1/*	$NetBSD: poly1305.c,v 1.6 2023/10/25 20:19:57 christos Exp $	*/
2/* $OpenBSD: poly1305.c,v 1.4 2023/07/17 05:26:38 djm Exp $ */
3/*
4 * Public Domain poly1305 from Andrew Moon
5 * poly1305-donna-unrolled.c from https://github.com/floodyberry/poly1305-donna
6 */
7
8#include "includes.h"
9__RCSID("$NetBSD: poly1305.c,v 1.6 2023/10/25 20:19:57 christos Exp $");
10
11#include <sys/types.h>
12#include <stdint.h>
13
14#include "poly1305.h"
15
16#define mul32x32_64(a,b) ((uint64_t)(a) * (b))
17
18#define U8TO32_LE(p) \
19	(((uint32_t)((p)[0])) | \
20	 ((uint32_t)((p)[1]) <<  8) | \
21	 ((uint32_t)((p)[2]) << 16) | \
22	 ((uint32_t)((p)[3]) << 24))
23
24#define U32TO8_LE(p, v) \
25	do { \
26		(p)[0] = (uint8_t)((v)); \
27		(p)[1] = (uint8_t)((v) >>  8); \
28		(p)[2] = (uint8_t)((v) >> 16); \
29		(p)[3] = (uint8_t)((v) >> 24); \
30	} while (0)
31
32void
33poly1305_auth(unsigned char out[POLY1305_TAGLEN], const unsigned char *m, size_t inlen, const unsigned char key[POLY1305_KEYLEN]) {
34	uint32_t t0,t1,t2,t3;
35	uint32_t h0,h1,h2,h3,h4;
36	uint32_t r0,r1,r2,r3,r4;
37	uint32_t s1,s2,s3,s4;
38	uint32_t b, nb;
39	size_t j;
40	uint64_t t[5];
41	uint64_t f0,f1,f2,f3;
42	uint32_t g0,g1,g2,g3,g4;
43	uint64_t c;
44	unsigned char mp[16];
45
46	/* clamp key */
47	t0 = U8TO32_LE(key+0);
48	t1 = U8TO32_LE(key+4);
49	t2 = U8TO32_LE(key+8);
50	t3 = U8TO32_LE(key+12);
51
52	/* precompute multipliers */
53	r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6;
54	r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12;
55	r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18;
56	r3 = t2 & 0x3f03fff; t3 >>= 8;
57	r4 = t3 & 0x00fffff;
58
59	s1 = r1 * 5;
60	s2 = r2 * 5;
61	s3 = r3 * 5;
62	s4 = r4 * 5;
63
64	/* init state */
65	h0 = 0;
66	h1 = 0;
67	h2 = 0;
68	h3 = 0;
69	h4 = 0;
70
71	/* full blocks */
72	if (inlen < 16) goto poly1305_donna_atmost15bytes;
73poly1305_donna_16bytes:
74	m += 16;
75	inlen -= 16;
76
77	t0 = U8TO32_LE(m-16);
78	t1 = U8TO32_LE(m-12);
79	t2 = U8TO32_LE(m-8);
80	t3 = U8TO32_LE(m-4);
81
82	h0 += t0 & 0x3ffffff;
83	h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
84	h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
85	h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
86	h4 += (t3 >> 8) | (1 << 24);
87
88
89poly1305_donna_mul:
90	t[0]  = mul32x32_64(h0,r0) + mul32x32_64(h1,s4) + mul32x32_64(h2,s3) + mul32x32_64(h3,s2) + mul32x32_64(h4,s1);
91	t[1]  = mul32x32_64(h0,r1) + mul32x32_64(h1,r0) + mul32x32_64(h2,s4) + mul32x32_64(h3,s3) + mul32x32_64(h4,s2);
92	t[2]  = mul32x32_64(h0,r2) + mul32x32_64(h1,r1) + mul32x32_64(h2,r0) + mul32x32_64(h3,s4) + mul32x32_64(h4,s3);
93	t[3]  = mul32x32_64(h0,r3) + mul32x32_64(h1,r2) + mul32x32_64(h2,r1) + mul32x32_64(h3,r0) + mul32x32_64(h4,s4);
94	t[4]  = mul32x32_64(h0,r4) + mul32x32_64(h1,r3) + mul32x32_64(h2,r2) + mul32x32_64(h3,r1) + mul32x32_64(h4,r0);
95
96	                h0 = (uint32_t)t[0] & 0x3ffffff; c =           (t[0] >> 26);
97	t[1] += c;      h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >> 26);
98	t[2] += b;      h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >> 26);
99	t[3] += b;      h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >> 26);
100	t[4] += b;      h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >> 26);
101	h0 += b * 5;
102
103	if (inlen >= 16) goto poly1305_donna_16bytes;
104
105	/* final bytes */
106poly1305_donna_atmost15bytes:
107	if (!inlen) goto poly1305_donna_finish;
108
109	for (j = 0; j < inlen; j++) mp[j] = m[j];
110	mp[j++] = 1;
111	for (; j < 16; j++)	mp[j] = 0;
112	inlen = 0;
113
114	t0 = U8TO32_LE(mp+0);
115	t1 = U8TO32_LE(mp+4);
116	t2 = U8TO32_LE(mp+8);
117	t3 = U8TO32_LE(mp+12);
118
119	h0 += t0 & 0x3ffffff;
120	h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff;
121	h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff;
122	h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff;
123	h4 += (t3 >> 8);
124
125	goto poly1305_donna_mul;
126
127poly1305_donna_finish:
128	             b = h0 >> 26; h0 = h0 & 0x3ffffff;
129	h1 +=     b; b = h1 >> 26; h1 = h1 & 0x3ffffff;
130	h2 +=     b; b = h2 >> 26; h2 = h2 & 0x3ffffff;
131	h3 +=     b; b = h3 >> 26; h3 = h3 & 0x3ffffff;
132	h4 +=     b; b = h4 >> 26; h4 = h4 & 0x3ffffff;
133	h0 += b * 5; b = h0 >> 26; h0 = h0 & 0x3ffffff;
134	h1 +=     b;
135
136	g0 = h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff;
137	g1 = h1 + b; b = g1 >> 26; g1 &= 0x3ffffff;
138	g2 = h2 + b; b = g2 >> 26; g2 &= 0x3ffffff;
139	g3 = h3 + b; b = g3 >> 26; g3 &= 0x3ffffff;
140	g4 = h4 + b - (1 << 26);
141
142	b = (g4 >> 31) - 1;
143	nb = ~b;
144	h0 = (h0 & nb) | (g0 & b);
145	h1 = (h1 & nb) | (g1 & b);
146	h2 = (h2 & nb) | (g2 & b);
147	h3 = (h3 & nb) | (g3 & b);
148	h4 = (h4 & nb) | (g4 & b);
149
150	f0 = ((h0      ) | (h1 << 26)) + (uint64_t)U8TO32_LE(&key[16]);
151	f1 = ((h1 >>  6) | (h2 << 20)) + (uint64_t)U8TO32_LE(&key[20]);
152	f2 = ((h2 >> 12) | (h3 << 14)) + (uint64_t)U8TO32_LE(&key[24]);
153	f3 = ((h3 >> 18) | (h4 <<  8)) + (uint64_t)U8TO32_LE(&key[28]);
154
155	U32TO8_LE(&out[ 0], f0); f1 += (f0 >> 32);
156	U32TO8_LE(&out[ 4], f1); f2 += (f1 >> 32);
157	U32TO8_LE(&out[ 8], f2); f3 += (f2 >> 32);
158	U32TO8_LE(&out[12], f3);
159}
160