chiark / gitweb /
configure.ac, symm/rijndael*: Use ARMv8 AES instructions where available.
[catacomb] / symm / chacha.c
1 /* -*-c-*-
2  *
3  * ChaCha stream cipher
4  *
5  * (c) 2015 Straylight/Edgeware
6  */
7
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
28 /*----- Header files ------------------------------------------------------*/
29
30 #include "config.h"
31
32 #include <stdarg.h>
33
34 #include <mLib/bits.h>
35
36 #include "arena.h"
37 #include "chacha.h"
38 #include "chacha-core.h"
39 #include "dispatch.h"
40 #include "gcipher.h"
41 #include "grand.h"
42 #include "keysz.h"
43 #include "paranoia.h"
44
45 /*----- Global variables --------------------------------------------------*/
46
47 const octet chacha_keysz[] = { KSZ_SET, 32, 16, 10, 0 };
48
49 /*----- The ChaCha core function and utilities ----------------------------*/
50
51 /* --- @core@ --- *
52  *
53  * Arguments:   @unsigned r@ = number of rounds
54  *              @const chacha_matrix src@ = input matrix
55  *              @chacha_matrix dest@ = where to put the output
56  *
57  * Returns:     ---
58  *
59  *
60  * Use:         Apply the ChaCha/r core function to @src@, writing the
61  *              result to @dest@.  This consists of @r@ rounds followed by
62  *              the feedforward step.
63  */
64
65 CPU_DISPATCH(static, (void), void, core,
66              (unsigned r, const chacha_matrix src, chacha_matrix dest),
67              (r, src, dest), pick_core, simple_core);
68
69 static void simple_core(unsigned r, const chacha_matrix src,
70                         chacha_matrix dest)
71   { CHACHA_nR(dest, src, r); CHACHA_FFWD(dest, src); }
72
73 #if CPUFAM_X86 || CPUFAM_AMD64
74 extern core__functype chacha_core_x86ish_sse2;
75 #endif
76
77 #if CPUFAM_ARMEL
78 extern core__functype chacha_core_arm_neon;
79 #endif
80
81 static core__functype *pick_core(void)
82 {
83 #if CPUFAM_X86 || CPUFAM_AMD64
84   DISPATCH_PICK_COND(chacha_core, chacha_core_x86ish_sse2,
85                      cpu_feature_p(CPUFEAT_X86_SSE2));
86 #endif
87 #if CPUFAM_ARMEL
88   DISPATCH_PICK_COND(chacha_core, chacha_core_arm_neon,
89                      cpu_feature_p(CPUFEAT_ARM_NEON));
90 #endif
91   DISPATCH_PICK_FALLBACK(chacha_core, simple_core);
92 }
93
94 /* --- @populate@ --- *
95  *
96  * Arguments:   @chacha_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 ChaCha 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(chacha_matrix a, const void *key, size_t ksz)
108 {
109   const octet *k = key;
110
111   KSZ_ASSERT(chacha, ksz);
112
113   a[ 4] = LOAD32_L(k +  0);
114   a[ 5] = LOAD32_L(k +  4);
115   if (ksz == 10) {
116     a[ 6] = LOAD16_L(k +  8);
117     a[ 7] = 0;
118   } else {
119     a[ 6] = LOAD32_L(k +  8);
120     a[ 7] = LOAD32_L(k + 12);
121   }
122   if (ksz <= 16) {
123     a[ 8] = a[ 4];
124     a[ 9] = a[ 5];
125     a[10] = a[ 6];
126     a[11] = a[ 7];
127     a[ 0] = CHACHA_A128;
128     a[ 1] = CHACHA_B128;
129     a[ 2] = ksz == 10 ? CHACHA_C80 : CHACHA_C128;
130     a[ 3] = CHACHA_D128;
131   } else {
132     a[ 8] = LOAD32_L(k + 16);
133     a[ 9] = LOAD32_L(k + 20);
134     a[10] = LOAD32_L(k + 24);
135     a[11] = LOAD32_L(k + 28);
136     a[ 0] = CHACHA_A256;
137     a[ 1] = CHACHA_B256;
138     a[ 2] = CHACHA_C256;
139     a[ 3] = CHACHA_D256;
140   }
141 }
142
143 /*----- ChaCha implementation ---------------------------------------------*/
144
145 /* --- @chacha_init@ --- *
146  *
147  * Arguments:   @chacha_ctx *ctx@ = context to fill in
148  *              @const void *key@ = pointer to key material
149  *              @size_t ksz@ = size of key (either 32 or 16)
150  *              @const void *nonce@ = initial nonce, or null
151  *
152  * Returns:     ---
153  *
154  * Use:         Initializes a ChaCha context ready for use.
155  */
156
157 void chacha_init(chacha_ctx *ctx, const void *key, size_t ksz,
158                   const void *nonce)
159 {
160   static const octet zerononce[CHACHA_NONCESZ];
161
162   populate(ctx->a, key, ksz);
163   chacha_setnonce(ctx, nonce ? nonce : zerononce);
164 }
165
166 /* --- @chacha_setnonce@ --- *
167  *
168  * Arguments:   @chacha_ctx *ctx@ = pointer to context
169  *              @const void *nonce@ = the nonce (@CHACHA_NONCESZ@ bytes)
170  *
171  * Returns:     ---
172  *
173  * Use:         Set a new nonce in the context @ctx@, e.g., for processing a
174  *              different message.  The stream position is reset to zero (see
175  *              @chacha_seek@ etc.).
176  */
177
178 void chacha_setnonce(chacha_ctx *ctx, const void *nonce)
179 {
180   const octet *n = nonce;
181
182   ctx->a[14] = LOAD32_L(n + 0);
183   ctx->a[15] = LOAD32_L(n + 4);
184   chacha_seek(ctx, 0);
185 }
186
187 /* --- @chacha_seek@, @chacha_seeku64@ --- *
188  *
189  * Arguments:   @chacha_ctx *ctx@ = pointer to context
190  *              @unsigned long i@, @kludge64 i@ = new position to set
191  *
192  * Returns:     ---
193  *
194  * Use:         Sets a new stream position, in units of Chacha output
195  *              blocks, which are @CHACHA_OUTSZ@ bytes each.  Byte
196  *              granularity can be achieved by calling @chachaR_encrypt@
197  *              appropriately.
198  */
199
200 void chacha_seek(chacha_ctx *ctx, unsigned long i)
201   { kludge64 ii; ASSIGN64(ii, i); chacha_seeku64(ctx, ii); }
202
203 void chacha_seeku64(chacha_ctx *ctx, kludge64 i)
204 {
205   ctx->a[12] = LO64(i); ctx->a[13] = HI64(i);
206   ctx->bufi = CHACHA_OUTSZ;
207 }
208
209 /* --- @chacha_tell@, @chacha_tellu64@ --- *
210  *
211  * Arguments:   @chacha_ctx *ctx@ = pointer to context
212  *
213  * Returns:     The current position in the output stream, in blocks,
214  *              rounding upwards.
215  */
216
217 unsigned long chacha_tell(chacha_ctx *ctx)
218   { kludge64 i = chacha_tellu64(ctx); return (GET64(unsigned long, i)); }
219
220 kludge64 chacha_tellu64(chacha_ctx *ctx)
221   { kludge64 i; SET64(i, ctx->a[9], ctx->a[8]); return (i); }
222
223 /* --- @chacha{,12,8}_encrypt@ --- *
224  *
225  * Arguments:   @chacha_ctx *ctx@ = pointer to context
226  *              @const void *src@ = source buffer (or null)
227  *              @void *dest@ = destination buffer (or null)
228  *              @size_t sz@ = size of the buffers
229  *
230  * Returns:     ---
231  *
232  * Use:         Encrypts or decrypts @sz@ bytes of data from @src@ to @dest@.
233  *              ChaCha works by XORing plaintext with a keystream, so
234  *              encryption and decryption are the same operation.  If @dest@
235  *              is null then ignore @src@ and skip @sz@ bytes of the
236  *              keystream.  If @src@ is null, then just write the keystream
237  *              to @dest@.
238  */
239
240 #define CHACHA_ENCRYPT(r, ctx, src, dest, sz)                           \
241   chacha##r##_encrypt(ctx, src, dest, sz)
242 #define DEFENCRYPT(r)                                                   \
243   void CHACHA_ENCRYPT(r, chacha_ctx *ctx, const void *src,              \
244                       void *dest, size_t sz)                            \
245   {                                                                     \
246     chacha_matrix b;                                                    \
247     const octet *s = src;                                               \
248     octet *d = dest;                                                    \
249     size_t n;                                                           \
250     kludge64 pos, delta;                                                \
251                                                                         \
252     SALSA20_OUTBUF(ctx, d, s, sz);                                      \
253     if (!sz) return;                                                    \
254                                                                         \
255     if (!dest) {                                                        \
256       n = sz/CHACHA_OUTSZ;                                              \
257       pos = chacha_tellu64(ctx);                                        \
258       ASSIGN64(delta, n);                                               \
259       ADD64(pos, pos, delta);                                           \
260       chacha_seeku64(ctx, pos);                                         \
261       sz = sz%CHACHA_OUTSZ;                                             \
262     } else if (!src) {                                                  \
263       while (sz >= CHACHA_OUTSZ) {                                      \
264         core(r, ctx->a, b);                                             \
265         CHACHA_STEP(ctx->a);                                            \
266         SALSA20_GENFULL(b, d);                                          \
267         sz -= CHACHA_OUTSZ;                                             \
268       }                                                                 \
269     } else {                                                            \
270       while (sz >= CHACHA_OUTSZ) {                                      \
271         core(r, ctx->a, b);                                             \
272         CHACHA_STEP(ctx->a);                                            \
273         SALSA20_MIXFULL(b, d, s);                                       \
274         sz -= CHACHA_OUTSZ;                                             \
275       }                                                                 \
276     }                                                                   \
277                                                                         \
278     if (sz) {                                                           \
279       core(r, ctx->a, b);                                               \
280       CHACHA_STEP(ctx->a);                                              \
281       SALSA20_PREPBUF(ctx, b);                                          \
282       SALSA20_OUTBUF(ctx, d, s, sz);                                    \
283       assert(!sz);                                                      \
284     }                                                                   \
285   }
286 CHACHA_VARS(DEFENCRYPT)
287
288 /*----- HChaCha implementation --------------------------------------------*/
289
290 #define HCHACHA_RAW(r, ctx, src, dest) hchacha##r##_raw(ctx, src, dest)
291 #define HCHACHA_PRF(r, ctx, src, dest) hchacha##r##_prf(ctx, src, dest)
292
293 /* --- @hchacha{20,12,8}_prf@ --- *
294  *
295  * Arguments:   @chacha_ctx *ctx@ = pointer to context
296  *              @const void *src@ = the input (@HCHACHA_INSZ@ bytes)
297  *              @void *dest@ = the output (@HCHACHA_OUTSZ@ bytes)
298  *
299  * Returns:     ---
300  *
301  * Use:         Apply the HChacha/r pseudorandom function to @src@, writing
302  *              the result to @out@.
303  */
304
305 #define DEFHCHACHA(r)                                                   \
306   static void HCHACHA_RAW(r, chacha_matrix k,                           \
307                           const uint32 *src, uint32 *dest)              \
308   {                                                                     \
309     chacha_matrix a;                                                    \
310     int i;                                                              \
311                                                                         \
312     /* --- HChaCha, computed from full ChaCha --- *                     \
313      *                                                                  \
314      * The security proof makes use of the fact that HChaCha (i.e.,     \
315      * without the final feedforward step) can be computed from full    \
316      * ChaCha using only knowledge of the non-secret input.  I don't    \
317      * want to compromise the performance of the main function by       \
318      * making the feedforward step separate, but this operation is less \
319      * speed critical, so we do it the harder way.                      \
320      */                                                                 \
321                                                                         \
322     for (i = 0; i < 4; i++) k[12 + i] = src[i];                         \
323     core(r, k, a);                                                      \
324     for (i = 0; i < 8; i++) dest[i] = a[(i + 4)^4] - k[(i + 4)^4];      \
325   }                                                                     \
326                                                                         \
327   void HCHACHA_PRF(r, chacha_ctx *ctx, const void *src, void *dest)     \
328   {                                                                     \
329     const octet *s = src;                                               \
330     octet *d = dest;                                                    \
331     uint32 in[4], out[8];                                               \
332     int i;                                                              \
333                                                                         \
334     for (i = 0; i < 4; i++) in[i] = LOAD32_L(s + 4*i);                  \
335     HCHACHA_RAW(r, ctx->a, in, out);                                    \
336     for (i = 0; i < 8; i++) STORE32_L(d + 4*i, out[i]);                 \
337   }
338 CHACHA_VARS(DEFHCHACHA)
339
340 /*----- XChaCha implementation -------------------------------------------*/
341
342 /* --- Some convenient macros for naming functions --- *
343  *
344  * Because the crypto core is involved in XChaCha/r's per-nonce setup, we
345  * need to take an interest in the number of rounds in most of the various
346  * functions, and it will probably help if we distinguish the context
347  * structures for the various versions.
348  */
349
350 #define XCHACHA_CTX(r) xchacha##r##_ctx
351 #define XCHACHA_INIT(r, ctx, k, ksz, n) xchacha##r##_init(ctx, k, ksz, n)
352 #define XCHACHA_SETNONCE(r, ctx, n) xchacha##r##_setnonce(ctx, n)
353 #define XCHACHA_SEEK(r, ctx, i) xchacha##r##_seek(ctx, i)
354 #define XCHACHA_SEEKU64(r, ctx, i) xchacha##r##_seeku64(ctx, i)
355 #define XCHACHA_TELL(r, ctx) xchacha##r##_tell(ctx)
356 #define XCHACHA_TELLU64(r, ctx) xchacha##r##_tellu64(ctx)
357 #define XCHACHA_ENCRYPT(r, ctx, src, dest, sz)                          \
358   xchacha##r##_encrypt(ctx, src, dest, sz)
359
360 /* --- @xchacha{20,12,8}_init@ --- *
361  *
362  * Arguments:   @xchachaR_ctx *ctx@ = the context to fill in
363  *              @const void *key@ = pointer to key material
364  *              @size_t ksz@ = size of key (either 32 or 16)
365  *              @const void *nonce@ = initial nonce, or null
366  *
367  * Returns:     ---
368  *
369  * Use:         Initializes an XChaCha/r context ready for use.
370  *
371  *              There is a different function for each number of rounds,
372  *              unlike for plain ChaCha.
373  */
374
375 #define DEFXINIT(r)                                                     \
376   void XCHACHA_INIT(r, XCHACHA_CTX(r) *ctx,                             \
377                     const void *key, size_t ksz, const void *nonce)     \
378   {                                                                     \
379     static const octet zerononce[XCHACHA_NONCESZ];                      \
380                                                                         \
381     populate(ctx->k, key, ksz);                                         \
382     ctx->s.a[ 0] = CHACHA_A256;                                         \
383     ctx->s.a[ 1] = CHACHA_B256;                                         \
384     ctx->s.a[ 2] = CHACHA_C256;                                         \
385     ctx->s.a[ 3] = CHACHA_D256;                                         \
386     XCHACHA_SETNONCE(r, ctx, nonce ? nonce : zerononce);                \
387   }
388 CHACHA_VARS(DEFXINIT)
389
390 /* --- @xchacha{20,12,8}_setnonce@ --- *
391  *
392  * Arguments:   @xchachaR_ctx *ctx@ = pointer to context
393  *              @const void *nonce@ = the nonce (@XCHACHA_NONCESZ@ bytes)
394  *
395  * Returns:     ---
396  *
397  * Use:         Set a new nonce in the context @ctx@, e.g., for processing a
398  *              different message.  The stream position is reset to zero (see
399  *              @chacha_seek@ etc.).
400  *
401  *              There is a different function for each number of rounds,
402  *              unlike for plain ChaCha.
403  */
404
405 #define DEFXNONCE(r)                                                    \
406   void XCHACHA_SETNONCE(r, XCHACHA_CTX(r) *ctx, const void *nonce)      \
407   {                                                                     \
408     const octet *n = nonce;                                             \
409     uint32 in[4];                                                       \
410     int i;                                                              \
411                                                                         \
412     for (i = 0; i < 4; i++) in[i] = LOAD32_L(n + 4*i);                  \
413     HCHACHA_RAW(r, ctx->k, in, ctx->s.a + 4);                           \
414     chacha_setnonce(&ctx->s, n + 16);                                   \
415   }
416 CHACHA_VARS(DEFXNONCE)
417
418 /* --- @xchacha{20,12,8}_seek@, @xchacha{20,12,8}_seeku64@ --- *
419  *
420  * Arguments:   @xchachaR_ctx *ctx@ = pointer to context
421  *              @unsigned long i@, @kludge64 i@ = new position to set
422  *
423  * Returns:     ---
424  *
425  * Use:         Sets a new stream position, in units of ChaCha output
426  *              blocks, which are @XCHACHA_OUTSZ@ bytes each.  Byte
427  *              granularity can be achieved by calling @xchachaR_encrypt@
428  *              appropriately.
429  *
430  *              There is a different function for each number of rounds,
431  *              unlike for plain ChaCha, because the context structures are
432  *              different.
433  */
434
435 /* --- @xchacha{20,12,8}_tell@, @xchacha{20,12,8}_tellu64@ --- *
436  *
437  * Arguments:   @chacha_ctx *ctx@ = pointer to context
438  *
439  * Returns:     The current position in the output stream, in blocks,
440  *              rounding upwards.
441  *
442  *              There is a different function for each number of rounds,
443  *              unlike for plain ChaCha, because the context structures are
444  *              different.
445  */
446
447 /* --- @xchacha{,12,8}_encrypt@ --- *
448  *
449  * Arguments:   @xchachaR_ctx *ctx@ = pointer to context
450  *              @const void *src@ = source buffer (or null)
451  *              @void *dest@ = destination buffer (or null)
452  *              @size_t sz@ = size of the buffers
453  *
454  * Returns:     ---
455  *
456  * Use:         Encrypts or decrypts @sz@ bytes of data from @src@ to @dest@.
457  *              XChaCha works by XORing plaintext with a keystream, so
458  *              encryption and decryption are the same operation.  If @dest@
459  *              is null then ignore @src@ and skip @sz@ bytes of the
460  *              keystream.  If @src@ is null, then just write the keystream
461  *              to @dest@.
462  */
463
464 #define DEFXPASSTHRU(r)                                                 \
465   void XCHACHA_SEEK(r, XCHACHA_CTX(r) *ctx, unsigned long i)            \
466     { chacha_seek(&ctx->s, i); }                                        \
467   void XCHACHA_SEEKU64(r, XCHACHA_CTX(r) *ctx, kludge64 i)              \
468     { chacha_seeku64(&ctx->s, i); }                                     \
469   unsigned long XCHACHA_TELL(r, XCHACHA_CTX(r) *ctx)                    \
470     { return chacha_tell(&ctx->s); }                                    \
471   kludge64 XCHACHA_TELLU64(r, XCHACHA_CTX(r) *ctx)                      \
472     { return chacha_tellu64(&ctx->s); }                                 \
473   void XCHACHA_ENCRYPT(r, XCHACHA_CTX(r) *ctx,                          \
474                         const void *src, void *dest, size_t sz)         \
475     { CHACHA_ENCRYPT(r, &ctx->s, src, dest, sz); }
476 CHACHA_VARS(DEFXPASSTHRU)
477
478 /*----- Generic cipher interface ------------------------------------------*/
479
480 typedef struct gctx { gcipher c; chacha_ctx ctx; } gctx;
481
482 static void gsetiv(gcipher *c, const void *iv)
483   { gctx *g = (gctx *)c; chacha_setnonce(&g->ctx, iv); }
484
485 static void gdestroy(gcipher *c)
486   { gctx *g = (gctx *)c; BURN(*g); S_DESTROY(g); }
487
488 #define DEFGCIPHER(r)                                                   \
489                                                                         \
490   static const gcipher_ops gops_##r;                                    \
491                                                                         \
492   static gcipher *ginit_##r(const void *k, size_t sz)                   \
493   {                                                                     \
494     gctx *g = S_CREATE(gctx);                                           \
495     g->c.ops = &gops_##r;                                               \
496     chacha_init(&g->ctx, k, sz, 0);                                     \
497     return (&g->c);                                                     \
498   }                                                                     \
499                                                                         \
500   static void gencrypt_##r(gcipher *c, const void *s,                   \
501                            void *t, size_t sz)                          \
502     { gctx *g = (gctx *)c; CHACHA_ENCRYPT(r, &g->ctx, s, t, sz); }      \
503                                                                         \
504   static const gcipher_ops gops_##r = {                                 \
505     &chacha##r,                                                         \
506     gencrypt_##r, gencrypt_##r, gdestroy, gsetiv, 0                     \
507   };                                                                    \
508                                                                         \
509   const gccipher chacha##r = {                                          \
510     "chacha" #r, chacha_keysz,                                          \
511     CHACHA_NONCESZ, ginit_##r                                           \
512   };
513
514 CHACHA_VARS(DEFGCIPHER)
515
516 #define DEFGXCIPHER(r)                                                  \
517                                                                         \
518   typedef struct { gcipher c; XCHACHA_CTX(r) ctx; } gxctx_##r;          \
519                                                                         \
520   static void gxsetiv_##r(gcipher *c, const void *iv)                   \
521     { gxctx_##r *g = (gxctx_##r *)c; XCHACHA_SETNONCE(r, &g->ctx, iv); } \
522                                                                         \
523   static void gxdestroy_##r(gcipher *c)                                 \
524     { gxctx_##r *g = (gxctx_##r *)c; BURN(*g); S_DESTROY(g); }          \
525                                                                         \
526   static const gcipher_ops gxops_##r;                                   \
527                                                                         \
528   static gcipher *gxinit_##r(const void *k, size_t sz)                  \
529   {                                                                     \
530     gxctx_##r *g = S_CREATE(gxctx_##r);                                 \
531     g->c.ops = &gxops_##r;                                              \
532     XCHACHA_INIT(r, &g->ctx, k, sz, 0);                                 \
533     return (&g->c);                                                     \
534   }                                                                     \
535                                                                         \
536   static void gxencrypt_##r(gcipher *c, const void *s,                  \
537                             void *t, size_t sz)                         \
538   {                                                                     \
539     gxctx_##r *g = (gxctx_##r *)c;                                      \
540     XCHACHA_ENCRYPT(r, &g->ctx, s, t, sz);                              \
541   }                                                                     \
542                                                                         \
543   static const gcipher_ops gxops_##r = {                                \
544     &xchacha##r,                                                        \
545     gxencrypt_##r, gxencrypt_##r, gxdestroy_##r, gxsetiv_##r, 0         \
546   };                                                                    \
547                                                                         \
548   const gccipher xchacha##r = {                                         \
549     "xchacha" #r, chacha_keysz,                                         \
550     CHACHA_NONCESZ, gxinit_##r                                          \
551   };
552
553 CHACHA_VARS(DEFGXCIPHER)
554
555 /*----- Generic random number generator interface -------------------------*/
556
557 typedef struct grops {
558   size_t noncesz;
559   void (*seek)(void *, kludge64);
560   kludge64 (*tell)(void *);
561   void (*setnonce)(void *, const void *);
562   void (*generate)(void *, void *, size_t);
563 } grops;
564
565 typedef struct grbasectx {
566   grand r;
567   const grops *ops;
568 } grbasectx;
569
570 static int grmisc(grand *r, unsigned op, ...)
571 {
572   octet buf[XCHACHA_NONCESZ];
573   grbasectx *g = (grbasectx *)r;
574   grand *rr;
575   const octet *p;
576   size_t sz;
577   uint32 i;
578   unsigned long ul;
579   kludge64 pos;
580   va_list ap;
581   int rc = 0;
582
583   va_start(ap, op);
584
585   switch (op) {
586     case GRAND_CHECK:
587       switch (va_arg(ap, unsigned)) {
588         case GRAND_CHECK:
589         case GRAND_SEEDINT:
590         case GRAND_SEEDUINT32:
591         case GRAND_SEEDBLOCK:
592         case GRAND_SEEDRAND:
593         case CHACHA_SEEK:
594         case CHACHA_SEEKU64:
595         case CHACHA_TELL:
596         case CHACHA_TELLU64:
597           rc = 1;
598           break;
599         default:
600           rc = 0;
601           break;
602       }
603       break;
604
605     case GRAND_SEEDINT:
606       i = va_arg(ap, unsigned); STORE32_L(buf, i);
607       memset(buf + 4, 0, g->ops->noncesz - 4);
608       g->ops->setnonce(g, buf);
609       break;
610     case GRAND_SEEDUINT32:
611       i = va_arg(ap, uint32); STORE32_L(buf, i);
612       memset(buf + 4, 0, g->ops->noncesz - 4);
613       g->ops->setnonce(g, buf);
614       break;
615     case GRAND_SEEDBLOCK:
616       p = va_arg(ap, const void *);
617       sz = va_arg(ap, size_t);
618       if (sz < g->ops->noncesz) {
619         memcpy(buf, p, sz);
620         memset(buf + sz, 0, g->ops->noncesz - sz);
621         p = buf;
622       }
623       g->ops->setnonce(g, p);
624       break;
625     case GRAND_SEEDRAND:
626       rr = va_arg(ap, grand *);
627       rr->ops->fill(rr, buf, g->ops->noncesz);
628       g->ops->setnonce(g, buf);
629       break;
630     case CHACHA_SEEK:
631       ul = va_arg(ap, unsigned long); ASSIGN64(pos, ul);
632       g->ops->seek(g, pos);
633       break;
634     case CHACHA_SEEKU64:
635       pos = va_arg(ap, kludge64);
636       g->ops->seek(g, pos);
637       break;
638     case CHACHA_TELL:
639       pos = g->ops->tell(g);
640       *va_arg(ap, unsigned long *) = GET64(unsigned long, pos);
641       break;
642     case CHACHA_TELLU64:
643       *va_arg(ap, kludge64 *) = g->ops->tell(g);
644       break;
645     default:
646       GRAND_BADOP;
647       break;
648   }
649
650   return (rc);
651 }
652
653 static octet grbyte(grand *r)
654 {
655   grbasectx *g = (grbasectx *)r;
656   octet o;
657   g->ops->generate(g, &o, 1);
658   return (o);
659 }
660
661 static uint32 grword(grand *r)
662 {
663   grbasectx *g = (grbasectx *)r;
664   octet b[4];
665   g->ops->generate(g, b, sizeof(b));
666   return (LOAD32_L(b));
667 }
668
669 static void grfill(grand *r, void *p, size_t sz)
670 {
671   grbasectx *g = (grbasectx *)r;
672   g->ops->generate(r, p, sz);
673 }
674
675 typedef struct grctx {
676   grbasectx r;
677   chacha_ctx ctx;
678 } grctx;
679
680 static void gr_seek(void *r, kludge64 pos)
681   { grctx *g = r; chacha_seeku64(&g->ctx, pos); }
682
683 static kludge64 gr_tell(void *r)
684   { grctx *g = r; return (chacha_tellu64(&g->ctx)); }
685
686 static void gr_setnonce(void *r, const void *n)
687   { grctx *g = r; chacha_setnonce(&g->ctx, n); }
688
689 static void grdestroy(grand *r)
690   { grctx *g = (grctx *)r; BURN(*g); S_DESTROY(g); }
691
692 #define DEFGRAND(rr)                                                    \
693                                                                         \
694   static void gr_generate_##rr(void *r, void *b, size_t sz)             \
695     { grctx *g = r; CHACHA_ENCRYPT(rr, &g->ctx, 0, b, sz); }            \
696                                                                         \
697   static const grops grops_##rr =                                       \
698     { CHACHA_NONCESZ, gr_seek, gr_tell,                                 \
699       gr_setnonce, gr_generate_##rr };                                  \
700                                                                         \
701   static const grand_ops grops_rand_##rr = {                            \
702     "chacha" #rr, GRAND_CRYPTO, 0,                                      \
703     grmisc, grdestroy, grword,                                          \
704     grbyte, grword, grand_defaultrange, grfill                          \
705   };                                                                    \
706                                                                         \
707   grand *chacha##rr##_rand(const void *k, size_t ksz, const void *n)    \
708   {                                                                     \
709     grctx *g = S_CREATE(g);                                             \
710     g->r.r.ops = &grops_rand_##rr;                                      \
711     g->r.ops = &grops_##rr;                                             \
712     chacha_init(&g->ctx, k, ksz, n);                                    \
713     return (&g->r.r);                                                   \
714   }
715 CHACHA_VARS(DEFGRAND)
716
717 #define DEFXGRAND(rr)                                                   \
718                                                                         \
719   typedef struct grxctx_##rr {                                          \
720     grbasectx r;                                                        \
721     XCHACHA_CTX(rr) ctx;                                                \
722   } grxctx_##rr;                                                        \
723                                                                         \
724   static void grx_seek_##rr(void *r, kludge64 pos)                      \
725     { grxctx_##rr *g = r; XCHACHA_SEEKU64(rr, &g->ctx, pos); }          \
726                                                                         \
727   static kludge64 grx_tell_##rr(void *r)                                \
728     { grxctx_##rr *g = r; return (XCHACHA_TELLU64(rr, &g->ctx)); }      \
729                                                                         \
730   static void grx_setnonce_##rr(void *r, const void *n)                 \
731     { grxctx_##rr *g = r; XCHACHA_SETNONCE(rr, &g->ctx, n); }           \
732                                                                         \
733   static void grxdestroy_##rr(grand *r)                                 \
734     { grxctx_##rr *g = (grxctx_##rr *)r; BURN(*g); S_DESTROY(g); }      \
735                                                                         \
736   static void grx_generate_##rr(void *r, void *b, size_t sz)            \
737     { grxctx_##rr *g = r; XCHACHA_ENCRYPT(rr, &g->ctx, 0, b, sz); }     \
738                                                                         \
739   static const grops grxops_##rr =                                      \
740     { XCHACHA_NONCESZ, grx_seek_##rr, grx_tell_##rr,                    \
741       grx_setnonce_##rr, grx_generate_##rr };                           \
742                                                                         \
743   static const grand_ops grxops_rand_##rr = {                           \
744     "xchacha" #rr, GRAND_CRYPTO, 0,                                     \
745     grmisc, grxdestroy_##rr, grword,                                    \
746     grbyte, grword, grand_defaultrange, grfill                          \
747   };                                                                    \
748                                                                         \
749   grand *xchacha##rr##_rand(const void *k, size_t ksz, const void *n)   \
750   {                                                                     \
751     grxctx_##rr *g = S_CREATE(g);                                       \
752     g->r.r.ops = &grxops_rand_##rr;                                     \
753     g->r.ops = &grxops_##rr;                                            \
754     XCHACHA_INIT(rr, &g->ctx, k, ksz, n);                               \
755     return (&g->r.r);                                                   \
756   }
757 CHACHA_VARS(DEFXGRAND)
758
759 /*----- Test rig ----------------------------------------------------------*/
760
761 #ifdef TEST_RIG
762
763 #include <stdio.h>
764 #include <string.h>
765
766 #include <mLib/quis.h>
767 #include <mLib/testrig.h>
768
769 #define DEFVCORE(r)                                                     \
770   static int v_core_##r(dstr *v)                                        \
771   {                                                                     \
772     chacha_matrix a, b;                                                 \
773     dstr d = DSTR_INIT;                                                 \
774     int i, n;                                                           \
775     int ok = 1;                                                         \
776                                                                         \
777     DENSURE(&d, CHACHA_OUTSZ); d.len = CHACHA_OUTSZ;                    \
778     n = *(int *)v[0].buf;                                               \
779     for (i = 0; i < CHACHA_OUTSZ/4; i++)                                \
780       a[i] = LOAD32_L(v[1].buf + 4*i);                                  \
781     for (i = 0; i < n; i++) {                                           \
782       core(r, a, b);                                                    \
783       memcpy(a, b, sizeof(a));                                          \
784     }                                                                   \
785     for (i = 0; i < CHACHA_OUTSZ/4; i++) STORE32_L(d.buf + 4*i, a[i]);  \
786                                                                         \
787     if (d.len != v[2].len || memcmp(d.buf, v[2].buf, v[2].len) != 0) {  \
788       ok = 0;                                                           \
789       printf("\nfail core:"                                             \
790              "\n\titerations = %d"                                      \
791              "\n\tin       = ", n);                                     \
792       type_hex.dump(&v[1], stdout);                                     \
793       printf("\n\texpected   = ");                                      \
794       type_hex.dump(&v[2], stdout);                                     \
795       printf("\n\tcalculated = ");                                      \
796       type_hex.dump(&d, stdout);                                        \
797       putchar('\n');                                                    \
798     }                                                                   \
799                                                                         \
800     dstr_destroy(&d);                                                   \
801     return (ok);                                                        \
802   }
803 CHACHA_VARS(DEFVCORE)
804
805 #define CHACHA_CTX(r) chacha_ctx
806 #define CHACHA_INIT(r, ctx, k, ksz, n) chacha_init(ctx, k, ksz, n)
807 #define CHACHA_SEEKU64(r, ctx, i) chacha_seeku64(ctx, i)
808 #define XCHACHA_SEEKU64(r, ctx, i) xchacha##r##_seeku64(ctx, i)
809
810 #define DEFxVENC(base, BASE, r)                                         \
811   static int v_encrypt_##base##_##r(dstr *v)                            \
812   {                                                                     \
813     BASE##_CTX(r) ctx;                                                  \
814     dstr d = DSTR_INIT;                                                 \
815     kludge64 pos;                                                       \
816     const octet *p, *p0;                                                \
817     octet *q;                                                           \
818     size_t sz, sz0, step;                                               \
819     unsigned long skip;                                                 \
820     int ok = 1;                                                         \
821                                                                         \
822     if (v[4].len) { p0 = (const octet *)v[4].buf; sz0 = v[4].len; }     \
823     else { p0 = 0; sz0 = v[5].len; }                                    \
824     DENSURE(&d, sz0); d.len = sz0;                                      \
825     skip = *(unsigned long *)v[3].buf;                                  \
826                                                                         \
827     step = 0;                                                           \
828     while (step < sz0 + skip) {                                         \
829       step = step ? 3*step + 4 : 1;                                     \
830       if (step > sz0 + skip) step = sz0 + skip;                         \
831       BASE##_INIT(r, &ctx, v[0].buf, v[0].len, v[1].buf);               \
832       if (v[2].len) {                                                   \
833         LOAD64_(pos, v[2].buf);                                         \
834         BASE##_SEEKU64(r, &ctx, pos);                                   \
835       }                                                                 \
836                                                                         \
837       for (sz = skip; sz >= step; sz -= step)                           \
838         BASE##_ENCRYPT(r, &ctx, 0, 0, step);                            \
839       if (sz) BASE##_ENCRYPT(r, &ctx, 0, 0, sz);                        \
840       for (p = p0, q = (octet *)d.buf, sz = sz0;                        \
841            sz >= step;                                                  \
842            sz -= step, q += step) {                                     \
843         BASE##_ENCRYPT(r, &ctx, p, q, step);                            \
844         if (p) p += step;                                               \
845       }                                                                 \
846       if (sz) BASE##_ENCRYPT(r, &ctx, p, q, sz);                        \
847                                                                         \
848       if (d.len != v[5].len || memcmp(d.buf, v[5].buf, v[5].len) != 0) { \
849         ok = 0;                                                         \
850         printf("\nfail encrypt:"                                        \
851                "\n\tstep           = %lu"                               \
852                "\n\tkey    = ", (unsigned long)step);                   \
853         type_hex.dump(&v[0], stdout);                                   \
854         printf("\n\tnonce          = ");                                \
855         type_hex.dump(&v[1], stdout);                                   \
856         printf("\n\tposition   = ");                                    \
857         type_hex.dump(&v[2], stdout);                                   \
858         printf("\n\tskip           = %lu", skip);                       \
859         printf("\n\tmessage    = ");                                    \
860         type_hex.dump(&v[4], stdout);                                   \
861         printf("\n\texpected   = ");                                    \
862         type_hex.dump(&v[5], stdout);                                   \
863         printf("\n\tcalculated = ");                                    \
864         type_hex.dump(&d, stdout);                                      \
865         putchar('\n');                                                  \
866       }                                                                 \
867     }                                                                   \
868                                                                         \
869     dstr_destroy(&d);                                                   \
870     return (ok);                                                        \
871   }
872 #define DEFVENC(r) DEFxVENC(chacha, CHACHA, r)
873 #define DEFXVENC(r) DEFxVENC(xchacha, XCHACHA, r)
874 CHACHA_VARS(DEFVENC)
875 CHACHA_VARS(DEFXVENC)
876
877 static test_chunk defs[] = {
878 #define DEFxTAB(base, r)                                                \
879   { #base #r, v_encrypt_##base##_##r,                                   \
880     { &type_hex, &type_hex, &type_hex, &type_ulong,                     \
881       &type_hex, &type_hex, 0 } },
882 #define DEFTAB(r)                                                       \
883   { "chacha" #r "-core", v_core_##r,                                    \
884     { &type_int, &type_hex, &type_hex, 0 } },                           \
885   DEFxTAB(chacha, r)
886 #define DEFXTAB(r) DEFxTAB(xchacha, r)
887 CHACHA_VARS(DEFTAB)
888 CHACHA_VARS(DEFXTAB)
889   { 0, 0, { 0 } }
890 };
891
892 int main(int argc, char *argv[])
893 {
894   test_run(argc, argv, defs, SRCDIR"/t/chacha");
895   return (0);
896 }
897
898 #endif
899
900 /*----- That's all, folks -------------------------------------------------*/