2 * Glue Code for AVX assembler versions of Serpent Cipher
4 * Copyright (C) 2012 Johannes Goetzfried
5 * <Johannes.Goetzfried@informatik.stud.uni-erlangen.de>
7 * Copyright © 2011-2013 Jussi Kivilinna <jussi.kivilinna@iki.fi>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
26 #include <linux/module.h>
27 #include <linux/hardirq.h>
28 #include <linux/types.h>
29 #include <linux/crypto.h>
30 #include <linux/err.h>
31 #include <crypto/algapi.h>
32 #include <crypto/serpent.h>
33 #include <crypto/cryptd.h>
34 #include <crypto/b128ops.h>
35 #include <crypto/ctr.h>
36 #include <crypto/lrw.h>
37 #include <crypto/xts.h>
39 #include <asm/xsave.h>
40 #include <asm/crypto/serpent-avx.h>
41 #include <asm/crypto/ablk_helper.h>
42 #include <asm/crypto/glue_helper.h>
44 /* 8-way parallel cipher functions */
45 asmlinkage void serpent_ecb_enc_8way_avx(struct serpent_ctx *ctx, u8 *dst,
47 EXPORT_SYMBOL_GPL(serpent_ecb_enc_8way_avx);
49 asmlinkage void serpent_ecb_dec_8way_avx(struct serpent_ctx *ctx, u8 *dst,
51 EXPORT_SYMBOL_GPL(serpent_ecb_dec_8way_avx);
53 asmlinkage void serpent_cbc_dec_8way_avx(struct serpent_ctx *ctx, u8 *dst,
55 EXPORT_SYMBOL_GPL(serpent_cbc_dec_8way_avx);
57 asmlinkage void serpent_ctr_8way_avx(struct serpent_ctx *ctx, u8 *dst,
58 const u8 *src, le128 *iv);
59 EXPORT_SYMBOL_GPL(serpent_ctr_8way_avx);
61 asmlinkage void serpent_xts_enc_8way_avx(struct serpent_ctx *ctx, u8 *dst,
62 const u8 *src, le128 *iv);
63 EXPORT_SYMBOL_GPL(serpent_xts_enc_8way_avx);
65 asmlinkage void serpent_xts_dec_8way_avx(struct serpent_ctx *ctx, u8 *dst,
66 const u8 *src, le128 *iv);
67 EXPORT_SYMBOL_GPL(serpent_xts_dec_8way_avx);
69 void __serpent_crypt_ctr(void *ctx, u128 *dst, const u128 *src, le128 *iv)
73 le128_to_be128(&ctrblk, iv);
76 __serpent_encrypt(ctx, (u8 *)&ctrblk, (u8 *)&ctrblk);
77 u128_xor(dst, src, (u128 *)&ctrblk);
79 EXPORT_SYMBOL_GPL(__serpent_crypt_ctr);
81 void serpent_xts_enc(void *ctx, u128 *dst, const u128 *src, le128 *iv)
83 glue_xts_crypt_128bit_one(ctx, dst, src, iv,
84 GLUE_FUNC_CAST(__serpent_encrypt));
86 EXPORT_SYMBOL_GPL(serpent_xts_enc);
88 void serpent_xts_dec(void *ctx, u128 *dst, const u128 *src, le128 *iv)
90 glue_xts_crypt_128bit_one(ctx, dst, src, iv,
91 GLUE_FUNC_CAST(__serpent_decrypt));
93 EXPORT_SYMBOL_GPL(serpent_xts_dec);
96 static const struct common_glue_ctx serpent_enc = {
98 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
101 .num_blocks = SERPENT_PARALLEL_BLOCKS,
102 .fn_u = { .ecb = GLUE_FUNC_CAST(serpent_ecb_enc_8way_avx) }
105 .fn_u = { .ecb = GLUE_FUNC_CAST(__serpent_encrypt) }
109 static const struct common_glue_ctx serpent_ctr = {
111 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
114 .num_blocks = SERPENT_PARALLEL_BLOCKS,
115 .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(serpent_ctr_8way_avx) }
118 .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(__serpent_crypt_ctr) }
122 static const struct common_glue_ctx serpent_enc_xts = {
124 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
127 .num_blocks = SERPENT_PARALLEL_BLOCKS,
128 .fn_u = { .xts = GLUE_XTS_FUNC_CAST(serpent_xts_enc_8way_avx) }
131 .fn_u = { .xts = GLUE_XTS_FUNC_CAST(serpent_xts_enc) }
135 static const struct common_glue_ctx serpent_dec = {
137 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
140 .num_blocks = SERPENT_PARALLEL_BLOCKS,
141 .fn_u = { .ecb = GLUE_FUNC_CAST(serpent_ecb_dec_8way_avx) }
144 .fn_u = { .ecb = GLUE_FUNC_CAST(__serpent_decrypt) }
148 static const struct common_glue_ctx serpent_dec_cbc = {
150 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
153 .num_blocks = SERPENT_PARALLEL_BLOCKS,
154 .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(serpent_cbc_dec_8way_avx) }
157 .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(__serpent_decrypt) }
161 static const struct common_glue_ctx serpent_dec_xts = {
163 .fpu_blocks_limit = SERPENT_PARALLEL_BLOCKS,
166 .num_blocks = SERPENT_PARALLEL_BLOCKS,
167 .fn_u = { .xts = GLUE_XTS_FUNC_CAST(serpent_xts_dec_8way_avx) }
170 .fn_u = { .xts = GLUE_XTS_FUNC_CAST(serpent_xts_dec) }
174 static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
175 struct scatterlist *src, unsigned int nbytes)
177 return glue_ecb_crypt_128bit(&serpent_enc, desc, dst, src, nbytes);
180 static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
181 struct scatterlist *src, unsigned int nbytes)
183 return glue_ecb_crypt_128bit(&serpent_dec, desc, dst, src, nbytes);
186 static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
187 struct scatterlist *src, unsigned int nbytes)
189 return glue_cbc_encrypt_128bit(GLUE_FUNC_CAST(__serpent_encrypt), desc,
193 static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
194 struct scatterlist *src, unsigned int nbytes)
196 return glue_cbc_decrypt_128bit(&serpent_dec_cbc, desc, dst, src,
200 static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst,
201 struct scatterlist *src, unsigned int nbytes)
203 return glue_ctr_crypt_128bit(&serpent_ctr, desc, dst, src, nbytes);
206 static inline bool serpent_fpu_begin(bool fpu_enabled, unsigned int nbytes)
208 return glue_fpu_begin(SERPENT_BLOCK_SIZE, SERPENT_PARALLEL_BLOCKS,
209 NULL, fpu_enabled, nbytes);
212 static inline void serpent_fpu_end(bool fpu_enabled)
214 glue_fpu_end(fpu_enabled);
218 struct serpent_ctx *ctx;
222 static void encrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes)
224 const unsigned int bsize = SERPENT_BLOCK_SIZE;
225 struct crypt_priv *ctx = priv;
228 ctx->fpu_enabled = serpent_fpu_begin(ctx->fpu_enabled, nbytes);
230 if (nbytes == bsize * SERPENT_PARALLEL_BLOCKS) {
231 serpent_ecb_enc_8way_avx(ctx->ctx, srcdst, srcdst);
235 for (i = 0; i < nbytes / bsize; i++, srcdst += bsize)
236 __serpent_encrypt(ctx->ctx, srcdst, srcdst);
239 static void decrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes)
241 const unsigned int bsize = SERPENT_BLOCK_SIZE;
242 struct crypt_priv *ctx = priv;
245 ctx->fpu_enabled = serpent_fpu_begin(ctx->fpu_enabled, nbytes);
247 if (nbytes == bsize * SERPENT_PARALLEL_BLOCKS) {
248 serpent_ecb_dec_8way_avx(ctx->ctx, srcdst, srcdst);
252 for (i = 0; i < nbytes / bsize; i++, srcdst += bsize)
253 __serpent_decrypt(ctx->ctx, srcdst, srcdst);
256 int lrw_serpent_setkey(struct crypto_tfm *tfm, const u8 *key,
259 struct serpent_lrw_ctx *ctx = crypto_tfm_ctx(tfm);
262 err = __serpent_setkey(&ctx->serpent_ctx, key, keylen -
267 return lrw_init_table(&ctx->lrw_table, key + keylen -
270 EXPORT_SYMBOL_GPL(lrw_serpent_setkey);
272 static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
273 struct scatterlist *src, unsigned int nbytes)
275 struct serpent_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
276 be128 buf[SERPENT_PARALLEL_BLOCKS];
277 struct crypt_priv crypt_ctx = {
278 .ctx = &ctx->serpent_ctx,
279 .fpu_enabled = false,
281 struct lrw_crypt_req req = {
283 .tbuflen = sizeof(buf),
285 .table_ctx = &ctx->lrw_table,
286 .crypt_ctx = &crypt_ctx,
287 .crypt_fn = encrypt_callback,
291 desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
292 ret = lrw_crypt(desc, dst, src, nbytes, &req);
293 serpent_fpu_end(crypt_ctx.fpu_enabled);
298 static int lrw_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
299 struct scatterlist *src, unsigned int nbytes)
301 struct serpent_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
302 be128 buf[SERPENT_PARALLEL_BLOCKS];
303 struct crypt_priv crypt_ctx = {
304 .ctx = &ctx->serpent_ctx,
305 .fpu_enabled = false,
307 struct lrw_crypt_req req = {
309 .tbuflen = sizeof(buf),
311 .table_ctx = &ctx->lrw_table,
312 .crypt_ctx = &crypt_ctx,
313 .crypt_fn = decrypt_callback,
317 desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP;
318 ret = lrw_crypt(desc, dst, src, nbytes, &req);
319 serpent_fpu_end(crypt_ctx.fpu_enabled);
324 void lrw_serpent_exit_tfm(struct crypto_tfm *tfm)
326 struct serpent_lrw_ctx *ctx = crypto_tfm_ctx(tfm);
328 lrw_free_table(&ctx->lrw_table);
330 EXPORT_SYMBOL_GPL(lrw_serpent_exit_tfm);
332 int xts_serpent_setkey(struct crypto_tfm *tfm, const u8 *key,
335 struct serpent_xts_ctx *ctx = crypto_tfm_ctx(tfm);
336 u32 *flags = &tfm->crt_flags;
339 /* key consists of keys of equal size concatenated, therefore
340 * the length must be even
343 *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
347 /* first half of xts-key is for crypt */
348 err = __serpent_setkey(&ctx->crypt_ctx, key, keylen / 2);
352 /* second half of xts-key is for tweak */
353 return __serpent_setkey(&ctx->tweak_ctx, key + keylen / 2, keylen / 2);
355 EXPORT_SYMBOL_GPL(xts_serpent_setkey);
357 static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
358 struct scatterlist *src, unsigned int nbytes)
360 struct serpent_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
362 return glue_xts_crypt_128bit(&serpent_enc_xts, desc, dst, src, nbytes,
363 XTS_TWEAK_CAST(__serpent_encrypt),
364 &ctx->tweak_ctx, &ctx->crypt_ctx);
367 static int xts_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
368 struct scatterlist *src, unsigned int nbytes)
370 struct serpent_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
372 return glue_xts_crypt_128bit(&serpent_dec_xts, desc, dst, src, nbytes,
373 XTS_TWEAK_CAST(__serpent_encrypt),
374 &ctx->tweak_ctx, &ctx->crypt_ctx);
377 static struct crypto_alg serpent_algs[10] = { {
378 .cra_name = "__ecb-serpent-avx",
379 .cra_driver_name = "__driver-ecb-serpent-avx",
381 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
382 .cra_blocksize = SERPENT_BLOCK_SIZE,
383 .cra_ctxsize = sizeof(struct serpent_ctx),
385 .cra_type = &crypto_blkcipher_type,
386 .cra_module = THIS_MODULE,
389 .min_keysize = SERPENT_MIN_KEY_SIZE,
390 .max_keysize = SERPENT_MAX_KEY_SIZE,
391 .setkey = serpent_setkey,
392 .encrypt = ecb_encrypt,
393 .decrypt = ecb_decrypt,
397 .cra_name = "__cbc-serpent-avx",
398 .cra_driver_name = "__driver-cbc-serpent-avx",
400 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
401 .cra_blocksize = SERPENT_BLOCK_SIZE,
402 .cra_ctxsize = sizeof(struct serpent_ctx),
404 .cra_type = &crypto_blkcipher_type,
405 .cra_module = THIS_MODULE,
408 .min_keysize = SERPENT_MIN_KEY_SIZE,
409 .max_keysize = SERPENT_MAX_KEY_SIZE,
410 .setkey = serpent_setkey,
411 .encrypt = cbc_encrypt,
412 .decrypt = cbc_decrypt,
416 .cra_name = "__ctr-serpent-avx",
417 .cra_driver_name = "__driver-ctr-serpent-avx",
419 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
421 .cra_ctxsize = sizeof(struct serpent_ctx),
423 .cra_type = &crypto_blkcipher_type,
424 .cra_module = THIS_MODULE,
427 .min_keysize = SERPENT_MIN_KEY_SIZE,
428 .max_keysize = SERPENT_MAX_KEY_SIZE,
429 .ivsize = SERPENT_BLOCK_SIZE,
430 .setkey = serpent_setkey,
431 .encrypt = ctr_crypt,
432 .decrypt = ctr_crypt,
436 .cra_name = "__lrw-serpent-avx",
437 .cra_driver_name = "__driver-lrw-serpent-avx",
439 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
440 .cra_blocksize = SERPENT_BLOCK_SIZE,
441 .cra_ctxsize = sizeof(struct serpent_lrw_ctx),
443 .cra_type = &crypto_blkcipher_type,
444 .cra_module = THIS_MODULE,
445 .cra_exit = lrw_serpent_exit_tfm,
448 .min_keysize = SERPENT_MIN_KEY_SIZE +
450 .max_keysize = SERPENT_MAX_KEY_SIZE +
452 .ivsize = SERPENT_BLOCK_SIZE,
453 .setkey = lrw_serpent_setkey,
454 .encrypt = lrw_encrypt,
455 .decrypt = lrw_decrypt,
459 .cra_name = "__xts-serpent-avx",
460 .cra_driver_name = "__driver-xts-serpent-avx",
462 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
463 .cra_blocksize = SERPENT_BLOCK_SIZE,
464 .cra_ctxsize = sizeof(struct serpent_xts_ctx),
466 .cra_type = &crypto_blkcipher_type,
467 .cra_module = THIS_MODULE,
470 .min_keysize = SERPENT_MIN_KEY_SIZE * 2,
471 .max_keysize = SERPENT_MAX_KEY_SIZE * 2,
472 .ivsize = SERPENT_BLOCK_SIZE,
473 .setkey = xts_serpent_setkey,
474 .encrypt = xts_encrypt,
475 .decrypt = xts_decrypt,
479 .cra_name = "ecb(serpent)",
480 .cra_driver_name = "ecb-serpent-avx",
482 .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC,
483 .cra_blocksize = SERPENT_BLOCK_SIZE,
484 .cra_ctxsize = sizeof(struct async_helper_ctx),
486 .cra_type = &crypto_ablkcipher_type,
487 .cra_module = THIS_MODULE,
488 .cra_init = ablk_init,
489 .cra_exit = ablk_exit,
492 .min_keysize = SERPENT_MIN_KEY_SIZE,
493 .max_keysize = SERPENT_MAX_KEY_SIZE,
494 .setkey = ablk_set_key,
495 .encrypt = ablk_encrypt,
496 .decrypt = ablk_decrypt,
500 .cra_name = "cbc(serpent)",
501 .cra_driver_name = "cbc-serpent-avx",
503 .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC,
504 .cra_blocksize = SERPENT_BLOCK_SIZE,
505 .cra_ctxsize = sizeof(struct async_helper_ctx),
507 .cra_type = &crypto_ablkcipher_type,
508 .cra_module = THIS_MODULE,
509 .cra_init = ablk_init,
510 .cra_exit = ablk_exit,
513 .min_keysize = SERPENT_MIN_KEY_SIZE,
514 .max_keysize = SERPENT_MAX_KEY_SIZE,
515 .ivsize = SERPENT_BLOCK_SIZE,
516 .setkey = ablk_set_key,
517 .encrypt = __ablk_encrypt,
518 .decrypt = ablk_decrypt,
522 .cra_name = "ctr(serpent)",
523 .cra_driver_name = "ctr-serpent-avx",
525 .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC,
527 .cra_ctxsize = sizeof(struct async_helper_ctx),
529 .cra_type = &crypto_ablkcipher_type,
530 .cra_module = THIS_MODULE,
531 .cra_init = ablk_init,
532 .cra_exit = ablk_exit,
535 .min_keysize = SERPENT_MIN_KEY_SIZE,
536 .max_keysize = SERPENT_MAX_KEY_SIZE,
537 .ivsize = SERPENT_BLOCK_SIZE,
538 .setkey = ablk_set_key,
539 .encrypt = ablk_encrypt,
540 .decrypt = ablk_encrypt,
545 .cra_name = "lrw(serpent)",
546 .cra_driver_name = "lrw-serpent-avx",
548 .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC,
549 .cra_blocksize = SERPENT_BLOCK_SIZE,
550 .cra_ctxsize = sizeof(struct async_helper_ctx),
552 .cra_type = &crypto_ablkcipher_type,
553 .cra_module = THIS_MODULE,
554 .cra_init = ablk_init,
555 .cra_exit = ablk_exit,
558 .min_keysize = SERPENT_MIN_KEY_SIZE +
560 .max_keysize = SERPENT_MAX_KEY_SIZE +
562 .ivsize = SERPENT_BLOCK_SIZE,
563 .setkey = ablk_set_key,
564 .encrypt = ablk_encrypt,
565 .decrypt = ablk_decrypt,
569 .cra_name = "xts(serpent)",
570 .cra_driver_name = "xts-serpent-avx",
572 .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC,
573 .cra_blocksize = SERPENT_BLOCK_SIZE,
574 .cra_ctxsize = sizeof(struct async_helper_ctx),
576 .cra_type = &crypto_ablkcipher_type,
577 .cra_module = THIS_MODULE,
578 .cra_init = ablk_init,
579 .cra_exit = ablk_exit,
582 .min_keysize = SERPENT_MIN_KEY_SIZE * 2,
583 .max_keysize = SERPENT_MAX_KEY_SIZE * 2,
584 .ivsize = SERPENT_BLOCK_SIZE,
585 .setkey = ablk_set_key,
586 .encrypt = ablk_encrypt,
587 .decrypt = ablk_decrypt,
592 static int __init serpent_init(void)
596 if (!cpu_has_avx || !cpu_has_osxsave) {
597 printk(KERN_INFO "AVX instructions are not detected.\n");
601 xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK);
602 if ((xcr0 & (XSTATE_SSE | XSTATE_YMM)) != (XSTATE_SSE | XSTATE_YMM)) {
603 printk(KERN_INFO "AVX detected but unusable.\n");
607 return crypto_register_algs(serpent_algs, ARRAY_SIZE(serpent_algs));
610 static void __exit serpent_exit(void)
612 crypto_unregister_algs(serpent_algs, ARRAY_SIZE(serpent_algs));
615 module_init(serpent_init);
616 module_exit(serpent_exit);
618 MODULE_DESCRIPTION("Serpent Cipher Algorithm, AVX optimized");
619 MODULE_LICENSE("GPL");
620 MODULE_ALIAS("serpent");