/src/botan/src/lib/stream/chacha/chacha.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * ChaCha |
3 | | * (C) 2014,2018,2023 Jack Lloyd |
4 | | * |
5 | | * Botan is released under the Simplified BSD License (see license.txt) |
6 | | */ |
7 | | |
8 | | #include <botan/internal/chacha.h> |
9 | | |
10 | | #include <botan/exceptn.h> |
11 | | #include <botan/internal/fmt.h> |
12 | | #include <botan/internal/loadstor.h> |
13 | | #include <botan/internal/rotate.h> |
14 | | |
15 | | #if defined(BOTAN_HAS_CPUID) |
16 | | #include <botan/internal/cpuid.h> |
17 | | #endif |
18 | | |
19 | | namespace Botan { |
20 | | |
21 | | namespace { |
22 | | |
23 | 0 | inline void chacha_quarter_round(uint32_t& a, uint32_t& b, uint32_t& c, uint32_t& d) { |
24 | 0 | a += b; |
25 | 0 | d ^= a; |
26 | 0 | d = rotl<16>(d); |
27 | 0 | c += d; |
28 | 0 | b ^= c; |
29 | 0 | b = rotl<12>(b); |
30 | 0 | a += b; |
31 | 0 | d ^= a; |
32 | 0 | d = rotl<8>(d); |
33 | 0 | c += d; |
34 | 0 | b ^= c; |
35 | 0 | b = rotl<7>(b); |
36 | 0 | } |
37 | | |
38 | | /* |
39 | | * Generate HChaCha cipher stream (for XChaCha IV setup) |
40 | | */ |
41 | 0 | void hchacha(uint32_t output[8], const uint32_t input[16], size_t rounds) { |
42 | 0 | BOTAN_ASSERT(rounds % 2 == 0, "Valid rounds"); |
43 | |
|
44 | 0 | uint32_t x00 = input[0], x01 = input[1], x02 = input[2], x03 = input[3], x04 = input[4], x05 = input[5], |
45 | 0 | x06 = input[6], x07 = input[7], x08 = input[8], x09 = input[9], x10 = input[10], x11 = input[11], |
46 | 0 | x12 = input[12], x13 = input[13], x14 = input[14], x15 = input[15]; |
47 | |
|
48 | 0 | for(size_t i = 0; i != rounds / 2; ++i) { |
49 | 0 | chacha_quarter_round(x00, x04, x08, x12); |
50 | 0 | chacha_quarter_round(x01, x05, x09, x13); |
51 | 0 | chacha_quarter_round(x02, x06, x10, x14); |
52 | 0 | chacha_quarter_round(x03, x07, x11, x15); |
53 | |
|
54 | 0 | chacha_quarter_round(x00, x05, x10, x15); |
55 | 0 | chacha_quarter_round(x01, x06, x11, x12); |
56 | 0 | chacha_quarter_round(x02, x07, x08, x13); |
57 | 0 | chacha_quarter_round(x03, x04, x09, x14); |
58 | 0 | } |
59 | |
|
60 | 0 | output[0] = x00; |
61 | 0 | output[1] = x01; |
62 | 0 | output[2] = x02; |
63 | 0 | output[3] = x03; |
64 | 0 | output[4] = x12; |
65 | 0 | output[5] = x13; |
66 | 0 | output[6] = x14; |
67 | 0 | output[7] = x15; |
68 | 0 | } |
69 | | |
70 | | } // namespace |
71 | | |
72 | 7 | ChaCha::ChaCha(size_t rounds) : m_rounds(rounds) { |
73 | 7 | BOTAN_ARG_CHECK(m_rounds == 8 || m_rounds == 12 || m_rounds == 20, "ChaCha only supports 8, 12 or 20 rounds"); |
74 | 7 | } |
75 | | |
76 | 14 | size_t ChaCha::parallelism() { |
77 | 14 | #if defined(BOTAN_HAS_CHACHA_AVX512) |
78 | 14 | if(CPUID::has(CPUID::Feature::AVX512)) { |
79 | 0 | return 16; |
80 | 0 | } |
81 | 14 | #endif |
82 | | |
83 | 14 | #if defined(BOTAN_HAS_CHACHA_AVX2) |
84 | 14 | if(CPUID::has(CPUID::Feature::AVX2)) { |
85 | 14 | return 8; |
86 | 14 | } |
87 | 0 | #endif |
88 | | |
89 | 0 | return 4; |
90 | 14 | } |
91 | | |
92 | 0 | std::string ChaCha::provider() const { |
93 | 0 | #if defined(BOTAN_HAS_CHACHA_AVX512) |
94 | 0 | if(CPUID::has(CPUID::Feature::AVX512)) { |
95 | 0 | return "avx512"; |
96 | 0 | } |
97 | 0 | #endif |
98 | | |
99 | 0 | #if defined(BOTAN_HAS_CHACHA_AVX2) |
100 | 0 | if(CPUID::has(CPUID::Feature::AVX2)) { |
101 | 0 | return "avx2"; |
102 | 0 | } |
103 | 0 | #endif |
104 | | |
105 | 0 | #if defined(BOTAN_HAS_CHACHA_SIMD32) |
106 | 0 | if(CPUID::has_simd_4x32()) { |
107 | 0 | return "simd32"; |
108 | 0 | } |
109 | 0 | #endif |
110 | | |
111 | 0 | return "base"; |
112 | 0 | } |
113 | | |
114 | 12.7k | void ChaCha::chacha(uint8_t output[], size_t output_blocks, uint32_t state[16], size_t rounds) { |
115 | 12.7k | BOTAN_ASSERT(rounds % 2 == 0, "Valid rounds"); |
116 | | |
117 | 12.7k | #if defined(BOTAN_HAS_CHACHA_AVX512) |
118 | 12.7k | if(CPUID::has(CPUID::Feature::AVX512)) { |
119 | 0 | while(output_blocks >= 16) { |
120 | 0 | ChaCha::chacha_avx512_x16(output, state, rounds); |
121 | 0 | output += 16 * 64; |
122 | 0 | output_blocks -= 16; |
123 | 0 | } |
124 | 0 | } |
125 | 12.7k | #endif |
126 | | |
127 | 12.7k | #if defined(BOTAN_HAS_CHACHA_AVX2) |
128 | 12.7k | if(CPUID::has(CPUID::Feature::AVX2)) { |
129 | 25.5k | while(output_blocks >= 8) { |
130 | 12.7k | ChaCha::chacha_avx2_x8(output, state, rounds); |
131 | 12.7k | output += 8 * 64; |
132 | 12.7k | output_blocks -= 8; |
133 | 12.7k | } |
134 | 12.7k | } |
135 | 12.7k | #endif |
136 | | |
137 | 12.7k | #if defined(BOTAN_HAS_CHACHA_SIMD32) |
138 | 12.7k | if(CPUID::has_simd_4x32()) { |
139 | 12.7k | while(output_blocks >= 4) { |
140 | 0 | ChaCha::chacha_simd32_x4(output, state, rounds); |
141 | 0 | output += 4 * 64; |
142 | 0 | output_blocks -= 4; |
143 | 0 | } |
144 | 12.7k | } |
145 | 12.7k | #endif |
146 | | |
147 | | // TODO interleave rounds |
148 | 12.7k | for(size_t i = 0; i != output_blocks; ++i) { |
149 | 0 | uint32_t x00 = state[0], x01 = state[1], x02 = state[2], x03 = state[3], x04 = state[4], x05 = state[5], |
150 | 0 | x06 = state[6], x07 = state[7], x08 = state[8], x09 = state[9], x10 = state[10], x11 = state[11], |
151 | 0 | x12 = state[12], x13 = state[13], x14 = state[14], x15 = state[15]; |
152 | |
|
153 | 0 | for(size_t r = 0; r != rounds / 2; ++r) { |
154 | 0 | chacha_quarter_round(x00, x04, x08, x12); |
155 | 0 | chacha_quarter_round(x01, x05, x09, x13); |
156 | 0 | chacha_quarter_round(x02, x06, x10, x14); |
157 | 0 | chacha_quarter_round(x03, x07, x11, x15); |
158 | |
|
159 | 0 | chacha_quarter_round(x00, x05, x10, x15); |
160 | 0 | chacha_quarter_round(x01, x06, x11, x12); |
161 | 0 | chacha_quarter_round(x02, x07, x08, x13); |
162 | 0 | chacha_quarter_round(x03, x04, x09, x14); |
163 | 0 | } |
164 | |
|
165 | 0 | x00 += state[0]; |
166 | 0 | x01 += state[1]; |
167 | 0 | x02 += state[2]; |
168 | 0 | x03 += state[3]; |
169 | 0 | x04 += state[4]; |
170 | 0 | x05 += state[5]; |
171 | 0 | x06 += state[6]; |
172 | 0 | x07 += state[7]; |
173 | 0 | x08 += state[8]; |
174 | 0 | x09 += state[9]; |
175 | 0 | x10 += state[10]; |
176 | 0 | x11 += state[11]; |
177 | 0 | x12 += state[12]; |
178 | 0 | x13 += state[13]; |
179 | 0 | x14 += state[14]; |
180 | 0 | x15 += state[15]; |
181 | |
|
182 | 0 | store_le(x00, output + 64 * i + 4 * 0); |
183 | 0 | store_le(x01, output + 64 * i + 4 * 1); |
184 | 0 | store_le(x02, output + 64 * i + 4 * 2); |
185 | 0 | store_le(x03, output + 64 * i + 4 * 3); |
186 | 0 | store_le(x04, output + 64 * i + 4 * 4); |
187 | 0 | store_le(x05, output + 64 * i + 4 * 5); |
188 | 0 | store_le(x06, output + 64 * i + 4 * 6); |
189 | 0 | store_le(x07, output + 64 * i + 4 * 7); |
190 | 0 | store_le(x08, output + 64 * i + 4 * 8); |
191 | 0 | store_le(x09, output + 64 * i + 4 * 9); |
192 | 0 | store_le(x10, output + 64 * i + 4 * 10); |
193 | 0 | store_le(x11, output + 64 * i + 4 * 11); |
194 | 0 | store_le(x12, output + 64 * i + 4 * 12); |
195 | 0 | store_le(x13, output + 64 * i + 4 * 13); |
196 | 0 | store_le(x14, output + 64 * i + 4 * 14); |
197 | 0 | store_le(x15, output + 64 * i + 4 * 15); |
198 | |
|
199 | 0 | state[12]++; |
200 | 0 | state[13] += (state[12] == 0); |
201 | 0 | } |
202 | 12.7k | } |
203 | | |
204 | | /* |
205 | | * Combine cipher stream with message |
206 | | */ |
207 | 0 | void ChaCha::cipher_bytes(const uint8_t in[], uint8_t out[], size_t length) { |
208 | 0 | assert_key_material_set(); |
209 | |
|
210 | 0 | while(length >= m_buffer.size() - m_position) { |
211 | 0 | const size_t available = m_buffer.size() - m_position; |
212 | |
|
213 | 0 | xor_buf(out, in, &m_buffer[m_position], available); |
214 | 0 | chacha(m_buffer.data(), m_buffer.size() / 64, m_state.data(), m_rounds); |
215 | |
|
216 | 0 | length -= available; |
217 | 0 | in += available; |
218 | 0 | out += available; |
219 | 0 | m_position = 0; |
220 | 0 | } |
221 | |
|
222 | 0 | xor_buf(out, in, &m_buffer[m_position], length); |
223 | |
|
224 | 0 | m_position += length; |
225 | 0 | } |
226 | | |
227 | 165k | void ChaCha::generate_keystream(uint8_t out[], size_t length) { |
228 | 165k | assert_key_material_set(); |
229 | | |
230 | 178k | while(length >= m_buffer.size() - m_position) { |
231 | 12.7k | const size_t available = m_buffer.size() - m_position; |
232 | | |
233 | | // TODO: this could write directly to the output buffer |
234 | | // instead of bouncing it through m_buffer first |
235 | 12.7k | copy_mem(out, &m_buffer[m_position], available); |
236 | 12.7k | chacha(m_buffer.data(), m_buffer.size() / 64, m_state.data(), m_rounds); |
237 | | |
238 | 12.7k | length -= available; |
239 | 12.7k | out += available; |
240 | 12.7k | m_position = 0; |
241 | 12.7k | } |
242 | | |
243 | 165k | copy_mem(out, &m_buffer[m_position], length); |
244 | | |
245 | 165k | m_position += length; |
246 | 165k | } |
247 | | |
248 | 14 | void ChaCha::initialize_state() { |
249 | 14 | static const uint32_t TAU[] = {0x61707865, 0x3120646e, 0x79622d36, 0x6b206574}; |
250 | | |
251 | 14 | static const uint32_t SIGMA[] = {0x61707865, 0x3320646e, 0x79622d32, 0x6b206574}; |
252 | | |
253 | 14 | m_state[4] = m_key[0]; |
254 | 14 | m_state[5] = m_key[1]; |
255 | 14 | m_state[6] = m_key[2]; |
256 | 14 | m_state[7] = m_key[3]; |
257 | | |
258 | 14 | if(m_key.size() == 4) { |
259 | 0 | m_state[0] = TAU[0]; |
260 | 0 | m_state[1] = TAU[1]; |
261 | 0 | m_state[2] = TAU[2]; |
262 | 0 | m_state[3] = TAU[3]; |
263 | |
|
264 | 0 | m_state[8] = m_key[0]; |
265 | 0 | m_state[9] = m_key[1]; |
266 | 0 | m_state[10] = m_key[2]; |
267 | 0 | m_state[11] = m_key[3]; |
268 | 14 | } else { |
269 | 14 | m_state[0] = SIGMA[0]; |
270 | 14 | m_state[1] = SIGMA[1]; |
271 | 14 | m_state[2] = SIGMA[2]; |
272 | 14 | m_state[3] = SIGMA[3]; |
273 | | |
274 | 14 | m_state[8] = m_key[4]; |
275 | 14 | m_state[9] = m_key[5]; |
276 | 14 | m_state[10] = m_key[6]; |
277 | 14 | m_state[11] = m_key[7]; |
278 | 14 | } |
279 | | |
280 | 14 | m_state[12] = 0; |
281 | 14 | m_state[13] = 0; |
282 | 14 | m_state[14] = 0; |
283 | 14 | m_state[15] = 0; |
284 | | |
285 | 14 | m_position = 0; |
286 | 14 | } |
287 | | |
288 | 165k | bool ChaCha::has_keying_material() const { |
289 | 165k | return !m_state.empty(); |
290 | 165k | } |
291 | | |
292 | 0 | size_t ChaCha::buffer_size() const { |
293 | 0 | return 64; |
294 | 0 | } |
295 | | |
296 | | /* |
297 | | * ChaCha Key Schedule |
298 | | */ |
299 | 14 | void ChaCha::key_schedule(std::span<const uint8_t> key) { |
300 | 14 | m_key.resize(key.size() / 4); |
301 | 14 | load_le<uint32_t>(m_key.data(), key.data(), m_key.size()); |
302 | | |
303 | 14 | m_state.resize(16); |
304 | | |
305 | 14 | const size_t chacha_block = 64; |
306 | 14 | m_buffer.resize(parallelism() * chacha_block); |
307 | | |
308 | 14 | set_iv(nullptr, 0); |
309 | 14 | } |
310 | | |
311 | 0 | size_t ChaCha::default_iv_length() const { |
312 | 0 | return 24; |
313 | 0 | } |
314 | | |
315 | 14 | Key_Length_Specification ChaCha::key_spec() const { |
316 | 14 | return Key_Length_Specification(16, 32, 16); |
317 | 14 | } |
318 | | |
319 | 0 | std::unique_ptr<StreamCipher> ChaCha::new_object() const { |
320 | 0 | return std::make_unique<ChaCha>(m_rounds); |
321 | 0 | } |
322 | | |
323 | 14 | bool ChaCha::valid_iv_length(size_t iv_len) const { |
324 | 14 | return (iv_len == 0 || iv_len == 8 || iv_len == 12 || iv_len == 24); |
325 | 14 | } |
326 | | |
327 | 14 | void ChaCha::set_iv_bytes(const uint8_t iv[], size_t length) { |
328 | 14 | assert_key_material_set(); |
329 | | |
330 | 14 | if(!valid_iv_length(length)) { |
331 | 0 | throw Invalid_IV_Length(name(), length); |
332 | 0 | } |
333 | | |
334 | 14 | initialize_state(); |
335 | | |
336 | 14 | if(length == 0) { |
337 | | // Treat zero length IV same as an all-zero IV |
338 | 14 | m_state[14] = 0; |
339 | 14 | m_state[15] = 0; |
340 | 14 | } else if(length == 8) { |
341 | 0 | m_state[14] = load_le<uint32_t>(iv, 0); |
342 | 0 | m_state[15] = load_le<uint32_t>(iv, 1); |
343 | 0 | } else if(length == 12) { |
344 | 0 | m_state[13] = load_le<uint32_t>(iv, 0); |
345 | 0 | m_state[14] = load_le<uint32_t>(iv, 1); |
346 | 0 | m_state[15] = load_le<uint32_t>(iv, 2); |
347 | 0 | } else if(length == 24) { |
348 | 0 | m_state[12] = load_le<uint32_t>(iv, 0); |
349 | 0 | m_state[13] = load_le<uint32_t>(iv, 1); |
350 | 0 | m_state[14] = load_le<uint32_t>(iv, 2); |
351 | 0 | m_state[15] = load_le<uint32_t>(iv, 3); |
352 | |
|
353 | 0 | secure_vector<uint32_t> hc(8); |
354 | 0 | hchacha(hc.data(), m_state.data(), m_rounds); |
355 | |
|
356 | 0 | m_state[4] = hc[0]; |
357 | 0 | m_state[5] = hc[1]; |
358 | 0 | m_state[6] = hc[2]; |
359 | 0 | m_state[7] = hc[3]; |
360 | 0 | m_state[8] = hc[4]; |
361 | 0 | m_state[9] = hc[5]; |
362 | 0 | m_state[10] = hc[6]; |
363 | 0 | m_state[11] = hc[7]; |
364 | 0 | m_state[12] = 0; |
365 | 0 | m_state[13] = 0; |
366 | 0 | m_state[14] = load_le<uint32_t>(iv, 4); |
367 | 0 | m_state[15] = load_le<uint32_t>(iv, 5); |
368 | 0 | } |
369 | | |
370 | 14 | chacha(m_buffer.data(), m_buffer.size() / 64, m_state.data(), m_rounds); |
371 | 14 | m_position = 0; |
372 | 14 | } |
373 | | |
374 | 0 | void ChaCha::clear() { |
375 | 0 | zap(m_key); |
376 | 0 | zap(m_state); |
377 | 0 | zap(m_buffer); |
378 | 0 | m_position = 0; |
379 | 0 | } |
380 | | |
381 | 0 | std::string ChaCha::name() const { |
382 | 0 | return fmt("ChaCha({})", m_rounds); |
383 | 0 | } |
384 | | |
385 | 0 | void ChaCha::seek(uint64_t offset) { |
386 | 0 | assert_key_material_set(); |
387 | | |
388 | | // Find the block offset |
389 | 0 | const uint64_t counter = offset / 64; |
390 | |
|
391 | 0 | uint8_t out[8]; |
392 | |
|
393 | 0 | store_le(counter, out); |
394 | |
|
395 | 0 | m_state[12] = load_le<uint32_t>(out, 0); |
396 | 0 | m_state[13] += load_le<uint32_t>(out, 1); |
397 | |
|
398 | 0 | chacha(m_buffer.data(), m_buffer.size() / 64, m_state.data(), m_rounds); |
399 | 0 | m_position = offset % 64; |
400 | 0 | } |
401 | | } // namespace Botan |