 0);
71 block_words[1] = load32(block + 4 * 1);
72 block_words[2] = load32(block + 4 * 2);
73 block_words[3] = load32(block + 4 * 3);
74 block_words[4] = load32(block + 4 * 4);
75 block_words[5] = load32(block + 4 * 5);
76 block_words[6] = load32(block + 4 * 6);
77 block_words[7] = load32(block + 4 * 7);
78 block_words[8] = load32(block + 4 * 8);
79 block_words[9] = load32(block + 4 * 9);
80 block_words[10] = load32(block + 4 * 10);
81 block_words[11] = load32(block + 4 * 11);
82 block_words[12] = load32(block + 4 * 12);
83 block_words[13] = load32(block + 4 * 13);
84 block_words[14] = load32(block + 4 * 14);
85 block_words[15] = load32(block + 4 * 15);
87 state[0] = cv[0];
88 state[1] = cv[1];
89 state[2] = cv[2];
90 state[3] = cv[3];
91 state[4] = cv[4];
92 state[5] = cv[5];
93 state[6] = cv[6];
94 state[7] = cv[7];
95 state[8] = BLAKE3_IV[0];
96 state[9] = BLAKE3_IV[1];
97 state[10] = BLAKE3_IV[2];
98 state[11] = BLAKE3_IV[3];
99 state[12] = counter_low(counter);
100 state[13] = counter_high(counter);
101 state[14] = (uint32_t)block_len;
102 state[15] = (uint32_t)flags;
104 round_fn(state, &block_words[0], 0);
105 round_fn(state, &block_words[0], 1);
106 round_fn(state, &block_words[0], 2);
107 round_fn(state, &block_words[0], 3);
108 round_fn(state, &block_words[0], 4);
109 round_fn(state, &block_words[0], 5);
110 round_fn(state, &block_words[0], 6);
113 static inline void blake3_compress_in_place_generic(uint32_t cv[8],
114 const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
115 uint64_t counter, uint8_t flags)
117 uint32_t state[16];
118 compress_pre(state, cv, block, block_len, counter, flags);
119 cv[0] = state[0] ^ state[8];
120 cv[1] = state[1] ^ state[9];
121 cv[2] = state[2] ^ state[10];
122 cv[3] = state[3] ^ state[11];
123 cv[4] = state[4] ^ state[12];
124 cv[5] = state[5] ^ state[13];
125 cv[6] = state[6] ^ state[14];
126 cv[7] = state[7] ^ state[15];
129 static inline void hash_one_generic(const uint8_t *input, size_t blocks,
130 const uint32_t key[8], uint64_t counter, uint8_t flags,
131 uint8_t flags_start, uint8_t flags_end, uint8_t out[BLAKE3_OUT_LEN])
133 uint32_t cv[8];
134 memcpy(cv, key, BLAKE3_KEY_LEN);
135 uint8_t block_flags = flags | flags_start;
136 while (blocks > 0) {
137 if (blocks == 1) {
138 block_flags |= flags_end;
140 blake3_compress_in_place_generic(cv, input, BLAKE3_BLOCK_LEN,
141 counter, block_flags);
142 input = &input[BLAKE3_BLOCK_LEN];
143 blocks -= 1;
144 block_flags = flags;
146 store_cv_words(out, cv);
149 static inline void blake3_compress_xof_generic(const uint32_t cv[8],
150 const uint8_t block[BLAKE3_BLOCK_LEN], uint8_t block_len,
151 uint64_t counter, uint8_t flags, uint8_t out[64])
153 uint32_t state[16];
154 compress_pre(state, cv, block, block_len, counter, flags);
156 store32(&out[0 * 4], state[0] ^ state[8]);
157 store32(&out[1 * 4], state[1] ^ state[9]);
158 store32(&out[2 * 4], state[2] ^ state[10]);
159 store32(&out[3 * 4], state[3] ^ state[11]);
160 store32(&out[4 * 4], state[4] ^ state[12]);
161 store32(&out[5 * 4], state[5] ^ state[13]);
162 store32(&out[6 * 4], state[6] ^ state[14]);
163 store32(&out[7 * 4], state[7] ^ state[15]);
164 store32(&out[8 * 4], state[8] ^ cv[0]);
165 store32(&out[9 * 4], state[9] ^ cv[1]);
166 store32(&out[10 * 4], state[10] ^ cv[2]);
167 store32(&out[11 * 4], state[11] ^ cv[3]);
168 store32(&out[12 * 4], state[12] ^ cv[4]);
169 store32(&out[13 * 4], state[13] ^ cv[5]);
170 store32(&out[14 * 4], state[14] ^ cv[6]);
171 store32(&out[15 * 4], state[15] ^ cv[7]);
174 static inline void blake3_hash_many_generic(const uint8_t * const *inputs,
175 size_t num_inputs, size_t blocks, const uint32_t key[8], uint64_t counter,
176 boolean_t increment_counter, uint8_t flags, uint8_t flags_start,
177 uint8_t flags_end, uint8_t *out)
179 while (num_inputs > 0) {
180 hash_one_generic(inputs[0], blocks, key, counter, flags,
181 flags_start, flags_end, out);
182 if (increment_counter) {
183 counter += 1;
185 inputs += 1;
186 num_inputs -= 1;
187 out = &out[BLAKE3_OUT_LEN];
191 /* the generic implementation is always okay */
192 static boolean_t blake3_is_supported(void)
194 return (B_TRUE);
197 const blake3_ops_t blake3_generic_impl = {
198 .compress_in_place = blake3_compress_in_place_generic,
199 .compress_xof = blake3_compress_xof_generic,
200 .hash_many = blake3_hash_many_generic,
201 .is_supported = blake3_is_supported,
202 .degree = 4,
203 .name = "generic"