Annotation of src/usr.bin/ssh/poly1305.c, Revision 1.2
1.1 djm 1: /*
2: * Public Domain poly1305 from Andrew M.
3: * poly1305-donna-unrolled.c from https://github.com/floodyberry/poly1305-donna
4: */
5:
1.2 ! djm 6: /* $OpenBSD$ */
! 7:
1.1 djm 8: #include <sys/types.h>
9: #include <stdint.h>
10:
11: #include "poly1305.h"
12:
13: #define mul32x32_64(a,b) ((uint64_t)(a) * (b))
14:
15: #define U8TO32_LE(p) \
16: (((uint32_t)((p)[0])) | \
17: ((uint32_t)((p)[1]) << 8) | \
18: ((uint32_t)((p)[2]) << 16) | \
19: ((uint32_t)((p)[3]) << 24))
20:
21: #define U32TO8_LE(p, v) \
22: do { \
23: (p)[0] = (uint8_t)((v)); \
24: (p)[1] = (uint8_t)((v) >> 8); \
25: (p)[2] = (uint8_t)((v) >> 16); \
26: (p)[3] = (uint8_t)((v) >> 24); \
27: } while (0)
28:
29: void
30: poly1305_auth(unsigned char out[POLY1305_TAGLEN], const unsigned char *m, size_t inlen, const unsigned char key[POLY1305_KEYLEN]) {
31: uint32_t t0,t1,t2,t3;
32: uint32_t h0,h1,h2,h3,h4;
33: uint32_t r0,r1,r2,r3,r4;
34: uint32_t s1,s2,s3,s4;
35: uint32_t b, nb;
36: size_t j;
37: uint64_t t[5];
38: uint64_t f0,f1,f2,f3;
39: uint32_t g0,g1,g2,g3,g4;
40: uint64_t c;
41: unsigned char mp[16];
42:
43: /* clamp key */
44: t0 = U8TO32_LE(key+0);
45: t1 = U8TO32_LE(key+4);
46: t2 = U8TO32_LE(key+8);
47: t3 = U8TO32_LE(key+12);
48:
49: /* precompute multipliers */
50: r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
51: r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
52: r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
53: r3 = t2 & 0x3f03fff; t3 >>= 8;
54: r4 = t3 & 0x00fffff;
55:
56: s1 = r1 * 5;
57: s2 = r2 * 5;
58: s3 = r3 * 5;
59: s4 = r4 * 5;
60:
61: /* init state */
62: h0 = 0;
63: h1 = 0;
64: h2 = 0;
65: h3 = 0;
66: h4 = 0;
67:
68: /* full blocks */
69: if (inlen < 16) goto poly1305_donna_atmost15bytes;
70: poly1305_donna_16bytes:
71: m += 16;
72: inlen -= 16;
73:
74: t0 = U8TO32_LE(m-16);
75: t1 = U8TO32_LE(m-12);
76: t2 = U8TO32_LE(m-8);
77: t3 = U8TO32_LE(m-4);
78:
79: h0 += t0 & 0x3ffffff;
80: h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
81: h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
82: h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
83: h4 += (t3 >> 8) | (1 << 24);
84:
85:
86: poly1305_donna_mul:
87: t[0] = mul32x32_64(h0,r0) + mul32x32_64(h1,s4) + mul32x32_64(h2,s3) + mul32x32_64(h3,s2) + mul32x32_64(h4,s1);
88: t[1] = mul32x32_64(h0,r1) + mul32x32_64(h1,r0) + mul32x32_64(h2,s4) + mul32x32_64(h3,s3) + mul32x32_64(h4,s2);
89: t[2] = mul32x32_64(h0,r2) + mul32x32_64(h1,r1) + mul32x32_64(h2,r0) + mul32x32_64(h3,s4) + mul32x32_64(h4,s3);
90: t[3] = mul32x32_64(h0,r3) + mul32x32_64(h1,r2) + mul32x32_64(h2,r1) + mul32x32_64(h3,r0) + mul32x32_64(h4,s4);
91: t[4] = mul32x32_64(h0,r4) + mul32x32_64(h1,r3) + mul32x32_64(h2,r2) + mul32x32_64(h3,r1) + mul32x32_64(h4,r0);
92:
93: h0 = (uint32_t)t[0] & 0x3ffffff; c = (t[0] >> 26);
94: t[1] += c; h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >> 26);
95: t[2] += b; h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >> 26);
96: t[3] += b; h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >> 26);
97: t[4] += b; h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >> 26);
98: h0 += b * 5;
99:
100: if (inlen >= 16) goto poly1305_donna_16bytes;
101:
102: /* final bytes */
103: poly1305_donna_atmost15bytes:
104: if (!inlen) goto poly1305_donna_finish;
105:
106: for (j = 0; j < inlen; j++) mp[j] = m[j];
107: mp[j++] = 1;
108: for (; j < 16; j++) mp[j] = 0;
109: inlen = 0;
110:
111: t0 = U8TO32_LE(mp+0);
112: t1 = U8TO32_LE(mp+4);
113: t2 = U8TO32_LE(mp+8);
114: t3 = U8TO32_LE(mp+12);
115:
116: h0 += t0 & 0x3ffffff;
117: h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
118: h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
119: h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
120: h4 += (t3 >> 8);
121:
122: goto poly1305_donna_mul;
123:
124: poly1305_donna_finish:
125: b = h0 >> 26; h0 = h0 & 0x3ffffff;
126: h1 += b; b = h1 >> 26; h1 = h1 & 0x3ffffff;
127: h2 += b; b = h2 >> 26; h2 = h2 & 0x3ffffff;
128: h3 += b; b = h3 >> 26; h3 = h3 & 0x3ffffff;
129: h4 += b; b = h4 >> 26; h4 = h4 & 0x3ffffff;
130: h0 += b * 5; b = h0 >> 26; h0 = h0 & 0x3ffffff;
131: h1 += b;
132:
133: g0 = h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
134: g1 = h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
135: g2 = h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
136: g3 = h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
137: g4 = h4 + b - (1 << 26);
138:
139: b = (g4 >> 31) - 1;
140: nb = ~b;
141: h0 = (h0 & nb) | (g0 & b);
142: h1 = (h1 & nb) | (g1 & b);
143: h2 = (h2 & nb) | (g2 & b);
144: h3 = (h3 & nb) | (g3 & b);
145: h4 = (h4 & nb) | (g4 & b);
146:
147: f0 = ((h0 ) | (h1 << 26)) + (uint64_t)U8TO32_LE(&key[16]);
148: f1 = ((h1 >> 6) | (h2 << 20)) + (uint64_t)U8TO32_LE(&key[20]);
149: f2 = ((h2 >> 12) | (h3 << 14)) + (uint64_t)U8TO32_LE(&key[24]);
150: f3 = ((h3 >> 18) | (h4 << 8)) + (uint64_t)U8TO32_LE(&key[28]);
151:
152: U32TO8_LE(&out[ 0], f0); f1 += (f0 >> 32);
153: U32TO8_LE(&out[ 4], f1); f2 += (f1 >> 32);
154: U32TO8_LE(&out[ 8], f2); f3 += (f2 >> 32);
155: U32TO8_LE(&out[12], f3);
156: }