/src/openssl/crypto/blake2/blake2b.c
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * Copyright 2016-2017 The OpenSSL Project Authors. All Rights Reserved. |
3 | | * |
4 | | * Licensed under the OpenSSL license (the "License"). You may not use |
5 | | * this file except in compliance with the License. You can obtain a copy |
6 | | * in the file LICENSE in the source distribution or at |
7 | | * https://www.openssl.org/source/license.html |
8 | | */ |
9 | | |
10 | | /* |
11 | | * Derived from the BLAKE2 reference implementation written by Samuel Neves. |
12 | | * Copyright 2012, Samuel Neves <sneves@dei.uc.pt> |
13 | | * More information about the BLAKE2 hash function and its implementations |
14 | | * can be found at https://blake2.net. |
15 | | */ |
16 | | |
17 | | #include <assert.h> |
18 | | #include <string.h> |
19 | | #include <openssl/crypto.h> |
20 | | |
21 | | #include "blake2_locl.h" |
22 | | #include "blake2_impl.h" |
23 | | |
24 | | static const uint64_t blake2b_IV[8] = |
25 | | { |
26 | | 0x6a09e667f3bcc908U, 0xbb67ae8584caa73bU, |
27 | | 0x3c6ef372fe94f82bU, 0xa54ff53a5f1d36f1U, |
28 | | 0x510e527fade682d1U, 0x9b05688c2b3e6c1fU, |
29 | | 0x1f83d9abfb41bd6bU, 0x5be0cd19137e2179U |
30 | | }; |
31 | | |
32 | | static const uint8_t blake2b_sigma[12][16] = |
33 | | { |
34 | | { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 } , |
35 | | { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 } , |
36 | | { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 } , |
37 | | { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 } , |
38 | | { 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 } , |
39 | | { 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 } , |
40 | | { 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 } , |
41 | | { 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 } , |
42 | | { 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 } , |
43 | | { 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13 , 0 } , |
44 | | { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 } , |
45 | | { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 } |
46 | | }; |
47 | | |
48 | | /* Set that it's the last block we'll compress */ |
49 | | static ossl_inline void blake2b_set_lastblock(BLAKE2B_CTX *S) |
50 | 0 | { |
51 | 0 | S->f[0] = -1; |
52 | 0 | } |
53 | | |
54 | | /* Initialize the hashing state. */ |
55 | | static ossl_inline void blake2b_init0(BLAKE2B_CTX *S) |
56 | 0 | { |
57 | 0 | int i; |
58 | 0 |
|
59 | 0 | memset(S, 0, sizeof(BLAKE2B_CTX)); |
60 | 0 | for (i = 0; i < 8; ++i) { |
61 | 0 | S->h[i] = blake2b_IV[i]; |
62 | 0 | } |
63 | 0 | } |
64 | | |
65 | | /* init xors IV with input parameter block */ |
66 | | static void blake2b_init_param(BLAKE2B_CTX *S, const BLAKE2B_PARAM *P) |
67 | 0 | { |
68 | 0 | size_t i; |
69 | 0 | const uint8_t *p = (const uint8_t *)(P); |
70 | 0 | blake2b_init0(S); |
71 | 0 |
|
72 | 0 | /* The param struct is carefully hand packed, and should be 64 bytes on |
73 | 0 | * every platform. */ |
74 | 0 | assert(sizeof(BLAKE2B_PARAM) == 64); |
75 | 0 | /* IV XOR ParamBlock */ |
76 | 0 | for (i = 0; i < 8; ++i) { |
77 | 0 | S->h[i] ^= load64(p + sizeof(S->h[i]) * i); |
78 | 0 | } |
79 | 0 | } |
80 | | |
81 | | /* Initialize the hashing context. Always returns 1. */ |
82 | | int BLAKE2b_Init(BLAKE2B_CTX *c) |
83 | 0 | { |
84 | 0 | BLAKE2B_PARAM P[1]; |
85 | 0 | P->digest_length = BLAKE2B_DIGEST_LENGTH; |
86 | 0 | P->key_length = 0; |
87 | 0 | P->fanout = 1; |
88 | 0 | P->depth = 1; |
89 | 0 | store32(P->leaf_length, 0); |
90 | 0 | store64(P->node_offset, 0); |
91 | 0 | P->node_depth = 0; |
92 | 0 | P->inner_length = 0; |
93 | 0 | memset(P->reserved, 0, sizeof(P->reserved)); |
94 | 0 | memset(P->salt, 0, sizeof(P->salt)); |
95 | 0 | memset(P->personal, 0, sizeof(P->personal)); |
96 | 0 | blake2b_init_param(c, P); |
97 | 0 | return 1; |
98 | 0 | } |
99 | | |
100 | | /* Permute the state while xoring in the block of data. */ |
101 | | static void blake2b_compress(BLAKE2B_CTX *S, |
102 | | const uint8_t *blocks, |
103 | | size_t len) |
104 | 0 | { |
105 | 0 | uint64_t m[16]; |
106 | 0 | uint64_t v[16]; |
107 | 0 | int i; |
108 | 0 | size_t increment; |
109 | 0 |
|
110 | 0 | /* |
111 | 0 | * There are two distinct usage vectors for this function: |
112 | 0 | * |
113 | 0 | * a) BLAKE2b_Update uses it to process complete blocks, |
114 | 0 | * possibly more than one at a time; |
115 | 0 | * |
116 | 0 | * b) BLAK2b_Final uses it to process last block, always |
117 | 0 | * single but possibly incomplete, in which case caller |
118 | 0 | * pads input with zeros. |
119 | 0 | */ |
120 | 0 | assert(len < BLAKE2B_BLOCKBYTES || len % BLAKE2B_BLOCKBYTES == 0); |
121 | 0 |
|
122 | 0 | /* |
123 | 0 | * Since last block is always processed with separate call, |
124 | 0 | * |len| not being multiple of complete blocks can be observed |
125 | 0 | * only with |len| being less than BLAKE2B_BLOCKBYTES ("less" |
126 | 0 | * including even zero), which is why following assignment doesn't |
127 | 0 | * have to reside inside the main loop below. |
128 | 0 | */ |
129 | 0 | increment = len < BLAKE2B_BLOCKBYTES ? len : BLAKE2B_BLOCKBYTES; |
130 | 0 |
|
131 | 0 | for (i = 0; i < 8; ++i) { |
132 | 0 | v[i] = S->h[i]; |
133 | 0 | } |
134 | 0 |
|
135 | 0 | do { |
136 | 0 | for (i = 0; i < 16; ++i) { |
137 | 0 | m[i] = load64(blocks + i * sizeof(m[i])); |
138 | 0 | } |
139 | 0 |
|
140 | 0 | /* blake2b_increment_counter */ |
141 | 0 | S->t[0] += increment; |
142 | 0 | S->t[1] += (S->t[0] < increment); |
143 | 0 |
|
144 | 0 | v[8] = blake2b_IV[0]; |
145 | 0 | v[9] = blake2b_IV[1]; |
146 | 0 | v[10] = blake2b_IV[2]; |
147 | 0 | v[11] = blake2b_IV[3]; |
148 | 0 | v[12] = S->t[0] ^ blake2b_IV[4]; |
149 | 0 | v[13] = S->t[1] ^ blake2b_IV[5]; |
150 | 0 | v[14] = S->f[0] ^ blake2b_IV[6]; |
151 | 0 | v[15] = S->f[1] ^ blake2b_IV[7]; |
152 | 0 | #define G(r,i,a,b,c,d) \ |
153 | 0 | do { \ |
154 | 0 | a = a + b + m[blake2b_sigma[r][2*i+0]]; \ |
155 | 0 | d = rotr64(d ^ a, 32); \ |
156 | 0 | c = c + d; \ |
157 | 0 | b = rotr64(b ^ c, 24); \ |
158 | 0 | a = a + b + m[blake2b_sigma[r][2*i+1]]; \ |
159 | 0 | d = rotr64(d ^ a, 16); \ |
160 | 0 | c = c + d; \ |
161 | 0 | b = rotr64(b ^ c, 63); \ |
162 | 0 | } while (0) |
163 | 0 | #define ROUND(r) \ |
164 | 0 | do { \ |
165 | 0 | G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \ |
166 | 0 | G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \ |
167 | 0 | G(r,2,v[ 2],v[ 6],v[10],v[14]); \ |
168 | 0 | G(r,3,v[ 3],v[ 7],v[11],v[15]); \ |
169 | 0 | G(r,4,v[ 0],v[ 5],v[10],v[15]); \ |
170 | 0 | G(r,5,v[ 1],v[ 6],v[11],v[12]); \ |
171 | 0 | G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \ |
172 | 0 | G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \ |
173 | 0 | } while (0) |
174 | | #if defined(OPENSSL_SMALL_FOOTPRINT) |
175 | | /* 3x size reduction on x86_64, almost 7x on ARMv8, 9x on ARMv4 */ |
176 | | for (i = 0; i < 12; i++) { |
177 | | ROUND(i); |
178 | | } |
179 | | #else |
180 | 0 | ROUND(0); |
181 | 0 | ROUND(1); |
182 | 0 | ROUND(2); |
183 | 0 | ROUND(3); |
184 | 0 | ROUND(4); |
185 | 0 | ROUND(5); |
186 | 0 | ROUND(6); |
187 | 0 | ROUND(7); |
188 | 0 | ROUND(8); |
189 | 0 | ROUND(9); |
190 | 0 | ROUND(10); |
191 | 0 | ROUND(11); |
192 | 0 | #endif |
193 | 0 |
|
194 | 0 | for (i = 0; i < 8; ++i) { |
195 | 0 | S->h[i] = v[i] ^= v[i + 8] ^ S->h[i]; |
196 | 0 | } |
197 | 0 | #undef G |
198 | 0 | #undef ROUND |
199 | 0 | blocks += increment; |
200 | 0 | len -= increment; |
201 | 0 | } while (len); |
202 | 0 | } |
203 | | |
204 | | /* Absorb the input data into the hash state. Always returns 1. */ |
205 | | int BLAKE2b_Update(BLAKE2B_CTX *c, const void *data, size_t datalen) |
206 | 0 | { |
207 | 0 | const uint8_t *in = data; |
208 | 0 | size_t fill; |
209 | 0 |
|
210 | 0 | /* |
211 | 0 | * Intuitively one would expect intermediate buffer, c->buf, to |
212 | 0 | * store incomplete blocks. But in this case we are interested to |
213 | 0 | * temporarily stash even complete blocks, because last one in the |
214 | 0 | * stream has to be treated in special way, and at this point we |
215 | 0 | * don't know if last block in *this* call is last one "ever". This |
216 | 0 | * is the reason for why |datalen| is compared as >, and not >=. |
217 | 0 | */ |
218 | 0 | fill = sizeof(c->buf) - c->buflen; |
219 | 0 | if (datalen > fill) { |
220 | 0 | if (c->buflen) { |
221 | 0 | memcpy(c->buf + c->buflen, in, fill); /* Fill buffer */ |
222 | 0 | blake2b_compress(c, c->buf, BLAKE2B_BLOCKBYTES); |
223 | 0 | c->buflen = 0; |
224 | 0 | in += fill; |
225 | 0 | datalen -= fill; |
226 | 0 | } |
227 | 0 | if (datalen > BLAKE2B_BLOCKBYTES) { |
228 | 0 | size_t stashlen = datalen % BLAKE2B_BLOCKBYTES; |
229 | 0 | /* |
230 | 0 | * If |datalen| is a multiple of the blocksize, stash |
231 | 0 | * last complete block, it can be final one... |
232 | 0 | */ |
233 | 0 | stashlen = stashlen ? stashlen : BLAKE2B_BLOCKBYTES; |
234 | 0 | datalen -= stashlen; |
235 | 0 | blake2b_compress(c, in, datalen); |
236 | 0 | in += datalen; |
237 | 0 | datalen = stashlen; |
238 | 0 | } |
239 | 0 | } |
240 | 0 |
|
241 | 0 | assert(datalen <= BLAKE2B_BLOCKBYTES); |
242 | 0 |
|
243 | 0 | memcpy(c->buf + c->buflen, in, datalen); |
244 | 0 | c->buflen += datalen; /* Be lazy, do not compress */ |
245 | 0 |
|
246 | 0 | return 1; |
247 | 0 | } |
248 | | |
249 | | /* |
250 | | * Calculate the final hash and save it in md. |
251 | | * Always returns 1. |
252 | | */ |
253 | | int BLAKE2b_Final(unsigned char *md, BLAKE2B_CTX *c) |
254 | 0 | { |
255 | 0 | int i; |
256 | 0 |
|
257 | 0 | blake2b_set_lastblock(c); |
258 | 0 | /* Padding */ |
259 | 0 | memset(c->buf + c->buflen, 0, sizeof(c->buf) - c->buflen); |
260 | 0 | blake2b_compress(c, c->buf, c->buflen); |
261 | 0 |
|
262 | 0 | /* Output full hash to message digest */ |
263 | 0 | for (i = 0; i < 8; ++i) { |
264 | 0 | store64(md + sizeof(c->h[i]) * i, c->h[i]); |
265 | 0 | } |
266 | 0 |
|
267 | 0 | OPENSSL_cleanse(c, sizeof(BLAKE2B_CTX)); |
268 | 0 | return 1; |
269 | 0 | } |