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// This implementation of poly1305 is by Andrew Moon
16// (https://github.com/floodyberry/poly1305-donna) and released as public
17// domain.
18
19#include <openssl/poly1305.h>
20
21#include <string.h>
22
23#include <openssl/cpu.h>
24
25#include "internal.h"
26#include "../internal.h"
27
28
29#if defined(OPENSSL_WINDOWS) || !defined(OPENSSL_X86_64)
30
31// We can assume little-endian.
32static uint32_t U8TO32_LE(const uint8_t *m) {
33 uint32_t r;
34 OPENSSL_memcpy(&r, m, sizeof(r));
35 return r;
36}
37
38static void U32TO8_LE(uint8_t *m, uint32_t v) {
39 OPENSSL_memcpy(m, &v, sizeof(v));
40}
41
42static uint64_t mul32x32_64(uint32_t a, uint32_t b) { return (uint64_t)a * b; }
43
44struct poly1305_state_st {
45 uint32_t r0, r1, r2, r3, r4;
46 uint32_t s1, s2, s3, s4;
47 uint32_t h0, h1, h2, h3, h4;
48 uint8_t buf[16];
49 unsigned int buf_used;
50 uint8_t key[16];
51};
52
53static inline struct poly1305_state_st *poly1305_aligned_state(
54 poly1305_state *state) {
55 return (struct poly1305_state_st *)(((uintptr_t)state + 63) & ~63);
56}
57
58// poly1305_blocks updates |state| given some amount of input data. This
59// function may only be called with a |len| that is not a multiple of 16 at the
60// end of the data. Otherwise the input must be buffered into 16 byte blocks.
61static void poly1305_update(struct poly1305_state_st *state, const uint8_t *in,
62 size_t len) {
63 uint32_t t0, t1, t2, t3;
64 uint64_t t[5];
65 uint32_t b;
66 uint64_t c;
67 size_t j;
68 uint8_t mp[16];
69
70 if (len < 16) {
71 goto poly1305_donna_atmost15bytes;
72 }
73
74poly1305_donna_16bytes:
75 t0 = U8TO32_LE(in);
76 t1 = U8TO32_LE(in + 4);
77 t2 = U8TO32_LE(in + 8);
78 t3 = U8TO32_LE(in + 12);
79
80 in += 16;
81 len -= 16;
82
83 state->h0 += t0 & 0x3ffffff;
84 state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
85 state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
86 state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
87 state->h4 += (t3 >> 8) | (1 << 24);
88
89poly1305_donna_mul:
90 t[0] = mul32x32_64(state->h0, state->r0) + mul32x32_64(state->h1, state->s4) +
91 mul32x32_64(state->h2, state->s3) + mul32x32_64(state->h3, state->s2) +
92 mul32x32_64(state->h4, state->s1);
93 t[1] = mul32x32_64(state->h0, state->r1) + mul32x32_64(state->h1, state->r0) +
94 mul32x32_64(state->h2, state->s4) + mul32x32_64(state->h3, state->s3) +
95 mul32x32_64(state->h4, state->s2);
96 t[2] = mul32x32_64(state->h0, state->r2) + mul32x32_64(state->h1, state->r1) +
97 mul32x32_64(state->h2, state->r0) + mul32x32_64(state->h3, state->s4) +
98 mul32x32_64(state->h4, state->s3);
99 t[3] = mul32x32_64(state->h0, state->r3) + mul32x32_64(state->h1, state->r2) +
100 mul32x32_64(state->h2, state->r1) + mul32x32_64(state->h3, state->r0) +
101 mul32x32_64(state->h4, state->s4);
102 t[4] = mul32x32_64(state->h0, state->r4) + mul32x32_64(state->h1, state->r3) +
103 mul32x32_64(state->h2, state->r2) + mul32x32_64(state->h3, state->r1) +
104 mul32x32_64(state->h4, state->r0);
105
106 state->h0 = (uint32_t)t[0] & 0x3ffffff;
107 c = (t[0] >> 26);
108 t[1] += c;
109 state->h1 = (uint32_t)t[1] & 0x3ffffff;
110 b = (uint32_t)(t[1] >> 26);
111 t[2] += b;
112 state->h2 = (uint32_t)t[2] & 0x3ffffff;
113 b = (uint32_t)(t[2] >> 26);
114 t[3] += b;
115 state->h3 = (uint32_t)t[3] & 0x3ffffff;
116 b = (uint32_t)(t[3] >> 26);
117 t[4] += b;
118 state->h4 = (uint32_t)t[4] & 0x3ffffff;
119 b = (uint32_t)(t[4] >> 26);
120 state->h0 += b * 5;
121
122 if (len >= 16) {
123 goto poly1305_donna_16bytes;
124 }
125
126// final bytes
127poly1305_donna_atmost15bytes:
128 if (!len) {
129 return;
130 }
131
132 for (j = 0; j < len; j++) {
133 mp[j] = in[j];
134 }
135 mp[j++] = 1;
136 for (; j < 16; j++) {
137 mp[j] = 0;
138 }
139 len = 0;
140
141 t0 = U8TO32_LE(mp + 0);
142 t1 = U8TO32_LE(mp + 4);
143 t2 = U8TO32_LE(mp + 8);
144 t3 = U8TO32_LE(mp + 12);
145
146 state->h0 += t0 & 0x3ffffff;
147 state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
148 state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
149 state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
150 state->h4 += (t3 >> 8);
151
152 goto poly1305_donna_mul;
153}
154
155void CRYPTO_poly1305_init(poly1305_state *statep, const uint8_t key[32]) {
156 struct poly1305_state_st *state = poly1305_aligned_state(statep);
157 uint32_t t0, t1, t2, t3;
158
159#if defined(OPENSSL_POLY1305_NEON)
160 if (CRYPTO_is_NEON_capable()) {
161 CRYPTO_poly1305_init_neon(statep, key);
162 return;
163 }
164#endif
165
166 t0 = U8TO32_LE(key + 0);
167 t1 = U8TO32_LE(key + 4);
168 t2 = U8TO32_LE(key + 8);
169 t3 = U8TO32_LE(key + 12);
170
171 // precompute multipliers
172 state->r0 = t0 & 0x3ffffff;
173 t0 >>= 26;
174 t0 |= t1 << 6;
175 state->r1 = t0 & 0x3ffff03;
176 t1 >>= 20;
177 t1 |= t2 << 12;
178 state->r2 = t1 & 0x3ffc0ff;
179 t2 >>= 14;
180 t2 |= t3 << 18;
181 state->r3 = t2 & 0x3f03fff;
182 t3 >>= 8;
183 state->r4 = t3 & 0x00fffff;
184
185 state->s1 = state->r1 * 5;
186 state->s2 = state->r2 * 5;
187 state->s3 = state->r3 * 5;
188 state->s4 = state->r4 * 5;
189
190 // init state
191 state->h0 = 0;
192 state->h1 = 0;
193 state->h2 = 0;
194 state->h3 = 0;
195 state->h4 = 0;
196
197 state->buf_used = 0;
198 OPENSSL_memcpy(state->key, key + 16, sizeof(state->key));
199}
200
201void CRYPTO_poly1305_update(poly1305_state *statep, const uint8_t *in,
202 size_t in_len) {
203 unsigned int i;
204 struct poly1305_state_st *state = poly1305_aligned_state(statep);
205
206#if defined(OPENSSL_POLY1305_NEON)
207 if (CRYPTO_is_NEON_capable()) {
208 CRYPTO_poly1305_update_neon(statep, in, in_len);
209 return;
210 }
211#endif
212
213 if (state->buf_used) {
214 unsigned todo = 16 - state->buf_used;
215 if (todo > in_len) {
216 todo = (unsigned)in_len;
217 }
218 for (i = 0; i < todo; i++) {
219 state->buf[state->buf_used + i] = in[i];
220 }
221 state->buf_used += todo;
222 in_len -= todo;
223 in += todo;
224
225 if (state->buf_used == 16) {
226 poly1305_update(state, state->buf, 16);
227 state->buf_used = 0;
228 }
229 }
230
231 if (in_len >= 16) {
232 size_t todo = in_len & ~0xf;
233 poly1305_update(state, in, todo);
234 in += todo;
235 in_len &= 0xf;
236 }
237
238 if (in_len) {
239 for (i = 0; i < in_len; i++) {
240 state->buf[i] = in[i];
241 }
242 state->buf_used = (unsigned)in_len;
243 }
244}
245
246void CRYPTO_poly1305_finish(poly1305_state *statep, uint8_t mac[16]) {
247 struct poly1305_state_st *state = poly1305_aligned_state(statep);
248 uint64_t f0, f1, f2, f3;
249 uint32_t g0, g1, g2, g3, g4;
250 uint32_t b, nb;
251
252#if defined(OPENSSL_POLY1305_NEON)
253 if (CRYPTO_is_NEON_capable()) {
254 CRYPTO_poly1305_finish_neon(statep, mac);
255 return;
256 }
257#endif
258
259 if (state->buf_used) {
260 poly1305_update(state, state->buf, state->buf_used);
261 }
262
263 b = state->h0 >> 26;
264 state->h0 = state->h0 & 0x3ffffff;
265 state->h1 += b;
266 b = state->h1 >> 26;
267 state->h1 = state->h1 & 0x3ffffff;
268 state->h2 += b;
269 b = state->h2 >> 26;
270 state->h2 = state->h2 & 0x3ffffff;
271 state->h3 += b;
272 b = state->h3 >> 26;
273 state->h3 = state->h3 & 0x3ffffff;
274 state->h4 += b;
275 b = state->h4 >> 26;
276 state->h4 = state->h4 & 0x3ffffff;
277 state->h0 += b * 5;
278
279 g0 = state->h0 + 5;
280 b = g0 >> 26;
281 g0 &= 0x3ffffff;
282 g1 = state->h1 + b;
283 b = g1 >> 26;
284 g1 &= 0x3ffffff;
285 g2 = state->h2 + b;
286 b = g2 >> 26;
287 g2 &= 0x3ffffff;
288 g3 = state->h3 + b;
289 b = g3 >> 26;
290 g3 &= 0x3ffffff;
291 g4 = state->h4 + b - (1 << 26);
292
293 b = (g4 >> 31) - 1;
294 nb = ~b;
295 state->h0 = (state->h0 & nb) | (g0 & b);
296 state->h1 = (state->h1 & nb) | (g1 & b);
297 state->h2 = (state->h2 & nb) | (g2 & b);
298 state->h3 = (state->h3 & nb) | (g3 & b);
299 state->h4 = (state->h4 & nb) | (g4 & b);
300
301 f0 = ((state->h0) | (state->h1 << 26)) + (uint64_t)U8TO32_LE(&state->key[0]);
302 f1 = ((state->h1 >> 6) | (state->h2 << 20)) +
303 (uint64_t)U8TO32_LE(&state->key[4]);
304 f2 = ((state->h2 >> 12) | (state->h3 << 14)) +
305 (uint64_t)U8TO32_LE(&state->key[8]);
306 f3 = ((state->h3 >> 18) | (state->h4 << 8)) +
307 (uint64_t)U8TO32_LE(&state->key[12]);
308
309 U32TO8_LE(&mac[0], f0);
310 f1 += (f0 >> 32);
311 U32TO8_LE(&mac[4], f1);
312 f2 += (f1 >> 32);
313 U32TO8_LE(&mac[8], f2);
314 f3 += (f2 >> 32);
315 U32TO8_LE(&mac[12], f3);
316}
317
318#endif // OPENSSL_WINDOWS || !OPENSSL_X86_64
319