Commit | Line | Data |
---|---|---|
194e93f2 MW |
1 | /* -*-c-*- |
2 | * | |
3 | * Salsa20 stream cipher | |
4 | * | |
5 | * (c) 2015 Straylight/Edgeware | |
6 | */ | |
7 | ||
0a847b11 MW |
8 | /*----- Licensing notice --------------------------------------------------* |
9 | * | |
10 | * This file is part of Catacomb. | |
11 | * | |
12 | * Catacomb is free software; you can redistribute it and/or modify | |
13 | * it under the terms of the GNU Library General Public License as | |
14 | * published by the Free Software Foundation; either version 2 of the | |
15 | * License, or (at your option) any later version. | |
16 | * | |
17 | * Catacomb is distributed in the hope that it will be useful, | |
18 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
19 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
20 | * GNU Library General Public License for more details. | |
21 | * | |
22 | * You should have received a copy of the GNU Library General Public | |
23 | * License along with Catacomb; if not, write to the Free | |
24 | * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, | |
25 | * MA 02111-1307, USA. | |
26 | */ | |
27 | ||
194e93f2 MW |
28 | /*----- Header files ------------------------------------------------------*/ |
29 | ||
e10e6494 MW |
30 | #include "config.h" |
31 | ||
194e93f2 MW |
32 | #include <stdarg.h> |
33 | ||
34 | #include <mLib/bits.h> | |
35 | ||
36 | #include "arena.h" | |
e10e6494 | 37 | #include "dispatch.h" |
194e93f2 MW |
38 | #include "gcipher.h" |
39 | #include "grand.h" | |
40 | #include "keysz.h" | |
41 | #include "paranoia.h" | |
42 | #include "salsa20.h" | |
43 | #include "salsa20-core.h" | |
44 | ||
45 | /*----- Global variables --------------------------------------------------*/ | |
46 | ||
47 | const octet salsa20_keysz[] = { KSZ_SET, 32, 16, 10, 0 }; | |
48 | ||
49 | /*----- The Salsa20 core function and utilities ---------------------------*/ | |
50 | ||
51 | /* --- @core@ --- * | |
52 | * | |
53 | * Arguments: @unsigned r@ = number of rounds | |
54 | * @const salsa20_matrix src@ = input matrix | |
55 | * @salsa20_matrix dest@ = where to put the output | |
56 | * | |
57 | * Returns: --- | |
58 | * | |
59 | * | |
60 | * Use: Apply the Salsa20/r core function to @src@, writing the | |
61 | * result to @dest@. This consists of @r@ rounds followed by | |
62 | * the feedforward step. | |
63 | */ | |
64 | ||
8a1aa284 MW |
65 | CPU_DISPATCH(static, (void), void, core, |
66 | (unsigned r, const salsa20_matrix src, salsa20_matrix dest), | |
67 | (r, src, dest), pick_core, simple_core); | |
e10e6494 MW |
68 | |
69 | static void simple_core(unsigned r, const salsa20_matrix src, | |
70 | salsa20_matrix dest) | |
194e93f2 MW |
71 | { SALSA20_nR(dest, src, r); SALSA20_FFWD(dest, src); } |
72 | ||
0f23f75f MW |
73 | #if CPUFAM_X86 || CPUFAM_AMD64 |
74 | extern core__functype salsa20_core_x86ish_sse2; | |
e10e6494 MW |
75 | #endif |
76 | ||
704d59c8 MW |
77 | #if CPUFAM_ARMEL |
78 | extern core__functype salsa20_core_arm_neon; | |
79 | #endif | |
80 | ||
e10e6494 MW |
81 | static core__functype *pick_core(void) |
82 | { | |
0f23f75f MW |
83 | #if CPUFAM_X86 || CPUFAM_AMD64 |
84 | DISPATCH_PICK_COND(salsa20_core, salsa20_core_x86ish_sse2, | |
fac645f7 | 85 | cpu_feature_p(CPUFEAT_X86_SSE2)); |
e10e6494 | 86 | #endif |
704d59c8 MW |
87 | #if CPUFAM_ARMEL |
88 | DISPATCH_PICK_COND(salsa20_core, salsa20_core_arm_neon, | |
89 | cpu_feature_p(CPUFEAT_ARM_NEON)); | |
90 | #endif | |
fac645f7 | 91 | DISPATCH_PICK_FALLBACK(salsa20_core, simple_core); |
e10e6494 MW |
92 | } |
93 | ||
194e93f2 MW |
94 | /* --- @populate@ --- * |
95 | * | |
96 | * Arguments: @salsa20_matrix a@ = a matrix to fill in | |
97 | * @const void *key@ = pointer to key material | |
98 | * @size_t ksz@ = size of key | |
99 | * | |
100 | * Returns: --- | |
101 | * | |
102 | * Use: Fills in a Salsa20 matrix from the key, setting the | |
103 | * appropriate constants according to the key length. The nonce | |
104 | * and position words are left uninitialized. | |
105 | */ | |
106 | ||
107 | static void populate(salsa20_matrix a, const void *key, size_t ksz) | |
108 | { | |
109 | const octet *k = key; | |
110 | ||
111 | KSZ_ASSERT(salsa20, ksz); | |
112 | ||
a4c2e267 MW |
113 | /* Here's the pattern of key, constant, nonce, and counter pieces in the |
114 | * matrix, before and after our permutation. | |
115 | * | |
116 | * [ C0 K0 K1 K2 ] [ C0 C1 C2 C3 ] | |
117 | * [ K3 C1 N0 N1 ] --> [ K3 T1 K7 K2 ] | |
118 | * [ T0 T1 C2 K4 ] [ T0 K6 K1 N1 ] | |
119 | * [ K5 K6 K7 C3 ] [ K5 K0 N0 K4 ] | |
120 | */ | |
121 | ||
122 | a[13] = LOAD32_L(k + 0); | |
123 | a[10] = LOAD32_L(k + 4); | |
194e93f2 | 124 | if (ksz == 10) { |
a4c2e267 | 125 | a[ 7] = LOAD16_L(k + 8); |
194e93f2 MW |
126 | a[ 4] = 0; |
127 | } else { | |
a4c2e267 | 128 | a[ 7] = LOAD32_L(k + 8); |
194e93f2 MW |
129 | a[ 4] = LOAD32_L(k + 12); |
130 | } | |
131 | if (ksz <= 16) { | |
a4c2e267 MW |
132 | a[15] = a[13]; |
133 | a[12] = a[10]; | |
134 | a[ 9] = a[ 7]; | |
135 | a[ 6] = a[ 4]; | |
194e93f2 | 136 | a[ 0] = SALSA20_A128; |
a4c2e267 MW |
137 | a[ 1] = SALSA20_B128; |
138 | a[ 2] = ksz == 10 ? SALSA20_C80 : SALSA20_C128; | |
139 | a[ 3] = SALSA20_D128; | |
194e93f2 | 140 | } else { |
a4c2e267 | 141 | a[15] = LOAD32_L(k + 16); |
194e93f2 | 142 | a[12] = LOAD32_L(k + 20); |
a4c2e267 MW |
143 | a[ 9] = LOAD32_L(k + 24); |
144 | a[ 6] = LOAD32_L(k + 28); | |
194e93f2 | 145 | a[ 0] = SALSA20_A256; |
a4c2e267 MW |
146 | a[ 1] = SALSA20_B256; |
147 | a[ 2] = SALSA20_C256; | |
148 | a[ 3] = SALSA20_D256; | |
194e93f2 MW |
149 | } |
150 | } | |
151 | ||
152 | /*----- Salsa20 implementation --------------------------------------------*/ | |
153 | ||
154 | /* --- @salsa20_init@ --- * | |
155 | * | |
156 | * Arguments: @salsa20_ctx *ctx@ = context to fill in | |
157 | * @const void *key@ = pointer to key material | |
158 | * @size_t ksz@ = size of key (either 32 or 16) | |
159 | * @const void *nonce@ = initial nonce, or null | |
160 | * | |
161 | * Returns: --- | |
162 | * | |
163 | * Use: Initializes a Salsa20 context ready for use. | |
164 | */ | |
165 | ||
166 | void salsa20_init(salsa20_ctx *ctx, const void *key, size_t ksz, | |
167 | const void *nonce) | |
168 | { | |
169 | static const octet zerononce[SALSA20_NONCESZ]; | |
170 | ||
171 | populate(ctx->a, key, ksz); | |
172 | salsa20_setnonce(ctx, nonce ? nonce : zerononce); | |
173 | } | |
174 | ||
175 | /* --- @salsa20_setnonce@ --- * | |
176 | * | |
177 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
178 | * @const void *nonce@ = the nonce (@SALSA20_NONCESZ@ bytes) | |
179 | * | |
180 | * Returns: --- | |
181 | * | |
182 | * Use: Set a new nonce in the context @ctx@, e.g., for processing a | |
183 | * different message. The stream position is reset to zero (see | |
184 | * @salsa20_seek@ etc.). | |
185 | */ | |
186 | ||
187 | void salsa20_setnonce(salsa20_ctx *ctx, const void *nonce) | |
188 | { | |
189 | const octet *n = nonce; | |
190 | ||
a4c2e267 MW |
191 | ctx->a[14] = LOAD32_L(n + 0); |
192 | ctx->a[11] = LOAD32_L(n + 4); | |
194e93f2 MW |
193 | salsa20_seek(ctx, 0); |
194 | } | |
195 | ||
196 | /* --- @salsa20_seek@, @salsa20_seeku64@ --- * | |
197 | * | |
198 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
199 | * @unsigned long i@, @kludge64 i@ = new position to set | |
200 | * | |
201 | * Returns: --- | |
202 | * | |
203 | * Use: Sets a new stream position, in units of Salsa20 output | |
204 | * blocks, which are @SALSA20_OUTSZ@ bytes each. Byte | |
205 | * granularity can be achieved by calling @salsa20R_encrypt@ | |
206 | * appropriately. | |
207 | */ | |
208 | ||
209 | void salsa20_seek(salsa20_ctx *ctx, unsigned long i) | |
210 | { kludge64 ii; ASSIGN64(ii, i); salsa20_seeku64(ctx, ii); } | |
211 | ||
212 | void salsa20_seeku64(salsa20_ctx *ctx, kludge64 i) | |
213 | { | |
a4c2e267 | 214 | ctx->a[8] = LO64(i); ctx->a[5] = HI64(i); |
194e93f2 MW |
215 | ctx->bufi = SALSA20_OUTSZ; |
216 | } | |
217 | ||
218 | /* --- @salsa20_tell@, @salsa20_tellu64@ --- * | |
219 | * | |
220 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
221 | * | |
222 | * Returns: The current position in the output stream, in blocks, | |
223 | * rounding upwards. | |
224 | */ | |
225 | ||
226 | unsigned long salsa20_tell(salsa20_ctx *ctx) | |
227 | { kludge64 i = salsa20_tellu64(ctx); return (GET64(unsigned long, i)); } | |
228 | ||
229 | kludge64 salsa20_tellu64(salsa20_ctx *ctx) | |
a4c2e267 | 230 | { kludge64 i; SET64(i, ctx->a[5], ctx->a[8]); return (i); } |
194e93f2 MW |
231 | |
232 | /* --- @salsa20{,12,8}_encrypt@ --- * | |
233 | * | |
234 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
235 | * @const void *src@ = source buffer (or null) | |
236 | * @void *dest@ = destination buffer (or null) | |
237 | * @size_t sz@ = size of the buffers | |
238 | * | |
239 | * Returns: --- | |
240 | * | |
241 | * Use: Encrypts or decrypts @sz@ bytes of data from @src@ to @dest@. | |
242 | * Salsa20 works by XORing plaintext with a keystream, so | |
243 | * encryption and decryption are the same operation. If @dest@ | |
244 | * is null then ignore @src@ and skip @sz@ bytes of the | |
245 | * keystream. If @src@ is null, then just write the keystream | |
246 | * to @dest@. | |
247 | */ | |
248 | ||
249 | #define SALSA20_ENCRYPT(r, ctx, src, dest, sz) \ | |
250 | SALSA20_DECOR(salsa20, r, _encrypt)(ctx, src, dest, sz) | |
251 | #define DEFENCRYPT(r) \ | |
252 | void SALSA20_ENCRYPT(r, salsa20_ctx *ctx, const void *src, \ | |
253 | void *dest, size_t sz) \ | |
254 | { \ | |
255 | salsa20_matrix b; \ | |
256 | const octet *s = src; \ | |
257 | octet *d = dest; \ | |
258 | size_t n; \ | |
259 | kludge64 pos, delta; \ | |
260 | \ | |
261 | SALSA20_OUTBUF(ctx, d, s, sz); \ | |
262 | if (!sz) return; \ | |
263 | \ | |
264 | if (!dest) { \ | |
265 | n = sz/SALSA20_OUTSZ; \ | |
266 | pos = salsa20_tellu64(ctx); \ | |
267 | ASSIGN64(delta, n); \ | |
268 | ADD64(pos, pos, delta); \ | |
269 | salsa20_seeku64(ctx, pos); \ | |
270 | sz = sz%SALSA20_OUTSZ; \ | |
271 | } else if (!src) { \ | |
272 | while (sz >= SALSA20_OUTSZ) { \ | |
273 | core(r, ctx->a, b); \ | |
274 | SALSA20_STEP(ctx->a); \ | |
275 | SALSA20_GENFULL(b, d); \ | |
276 | sz -= SALSA20_OUTSZ; \ | |
277 | } \ | |
278 | } else { \ | |
279 | while (sz >= SALSA20_OUTSZ) { \ | |
280 | core(r, ctx->a, b); \ | |
281 | SALSA20_STEP(ctx->a); \ | |
282 | SALSA20_MIXFULL(b, d, s); \ | |
283 | sz -= SALSA20_OUTSZ; \ | |
284 | } \ | |
285 | } \ | |
286 | \ | |
287 | if (sz) { \ | |
288 | core(r, ctx->a, b); \ | |
289 | SALSA20_STEP(ctx->a); \ | |
290 | SALSA20_PREPBUF(ctx, b); \ | |
291 | SALSA20_OUTBUF(ctx, d, s, sz); \ | |
292 | assert(!sz); \ | |
293 | } \ | |
294 | } | |
295 | SALSA20_VARS(DEFENCRYPT) | |
296 | ||
297 | /*----- HSalsa20 implementation -------------------------------------------*/ | |
298 | ||
299 | #define HSALSA20_RAW(r, ctx, src, dest) \ | |
300 | SALSA20_DECOR(hsalsa20, r, _raw)(ctx, src, dest) | |
301 | #define HSALSA20_PRF(r, ctx, src, dest) \ | |
302 | SALSA20_DECOR(hsalsa20, r, _prf)(ctx, src, dest) | |
303 | ||
304 | /* --- @hsalsa20{,12,8}_prf@ --- * | |
305 | * | |
306 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
307 | * @const void *src@ = the input (@HSALSA20_INSZ@ bytes) | |
308 | * @void *dest@ = the output (@HSALSA20_OUTSZ@ bytes) | |
309 | * | |
310 | * Returns: --- | |
311 | * | |
312 | * Use: Apply the HSalsa20/r pseudorandom function to @src@, writing | |
313 | * the result to @out@. | |
314 | */ | |
315 | ||
316 | #define DEFHSALSA20(r) \ | |
317 | static void HSALSA20_RAW(r, salsa20_matrix k, \ | |
318 | const uint32 *src, uint32 *dest) \ | |
319 | { \ | |
320 | salsa20_matrix a; \ | |
321 | int i; \ | |
322 | \ | |
323 | /* --- HSalsa20, computed from full Salsa20 --- * \ | |
324 | * \ | |
325 | * The security proof makes use of the fact that HSalsa20 (i.e., \ | |
326 | * without the final feedforward step) can be computed from full \ | |
327 | * Salsa20 using only knowledge of the non-secret input. I don't \ | |
328 | * want to compromise the performance of the main function by \ | |
329 | * making the feedforward step separate, but this operation is less \ | |
330 | * speed critical, so we do it the harder way. \ | |
331 | */ \ | |
332 | \ | |
a4c2e267 | 333 | for (i = 0; i < 4; i++) k[14 - 3*i] = src[i]; \ |
194e93f2 | 334 | core(r, k, a); \ |
a4c2e267 MW |
335 | for (i = 0; i < 4; i++) dest[i] = a[5*i] - k[i]; \ |
336 | for (i = 4; i < 8; i++) dest[i] = a[i + 2] - k[26 - 3*i]; \ | |
194e93f2 MW |
337 | } \ |
338 | \ | |
339 | void HSALSA20_PRF(r, salsa20_ctx *ctx, const void *src, void *dest) \ | |
340 | { \ | |
341 | const octet *s = src; \ | |
342 | octet *d = dest; \ | |
343 | uint32 in[4], out[8]; \ | |
344 | int i; \ | |
345 | \ | |
346 | for (i = 0; i < 4; i++) in[i] = LOAD32_L(s + 4*i); \ | |
347 | HSALSA20_RAW(r, ctx->a, in, out); \ | |
348 | for (i = 0; i < 8; i++) STORE32_L(d + 4*i, out[i]); \ | |
349 | } | |
350 | SALSA20_VARS(DEFHSALSA20) | |
351 | ||
352 | /*----- XSalsa20 implementation -------------------------------------------*/ | |
353 | ||
354 | /* --- Some convenient macros for naming functions --- * | |
355 | * | |
356 | * Because the crypto core is involved in XSalsa20/r's per-nonce setup, we | |
357 | * need to take an interest in the number of rounds in most of the various | |
358 | * functions, and it will probably help if we distinguish the context | |
359 | * structures for the various versions. | |
360 | */ | |
361 | ||
362 | #define XSALSA20_CTX(r) SALSA20_DECOR(xsalsa20, r, _ctx) | |
363 | #define XSALSA20_INIT(r, ctx, k, ksz, n) \ | |
364 | SALSA20_DECOR(xsalsa20, r, _init)(ctx, k, ksz, n) | |
365 | #define XSALSA20_SETNONCE(r, ctx, n) \ | |
366 | SALSA20_DECOR(xsalsa20, r, _setnonce)(ctx, n) | |
367 | #define XSALSA20_SEEK(r, ctx, i) \ | |
368 | SALSA20_DECOR(xsalsa20, r, _seek)(ctx, i) | |
369 | #define XSALSA20_SEEKU64(r, ctx, i) \ | |
370 | SALSA20_DECOR(xsalsa20, r, _seeku64)(ctx, i) | |
371 | #define XSALSA20_TELL(r, ctx) \ | |
372 | SALSA20_DECOR(xsalsa20, r, _tell)(ctx) | |
373 | #define XSALSA20_TELLU64(r, ctx) \ | |
374 | SALSA20_DECOR(xsalsa20, r, _tellu64)(ctx) | |
375 | #define XSALSA20_ENCRYPT(r, ctx, src, dest, sz) \ | |
376 | SALSA20_DECOR(xsalsa20, r, _encrypt)(ctx, src, dest, sz) | |
377 | ||
378 | /* --- @xsalsa20{,12,8}_init@ --- * | |
379 | * | |
380 | * Arguments: @xsalsa20R_ctx *ctx@ = the context to fill in | |
381 | * @const void *key@ = pointer to key material | |
382 | * @size_t ksz@ = size of key (either 32 or 16) | |
383 | * @const void *nonce@ = initial nonce, or null | |
384 | * | |
385 | * Returns: --- | |
386 | * | |
387 | * Use: Initializes an XSalsa20/r context ready for use. | |
388 | * | |
389 | * There is a different function for each number of rounds, | |
390 | * unlike for plain Salsa20. | |
391 | */ | |
392 | ||
393 | #define DEFXINIT(r) \ | |
394 | void XSALSA20_INIT(r, XSALSA20_CTX(r) *ctx, \ | |
395 | const void *key, size_t ksz, const void *nonce) \ | |
396 | { \ | |
397 | static const octet zerononce[XSALSA20_NONCESZ]; \ | |
398 | \ | |
399 | populate(ctx->k, key, ksz); \ | |
400 | ctx->s.a[ 0] = SALSA20_A256; \ | |
a4c2e267 MW |
401 | ctx->s.a[ 1] = SALSA20_B256; \ |
402 | ctx->s.a[ 2] = SALSA20_C256; \ | |
403 | ctx->s.a[ 3] = SALSA20_D256; \ | |
194e93f2 MW |
404 | XSALSA20_SETNONCE(r, ctx, nonce ? nonce : zerononce); \ |
405 | } | |
406 | SALSA20_VARS(DEFXINIT) | |
407 | ||
408 | /* --- @xsalsa20{,12,8}_setnonce@ --- * | |
409 | * | |
410 | * Arguments: @xsalsa20R_ctx *ctx@ = pointer to context | |
411 | * @const void *nonce@ = the nonce (@XSALSA20_NONCESZ@ bytes) | |
412 | * | |
413 | * Returns: --- | |
414 | * | |
415 | * Use: Set a new nonce in the context @ctx@, e.g., for processing a | |
416 | * different message. The stream position is reset to zero (see | |
417 | * @salsa20_seek@ etc.). | |
418 | * | |
419 | * There is a different function for each number of rounds, | |
420 | * unlike for plain Salsa20. | |
421 | */ | |
422 | ||
423 | #define DEFXNONCE(r) \ | |
424 | void XSALSA20_SETNONCE(r, XSALSA20_CTX(r) *ctx, const void *nonce) \ | |
425 | { \ | |
426 | const octet *n = nonce; \ | |
427 | uint32 in[4], out[8]; \ | |
428 | int i; \ | |
429 | \ | |
430 | for (i = 0; i < 4; i++) in[i] = LOAD32_L(n + 4*i); \ | |
431 | HSALSA20_RAW(r, ctx->k, in, out); \ | |
a4c2e267 MW |
432 | for (i = 0; i < 4; i++) ctx->s.a[13 - 3*i] = out[i]; \ |
433 | for (i = 4; i < 8; i++) ctx->s.a[27 - 3*i] = out[i]; \ | |
194e93f2 MW |
434 | salsa20_setnonce(&ctx->s, n + 16); \ |
435 | } | |
436 | SALSA20_VARS(DEFXNONCE) | |
437 | ||
438 | /* --- @xsalsa20{,12,8}_seek@, @xsalsa20{,12,8}_seeku64@ --- * | |
439 | * | |
440 | * Arguments: @xsalsa20R_ctx *ctx@ = pointer to context | |
441 | * @unsigned long i@, @kludge64 i@ = new position to set | |
442 | * | |
443 | * Returns: --- | |
444 | * | |
445 | * Use: Sets a new stream position, in units of Salsa20 output | |
446 | * blocks, which are @XSALSA20_OUTSZ@ bytes each. Byte | |
447 | * granularity can be achieved by calling @xsalsa20R_encrypt@ | |
448 | * appropriately. | |
449 | * | |
450 | * There is a different function for each number of rounds, | |
451 | * unlike for plain Salsa20, because the context structures are | |
452 | * different. | |
453 | */ | |
454 | ||
455 | /* --- @xsalsa20{,12,8}_tell@, @xsalsa20{,12,8}_tellu64@ --- * | |
456 | * | |
457 | * Arguments: @salsa20_ctx *ctx@ = pointer to context | |
458 | * | |
459 | * Returns: The current position in the output stream, in blocks, | |
460 | * rounding upwards. | |
461 | * | |
462 | * There is a different function for each number of rounds, | |
463 | * unlike for plain Salsa20, because the context structures are | |
464 | * different. | |
465 | */ | |
466 | ||
467 | /* --- @xsalsa20{,12,8}_encrypt@ --- * | |
468 | * | |
469 | * Arguments: @xsalsa20R_ctx *ctx@ = pointer to context | |
470 | * @const void *src@ = source buffer (or null) | |
471 | * @void *dest@ = destination buffer (or null) | |
472 | * @size_t sz@ = size of the buffers | |
473 | * | |
474 | * Returns: --- | |
475 | * | |
476 | * Use: Encrypts or decrypts @sz@ bytes of data from @src@ to @dest@. | |
477 | * XSalsa20 works by XORing plaintext with a keystream, so | |
478 | * encryption and decryption are the same operation. If @dest@ | |
479 | * is null then ignore @src@ and skip @sz@ bytes of the | |
480 | * keystream. If @src@ is null, then just write the keystream | |
481 | * to @dest@. | |
482 | */ | |
483 | ||
484 | #define DEFXPASSTHRU(r) \ | |
485 | void XSALSA20_SEEK(r, XSALSA20_CTX(r) *ctx, unsigned long i) \ | |
486 | { salsa20_seek(&ctx->s, i); } \ | |
487 | void XSALSA20_SEEKU64(r, XSALSA20_CTX(r) *ctx, kludge64 i) \ | |
488 | { salsa20_seeku64(&ctx->s, i); } \ | |
489 | unsigned long XSALSA20_TELL(r, XSALSA20_CTX(r) *ctx) \ | |
490 | { return salsa20_tell(&ctx->s); } \ | |
491 | kludge64 XSALSA20_TELLU64(r, XSALSA20_CTX(r) *ctx) \ | |
492 | { return salsa20_tellu64(&ctx->s); } \ | |
493 | void XSALSA20_ENCRYPT(r, XSALSA20_CTX(r) *ctx, \ | |
494 | const void *src, void *dest, size_t sz) \ | |
495 | { SALSA20_ENCRYPT(r, &ctx->s, src, dest, sz); } | |
496 | SALSA20_VARS(DEFXPASSTHRU) | |
497 | ||
498 | /*----- Generic cipher interface ------------------------------------------*/ | |
499 | ||
500 | typedef struct gctx { gcipher c; salsa20_ctx ctx; } gctx; | |
501 | ||
502 | static void gsetiv(gcipher *c, const void *iv) | |
503 | { gctx *g = (gctx *)c; salsa20_setnonce(&g->ctx, iv); } | |
504 | ||
505 | static void gdestroy(gcipher *c) | |
506 | { gctx *g = (gctx *)c; BURN(*g); S_DESTROY(g); } | |
507 | ||
508 | #define DEFGCIPHER(r) \ | |
509 | \ | |
510 | static const gcipher_ops gops_##r; \ | |
511 | \ | |
512 | static gcipher *ginit_##r(const void *k, size_t sz) \ | |
513 | { \ | |
514 | gctx *g = S_CREATE(gctx); \ | |
515 | g->c.ops = &gops_##r; \ | |
516 | salsa20_init(&g->ctx, k, sz, 0); \ | |
517 | return (&g->c); \ | |
518 | } \ | |
519 | \ | |
520 | static void gencrypt_##r(gcipher *c, const void *s, \ | |
521 | void *t, size_t sz) \ | |
522 | { gctx *g = (gctx *)c; SALSA20_ENCRYPT(r, &g->ctx, s, t, sz); } \ | |
523 | \ | |
524 | static const gcipher_ops gops_##r = { \ | |
525 | &SALSA20_DECOR(salsa20, r, ), \ | |
526 | gencrypt_##r, gencrypt_##r, gdestroy, gsetiv, 0 \ | |
527 | }; \ | |
528 | \ | |
529 | const gccipher SALSA20_DECOR(salsa20, r, ) = { \ | |
530 | SALSA20_NAME_##r, salsa20_keysz, \ | |
531 | SALSA20_NONCESZ, ginit_##r \ | |
532 | }; | |
533 | ||
534 | SALSA20_VARS(DEFGCIPHER) | |
535 | ||
536 | #define DEFGXCIPHER(r) \ | |
537 | \ | |
538 | typedef struct { gcipher c; XSALSA20_CTX(r) ctx; } gxctx_##r; \ | |
539 | \ | |
540 | static void gxsetiv_##r(gcipher *c, const void *iv) \ | |
541 | { gxctx_##r *g = (gxctx_##r *)c; XSALSA20_SETNONCE(r, &g->ctx, iv); } \ | |
542 | \ | |
543 | static void gxdestroy_##r(gcipher *c) \ | |
544 | { gxctx_##r *g = (gxctx_##r *)c; BURN(*g); S_DESTROY(g); } \ | |
545 | \ | |
546 | static const gcipher_ops gxops_##r; \ | |
547 | \ | |
548 | static gcipher *gxinit_##r(const void *k, size_t sz) \ | |
549 | { \ | |
550 | gxctx_##r *g = S_CREATE(gxctx_##r); \ | |
551 | g->c.ops = &gxops_##r; \ | |
552 | XSALSA20_INIT(r, &g->ctx, k, sz, 0); \ | |
553 | return (&g->c); \ | |
554 | } \ | |
555 | \ | |
556 | static void gxencrypt_##r(gcipher *c, const void *s, \ | |
557 | void *t, size_t sz) \ | |
558 | { \ | |
559 | gxctx_##r *g = (gxctx_##r *)c; \ | |
560 | XSALSA20_ENCRYPT(r, &g->ctx, s, t, sz); \ | |
561 | } \ | |
562 | \ | |
563 | static const gcipher_ops gxops_##r = { \ | |
564 | &SALSA20_DECOR(xsalsa20, r, ), \ | |
565 | gxencrypt_##r, gxencrypt_##r, gxdestroy_##r, gxsetiv_##r, 0 \ | |
566 | }; \ | |
567 | \ | |
568 | const gccipher SALSA20_DECOR(xsalsa20, r, ) = { \ | |
569 | "x" SALSA20_NAME_##r, salsa20_keysz, \ | |
570 | XSALSA20_NONCESZ, gxinit_##r \ | |
571 | }; | |
572 | ||
573 | SALSA20_VARS(DEFGXCIPHER) | |
574 | ||
575 | /*----- Generic random number generator interface -------------------------*/ | |
576 | ||
577 | typedef struct grops { | |
578 | size_t noncesz; | |
579 | void (*seek)(void *, kludge64); | |
580 | kludge64 (*tell)(void *); | |
581 | void (*setnonce)(void *, const void *); | |
582 | void (*generate)(void *, void *, size_t); | |
583 | } grops; | |
584 | ||
585 | typedef struct grbasectx { | |
586 | grand r; | |
587 | const grops *ops; | |
588 | } grbasectx; | |
589 | ||
590 | static int grmisc(grand *r, unsigned op, ...) | |
591 | { | |
592 | octet buf[XSALSA20_NONCESZ]; | |
593 | grbasectx *g = (grbasectx *)r; | |
594 | grand *rr; | |
595 | const octet *p; | |
596 | size_t sz; | |
597 | uint32 i; | |
598 | unsigned long ul; | |
599 | kludge64 pos; | |
600 | va_list ap; | |
601 | int rc = 0; | |
602 | ||
603 | va_start(ap, op); | |
604 | ||
605 | switch (op) { | |
606 | case GRAND_CHECK: | |
607 | switch (va_arg(ap, unsigned)) { | |
608 | case GRAND_CHECK: | |
609 | case GRAND_SEEDINT: | |
610 | case GRAND_SEEDUINT32: | |
611 | case GRAND_SEEDBLOCK: | |
612 | case GRAND_SEEDRAND: | |
613 | case SALSA20_SEEK: | |
614 | case SALSA20_SEEKU64: | |
615 | case SALSA20_TELL: | |
616 | case SALSA20_TELLU64: | |
617 | rc = 1; | |
618 | break; | |
619 | default: | |
620 | rc = 0; | |
621 | break; | |
622 | } | |
623 | break; | |
624 | ||
625 | case GRAND_SEEDINT: | |
626 | i = va_arg(ap, unsigned); STORE32_L(buf, i); | |
627 | memset(buf + 4, 0, g->ops->noncesz - 4); | |
628 | g->ops->setnonce(g, buf); | |
629 | break; | |
630 | case GRAND_SEEDUINT32: | |
631 | i = va_arg(ap, uint32); STORE32_L(buf, i); | |
632 | memset(buf + 4, 0, g->ops->noncesz - 4); | |
633 | g->ops->setnonce(g, buf); | |
634 | break; | |
635 | case GRAND_SEEDBLOCK: | |
636 | p = va_arg(ap, const void *); | |
637 | sz = va_arg(ap, size_t); | |
638 | if (sz < g->ops->noncesz) { | |
639 | memcpy(buf, p, sz); | |
640 | memset(buf + sz, 0, g->ops->noncesz - sz); | |
641 | p = buf; | |
642 | } | |
643 | g->ops->setnonce(g, p); | |
644 | break; | |
645 | case GRAND_SEEDRAND: | |
646 | rr = va_arg(ap, grand *); | |
647 | rr->ops->fill(rr, buf, g->ops->noncesz); | |
648 | g->ops->setnonce(g, buf); | |
649 | break; | |
650 | case SALSA20_SEEK: | |
651 | ul = va_arg(ap, unsigned long); ASSIGN64(pos, ul); | |
652 | g->ops->seek(g, pos); | |
653 | break; | |
654 | case SALSA20_SEEKU64: | |
655 | pos = va_arg(ap, kludge64); | |
656 | g->ops->seek(g, pos); | |
657 | break; | |
658 | case SALSA20_TELL: | |
659 | pos = g->ops->tell(g); | |
660 | *va_arg(ap, unsigned long *) = GET64(unsigned long, pos); | |
661 | break; | |
662 | case SALSA20_TELLU64: | |
663 | *va_arg(ap, kludge64 *) = g->ops->tell(g); | |
664 | break; | |
665 | default: | |
666 | GRAND_BADOP; | |
667 | break; | |
668 | } | |
669 | ||
670 | return (rc); | |
671 | } | |
672 | ||
673 | static octet grbyte(grand *r) | |
674 | { | |
675 | grbasectx *g = (grbasectx *)r; | |
676 | octet o; | |
677 | g->ops->generate(g, &o, 1); | |
678 | return (o); | |
679 | } | |
680 | ||
681 | static uint32 grword(grand *r) | |
682 | { | |
683 | grbasectx *g = (grbasectx *)r; | |
684 | octet b[4]; | |
685 | g->ops->generate(g, b, sizeof(b)); | |
686 | return (LOAD32_L(b)); | |
687 | } | |
688 | ||
689 | static void grfill(grand *r, void *p, size_t sz) | |
690 | { | |
691 | grbasectx *g = (grbasectx *)r; | |
692 | g->ops->generate(r, p, sz); | |
693 | } | |
694 | ||
695 | typedef struct grctx { | |
696 | grbasectx r; | |
697 | salsa20_ctx ctx; | |
698 | } grctx; | |
699 | ||
700 | static void gr_seek(void *r, kludge64 pos) | |
701 | { grctx *g = r; salsa20_seeku64(&g->ctx, pos); } | |
702 | ||
703 | static kludge64 gr_tell(void *r) | |
704 | { grctx *g = r; return (salsa20_tellu64(&g->ctx)); } | |
705 | ||
706 | static void gr_setnonce(void *r, const void *n) | |
707 | { grctx *g = r; salsa20_setnonce(&g->ctx, n); } | |
708 | ||
709 | static void grdestroy(grand *r) | |
710 | { grctx *g = (grctx *)r; BURN(*g); S_DESTROY(g); } | |
711 | ||
712 | #define DEFGRAND(rr) \ | |
713 | \ | |
714 | static void gr_generate_##rr(void *r, void *b, size_t sz) \ | |
715 | { grctx *g = r; SALSA20_ENCRYPT(rr, &g->ctx, 0, b, sz); } \ | |
716 | \ | |
717 | static const grops grops_##rr = \ | |
718 | { SALSA20_NONCESZ, gr_seek, gr_tell, \ | |
719 | gr_setnonce, gr_generate_##rr }; \ | |
720 | \ | |
721 | static const grand_ops grops_rand_##rr = { \ | |
722 | SALSA20_NAME_##rr, GRAND_CRYPTO, 0, \ | |
723 | grmisc, grdestroy, grword, \ | |
44ff6c11 | 724 | grbyte, grword, grand_defaultrange, grfill \ |
194e93f2 MW |
725 | }; \ |
726 | \ | |
727 | grand *SALSA20_DECOR(salsa20, rr, _rand) \ | |
728 | (const void *k, size_t ksz, const void *n) \ | |
729 | { \ | |
f9fe9910 | 730 | grctx *g = S_CREATE(grctx); \ |
194e93f2 MW |
731 | g->r.r.ops = &grops_rand_##rr; \ |
732 | g->r.ops = &grops_##rr; \ | |
733 | salsa20_init(&g->ctx, k, ksz, n); \ | |
734 | return (&g->r.r); \ | |
735 | } | |
736 | SALSA20_VARS(DEFGRAND) | |
737 | ||
738 | #define DEFXGRAND(rr) \ | |
739 | \ | |
740 | typedef struct grxctx_##rr { \ | |
741 | grbasectx r; \ | |
742 | XSALSA20_CTX(rr) ctx; \ | |
743 | } grxctx_##rr; \ | |
744 | \ | |
745 | static void grx_seek_##rr(void *r, kludge64 pos) \ | |
746 | { grxctx_##rr *g = r; XSALSA20_SEEKU64(rr, &g->ctx, pos); } \ | |
747 | \ | |
748 | static kludge64 grx_tell_##rr(void *r) \ | |
749 | { grxctx_##rr *g = r; return (XSALSA20_TELLU64(rr, &g->ctx)); } \ | |
750 | \ | |
751 | static void grx_setnonce_##rr(void *r, const void *n) \ | |
752 | { grxctx_##rr *g = r; XSALSA20_SETNONCE(rr, &g->ctx, n); } \ | |
753 | \ | |
754 | static void grxdestroy_##rr(grand *r) \ | |
755 | { grxctx_##rr *g = (grxctx_##rr *)r; BURN(*g); S_DESTROY(g); } \ | |
756 | \ | |
757 | static void grx_generate_##rr(void *r, void *b, size_t sz) \ | |
758 | { grxctx_##rr *g = r; XSALSA20_ENCRYPT(rr, &g->ctx, 0, b, sz); } \ | |
759 | \ | |
760 | static const grops grxops_##rr = \ | |
761 | { XSALSA20_NONCESZ, grx_seek_##rr, grx_tell_##rr, \ | |
762 | grx_setnonce_##rr, grx_generate_##rr }; \ | |
763 | \ | |
764 | static const grand_ops grxops_rand_##rr = { \ | |
765 | "x" SALSA20_NAME_##rr, GRAND_CRYPTO, 0, \ | |
766 | grmisc, grxdestroy_##rr, grword, \ | |
44ff6c11 | 767 | grbyte, grword, grand_defaultrange, grfill \ |
194e93f2 MW |
768 | }; \ |
769 | \ | |
770 | grand *SALSA20_DECOR(xsalsa20, rr, _rand) \ | |
771 | (const void *k, size_t ksz, const void *n) \ | |
772 | { \ | |
f9fe9910 | 773 | grxctx_##rr *g = S_CREATE(grxctx_##rr); \ |
194e93f2 MW |
774 | g->r.r.ops = &grxops_rand_##rr; \ |
775 | g->r.ops = &grxops_##rr; \ | |
776 | XSALSA20_INIT(rr, &g->ctx, k, ksz, n); \ | |
777 | return (&g->r.r); \ | |
778 | } | |
779 | SALSA20_VARS(DEFXGRAND) | |
780 | ||
781 | /*----- Test rig ----------------------------------------------------------*/ | |
782 | ||
783 | #ifdef TEST_RIG | |
784 | ||
785 | #include <stdio.h> | |
786 | #include <string.h> | |
787 | ||
788 | #include <mLib/quis.h> | |
789 | #include <mLib/testrig.h> | |
790 | ||
a4c2e267 MW |
791 | static const int perm[] = { |
792 | 0, 13, 10, 7, | |
793 | 4, 1, 14, 11, | |
794 | 8, 5, 2, 15, | |
795 | 12, 9, 6, 3 | |
796 | }; | |
797 | ||
194e93f2 MW |
798 | #define DEFVCORE(r) \ |
799 | static int v_core_##r(dstr *v) \ | |
800 | { \ | |
801 | salsa20_matrix a, b; \ | |
802 | dstr d = DSTR_INIT; \ | |
a4c2e267 | 803 | int i, j, n; \ |
194e93f2 MW |
804 | int ok = 1; \ |
805 | \ | |
806 | DENSURE(&d, SALSA20_OUTSZ); d.len = SALSA20_OUTSZ; \ | |
807 | n = *(int *)v[0].buf; \ | |
808 | for (i = 0; i < SALSA20_OUTSZ/4; i++) \ | |
a4c2e267 | 809 | b[i] = LOAD32_L(v[1].buf + 4*i); \ |
194e93f2 | 810 | for (i = 0; i < n; i++) { \ |
a4c2e267 | 811 | for (j = 0; j < 16; j++) a[perm[j]] = b[j]; \ |
194e93f2 MW |
812 | core(r, a, b); \ |
813 | memcpy(a, b, sizeof(a)); \ | |
814 | } \ | |
a4c2e267 | 815 | for (i = 0; i < SALSA20_OUTSZ/4; i++) STORE32_L(d.buf + 4*i, b[i]); \ |
194e93f2 MW |
816 | \ |
817 | if (d.len != v[2].len || memcmp(d.buf, v[2].buf, v[2].len) != 0) { \ | |
818 | ok = 0; \ | |
819 | printf("\nfail core:" \ | |
820 | "\n\titerations = %d" \ | |
821 | "\n\tin = ", n); \ | |
822 | type_hex.dump(&v[1], stdout); \ | |
823 | printf("\n\texpected = "); \ | |
824 | type_hex.dump(&v[2], stdout); \ | |
825 | printf("\n\tcalculated = "); \ | |
826 | type_hex.dump(&d, stdout); \ | |
827 | putchar('\n'); \ | |
828 | } \ | |
829 | \ | |
830 | dstr_destroy(&d); \ | |
831 | return (ok); \ | |
832 | } | |
833 | SALSA20_VARS(DEFVCORE) | |
834 | ||
835 | #define SALSA20_CTX(r) salsa20_ctx | |
836 | #define SALSA20_INIT(r, ctx, k, ksz, n) salsa20_init(ctx, k, ksz, n) | |
837 | #define SALSA20_SEEKU64(r, ctx, i) salsa20_seeku64(ctx, i) | |
838 | ||
839 | #define DEFxVENC(base, BASE, r) \ | |
840 | static int v_encrypt_##base##_##r(dstr *v) \ | |
841 | { \ | |
842 | BASE##_CTX(r) ctx; \ | |
843 | dstr d = DSTR_INIT; \ | |
844 | kludge64 pos; \ | |
845 | const octet *p, *p0; \ | |
846 | octet *q; \ | |
847 | size_t sz, sz0, step; \ | |
848 | unsigned long skip; \ | |
849 | int ok = 1; \ | |
850 | \ | |
851 | if (v[4].len) { p0 = (const octet *)v[4].buf; sz0 = v[4].len; } \ | |
852 | else { p0 = 0; sz0 = v[5].len; } \ | |
853 | DENSURE(&d, sz0); d.len = sz0; \ | |
854 | skip = *(unsigned long *)v[3].buf; \ | |
855 | \ | |
856 | step = 0; \ | |
857 | while (step < sz0 + skip) { \ | |
858 | step = step ? 3*step + 4 : 1; \ | |
859 | if (step > sz0 + skip) step = sz0 + skip; \ | |
860 | BASE##_INIT(r, &ctx, v[0].buf, v[0].len, v[1].buf); \ | |
861 | if (v[2].len) { \ | |
862 | LOAD64_(pos, v[2].buf); \ | |
863 | BASE##_SEEKU64(r, &ctx, pos); \ | |
864 | } \ | |
865 | \ | |
866 | for (sz = skip; sz >= step; sz -= step) \ | |
867 | BASE##_ENCRYPT(r, &ctx, 0, 0, step); \ | |
868 | if (sz) BASE##_ENCRYPT(r, &ctx, 0, 0, sz); \ | |
869 | for (p = p0, q = (octet *)d.buf, sz = sz0; \ | |
870 | sz >= step; \ | |
871 | sz -= step, q += step) { \ | |
872 | BASE##_ENCRYPT(r, &ctx, p, q, step); \ | |
873 | if (p) p += step; \ | |
874 | } \ | |
875 | if (sz) BASE##_ENCRYPT(r, &ctx, p, q, sz); \ | |
876 | \ | |
877 | if (d.len != v[5].len || memcmp(d.buf, v[5].buf, v[5].len) != 0) { \ | |
878 | ok = 0; \ | |
879 | printf("\nfail encrypt:" \ | |
880 | "\n\tstep = %lu" \ | |
881 | "\n\tkey = ", (unsigned long)step); \ | |
882 | type_hex.dump(&v[0], stdout); \ | |
883 | printf("\n\tnonce = "); \ | |
884 | type_hex.dump(&v[1], stdout); \ | |
885 | printf("\n\tposition = "); \ | |
886 | type_hex.dump(&v[2], stdout); \ | |
887 | printf("\n\tskip = %lu", skip); \ | |
888 | printf("\n\tmessage = "); \ | |
889 | type_hex.dump(&v[4], stdout); \ | |
890 | printf("\n\texpected = "); \ | |
891 | type_hex.dump(&v[5], stdout); \ | |
892 | printf("\n\tcalculated = "); \ | |
893 | type_hex.dump(&d, stdout); \ | |
894 | putchar('\n'); \ | |
895 | } \ | |
896 | } \ | |
897 | \ | |
898 | dstr_destroy(&d); \ | |
899 | return (ok); \ | |
900 | } | |
901 | #define DEFVENC(r) DEFxVENC(salsa20, SALSA20, r) | |
902 | #define DEFXVENC(r) DEFxVENC(xsalsa20, XSALSA20, r) | |
903 | SALSA20_VARS(DEFVENC) | |
904 | SALSA20_VARS(DEFXVENC) | |
905 | ||
906 | static test_chunk defs[] = { | |
907 | #define DEFxTAB(pre, base, r) \ | |
908 | { pre SALSA20_NAME_##r, v_encrypt_##base##_##r, \ | |
909 | { &type_hex, &type_hex, &type_hex, &type_ulong, \ | |
910 | &type_hex, &type_hex, 0 } }, | |
911 | #define DEFTAB(r) \ | |
912 | { SALSA20_NAME_##r "-core", v_core_##r, \ | |
913 | { &type_int, &type_hex, &type_hex, 0 } }, \ | |
914 | DEFxTAB("", salsa20, r) | |
915 | #define DEFXTAB(r) DEFxTAB("x", xsalsa20, r) | |
916 | SALSA20_VARS(DEFTAB) | |
917 | SALSA20_VARS(DEFXTAB) | |
918 | { 0, 0, { 0 } } | |
919 | }; | |
920 | ||
921 | int main(int argc, char *argv[]) | |
922 | { | |
923 | test_run(argc, argv, defs, SRCDIR"/t/salsa20"); | |
924 | return (0); | |
925 | } | |
926 | ||
927 | #endif | |
928 | ||
929 | /*----- That's all, folks -------------------------------------------------*/ |