]>
Commit | Line | Data |
---|---|---|
107778b5 JG |
1 | /* |
2 | * Glue Code for AVX assembler version of Twofish Cipher | |
3 | * | |
4 | * Copyright (C) 2012 Johannes Goetzfried | |
5 | * <Johannes.Goetzfried@informatik.stud.uni-erlangen.de> | |
6 | * | |
7 | * Glue code based on serpent_sse2_glue.c by: | |
8 | * Copyright (C) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi> | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or modify | |
11 | * it under the terms of the GNU General Public License as published by | |
12 | * the Free Software Foundation; either version 2 of the License, or | |
13 | * (at your option) any later version. | |
14 | * | |
15 | * This program is distributed in the hope that it will be useful, | |
16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | * GNU General Public License for more details. | |
19 | * | |
20 | * You should have received a copy of the GNU General Public License | |
21 | * along with this program; if not, write to the Free Software | |
22 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 | |
23 | * USA | |
24 | * | |
25 | */ | |
26 | ||
27 | #include <linux/module.h> | |
28 | #include <linux/hardirq.h> | |
29 | #include <linux/types.h> | |
30 | #include <linux/crypto.h> | |
31 | #include <linux/err.h> | |
32 | #include <crypto/algapi.h> | |
33 | #include <crypto/twofish.h> | |
34 | #include <crypto/cryptd.h> | |
35 | #include <crypto/b128ops.h> | |
36 | #include <crypto/ctr.h> | |
37 | #include <crypto/lrw.h> | |
38 | #include <crypto/xts.h> | |
39 | #include <asm/i387.h> | |
40 | #include <asm/xcr.h> | |
41 | #include <asm/xsave.h> | |
30a04008 | 42 | #include <asm/crypto/ablk_helper.h> |
107778b5 JG |
43 | #include <crypto/scatterwalk.h> |
44 | #include <linux/workqueue.h> | |
45 | #include <linux/spinlock.h> | |
46 | ||
47 | ||
48 | #define TWOFISH_PARALLEL_BLOCKS 8 | |
49 | ||
50 | /* regular block cipher functions from twofish_x86_64 module */ | |
51 | asmlinkage void twofish_enc_blk(struct twofish_ctx *ctx, u8 *dst, | |
52 | const u8 *src); | |
53 | asmlinkage void twofish_dec_blk(struct twofish_ctx *ctx, u8 *dst, | |
54 | const u8 *src); | |
55 | ||
56 | /* 3-way parallel cipher functions from twofish_x86_64-3way module */ | |
57 | asmlinkage void __twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, | |
58 | const u8 *src, bool xor); | |
59 | asmlinkage void twofish_dec_blk_3way(struct twofish_ctx *ctx, u8 *dst, | |
60 | const u8 *src); | |
61 | ||
62 | static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, | |
63 | const u8 *src) | |
64 | { | |
65 | __twofish_enc_blk_3way(ctx, dst, src, false); | |
66 | } | |
67 | ||
68 | static inline void twofish_enc_blk_3way_xor(struct twofish_ctx *ctx, u8 *dst, | |
69 | const u8 *src) | |
70 | { | |
71 | __twofish_enc_blk_3way(ctx, dst, src, true); | |
72 | } | |
73 | ||
74 | /* 8-way parallel cipher functions */ | |
75 | asmlinkage void __twofish_enc_blk_8way(struct twofish_ctx *ctx, u8 *dst, | |
76 | const u8 *src, bool xor); | |
77 | asmlinkage void twofish_dec_blk_8way(struct twofish_ctx *ctx, u8 *dst, | |
78 | const u8 *src); | |
79 | ||
80 | static inline void twofish_enc_blk_xway(struct twofish_ctx *ctx, u8 *dst, | |
81 | const u8 *src) | |
82 | { | |
83 | __twofish_enc_blk_8way(ctx, dst, src, false); | |
84 | } | |
85 | ||
86 | static inline void twofish_enc_blk_xway_xor(struct twofish_ctx *ctx, u8 *dst, | |
87 | const u8 *src) | |
88 | { | |
89 | __twofish_enc_blk_8way(ctx, dst, src, true); | |
90 | } | |
91 | ||
92 | static inline void twofish_dec_blk_xway(struct twofish_ctx *ctx, u8 *dst, | |
93 | const u8 *src) | |
94 | { | |
95 | twofish_dec_blk_8way(ctx, dst, src); | |
96 | } | |
97 | ||
98 | ||
107778b5 JG |
99 | static inline bool twofish_fpu_begin(bool fpu_enabled, unsigned int nbytes) |
100 | { | |
101 | if (fpu_enabled) | |
102 | return true; | |
103 | ||
104 | /* AVX is only used when chunk to be processed is large enough, so | |
105 | * do not enable FPU until it is necessary. | |
106 | */ | |
107 | if (nbytes < TF_BLOCK_SIZE * TWOFISH_PARALLEL_BLOCKS) | |
108 | return false; | |
109 | ||
110 | kernel_fpu_begin(); | |
111 | return true; | |
112 | } | |
113 | ||
114 | static inline void twofish_fpu_end(bool fpu_enabled) | |
115 | { | |
116 | if (fpu_enabled) | |
117 | kernel_fpu_end(); | |
118 | } | |
119 | ||
120 | static int ecb_crypt(struct blkcipher_desc *desc, struct blkcipher_walk *walk, | |
121 | bool enc) | |
122 | { | |
123 | bool fpu_enabled = false; | |
124 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
125 | const unsigned int bsize = TF_BLOCK_SIZE; | |
126 | unsigned int nbytes; | |
127 | int err; | |
128 | ||
129 | err = blkcipher_walk_virt(desc, walk); | |
130 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
131 | ||
132 | while ((nbytes = walk->nbytes)) { | |
133 | u8 *wsrc = walk->src.virt.addr; | |
134 | u8 *wdst = walk->dst.virt.addr; | |
135 | ||
136 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | |
137 | ||
138 | /* Process multi-block batch */ | |
139 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | |
140 | do { | |
141 | if (enc) | |
142 | twofish_enc_blk_xway(ctx, wdst, wsrc); | |
143 | else | |
144 | twofish_dec_blk_xway(ctx, wdst, wsrc); | |
145 | ||
146 | wsrc += bsize * TWOFISH_PARALLEL_BLOCKS; | |
147 | wdst += bsize * TWOFISH_PARALLEL_BLOCKS; | |
148 | nbytes -= bsize * TWOFISH_PARALLEL_BLOCKS; | |
149 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | |
150 | ||
151 | if (nbytes < bsize) | |
152 | goto done; | |
153 | } | |
154 | ||
155 | /* Process three block batch */ | |
156 | if (nbytes >= bsize * 3) { | |
157 | do { | |
158 | if (enc) | |
159 | twofish_enc_blk_3way(ctx, wdst, wsrc); | |
160 | else | |
161 | twofish_dec_blk_3way(ctx, wdst, wsrc); | |
162 | ||
163 | wsrc += bsize * 3; | |
164 | wdst += bsize * 3; | |
165 | nbytes -= bsize * 3; | |
166 | } while (nbytes >= bsize * 3); | |
167 | ||
168 | if (nbytes < bsize) | |
169 | goto done; | |
170 | } | |
171 | ||
172 | /* Handle leftovers */ | |
173 | do { | |
174 | if (enc) | |
175 | twofish_enc_blk(ctx, wdst, wsrc); | |
176 | else | |
177 | twofish_dec_blk(ctx, wdst, wsrc); | |
178 | ||
179 | wsrc += bsize; | |
180 | wdst += bsize; | |
181 | nbytes -= bsize; | |
182 | } while (nbytes >= bsize); | |
183 | ||
184 | done: | |
185 | err = blkcipher_walk_done(desc, walk, nbytes); | |
186 | } | |
187 | ||
188 | twofish_fpu_end(fpu_enabled); | |
189 | return err; | |
190 | } | |
191 | ||
192 | static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
193 | struct scatterlist *src, unsigned int nbytes) | |
194 | { | |
195 | struct blkcipher_walk walk; | |
196 | ||
197 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
198 | return ecb_crypt(desc, &walk, true); | |
199 | } | |
200 | ||
201 | static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
202 | struct scatterlist *src, unsigned int nbytes) | |
203 | { | |
204 | struct blkcipher_walk walk; | |
205 | ||
206 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
207 | return ecb_crypt(desc, &walk, false); | |
208 | } | |
209 | ||
210 | static unsigned int __cbc_encrypt(struct blkcipher_desc *desc, | |
211 | struct blkcipher_walk *walk) | |
212 | { | |
213 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
214 | const unsigned int bsize = TF_BLOCK_SIZE; | |
215 | unsigned int nbytes = walk->nbytes; | |
216 | u128 *src = (u128 *)walk->src.virt.addr; | |
217 | u128 *dst = (u128 *)walk->dst.virt.addr; | |
218 | u128 *iv = (u128 *)walk->iv; | |
219 | ||
220 | do { | |
221 | u128_xor(dst, src, iv); | |
222 | twofish_enc_blk(ctx, (u8 *)dst, (u8 *)dst); | |
223 | iv = dst; | |
224 | ||
225 | src += 1; | |
226 | dst += 1; | |
227 | nbytes -= bsize; | |
228 | } while (nbytes >= bsize); | |
229 | ||
230 | u128_xor((u128 *)walk->iv, (u128 *)walk->iv, iv); | |
231 | return nbytes; | |
232 | } | |
233 | ||
234 | static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
235 | struct scatterlist *src, unsigned int nbytes) | |
236 | { | |
237 | struct blkcipher_walk walk; | |
238 | int err; | |
239 | ||
240 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
241 | err = blkcipher_walk_virt(desc, &walk); | |
242 | ||
243 | while ((nbytes = walk.nbytes)) { | |
244 | nbytes = __cbc_encrypt(desc, &walk); | |
245 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
246 | } | |
247 | ||
248 | return err; | |
249 | } | |
250 | ||
251 | static unsigned int __cbc_decrypt(struct blkcipher_desc *desc, | |
252 | struct blkcipher_walk *walk) | |
253 | { | |
254 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
255 | const unsigned int bsize = TF_BLOCK_SIZE; | |
256 | unsigned int nbytes = walk->nbytes; | |
257 | u128 *src = (u128 *)walk->src.virt.addr; | |
258 | u128 *dst = (u128 *)walk->dst.virt.addr; | |
259 | u128 ivs[TWOFISH_PARALLEL_BLOCKS - 1]; | |
260 | u128 last_iv; | |
261 | int i; | |
262 | ||
263 | /* Start of the last block. */ | |
264 | src += nbytes / bsize - 1; | |
265 | dst += nbytes / bsize - 1; | |
266 | ||
267 | last_iv = *src; | |
268 | ||
269 | /* Process multi-block batch */ | |
270 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | |
271 | do { | |
272 | nbytes -= bsize * (TWOFISH_PARALLEL_BLOCKS - 1); | |
273 | src -= TWOFISH_PARALLEL_BLOCKS - 1; | |
274 | dst -= TWOFISH_PARALLEL_BLOCKS - 1; | |
275 | ||
276 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS - 1; i++) | |
277 | ivs[i] = src[i]; | |
278 | ||
279 | twofish_dec_blk_xway(ctx, (u8 *)dst, (u8 *)src); | |
280 | ||
281 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS - 1; i++) | |
282 | u128_xor(dst + (i + 1), dst + (i + 1), ivs + i); | |
283 | ||
284 | nbytes -= bsize; | |
285 | if (nbytes < bsize) | |
286 | goto done; | |
287 | ||
288 | u128_xor(dst, dst, src - 1); | |
289 | src -= 1; | |
290 | dst -= 1; | |
291 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | |
292 | ||
293 | if (nbytes < bsize) | |
294 | goto done; | |
295 | } | |
296 | ||
297 | /* Process three block batch */ | |
298 | if (nbytes >= bsize * 3) { | |
299 | do { | |
300 | nbytes -= bsize * (3 - 1); | |
301 | src -= 3 - 1; | |
302 | dst -= 3 - 1; | |
303 | ||
304 | ivs[0] = src[0]; | |
305 | ivs[1] = src[1]; | |
306 | ||
307 | twofish_dec_blk_3way(ctx, (u8 *)dst, (u8 *)src); | |
308 | ||
309 | u128_xor(dst + 1, dst + 1, ivs + 0); | |
310 | u128_xor(dst + 2, dst + 2, ivs + 1); | |
311 | ||
312 | nbytes -= bsize; | |
313 | if (nbytes < bsize) | |
314 | goto done; | |
315 | ||
316 | u128_xor(dst, dst, src - 1); | |
317 | src -= 1; | |
318 | dst -= 1; | |
319 | } while (nbytes >= bsize * 3); | |
320 | ||
321 | if (nbytes < bsize) | |
322 | goto done; | |
323 | } | |
324 | ||
325 | /* Handle leftovers */ | |
326 | for (;;) { | |
327 | twofish_dec_blk(ctx, (u8 *)dst, (u8 *)src); | |
328 | ||
329 | nbytes -= bsize; | |
330 | if (nbytes < bsize) | |
331 | break; | |
332 | ||
333 | u128_xor(dst, dst, src - 1); | |
334 | src -= 1; | |
335 | dst -= 1; | |
336 | } | |
337 | ||
338 | done: | |
339 | u128_xor(dst, dst, (u128 *)walk->iv); | |
340 | *(u128 *)walk->iv = last_iv; | |
341 | ||
342 | return nbytes; | |
343 | } | |
344 | ||
345 | static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
346 | struct scatterlist *src, unsigned int nbytes) | |
347 | { | |
348 | bool fpu_enabled = false; | |
349 | struct blkcipher_walk walk; | |
350 | int err; | |
351 | ||
352 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
353 | err = blkcipher_walk_virt(desc, &walk); | |
354 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
355 | ||
356 | while ((nbytes = walk.nbytes)) { | |
357 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | |
358 | nbytes = __cbc_decrypt(desc, &walk); | |
359 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
360 | } | |
361 | ||
362 | twofish_fpu_end(fpu_enabled); | |
363 | return err; | |
364 | } | |
365 | ||
366 | static inline void u128_to_be128(be128 *dst, const u128 *src) | |
367 | { | |
368 | dst->a = cpu_to_be64(src->a); | |
369 | dst->b = cpu_to_be64(src->b); | |
370 | } | |
371 | ||
372 | static inline void be128_to_u128(u128 *dst, const be128 *src) | |
373 | { | |
374 | dst->a = be64_to_cpu(src->a); | |
375 | dst->b = be64_to_cpu(src->b); | |
376 | } | |
377 | ||
378 | static inline void u128_inc(u128 *i) | |
379 | { | |
380 | i->b++; | |
381 | if (!i->b) | |
382 | i->a++; | |
383 | } | |
384 | ||
385 | static void ctr_crypt_final(struct blkcipher_desc *desc, | |
386 | struct blkcipher_walk *walk) | |
387 | { | |
388 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
389 | u8 *ctrblk = walk->iv; | |
390 | u8 keystream[TF_BLOCK_SIZE]; | |
391 | u8 *src = walk->src.virt.addr; | |
392 | u8 *dst = walk->dst.virt.addr; | |
393 | unsigned int nbytes = walk->nbytes; | |
394 | ||
395 | twofish_enc_blk(ctx, keystream, ctrblk); | |
396 | crypto_xor(keystream, src, nbytes); | |
397 | memcpy(dst, keystream, nbytes); | |
398 | ||
399 | crypto_inc(ctrblk, TF_BLOCK_SIZE); | |
400 | } | |
401 | ||
402 | static unsigned int __ctr_crypt(struct blkcipher_desc *desc, | |
403 | struct blkcipher_walk *walk) | |
404 | { | |
405 | struct twofish_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
406 | const unsigned int bsize = TF_BLOCK_SIZE; | |
407 | unsigned int nbytes = walk->nbytes; | |
408 | u128 *src = (u128 *)walk->src.virt.addr; | |
409 | u128 *dst = (u128 *)walk->dst.virt.addr; | |
410 | u128 ctrblk; | |
411 | be128 ctrblocks[TWOFISH_PARALLEL_BLOCKS]; | |
412 | int i; | |
413 | ||
414 | be128_to_u128(&ctrblk, (be128 *)walk->iv); | |
415 | ||
416 | /* Process multi-block batch */ | |
417 | if (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS) { | |
418 | do { | |
419 | /* create ctrblks for parallel encrypt */ | |
420 | for (i = 0; i < TWOFISH_PARALLEL_BLOCKS; i++) { | |
421 | if (dst != src) | |
422 | dst[i] = src[i]; | |
423 | ||
424 | u128_to_be128(&ctrblocks[i], &ctrblk); | |
425 | u128_inc(&ctrblk); | |
426 | } | |
427 | ||
428 | twofish_enc_blk_xway_xor(ctx, (u8 *)dst, | |
429 | (u8 *)ctrblocks); | |
430 | ||
431 | src += TWOFISH_PARALLEL_BLOCKS; | |
432 | dst += TWOFISH_PARALLEL_BLOCKS; | |
433 | nbytes -= bsize * TWOFISH_PARALLEL_BLOCKS; | |
434 | } while (nbytes >= bsize * TWOFISH_PARALLEL_BLOCKS); | |
435 | ||
436 | if (nbytes < bsize) | |
437 | goto done; | |
438 | } | |
439 | ||
440 | /* Process three block batch */ | |
441 | if (nbytes >= bsize * 3) { | |
442 | do { | |
443 | if (dst != src) { | |
444 | dst[0] = src[0]; | |
445 | dst[1] = src[1]; | |
446 | dst[2] = src[2]; | |
447 | } | |
448 | ||
449 | /* create ctrblks for parallel encrypt */ | |
450 | u128_to_be128(&ctrblocks[0], &ctrblk); | |
451 | u128_inc(&ctrblk); | |
452 | u128_to_be128(&ctrblocks[1], &ctrblk); | |
453 | u128_inc(&ctrblk); | |
454 | u128_to_be128(&ctrblocks[2], &ctrblk); | |
455 | u128_inc(&ctrblk); | |
456 | ||
457 | twofish_enc_blk_3way_xor(ctx, (u8 *)dst, | |
458 | (u8 *)ctrblocks); | |
459 | ||
460 | src += 3; | |
461 | dst += 3; | |
462 | nbytes -= bsize * 3; | |
463 | } while (nbytes >= bsize * 3); | |
464 | ||
465 | if (nbytes < bsize) | |
466 | goto done; | |
467 | } | |
468 | ||
469 | /* Handle leftovers */ | |
470 | do { | |
471 | if (dst != src) | |
472 | *dst = *src; | |
473 | ||
474 | u128_to_be128(&ctrblocks[0], &ctrblk); | |
475 | u128_inc(&ctrblk); | |
476 | ||
477 | twofish_enc_blk(ctx, (u8 *)ctrblocks, (u8 *)ctrblocks); | |
478 | u128_xor(dst, dst, (u128 *)ctrblocks); | |
479 | ||
480 | src += 1; | |
481 | dst += 1; | |
482 | nbytes -= bsize; | |
483 | } while (nbytes >= bsize); | |
484 | ||
485 | done: | |
486 | u128_to_be128((be128 *)walk->iv, &ctrblk); | |
487 | return nbytes; | |
488 | } | |
489 | ||
490 | static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
491 | struct scatterlist *src, unsigned int nbytes) | |
492 | { | |
493 | bool fpu_enabled = false; | |
494 | struct blkcipher_walk walk; | |
495 | int err; | |
496 | ||
497 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
498 | err = blkcipher_walk_virt_block(desc, &walk, TF_BLOCK_SIZE); | |
499 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
500 | ||
501 | while ((nbytes = walk.nbytes) >= TF_BLOCK_SIZE) { | |
502 | fpu_enabled = twofish_fpu_begin(fpu_enabled, nbytes); | |
503 | nbytes = __ctr_crypt(desc, &walk); | |
504 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
505 | } | |
506 | ||
507 | twofish_fpu_end(fpu_enabled); | |
508 | ||
509 | if (walk.nbytes) { | |
510 | ctr_crypt_final(desc, &walk); | |
511 | err = blkcipher_walk_done(desc, &walk, 0); | |
512 | } | |
513 | ||
514 | return err; | |
515 | } | |
516 | ||
517 | struct crypt_priv { | |
518 | struct twofish_ctx *ctx; | |
519 | bool fpu_enabled; | |
520 | }; | |
521 | ||
522 | static void encrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes) | |
523 | { | |
524 | const unsigned int bsize = TF_BLOCK_SIZE; | |
525 | struct crypt_priv *ctx = priv; | |
526 | int i; | |
527 | ||
528 | ctx->fpu_enabled = twofish_fpu_begin(ctx->fpu_enabled, nbytes); | |
529 | ||
530 | if (nbytes == bsize * TWOFISH_PARALLEL_BLOCKS) { | |
531 | twofish_enc_blk_xway(ctx->ctx, srcdst, srcdst); | |
532 | return; | |
533 | } | |
534 | ||
535 | for (i = 0; i < nbytes / (bsize * 3); i++, srcdst += bsize * 3) | |
536 | twofish_enc_blk_3way(ctx->ctx, srcdst, srcdst); | |
537 | ||
538 | nbytes %= bsize * 3; | |
539 | ||
540 | for (i = 0; i < nbytes / bsize; i++, srcdst += bsize) | |
541 | twofish_enc_blk(ctx->ctx, srcdst, srcdst); | |
542 | } | |
543 | ||
544 | static void decrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes) | |
545 | { | |
546 | const unsigned int bsize = TF_BLOCK_SIZE; | |
547 | struct crypt_priv *ctx = priv; | |
548 | int i; | |
549 | ||
550 | ctx->fpu_enabled = twofish_fpu_begin(ctx->fpu_enabled, nbytes); | |
551 | ||
552 | if (nbytes == bsize * TWOFISH_PARALLEL_BLOCKS) { | |
553 | twofish_dec_blk_xway(ctx->ctx, srcdst, srcdst); | |
554 | return; | |
555 | } | |
556 | ||
557 | for (i = 0; i < nbytes / (bsize * 3); i++, srcdst += bsize * 3) | |
558 | twofish_dec_blk_3way(ctx->ctx, srcdst, srcdst); | |
559 | ||
560 | nbytes %= bsize * 3; | |
561 | ||
562 | for (i = 0; i < nbytes / bsize; i++, srcdst += bsize) | |
563 | twofish_dec_blk(ctx->ctx, srcdst, srcdst); | |
564 | } | |
565 | ||
566 | struct twofish_lrw_ctx { | |
567 | struct lrw_table_ctx lrw_table; | |
568 | struct twofish_ctx twofish_ctx; | |
569 | }; | |
570 | ||
571 | static int lrw_twofish_setkey(struct crypto_tfm *tfm, const u8 *key, | |
572 | unsigned int keylen) | |
573 | { | |
574 | struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm); | |
575 | int err; | |
576 | ||
577 | err = __twofish_setkey(&ctx->twofish_ctx, key, | |
578 | keylen - TF_BLOCK_SIZE, &tfm->crt_flags); | |
579 | if (err) | |
580 | return err; | |
581 | ||
582 | return lrw_init_table(&ctx->lrw_table, key + keylen - | |
583 | TF_BLOCK_SIZE); | |
584 | } | |
585 | ||
586 | static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
587 | struct scatterlist *src, unsigned int nbytes) | |
588 | { | |
589 | struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
590 | be128 buf[TWOFISH_PARALLEL_BLOCKS]; | |
591 | struct crypt_priv crypt_ctx = { | |
592 | .ctx = &ctx->twofish_ctx, | |
593 | .fpu_enabled = false, | |
594 | }; | |
595 | struct lrw_crypt_req req = { | |
596 | .tbuf = buf, | |
597 | .tbuflen = sizeof(buf), | |
598 | ||
599 | .table_ctx = &ctx->lrw_table, | |
600 | .crypt_ctx = &crypt_ctx, | |
601 | .crypt_fn = encrypt_callback, | |
602 | }; | |
603 | int ret; | |
604 | ||
605 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
606 | ret = lrw_crypt(desc, dst, src, nbytes, &req); | |
607 | twofish_fpu_end(crypt_ctx.fpu_enabled); | |
608 | ||
609 | return ret; | |
610 | } | |
611 | ||
612 | static int lrw_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
613 | struct scatterlist *src, unsigned int nbytes) | |
614 | { | |
615 | struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
616 | be128 buf[TWOFISH_PARALLEL_BLOCKS]; | |
617 | struct crypt_priv crypt_ctx = { | |
618 | .ctx = &ctx->twofish_ctx, | |
619 | .fpu_enabled = false, | |
620 | }; | |
621 | struct lrw_crypt_req req = { | |
622 | .tbuf = buf, | |
623 | .tbuflen = sizeof(buf), | |
624 | ||
625 | .table_ctx = &ctx->lrw_table, | |
626 | .crypt_ctx = &crypt_ctx, | |
627 | .crypt_fn = decrypt_callback, | |
628 | }; | |
629 | int ret; | |
630 | ||
631 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
632 | ret = lrw_crypt(desc, dst, src, nbytes, &req); | |
633 | twofish_fpu_end(crypt_ctx.fpu_enabled); | |
634 | ||
635 | return ret; | |
636 | } | |
637 | ||
638 | static void lrw_exit_tfm(struct crypto_tfm *tfm) | |
639 | { | |
640 | struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm); | |
641 | ||
642 | lrw_free_table(&ctx->lrw_table); | |
643 | } | |
644 | ||
645 | struct twofish_xts_ctx { | |
646 | struct twofish_ctx tweak_ctx; | |
647 | struct twofish_ctx crypt_ctx; | |
648 | }; | |
649 | ||
650 | static int xts_twofish_setkey(struct crypto_tfm *tfm, const u8 *key, | |
651 | unsigned int keylen) | |
652 | { | |
653 | struct twofish_xts_ctx *ctx = crypto_tfm_ctx(tfm); | |
654 | u32 *flags = &tfm->crt_flags; | |
655 | int err; | |
656 | ||
657 | /* key consists of keys of equal size concatenated, therefore | |
658 | * the length must be even | |
659 | */ | |
660 | if (keylen % 2) { | |
661 | *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN; | |
662 | return -EINVAL; | |
663 | } | |
664 | ||
665 | /* first half of xts-key is for crypt */ | |
666 | err = __twofish_setkey(&ctx->crypt_ctx, key, keylen / 2, flags); | |
667 | if (err) | |
668 | return err; | |
669 | ||
670 | /* second half of xts-key is for tweak */ | |
671 | return __twofish_setkey(&ctx->tweak_ctx, | |
672 | key + keylen / 2, keylen / 2, flags); | |
673 | } | |
674 | ||
675 | static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
676 | struct scatterlist *src, unsigned int nbytes) | |
677 | { | |
678 | struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
679 | be128 buf[TWOFISH_PARALLEL_BLOCKS]; | |
680 | struct crypt_priv crypt_ctx = { | |
681 | .ctx = &ctx->crypt_ctx, | |
682 | .fpu_enabled = false, | |
683 | }; | |
684 | struct xts_crypt_req req = { | |
685 | .tbuf = buf, | |
686 | .tbuflen = sizeof(buf), | |
687 | ||
688 | .tweak_ctx = &ctx->tweak_ctx, | |
689 | .tweak_fn = XTS_TWEAK_CAST(twofish_enc_blk), | |
690 | .crypt_ctx = &crypt_ctx, | |
691 | .crypt_fn = encrypt_callback, | |
692 | }; | |
693 | int ret; | |
694 | ||
695 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
696 | ret = xts_crypt(desc, dst, src, nbytes, &req); | |
697 | twofish_fpu_end(crypt_ctx.fpu_enabled); | |
698 | ||
699 | return ret; | |
700 | } | |
701 | ||
702 | static int xts_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, | |
703 | struct scatterlist *src, unsigned int nbytes) | |
704 | { | |
705 | struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); | |
706 | be128 buf[TWOFISH_PARALLEL_BLOCKS]; | |
707 | struct crypt_priv crypt_ctx = { | |
708 | .ctx = &ctx->crypt_ctx, | |
709 | .fpu_enabled = false, | |
710 | }; | |
711 | struct xts_crypt_req req = { | |
712 | .tbuf = buf, | |
713 | .tbuflen = sizeof(buf), | |
714 | ||
715 | .tweak_ctx = &ctx->tweak_ctx, | |
716 | .tweak_fn = XTS_TWEAK_CAST(twofish_enc_blk), | |
717 | .crypt_ctx = &crypt_ctx, | |
718 | .crypt_fn = decrypt_callback, | |
719 | }; | |
720 | int ret; | |
721 | ||
722 | desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; | |
723 | ret = xts_crypt(desc, dst, src, nbytes, &req); | |
724 | twofish_fpu_end(crypt_ctx.fpu_enabled); | |
725 | ||
726 | return ret; | |
727 | } | |
728 | ||
107778b5 JG |
729 | static struct crypto_alg twofish_algs[10] = { { |
730 | .cra_name = "__ecb-twofish-avx", | |
731 | .cra_driver_name = "__driver-ecb-twofish-avx", | |
732 | .cra_priority = 0, | |
733 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
734 | .cra_blocksize = TF_BLOCK_SIZE, | |
735 | .cra_ctxsize = sizeof(struct twofish_ctx), | |
736 | .cra_alignmask = 0, | |
737 | .cra_type = &crypto_blkcipher_type, | |
738 | .cra_module = THIS_MODULE, | |
739 | .cra_list = LIST_HEAD_INIT(twofish_algs[0].cra_list), | |
740 | .cra_u = { | |
741 | .blkcipher = { | |
742 | .min_keysize = TF_MIN_KEY_SIZE, | |
743 | .max_keysize = TF_MAX_KEY_SIZE, | |
744 | .setkey = twofish_setkey, | |
745 | .encrypt = ecb_encrypt, | |
746 | .decrypt = ecb_decrypt, | |
747 | }, | |
748 | }, | |
749 | }, { | |
750 | .cra_name = "__cbc-twofish-avx", | |
751 | .cra_driver_name = "__driver-cbc-twofish-avx", | |
752 | .cra_priority = 0, | |
753 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
754 | .cra_blocksize = TF_BLOCK_SIZE, | |
755 | .cra_ctxsize = sizeof(struct twofish_ctx), | |
756 | .cra_alignmask = 0, | |
757 | .cra_type = &crypto_blkcipher_type, | |
758 | .cra_module = THIS_MODULE, | |
759 | .cra_list = LIST_HEAD_INIT(twofish_algs[1].cra_list), | |
760 | .cra_u = { | |
761 | .blkcipher = { | |
762 | .min_keysize = TF_MIN_KEY_SIZE, | |
763 | .max_keysize = TF_MAX_KEY_SIZE, | |
764 | .setkey = twofish_setkey, | |
765 | .encrypt = cbc_encrypt, | |
766 | .decrypt = cbc_decrypt, | |
767 | }, | |
768 | }, | |
769 | }, { | |
770 | .cra_name = "__ctr-twofish-avx", | |
771 | .cra_driver_name = "__driver-ctr-twofish-avx", | |
772 | .cra_priority = 0, | |
773 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
774 | .cra_blocksize = 1, | |
775 | .cra_ctxsize = sizeof(struct twofish_ctx), | |
776 | .cra_alignmask = 0, | |
777 | .cra_type = &crypto_blkcipher_type, | |
778 | .cra_module = THIS_MODULE, | |
779 | .cra_list = LIST_HEAD_INIT(twofish_algs[2].cra_list), | |
780 | .cra_u = { | |
781 | .blkcipher = { | |
782 | .min_keysize = TF_MIN_KEY_SIZE, | |
783 | .max_keysize = TF_MAX_KEY_SIZE, | |
784 | .ivsize = TF_BLOCK_SIZE, | |
785 | .setkey = twofish_setkey, | |
786 | .encrypt = ctr_crypt, | |
787 | .decrypt = ctr_crypt, | |
788 | }, | |
789 | }, | |
790 | }, { | |
791 | .cra_name = "__lrw-twofish-avx", | |
792 | .cra_driver_name = "__driver-lrw-twofish-avx", | |
793 | .cra_priority = 0, | |
794 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
795 | .cra_blocksize = TF_BLOCK_SIZE, | |
796 | .cra_ctxsize = sizeof(struct twofish_lrw_ctx), | |
797 | .cra_alignmask = 0, | |
798 | .cra_type = &crypto_blkcipher_type, | |
799 | .cra_module = THIS_MODULE, | |
800 | .cra_list = LIST_HEAD_INIT(twofish_algs[3].cra_list), | |
801 | .cra_exit = lrw_exit_tfm, | |
802 | .cra_u = { | |
803 | .blkcipher = { | |
804 | .min_keysize = TF_MIN_KEY_SIZE + | |
805 | TF_BLOCK_SIZE, | |
806 | .max_keysize = TF_MAX_KEY_SIZE + | |
807 | TF_BLOCK_SIZE, | |
808 | .ivsize = TF_BLOCK_SIZE, | |
809 | .setkey = lrw_twofish_setkey, | |
810 | .encrypt = lrw_encrypt, | |
811 | .decrypt = lrw_decrypt, | |
812 | }, | |
813 | }, | |
814 | }, { | |
815 | .cra_name = "__xts-twofish-avx", | |
816 | .cra_driver_name = "__driver-xts-twofish-avx", | |
817 | .cra_priority = 0, | |
818 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
819 | .cra_blocksize = TF_BLOCK_SIZE, | |
820 | .cra_ctxsize = sizeof(struct twofish_xts_ctx), | |
821 | .cra_alignmask = 0, | |
822 | .cra_type = &crypto_blkcipher_type, | |
823 | .cra_module = THIS_MODULE, | |
824 | .cra_list = LIST_HEAD_INIT(twofish_algs[4].cra_list), | |
825 | .cra_u = { | |
826 | .blkcipher = { | |
827 | .min_keysize = TF_MIN_KEY_SIZE * 2, | |
828 | .max_keysize = TF_MAX_KEY_SIZE * 2, | |
829 | .ivsize = TF_BLOCK_SIZE, | |
830 | .setkey = xts_twofish_setkey, | |
831 | .encrypt = xts_encrypt, | |
832 | .decrypt = xts_decrypt, | |
833 | }, | |
834 | }, | |
835 | }, { | |
836 | .cra_name = "ecb(twofish)", | |
837 | .cra_driver_name = "ecb-twofish-avx", | |
838 | .cra_priority = 400, | |
839 | .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, | |
840 | .cra_blocksize = TF_BLOCK_SIZE, | |
30a04008 | 841 | .cra_ctxsize = sizeof(struct async_helper_ctx), |
107778b5 JG |
842 | .cra_alignmask = 0, |
843 | .cra_type = &crypto_ablkcipher_type, | |
844 | .cra_module = THIS_MODULE, | |
845 | .cra_list = LIST_HEAD_INIT(twofish_algs[5].cra_list), | |
846 | .cra_init = ablk_init, | |
847 | .cra_exit = ablk_exit, | |
848 | .cra_u = { | |
849 | .ablkcipher = { | |
850 | .min_keysize = TF_MIN_KEY_SIZE, | |
851 | .max_keysize = TF_MAX_KEY_SIZE, | |
852 | .setkey = ablk_set_key, | |
853 | .encrypt = ablk_encrypt, | |
854 | .decrypt = ablk_decrypt, | |
855 | }, | |
856 | }, | |
857 | }, { | |
858 | .cra_name = "cbc(twofish)", | |
859 | .cra_driver_name = "cbc-twofish-avx", | |
860 | .cra_priority = 400, | |
861 | .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, | |
862 | .cra_blocksize = TF_BLOCK_SIZE, | |
30a04008 | 863 | .cra_ctxsize = sizeof(struct async_helper_ctx), |
107778b5 JG |
864 | .cra_alignmask = 0, |
865 | .cra_type = &crypto_ablkcipher_type, | |
866 | .cra_module = THIS_MODULE, | |
867 | .cra_list = LIST_HEAD_INIT(twofish_algs[6].cra_list), | |
868 | .cra_init = ablk_init, | |
869 | .cra_exit = ablk_exit, | |
870 | .cra_u = { | |
871 | .ablkcipher = { | |
872 | .min_keysize = TF_MIN_KEY_SIZE, | |
873 | .max_keysize = TF_MAX_KEY_SIZE, | |
874 | .ivsize = TF_BLOCK_SIZE, | |
875 | .setkey = ablk_set_key, | |
876 | .encrypt = __ablk_encrypt, | |
877 | .decrypt = ablk_decrypt, | |
878 | }, | |
879 | }, | |
880 | }, { | |
881 | .cra_name = "ctr(twofish)", | |
882 | .cra_driver_name = "ctr-twofish-avx", | |
883 | .cra_priority = 400, | |
884 | .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, | |
885 | .cra_blocksize = 1, | |
30a04008 | 886 | .cra_ctxsize = sizeof(struct async_helper_ctx), |
107778b5 JG |
887 | .cra_alignmask = 0, |
888 | .cra_type = &crypto_ablkcipher_type, | |
889 | .cra_module = THIS_MODULE, | |
890 | .cra_list = LIST_HEAD_INIT(twofish_algs[7].cra_list), | |
891 | .cra_init = ablk_init, | |
892 | .cra_exit = ablk_exit, | |
893 | .cra_u = { | |
894 | .ablkcipher = { | |
895 | .min_keysize = TF_MIN_KEY_SIZE, | |
896 | .max_keysize = TF_MAX_KEY_SIZE, | |
897 | .ivsize = TF_BLOCK_SIZE, | |
898 | .setkey = ablk_set_key, | |
899 | .encrypt = ablk_encrypt, | |
900 | .decrypt = ablk_encrypt, | |
901 | .geniv = "chainiv", | |
902 | }, | |
903 | }, | |
904 | }, { | |
905 | .cra_name = "lrw(twofish)", | |
906 | .cra_driver_name = "lrw-twofish-avx", | |
907 | .cra_priority = 400, | |
908 | .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, | |
909 | .cra_blocksize = TF_BLOCK_SIZE, | |
30a04008 | 910 | .cra_ctxsize = sizeof(struct async_helper_ctx), |
107778b5 JG |
911 | .cra_alignmask = 0, |
912 | .cra_type = &crypto_ablkcipher_type, | |
913 | .cra_module = THIS_MODULE, | |
914 | .cra_list = LIST_HEAD_INIT(twofish_algs[8].cra_list), | |
915 | .cra_init = ablk_init, | |
916 | .cra_exit = ablk_exit, | |
917 | .cra_u = { | |
918 | .ablkcipher = { | |
919 | .min_keysize = TF_MIN_KEY_SIZE + | |
920 | TF_BLOCK_SIZE, | |
921 | .max_keysize = TF_MAX_KEY_SIZE + | |
922 | TF_BLOCK_SIZE, | |
923 | .ivsize = TF_BLOCK_SIZE, | |
924 | .setkey = ablk_set_key, | |
925 | .encrypt = ablk_encrypt, | |
926 | .decrypt = ablk_decrypt, | |
927 | }, | |
928 | }, | |
929 | }, { | |
930 | .cra_name = "xts(twofish)", | |
931 | .cra_driver_name = "xts-twofish-avx", | |
932 | .cra_priority = 400, | |
933 | .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, | |
934 | .cra_blocksize = TF_BLOCK_SIZE, | |
30a04008 | 935 | .cra_ctxsize = sizeof(struct async_helper_ctx), |
107778b5 JG |
936 | .cra_alignmask = 0, |
937 | .cra_type = &crypto_ablkcipher_type, | |
938 | .cra_module = THIS_MODULE, | |
939 | .cra_list = LIST_HEAD_INIT(twofish_algs[9].cra_list), | |
940 | .cra_init = ablk_init, | |
941 | .cra_exit = ablk_exit, | |
942 | .cra_u = { | |
943 | .ablkcipher = { | |
944 | .min_keysize = TF_MIN_KEY_SIZE * 2, | |
945 | .max_keysize = TF_MAX_KEY_SIZE * 2, | |
946 | .ivsize = TF_BLOCK_SIZE, | |
947 | .setkey = ablk_set_key, | |
948 | .encrypt = ablk_encrypt, | |
949 | .decrypt = ablk_decrypt, | |
950 | }, | |
951 | }, | |
952 | } }; | |
953 | ||
954 | static int __init twofish_init(void) | |
955 | { | |
956 | u64 xcr0; | |
957 | ||
958 | if (!cpu_has_avx || !cpu_has_osxsave) { | |
959 | printk(KERN_INFO "AVX instructions are not detected.\n"); | |
960 | return -ENODEV; | |
961 | } | |
962 | ||
963 | xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK); | |
964 | if ((xcr0 & (XSTATE_SSE | XSTATE_YMM)) != (XSTATE_SSE | XSTATE_YMM)) { | |
965 | printk(KERN_INFO "AVX detected but unusable.\n"); | |
966 | return -ENODEV; | |
967 | } | |
968 | ||
969 | return crypto_register_algs(twofish_algs, ARRAY_SIZE(twofish_algs)); | |
970 | } | |
971 | ||
972 | static void __exit twofish_exit(void) | |
973 | { | |
974 | crypto_unregister_algs(twofish_algs, ARRAY_SIZE(twofish_algs)); | |
975 | } | |
976 | ||
977 | module_init(twofish_init); | |
978 | module_exit(twofish_exit); | |
979 | ||
980 | MODULE_DESCRIPTION("Twofish Cipher Algorithm, AVX optimized"); | |
981 | MODULE_LICENSE("GPL"); | |
982 | MODULE_ALIAS("twofish"); |