/src/botan/src/lib/mac/poly1305/poly1305.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * Derived from poly1305-donna-64.h by Andrew Moon <liquidsun@gmail.com> |
3 | | * in https://github.com/floodyberry/poly1305-donna |
4 | | * |
5 | | * (C) 2014 Andrew Moon |
6 | | * (C) 2014 Jack Lloyd |
7 | | * |
8 | | * Botan is released under the Simplified BSD License (see license.txt) |
9 | | */ |
10 | | |
11 | | #include <botan/internal/poly1305.h> |
12 | | #include <botan/internal/loadstor.h> |
13 | | #include <botan/internal/mul128.h> |
14 | | #include <botan/internal/donna128.h> |
15 | | #include <botan/internal/ct_utils.h> |
16 | | |
17 | | namespace Botan { |
18 | | |
19 | | namespace { |
20 | | |
21 | | void poly1305_init(secure_vector<uint64_t>& X, const uint8_t key[32]) |
22 | 61 | { |
23 | | /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */ |
24 | 61 | const uint64_t t0 = load_le<uint64_t>(key, 0); |
25 | 61 | const uint64_t t1 = load_le<uint64_t>(key, 1); |
26 | | |
27 | 61 | X[0] = ( t0 ) & 0xffc0fffffff; |
28 | 61 | X[1] = ((t0 >> 44) | (t1 << 20)) & 0xfffffc0ffff; |
29 | 61 | X[2] = ((t1 >> 24) ) & 0x00ffffffc0f; |
30 | | |
31 | | /* h = 0 */ |
32 | 61 | X[3] = 0; |
33 | 61 | X[4] = 0; |
34 | 61 | X[5] = 0; |
35 | | |
36 | | /* save pad for later */ |
37 | 61 | X[6] = load_le<uint64_t>(key, 2); |
38 | 61 | X[7] = load_le<uint64_t>(key, 3); |
39 | 61 | } |
40 | | |
41 | | void poly1305_blocks(secure_vector<uint64_t>& X, const uint8_t *m, size_t blocks, bool is_final = false) |
42 | 213 | { |
43 | | #if !defined(BOTAN_TARGET_HAS_NATIVE_UINT128) |
44 | | typedef donna128 uint128_t; |
45 | | #endif |
46 | | |
47 | 213 | const uint64_t hibit = is_final ? 0 : (static_cast<uint64_t>(1) << 40); /* 1 << 128 */ |
48 | | |
49 | 213 | const uint64_t r0 = X[0]; |
50 | 213 | const uint64_t r1 = X[1]; |
51 | 213 | const uint64_t r2 = X[2]; |
52 | | |
53 | 213 | const uint64_t M44 = 0xFFFFFFFFFFF; |
54 | 213 | const uint64_t M42 = 0x3FFFFFFFFFF; |
55 | | |
56 | 213 | uint64_t h0 = X[3+0]; |
57 | 213 | uint64_t h1 = X[3+1]; |
58 | 213 | uint64_t h2 = X[3+2]; |
59 | | |
60 | 213 | const uint64_t s1 = r1 * 20; |
61 | 213 | const uint64_t s2 = r2 * 20; |
62 | | |
63 | 2.46k | for(size_t i = 0; i != blocks; ++i) |
64 | 2.25k | { |
65 | 2.25k | const uint64_t t0 = load_le<uint64_t>(m, 0); |
66 | 2.25k | const uint64_t t1 = load_le<uint64_t>(m, 1); |
67 | | |
68 | 2.25k | h0 += (( t0 ) & M44); |
69 | 2.25k | h1 += (((t0 >> 44) | (t1 << 20)) & M44); |
70 | 2.25k | h2 += (((t1 >> 24) ) & M42) | hibit; |
71 | | |
72 | 2.25k | const uint128_t d0 = uint128_t(h0) * r0 + uint128_t(h1) * s2 + uint128_t(h2) * s1; |
73 | 2.25k | const uint64_t c0 = carry_shift(d0, 44); |
74 | | |
75 | 2.25k | const uint128_t d1 = uint128_t(h0) * r1 + uint128_t(h1) * r0 + uint128_t(h2) * s2 + c0; |
76 | 2.25k | const uint64_t c1 = carry_shift(d1, 44); |
77 | | |
78 | 2.25k | const uint128_t d2 = uint128_t(h0) * r2 + uint128_t(h1) * r1 + uint128_t(h2) * r0 + c1; |
79 | 2.25k | const uint64_t c2 = carry_shift(d2, 42); |
80 | | |
81 | 2.25k | h0 = d0 & M44; |
82 | 2.25k | h1 = d1 & M44; |
83 | 2.25k | h2 = d2 & M42; |
84 | | |
85 | 2.25k | h0 += c2 * 5; |
86 | 2.25k | h1 += carry_shift(h0, 44); |
87 | 2.25k | h0 = h0 & M44; |
88 | | |
89 | 2.25k | m += 16; |
90 | 2.25k | } |
91 | | |
92 | 213 | X[3+0] = h0; |
93 | 213 | X[3+1] = h1; |
94 | 213 | X[3+2] = h2; |
95 | 213 | } |
96 | | |
97 | | void poly1305_finish(secure_vector<uint64_t>& X, uint8_t mac[16]) |
98 | 61 | { |
99 | 61 | const uint64_t M44 = 0xFFFFFFFFFFF; |
100 | 61 | const uint64_t M42 = 0x3FFFFFFFFFF; |
101 | | |
102 | | /* fully carry h */ |
103 | 61 | uint64_t h0 = X[3+0]; |
104 | 61 | uint64_t h1 = X[3+1]; |
105 | 61 | uint64_t h2 = X[3+2]; |
106 | | |
107 | 61 | uint64_t c; |
108 | 61 | c = (h1 >> 44); h1 &= M44; |
109 | 61 | h2 += c; c = (h2 >> 42); h2 &= M42; |
110 | 61 | h0 += c * 5; c = (h0 >> 44); h0 &= M44; |
111 | 61 | h1 += c; c = (h1 >> 44); h1 &= M44; |
112 | 61 | h2 += c; c = (h2 >> 42); h2 &= M42; |
113 | 61 | h0 += c * 5; c = (h0 >> 44); h0 &= M44; |
114 | 61 | h1 += c; |
115 | | |
116 | | /* compute h + -p */ |
117 | 61 | uint64_t g0 = h0 + 5; c = (g0 >> 44); g0 &= M44; |
118 | 61 | uint64_t g1 = h1 + c; c = (g1 >> 44); g1 &= M44; |
119 | 61 | uint64_t g2 = h2 + c - (static_cast<uint64_t>(1) << 42); |
120 | | |
121 | | /* select h if h < p, or h + -p if h >= p */ |
122 | 61 | const auto c_mask = CT::Mask<uint64_t>::expand(c); |
123 | 61 | h0 = c_mask.select(g0, h0); |
124 | 61 | h1 = c_mask.select(g1, h1); |
125 | 61 | h2 = c_mask.select(g2, h2); |
126 | | |
127 | | /* h = (h + pad) */ |
128 | 61 | const uint64_t t0 = X[6]; |
129 | 61 | const uint64_t t1 = X[7]; |
130 | | |
131 | 61 | h0 += (( t0 ) & M44) ; c = (h0 >> 44); h0 &= M44; |
132 | 61 | h1 += (((t0 >> 44) | (t1 << 20)) & M44) + c; c = (h1 >> 44); h1 &= M44; |
133 | 61 | h2 += (((t1 >> 24) ) & M42) + c; h2 &= M42; |
134 | | |
135 | | /* mac = h % (2^128) */ |
136 | 61 | h0 = ((h0 ) | (h1 << 44)); |
137 | 61 | h1 = ((h1 >> 20) | (h2 << 24)); |
138 | | |
139 | 61 | store_le(mac, h0, h1); |
140 | | |
141 | | /* zero out the state */ |
142 | 61 | clear_mem(X.data(), X.size()); |
143 | 61 | } |
144 | | |
145 | | } |
146 | | |
147 | | void Poly1305::clear() |
148 | 0 | { |
149 | 0 | zap(m_poly); |
150 | 0 | zap(m_buf); |
151 | 0 | m_buf_pos = 0; |
152 | 0 | } |
153 | | |
154 | | void Poly1305::key_schedule(const uint8_t key[], size_t /*length*/) |
155 | 61 | { |
156 | 61 | m_buf_pos = 0; |
157 | 61 | m_buf.resize(16); |
158 | 61 | m_poly.resize(8); |
159 | | |
160 | 61 | poly1305_init(m_poly, key); |
161 | 61 | } |
162 | | |
163 | | void Poly1305::add_data(const uint8_t input[], size_t length) |
164 | 349 | { |
165 | 349 | verify_key_set(m_poly.size() == 8); |
166 | | |
167 | 349 | if(m_buf_pos) |
168 | 167 | { |
169 | 167 | buffer_insert(m_buf, m_buf_pos, input, length); |
170 | | |
171 | 167 | if(m_buf_pos + length >= m_buf.size()) |
172 | 167 | { |
173 | 167 | poly1305_blocks(m_poly, m_buf.data(), 1); |
174 | 167 | input += (m_buf.size() - m_buf_pos); |
175 | 167 | length -= (m_buf.size() - m_buf_pos); |
176 | 167 | m_buf_pos = 0; |
177 | 167 | } |
178 | 167 | } |
179 | | |
180 | 349 | const size_t full_blocks = length / m_buf.size(); |
181 | 349 | const size_t remaining = length % m_buf.size(); |
182 | | |
183 | 349 | if(full_blocks) |
184 | 46 | poly1305_blocks(m_poly, input, full_blocks); |
185 | | |
186 | 349 | buffer_insert(m_buf, m_buf_pos, input + full_blocks * m_buf.size(), remaining); |
187 | 349 | m_buf_pos += remaining; |
188 | 349 | } |
189 | | |
190 | | void Poly1305::final_result(uint8_t out[]) |
191 | 61 | { |
192 | 61 | verify_key_set(m_poly.size() == 8); |
193 | | |
194 | 61 | if(m_buf_pos != 0) |
195 | 0 | { |
196 | 0 | m_buf[m_buf_pos] = 1; |
197 | 0 | const size_t len = m_buf.size() - m_buf_pos - 1; |
198 | 0 | if (len > 0) |
199 | 0 | { |
200 | 0 | clear_mem(&m_buf[m_buf_pos+1], len); |
201 | 0 | } |
202 | 0 | poly1305_blocks(m_poly, m_buf.data(), 1, true); |
203 | 0 | } |
204 | | |
205 | 61 | poly1305_finish(m_poly, out); |
206 | | |
207 | 61 | m_poly.clear(); |
208 | 61 | m_buf_pos = 0; |
209 | 61 | } |
210 | | |
211 | | } |