poly1305.c revision 1.5.12.2 1 1.5.12.2 snj /*
2 1.5.12.2 snj * Public Domain poly1305 from Andrew Moon
3 1.5.12.2 snj * poly1305-donna-unrolled.c from https://github.com/floodyberry/poly1305-donna
4 1.5.12.2 snj */
5 1.5.12.2 snj
6 1.5.12.2 snj /* $OpenBSD: poly1305.c,v 1.3 2013/12/19 22:57:13 djm Exp $ */
7 1.5.12.2 snj #include "includes.h"
8 1.5.12.2 snj __RCSID("$NetBSD: poly1305.c,v 1.5.12.2 2017/08/15 05:27:52 snj Exp $");
9 1.5.12.2 snj
10 1.5.12.2 snj #include <sys/types.h>
11 1.5.12.2 snj #include <stdint.h>
12 1.5.12.2 snj
13 1.5.12.2 snj #include "poly1305.h"
14 1.5.12.2 snj
15 1.5.12.2 snj #define mul32x32_64(a,b) ((uint64_t)(a) * (b))
16 1.5.12.2 snj
17 1.5.12.2 snj #define U8TO32_LE(p) \
18 1.5.12.2 snj (((uint32_t)((p)[0])) | \
19 1.5.12.2 snj ((uint32_t)((p)[1]) << 8) | \
20 1.5.12.2 snj ((uint32_t)((p)[2]) << 16) | \
21 1.5.12.2 snj ((uint32_t)((p)[3]) << 24))
22 1.5.12.2 snj
23 1.5.12.2 snj #define U32TO8_LE(p, v) \
24 1.5.12.2 snj do { \
25 1.5.12.2 snj (p)[0] = (uint8_t)((v)); \
26 1.5.12.2 snj (p)[1] = (uint8_t)((v) >> 8); \
27 1.5.12.2 snj (p)[2] = (uint8_t)((v) >> 16); \
28 1.5.12.2 snj (p)[3] = (uint8_t)((v) >> 24); \
29 1.5.12.2 snj } while (0)
30 1.5.12.2 snj
31 1.5.12.2 snj void
32 1.5.12.2 snj poly1305_auth(unsigned char out[POLY1305_TAGLEN], const unsigned char *m, size_t inlen, const unsigned char key[POLY1305_KEYLEN]) {
33 1.5.12.2 snj uint32_t t0,t1,t2,t3;
34 1.5.12.2 snj uint32_t h0,h1,h2,h3,h4;
35 1.5.12.2 snj uint32_t r0,r1,r2,r3,r4;
36 1.5.12.2 snj uint32_t s1,s2,s3,s4;
37 1.5.12.2 snj uint32_t b, nb;
38 1.5.12.2 snj size_t j;
39 1.5.12.2 snj uint64_t t[5];
40 1.5.12.2 snj uint64_t f0,f1,f2,f3;
41 1.5.12.2 snj uint32_t g0,g1,g2,g3,g4;
42 1.5.12.2 snj uint64_t c;
43 1.5.12.2 snj unsigned char mp[16];
44 1.5.12.2 snj
45 1.5.12.2 snj /* clamp key */
46 1.5.12.2 snj t0 = U8TO32_LE(key+0);
47 1.5.12.2 snj t1 = U8TO32_LE(key+4);
48 1.5.12.2 snj t2 = U8TO32_LE(key+8);
49 1.5.12.2 snj t3 = U8TO32_LE(key+12);
50 1.5.12.2 snj
51 1.5.12.2 snj /* precompute multipliers */
52 1.5.12.2 snj r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
53 1.5.12.2 snj r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
54 1.5.12.2 snj r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
55 1.5.12.2 snj r3 = t2 & 0x3f03fff; t3 >>= 8;
56 1.5.12.2 snj r4 = t3 & 0x00fffff;
57 1.5.12.2 snj
58 1.5.12.2 snj s1 = r1 * 5;
59 1.5.12.2 snj s2 = r2 * 5;
60 1.5.12.2 snj s3 = r3 * 5;
61 1.5.12.2 snj s4 = r4 * 5;
62 1.5.12.2 snj
63 1.5.12.2 snj /* init state */
64 1.5.12.2 snj h0 = 0;
65 1.5.12.2 snj h1 = 0;
66 1.5.12.2 snj h2 = 0;
67 1.5.12.2 snj h3 = 0;
68 1.5.12.2 snj h4 = 0;
69 1.5.12.2 snj
70 1.5.12.2 snj /* full blocks */
71 1.5.12.2 snj if (inlen < 16) goto poly1305_donna_atmost15bytes;
72 1.5.12.2 snj poly1305_donna_16bytes:
73 1.5.12.2 snj m += 16;
74 1.5.12.2 snj inlen -= 16;
75 1.5.12.2 snj
76 1.5.12.2 snj t0 = U8TO32_LE(m-16);
77 1.5.12.2 snj t1 = U8TO32_LE(m-12);
78 1.5.12.2 snj t2 = U8TO32_LE(m-8);
79 1.5.12.2 snj t3 = U8TO32_LE(m-4);
80 1.5.12.2 snj
81 1.5.12.2 snj h0 += t0 & 0x3ffffff;
82 1.5.12.2 snj h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
83 1.5.12.2 snj h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
84 1.5.12.2 snj h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
85 1.5.12.2 snj h4 += (t3 >> 8) | (1 << 24);
86 1.5.12.2 snj
87 1.5.12.2 snj
88 1.5.12.2 snj poly1305_donna_mul:
89 1.5.12.2 snj t[0] = mul32x32_64(h0,r0) + mul32x32_64(h1,s4) + mul32x32_64(h2,s3) + mul32x32_64(h3,s2) + mul32x32_64(h4,s1);
90 1.5.12.2 snj t[1] = mul32x32_64(h0,r1) + mul32x32_64(h1,r0) + mul32x32_64(h2,s4) + mul32x32_64(h3,s3) + mul32x32_64(h4,s2);
91 1.5.12.2 snj t[2] = mul32x32_64(h0,r2) + mul32x32_64(h1,r1) + mul32x32_64(h2,r0) + mul32x32_64(h3,s4) + mul32x32_64(h4,s3);
92 1.5.12.2 snj t[3] = mul32x32_64(h0,r3) + mul32x32_64(h1,r2) + mul32x32_64(h2,r1) + mul32x32_64(h3,r0) + mul32x32_64(h4,s4);
93 1.5.12.2 snj t[4] = mul32x32_64(h0,r4) + mul32x32_64(h1,r3) + mul32x32_64(h2,r2) + mul32x32_64(h3,r1) + mul32x32_64(h4,r0);
94 1.5.12.2 snj
95 1.5.12.2 snj h0 = (uint32_t)t[0] & 0x3ffffff; c = (t[0] >> 26);
96 1.5.12.2 snj t[1] += c; h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >> 26);
97 1.5.12.2 snj t[2] += b; h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >> 26);
98 1.5.12.2 snj t[3] += b; h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >> 26);
99 1.5.12.2 snj t[4] += b; h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >> 26);
100 1.5.12.2 snj h0 += b * 5;
101 1.5.12.2 snj
102 1.5.12.2 snj if (inlen >= 16) goto poly1305_donna_16bytes;
103 1.5.12.2 snj
104 1.5.12.2 snj /* final bytes */
105 1.5.12.2 snj poly1305_donna_atmost15bytes:
106 1.5.12.2 snj if (!inlen) goto poly1305_donna_finish;
107 1.5.12.2 snj
108 1.5.12.2 snj for (j = 0; j < inlen; j++) mp[j] = m[j];
109 1.5.12.2 snj mp[j++] = 1;
110 1.5.12.2 snj for (; j < 16; j++) mp[j] = 0;
111 1.5.12.2 snj inlen = 0;
112 1.5.12.2 snj
113 1.5.12.2 snj t0 = U8TO32_LE(mp+0);
114 1.5.12.2 snj t1 = U8TO32_LE(mp+4);
115 1.5.12.2 snj t2 = U8TO32_LE(mp+8);
116 1.5.12.2 snj t3 = U8TO32_LE(mp+12);
117 1.5.12.2 snj
118 1.5.12.2 snj h0 += t0 & 0x3ffffff;
119 1.5.12.2 snj h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
120 1.5.12.2 snj h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
121 1.5.12.2 snj h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
122 1.5.12.2 snj h4 += (t3 >> 8);
123 1.5.12.2 snj
124 1.5.12.2 snj goto poly1305_donna_mul;
125 1.5.12.2 snj
126 1.5.12.2 snj poly1305_donna_finish:
127 1.5.12.2 snj b = h0 >> 26; h0 = h0 & 0x3ffffff;
128 1.5.12.2 snj h1 += b; b = h1 >> 26; h1 = h1 & 0x3ffffff;
129 1.5.12.2 snj h2 += b; b = h2 >> 26; h2 = h2 & 0x3ffffff;
130 1.5.12.2 snj h3 += b; b = h3 >> 26; h3 = h3 & 0x3ffffff;
131 1.5.12.2 snj h4 += b; b = h4 >> 26; h4 = h4 & 0x3ffffff;
132 1.5.12.2 snj h0 += b * 5; b = h0 >> 26; h0 = h0 & 0x3ffffff;
133 1.5.12.2 snj h1 += b;
134 1.5.12.2 snj
135 1.5.12.2 snj g0 = h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
136 1.5.12.2 snj g1 = h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
137 1.5.12.2 snj g2 = h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
138 1.5.12.2 snj g3 = h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
139 1.5.12.2 snj g4 = h4 + b - (1 << 26);
140 1.5.12.2 snj
141 1.5.12.2 snj b = (g4 >> 31) - 1;
142 1.5.12.2 snj nb = ~b;
143 1.5.12.2 snj h0 = (h0 & nb) | (g0 & b);
144 1.5.12.2 snj h1 = (h1 & nb) | (g1 & b);
145 1.5.12.2 snj h2 = (h2 & nb) | (g2 & b);
146 1.5.12.2 snj h3 = (h3 & nb) | (g3 & b);
147 1.5.12.2 snj h4 = (h4 & nb) | (g4 & b);
148 1.5.12.2 snj
149 1.5.12.2 snj f0 = ((h0 ) | (h1 << 26)) + (uint64_t)U8TO32_LE(&key[16]);
150 1.5.12.2 snj f1 = ((h1 >> 6) | (h2 << 20)) + (uint64_t)U8TO32_LE(&key[20]);
151 1.5.12.2 snj f2 = ((h2 >> 12) | (h3 << 14)) + (uint64_t)U8TO32_LE(&key[24]);
152 1.5.12.2 snj f3 = ((h3 >> 18) | (h4 << 8)) + (uint64_t)U8TO32_LE(&key[28]);
153 1.5.12.2 snj
154 1.5.12.2 snj U32TO8_LE(&out[ 0], f0); f1 += (f0 >> 32);
155 1.5.12.2 snj U32TO8_LE(&out[ 4], f1); f2 += (f1 >> 32);
156 1.5.12.2 snj U32TO8_LE(&out[ 8], f2); f3 += (f2 >> 32);
157 1.5.12.2 snj U32TO8_LE(&out[12], f3);
158 1.5.12.2 snj }
159