xref: /freebsd/contrib/bearssl/src/symcipher/poly1305_ctmul32.c (revision f7c32ed617858bcd22f8d1b03199099d50125721)
1 /*
2  * Copyright (c) 2017 Thomas Pornin <pornin@bolet.org>
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining
5  * a copy of this software and associated documentation files (the
6  * "Software"), to deal in the Software without restriction, including
7  * without limitation the rights to use, copy, modify, merge, publish,
8  * distribute, sublicense, and/or sell copies of the Software, and to
9  * permit persons to whom the Software is furnished to do so, subject to
10  * the following conditions:
11  *
12  * The above copyright notice and this permission notice shall be
13  * included in all copies or substantial portions of the Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
19  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
20  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
21  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 
25 #include "inner.h"
26 
27 /*
28  * Perform the inner processing of blocks for Poly1305.
29  */
30 static void
31 poly1305_inner(uint32_t *a, const uint32_t *r, const void *data, size_t len)
32 {
33 	/*
34 	 * Implementation notes: we split the 130-bit values into ten
35 	 * 13-bit words. This gives us some space for carries and allows
36 	 * using only 32x32->32 multiplications, which are way faster than
37 	 * 32x32->64 multiplications on the ARM Cortex-M0/M0+, and also
38 	 * help in making constant-time code on the Cortex-M3.
39 	 *
40 	 * Since we compute modulo 2^130-5, the "upper words" become
41 	 * low words with a factor of 5; that is, x*2^130 = x*5 mod p.
42 	 * This has already been integrated in the r[] array, which
43 	 * is extended to the 0..18 range.
44 	 *
45 	 * In each loop iteration, a[] and r[] words are 13-bit each,
46 	 * except a[1] which may use 14 bits.
47 	 */
48 	const unsigned char *buf;
49 
50 	buf = data;
51 	while (len > 0) {
52 		unsigned char tmp[16];
53 		uint32_t b[10];
54 		unsigned u, v;
55 		uint32_t z, cc1, cc2;
56 
57 		/*
58 		 * If there is a partial block, right-pad it with zeros.
59 		 */
60 		if (len < 16) {
61 			memset(tmp, 0, sizeof tmp);
62 			memcpy(tmp, buf, len);
63 			buf = tmp;
64 			len = 16;
65 		}
66 
67 		/*
68 		 * Decode next block and apply the "high bit"; that value
69 		 * is added to the accumulator.
70 		 */
71 		v = br_dec16le(buf);
72 		a[0] += v & 0x01FFF;
73 		v >>= 13;
74 		v |= buf[2] << 3;
75 		v |= buf[3] << 11;
76 		a[1] += v & 0x01FFF;
77 		v >>= 13;
78 		v |= buf[4] << 6;
79 		a[2] += v & 0x01FFF;
80 		v >>= 13;
81 		v |= buf[5] << 1;
82 		v |= buf[6] << 9;
83 		a[3] += v & 0x01FFF;
84 		v >>= 13;
85 		v |= buf[7] << 4;
86 		v |= buf[8] << 12;
87 		a[4] += v & 0x01FFF;
88 		v >>= 13;
89 		v |= buf[9] << 7;
90 		a[5] += v & 0x01FFF;
91 		v >>= 13;
92 		v |= buf[10] << 2;
93 		v |= buf[11] << 10;
94 		a[6] += v & 0x01FFF;
95 		v >>= 13;
96 		v |= buf[12] << 5;
97 		a[7] += v & 0x01FFF;
98 		v = br_dec16le(buf + 13);
99 		a[8] += v & 0x01FFF;
100 		v >>= 13;
101 		v |= buf[15] << 3;
102 		a[9] += v | 0x00800;
103 
104 		/*
105 		 * At that point, all a[] values fit on 14 bits, while
106 		 * all r[] values fit on 13 bits. Thus products fit on
107 		 * 27 bits, and we can accumulate up to 31 of them in
108 		 * a 32-bit word and still have some room for carries.
109 		 */
110 
111 		/*
112 		 * Now a[] contains words with values up to 14 bits each.
113 		 * We perform the multiplication with r[].
114 		 *
115 		 * The extended words of r[] may be larger than 13 bits
116 		 * (they are 5 times a 13-bit word) so the full summation
117 		 * may yield values up to 46 times a 27-bit word, which
118 		 * does not fit on a 32-bit word. To avoid that issue, we
119 		 * must split the loop below in two, with a carry
120 		 * propagation operation in the middle.
121 		 */
122 		cc1 = 0;
123 		for (u = 0; u < 10; u ++) {
124 			uint32_t s;
125 
126 			s = cc1
127 				+ MUL15(a[0], r[u + 9 - 0])
128 				+ MUL15(a[1], r[u + 9 - 1])
129 				+ MUL15(a[2], r[u + 9 - 2])
130 				+ MUL15(a[3], r[u + 9 - 3])
131 				+ MUL15(a[4], r[u + 9 - 4]);
132 			b[u] = s & 0x1FFF;
133 			cc1 = s >> 13;
134 		}
135 		cc2 = 0;
136 		for (u = 0; u < 10; u ++) {
137 			uint32_t s;
138 
139 			s = b[u] + cc2
140 				+ MUL15(a[5], r[u + 9 - 5])
141 				+ MUL15(a[6], r[u + 9 - 6])
142 				+ MUL15(a[7], r[u + 9 - 7])
143 				+ MUL15(a[8], r[u + 9 - 8])
144 				+ MUL15(a[9], r[u + 9 - 9]);
145 			b[u] = s & 0x1FFF;
146 			cc2 = s >> 13;
147 		}
148 		memcpy(a, b, sizeof b);
149 
150 		/*
151 		 * The two carries "loop back" with a factor of 5. We
152 		 * propagate them into a[0] and a[1].
153 		 */
154 		z = cc1 + cc2;
155 		z += (z << 2) + a[0];
156 		a[0] = z & 0x1FFF;
157 		a[1] += z >> 13;
158 
159 		buf += 16;
160 		len -= 16;
161 	}
162 }
163 
164 /* see bearssl_block.h */
165 void
166 br_poly1305_ctmul32_run(const void *key, const void *iv,
167 	void *data, size_t len, const void *aad, size_t aad_len,
168 	void *tag, br_chacha20_run ichacha, int encrypt)
169 {
170 	unsigned char pkey[32], foot[16];
171 	uint32_t z, r[19], acc[10], cc, ctl;
172 	int i;
173 
174 	/*
175 	 * Compute the MAC key. The 'r' value is the first 16 bytes of
176 	 * pkey[].
177 	 */
178 	memset(pkey, 0, sizeof pkey);
179 	ichacha(key, iv, 0, pkey, sizeof pkey);
180 
181 	/*
182 	 * If encrypting, ChaCha20 must run first, followed by Poly1305.
183 	 * When decrypting, the operations are reversed.
184 	 */
185 	if (encrypt) {
186 		ichacha(key, iv, 1, data, len);
187 	}
188 
189 	/*
190 	 * Run Poly1305. We must process the AAD, then ciphertext, then
191 	 * the footer (with the lengths). Note that the AAD and ciphertext
192 	 * are meant to be padded with zeros up to the next multiple of 16,
193 	 * and the length of the footer is 16 bytes as well.
194 	 */
195 
196 	/*
197 	 * Decode the 'r' value into 13-bit words, with the "clamping"
198 	 * operation applied.
199 	 */
200 	z = br_dec32le(pkey) & 0x03FFFFFF;
201 	r[9] = z & 0x1FFF;
202 	r[10] = z >> 13;
203 	z = (br_dec32le(pkey +  3) >> 2) & 0x03FFFF03;
204 	r[11] = z & 0x1FFF;
205 	r[12] = z >> 13;
206 	z = (br_dec32le(pkey +  6) >> 4) & 0x03FFC0FF;
207 	r[13] = z & 0x1FFF;
208 	r[14] = z >> 13;
209 	z = (br_dec32le(pkey +  9) >> 6) & 0x03F03FFF;
210 	r[15] = z & 0x1FFF;
211 	r[16] = z >> 13;
212 	z = (br_dec32le(pkey + 12) >> 8) & 0x000FFFFF;
213 	r[17] = z & 0x1FFF;
214 	r[18] = z >> 13;
215 
216 	/*
217 	 * Extend r[] with the 5x factor pre-applied.
218 	 */
219 	for (i = 0; i < 9; i ++) {
220 		r[i] = MUL15(5, r[i + 10]);
221 	}
222 
223 	/*
224 	 * Accumulator is 0.
225 	 */
226 	memset(acc, 0, sizeof acc);
227 
228 	/*
229 	 * Process the additional authenticated data, ciphertext, and
230 	 * footer in due order.
231 	 */
232 	br_enc64le(foot, (uint64_t)aad_len);
233 	br_enc64le(foot + 8, (uint64_t)len);
234 	poly1305_inner(acc, r, aad, aad_len);
235 	poly1305_inner(acc, r, data, len);
236 	poly1305_inner(acc, r, foot, sizeof foot);
237 
238 	/*
239 	 * Finalise modular reduction. This is done with carry propagation
240 	 * and applying the '2^130 = -5 mod p' rule. Note that the output
241 	 * of poly1035_inner() is already mostly reduced, since only
242 	 * acc[1] may be (very slightly) above 2^13. A single loop back
243 	 * to acc[1] will be enough to make the value fit in 130 bits.
244 	 */
245 	cc = 0;
246 	for (i = 1; i < 10; i ++) {
247 		z = acc[i] + cc;
248 		acc[i] = z & 0x1FFF;
249 		cc = z >> 13;
250 	}
251 	z = acc[0] + cc + (cc << 2);
252 	acc[0] = z & 0x1FFF;
253 	acc[1] += z >> 13;
254 
255 	/*
256 	 * We may still have a value in the 2^130-5..2^130-1 range, in
257 	 * which case we must reduce it again. The code below selects,
258 	 * in constant-time, between 'acc' and 'acc-p',
259 	 */
260 	ctl = GT(acc[0], 0x1FFA);
261 	for (i = 1; i < 10; i ++) {
262 		ctl &= EQ(acc[i], 0x1FFF);
263 	}
264 	acc[0] = MUX(ctl, acc[0] - 0x1FFB, acc[0]);
265 	for (i = 1; i < 10; i ++) {
266 		acc[i] &= ~(-ctl);
267 	}
268 
269 	/*
270 	 * Convert back the accumulator to 32-bit words, and add the
271 	 * 's' value (second half of pkey[]). That addition is done
272 	 * modulo 2^128.
273 	 */
274 	z = acc[0] + (acc[1] << 13) + br_dec16le(pkey + 16);
275 	br_enc16le((unsigned char *)tag, z & 0xFFFF);
276 	z = (z >> 16) + (acc[2] << 10) + br_dec16le(pkey + 18);
277 	br_enc16le((unsigned char *)tag + 2, z & 0xFFFF);
278 	z = (z >> 16) + (acc[3] << 7) + br_dec16le(pkey + 20);
279 	br_enc16le((unsigned char *)tag + 4, z & 0xFFFF);
280 	z = (z >> 16) + (acc[4] << 4) + br_dec16le(pkey + 22);
281 	br_enc16le((unsigned char *)tag + 6, z & 0xFFFF);
282 	z = (z >> 16) + (acc[5] << 1) + (acc[6] << 14) + br_dec16le(pkey + 24);
283 	br_enc16le((unsigned char *)tag + 8, z & 0xFFFF);
284 	z = (z >> 16) + (acc[7] << 11) + br_dec16le(pkey + 26);
285 	br_enc16le((unsigned char *)tag + 10, z & 0xFFFF);
286 	z = (z >> 16) + (acc[8] << 8) + br_dec16le(pkey + 28);
287 	br_enc16le((unsigned char *)tag + 12, z & 0xFFFF);
288 	z = (z >> 16) + (acc[9] << 5) + br_dec16le(pkey + 30);
289 	br_enc16le((unsigned char *)tag + 14, z & 0xFFFF);
290 
291 	/*
292 	 * If decrypting, then ChaCha20 runs _after_ Poly1305.
293 	 */
294 	if (!encrypt) {
295 		ichacha(key, iv, 1, data, len);
296 	}
297 }
298