1 /* Copyright (c) 2014, Google Inc.
2 *
3 * Permission to use, copy, modify, and/or distribute this software for any
4 * purpose with or without fee is hereby granted, provided that the above
5 * copyright notice and this permission notice appear in all copies.
6 *
7 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
8 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
9 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
10 * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
11 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
12 * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
13 * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
14
15 /* Adapted from the public domain, estream code by D. Bernstein. */
16
17 #include <openssl/chacha.h>
18
19 #include <string.h>
20
21 #include <openssl/cpu.h>
22
23
24 #if defined(OPENSSL_WINDOWS) || (!defined(OPENSSL_X86_64) && !defined(OPENSSL_X86)) || !defined(__SSE2__)
25
26 /* sigma contains the ChaCha constants, which happen to be an ASCII string. */
27 static const uint8_t sigma[16] = { 'e', 'x', 'p', 'a', 'n', 'd', ' ', '3',
28 '2', '-', 'b', 'y', 't', 'e', ' ', 'k' };
29
30 #define ROTATE(v, n) (((v) << (n)) | ((v) >> (32 - (n))))
31 #define XOR(v, w) ((v) ^ (w))
32 #define PLUS(x, y) ((x) + (y))
33 #define PLUSONE(v) (PLUS((v), 1))
34
35 #define U32TO8_LITTLE(p, v) \
36 { \
37 (p)[0] = (v >> 0) & 0xff; \
38 (p)[1] = (v >> 8) & 0xff; \
39 (p)[2] = (v >> 16) & 0xff; \
40 (p)[3] = (v >> 24) & 0xff; \
41 }
42
43 #define U8TO32_LITTLE(p) \
44 (((uint32_t)((p)[0])) | ((uint32_t)((p)[1]) << 8) | \
45 ((uint32_t)((p)[2]) << 16) | ((uint32_t)((p)[3]) << 24))
46
47 /* QUARTERROUND updates a, b, c, d with a ChaCha "quarter" round. */
48 #define QUARTERROUND(a,b,c,d) \
49 x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]),16); \
50 x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]),12); \
51 x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]), 8); \
52 x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]), 7);
53
54 #if defined(OPENSSL_ARM) && !defined(OPENSSL_NO_ASM)
55 /* Defined in chacha_vec.c */
56 void CRYPTO_chacha_20_neon(uint8_t *out, const uint8_t *in, size_t in_len,
57 const uint8_t key[32], const uint8_t nonce[12],
58 uint32_t counter);
59 #endif
60
61 /* chacha_core performs 20 rounds of ChaCha on the input words in
62 * |input| and writes the 64 output bytes to |output|. */
chacha_core(uint8_t output[64],const uint32_t input[16])63 static void chacha_core(uint8_t output[64], const uint32_t input[16]) {
64 uint32_t x[16];
65 int i;
66
67 memcpy(x, input, sizeof(uint32_t) * 16);
68 for (i = 20; i > 0; i -= 2) {
69 QUARTERROUND(0, 4, 8, 12)
70 QUARTERROUND(1, 5, 9, 13)
71 QUARTERROUND(2, 6, 10, 14)
72 QUARTERROUND(3, 7, 11, 15)
73 QUARTERROUND(0, 5, 10, 15)
74 QUARTERROUND(1, 6, 11, 12)
75 QUARTERROUND(2, 7, 8, 13)
76 QUARTERROUND(3, 4, 9, 14)
77 }
78
79 for (i = 0; i < 16; ++i) {
80 x[i] = PLUS(x[i], input[i]);
81 }
82 for (i = 0; i < 16; ++i) {
83 U32TO8_LITTLE(output + 4 * i, x[i]);
84 }
85 }
86
CRYPTO_chacha_20(uint8_t * out,const uint8_t * in,size_t in_len,const uint8_t key[32],const uint8_t nonce[12],uint32_t counter)87 void CRYPTO_chacha_20(uint8_t *out, const uint8_t *in, size_t in_len,
88 const uint8_t key[32], const uint8_t nonce[12],
89 uint32_t counter) {
90 uint32_t input[16];
91 uint8_t buf[64];
92 size_t todo, i;
93
94 #if defined(OPENSSL_ARM) && !defined(OPENSSL_NO_ASM)
95 if (CRYPTO_is_NEON_capable()) {
96 CRYPTO_chacha_20_neon(out, in, in_len, key, nonce, counter);
97 return;
98 }
99 #endif
100
101 input[0] = U8TO32_LITTLE(sigma + 0);
102 input[1] = U8TO32_LITTLE(sigma + 4);
103 input[2] = U8TO32_LITTLE(sigma + 8);
104 input[3] = U8TO32_LITTLE(sigma + 12);
105
106 input[4] = U8TO32_LITTLE(key + 0);
107 input[5] = U8TO32_LITTLE(key + 4);
108 input[6] = U8TO32_LITTLE(key + 8);
109 input[7] = U8TO32_LITTLE(key + 12);
110
111 input[8] = U8TO32_LITTLE(key + 16);
112 input[9] = U8TO32_LITTLE(key + 20);
113 input[10] = U8TO32_LITTLE(key + 24);
114 input[11] = U8TO32_LITTLE(key + 28);
115
116 input[12] = counter;
117 input[13] = U8TO32_LITTLE(nonce + 0);
118 input[14] = U8TO32_LITTLE(nonce + 4);
119 input[15] = U8TO32_LITTLE(nonce + 8);
120
121 while (in_len > 0) {
122 todo = sizeof(buf);
123 if (in_len < todo) {
124 todo = in_len;
125 }
126
127 chacha_core(buf, input);
128 for (i = 0; i < todo; i++) {
129 out[i] = in[i] ^ buf[i];
130 }
131
132 out += todo;
133 in += todo;
134 in_len -= todo;
135
136 input[12]++;
137 }
138 }
139
140 #endif /* OPENSSL_WINDOWS || !OPENSSL_X86_64 && !OPENSSL_X86 || !__SSE2__ */
141