1 /*
2 * Glue Code for 3-way parallel assembler optimized version of Twofish
3 *
4 * Copyright (c) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
19 * USA
20 *
21 */
22
23 #include <asm/processor.h>
24 #include <linux/crypto.h>
25 #include <linux/init.h>
26 #include <linux/module.h>
27 #include <linux/types.h>
28 #include <crypto/algapi.h>
29 #include <crypto/twofish.h>
30 #include <crypto/b128ops.h>
31 #include <asm/crypto/twofish.h>
32 #include <asm/crypto/glue_helper.h>
33 #include <crypto/lrw.h>
34 #include <crypto/xts.h>
35
36 EXPORT_SYMBOL_GPL(__twofish_enc_blk_3way);
37 EXPORT_SYMBOL_GPL(twofish_dec_blk_3way);
38
twofish_enc_blk_3way(struct twofish_ctx * ctx,u8 * dst,const u8 * src)39 static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst,
40 const u8 *src)
41 {
42 __twofish_enc_blk_3way(ctx, dst, src, false);
43 }
44
twofish_enc_blk_xor_3way(struct twofish_ctx * ctx,u8 * dst,const u8 * src)45 static inline void twofish_enc_blk_xor_3way(struct twofish_ctx *ctx, u8 *dst,
46 const u8 *src)
47 {
48 __twofish_enc_blk_3way(ctx, dst, src, true);
49 }
50
twofish_dec_blk_cbc_3way(void * ctx,u128 * dst,const u128 * src)51 void twofish_dec_blk_cbc_3way(void *ctx, u128 *dst, const u128 *src)
52 {
53 u128 ivs[2];
54
55 ivs[0] = src[0];
56 ivs[1] = src[1];
57
58 twofish_dec_blk_3way(ctx, (u8 *)dst, (u8 *)src);
59
60 u128_xor(&dst[1], &dst[1], &ivs[0]);
61 u128_xor(&dst[2], &dst[2], &ivs[1]);
62 }
63 EXPORT_SYMBOL_GPL(twofish_dec_blk_cbc_3way);
64
twofish_enc_blk_ctr(void * ctx,u128 * dst,const u128 * src,le128 * iv)65 void twofish_enc_blk_ctr(void *ctx, u128 *dst, const u128 *src, le128 *iv)
66 {
67 be128 ctrblk;
68
69 if (dst != src)
70 *dst = *src;
71
72 le128_to_be128(&ctrblk, iv);
73 le128_inc(iv);
74
75 twofish_enc_blk(ctx, (u8 *)&ctrblk, (u8 *)&ctrblk);
76 u128_xor(dst, dst, (u128 *)&ctrblk);
77 }
78 EXPORT_SYMBOL_GPL(twofish_enc_blk_ctr);
79
twofish_enc_blk_ctr_3way(void * ctx,u128 * dst,const u128 * src,le128 * iv)80 void twofish_enc_blk_ctr_3way(void *ctx, u128 *dst, const u128 *src,
81 le128 *iv)
82 {
83 be128 ctrblks[3];
84
85 if (dst != src) {
86 dst[0] = src[0];
87 dst[1] = src[1];
88 dst[2] = src[2];
89 }
90
91 le128_to_be128(&ctrblks[0], iv);
92 le128_inc(iv);
93 le128_to_be128(&ctrblks[1], iv);
94 le128_inc(iv);
95 le128_to_be128(&ctrblks[2], iv);
96 le128_inc(iv);
97
98 twofish_enc_blk_xor_3way(ctx, (u8 *)dst, (u8 *)ctrblks);
99 }
100 EXPORT_SYMBOL_GPL(twofish_enc_blk_ctr_3way);
101
102 static const struct common_glue_ctx twofish_enc = {
103 .num_funcs = 2,
104 .fpu_blocks_limit = -1,
105
106 .funcs = { {
107 .num_blocks = 3,
108 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_3way) }
109 }, {
110 .num_blocks = 1,
111 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk) }
112 } }
113 };
114
115 static const struct common_glue_ctx twofish_ctr = {
116 .num_funcs = 2,
117 .fpu_blocks_limit = -1,
118
119 .funcs = { {
120 .num_blocks = 3,
121 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_ctr_3way) }
122 }, {
123 .num_blocks = 1,
124 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_ctr) }
125 } }
126 };
127
128 static const struct common_glue_ctx twofish_dec = {
129 .num_funcs = 2,
130 .fpu_blocks_limit = -1,
131
132 .funcs = { {
133 .num_blocks = 3,
134 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk_3way) }
135 }, {
136 .num_blocks = 1,
137 .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk) }
138 } }
139 };
140
141 static const struct common_glue_ctx twofish_dec_cbc = {
142 .num_funcs = 2,
143 .fpu_blocks_limit = -1,
144
145 .funcs = { {
146 .num_blocks = 3,
147 .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk_cbc_3way) }
148 }, {
149 .num_blocks = 1,
150 .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk) }
151 } }
152 };
153
ecb_encrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)154 static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
155 struct scatterlist *src, unsigned int nbytes)
156 {
157 return glue_ecb_crypt_128bit(&twofish_enc, desc, dst, src, nbytes);
158 }
159
ecb_decrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)160 static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
161 struct scatterlist *src, unsigned int nbytes)
162 {
163 return glue_ecb_crypt_128bit(&twofish_dec, desc, dst, src, nbytes);
164 }
165
cbc_encrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)166 static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
167 struct scatterlist *src, unsigned int nbytes)
168 {
169 return glue_cbc_encrypt_128bit(GLUE_FUNC_CAST(twofish_enc_blk), desc,
170 dst, src, nbytes);
171 }
172
cbc_decrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)173 static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
174 struct scatterlist *src, unsigned int nbytes)
175 {
176 return glue_cbc_decrypt_128bit(&twofish_dec_cbc, desc, dst, src,
177 nbytes);
178 }
179
ctr_crypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)180 static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst,
181 struct scatterlist *src, unsigned int nbytes)
182 {
183 return glue_ctr_crypt_128bit(&twofish_ctr, desc, dst, src, nbytes);
184 }
185
encrypt_callback(void * priv,u8 * srcdst,unsigned int nbytes)186 static void encrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes)
187 {
188 const unsigned int bsize = TF_BLOCK_SIZE;
189 struct twofish_ctx *ctx = priv;
190 int i;
191
192 if (nbytes == 3 * bsize) {
193 twofish_enc_blk_3way(ctx, srcdst, srcdst);
194 return;
195 }
196
197 for (i = 0; i < nbytes / bsize; i++, srcdst += bsize)
198 twofish_enc_blk(ctx, srcdst, srcdst);
199 }
200
decrypt_callback(void * priv,u8 * srcdst,unsigned int nbytes)201 static void decrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes)
202 {
203 const unsigned int bsize = TF_BLOCK_SIZE;
204 struct twofish_ctx *ctx = priv;
205 int i;
206
207 if (nbytes == 3 * bsize) {
208 twofish_dec_blk_3way(ctx, srcdst, srcdst);
209 return;
210 }
211
212 for (i = 0; i < nbytes / bsize; i++, srcdst += bsize)
213 twofish_dec_blk(ctx, srcdst, srcdst);
214 }
215
lrw_twofish_setkey(struct crypto_tfm * tfm,const u8 * key,unsigned int keylen)216 int lrw_twofish_setkey(struct crypto_tfm *tfm, const u8 *key,
217 unsigned int keylen)
218 {
219 struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm);
220 int err;
221
222 err = __twofish_setkey(&ctx->twofish_ctx, key, keylen - TF_BLOCK_SIZE,
223 &tfm->crt_flags);
224 if (err)
225 return err;
226
227 return lrw_init_table(&ctx->lrw_table, key + keylen - TF_BLOCK_SIZE);
228 }
229 EXPORT_SYMBOL_GPL(lrw_twofish_setkey);
230
lrw_encrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)231 static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
232 struct scatterlist *src, unsigned int nbytes)
233 {
234 struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
235 be128 buf[3];
236 struct lrw_crypt_req req = {
237 .tbuf = buf,
238 .tbuflen = sizeof(buf),
239
240 .table_ctx = &ctx->lrw_table,
241 .crypt_ctx = &ctx->twofish_ctx,
242 .crypt_fn = encrypt_callback,
243 };
244
245 return lrw_crypt(desc, dst, src, nbytes, &req);
246 }
247
lrw_decrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)248 static int lrw_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
249 struct scatterlist *src, unsigned int nbytes)
250 {
251 struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
252 be128 buf[3];
253 struct lrw_crypt_req req = {
254 .tbuf = buf,
255 .tbuflen = sizeof(buf),
256
257 .table_ctx = &ctx->lrw_table,
258 .crypt_ctx = &ctx->twofish_ctx,
259 .crypt_fn = decrypt_callback,
260 };
261
262 return lrw_crypt(desc, dst, src, nbytes, &req);
263 }
264
lrw_twofish_exit_tfm(struct crypto_tfm * tfm)265 void lrw_twofish_exit_tfm(struct crypto_tfm *tfm)
266 {
267 struct twofish_lrw_ctx *ctx = crypto_tfm_ctx(tfm);
268
269 lrw_free_table(&ctx->lrw_table);
270 }
271 EXPORT_SYMBOL_GPL(lrw_twofish_exit_tfm);
272
xts_twofish_setkey(struct crypto_tfm * tfm,const u8 * key,unsigned int keylen)273 int xts_twofish_setkey(struct crypto_tfm *tfm, const u8 *key,
274 unsigned int keylen)
275 {
276 struct twofish_xts_ctx *ctx = crypto_tfm_ctx(tfm);
277 u32 *flags = &tfm->crt_flags;
278 int err;
279
280 /* key consists of keys of equal size concatenated, therefore
281 * the length must be even
282 */
283 if (keylen % 2) {
284 *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
285 return -EINVAL;
286 }
287
288 /* first half of xts-key is for crypt */
289 err = __twofish_setkey(&ctx->crypt_ctx, key, keylen / 2, flags);
290 if (err)
291 return err;
292
293 /* second half of xts-key is for tweak */
294 return __twofish_setkey(&ctx->tweak_ctx, key + keylen / 2, keylen / 2,
295 flags);
296 }
297 EXPORT_SYMBOL_GPL(xts_twofish_setkey);
298
xts_encrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)299 static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
300 struct scatterlist *src, unsigned int nbytes)
301 {
302 struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
303 be128 buf[3];
304 struct xts_crypt_req req = {
305 .tbuf = buf,
306 .tbuflen = sizeof(buf),
307
308 .tweak_ctx = &ctx->tweak_ctx,
309 .tweak_fn = XTS_TWEAK_CAST(twofish_enc_blk),
310 .crypt_ctx = &ctx->crypt_ctx,
311 .crypt_fn = encrypt_callback,
312 };
313
314 return xts_crypt(desc, dst, src, nbytes, &req);
315 }
316
xts_decrypt(struct blkcipher_desc * desc,struct scatterlist * dst,struct scatterlist * src,unsigned int nbytes)317 static int xts_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst,
318 struct scatterlist *src, unsigned int nbytes)
319 {
320 struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
321 be128 buf[3];
322 struct xts_crypt_req req = {
323 .tbuf = buf,
324 .tbuflen = sizeof(buf),
325
326 .tweak_ctx = &ctx->tweak_ctx,
327 .tweak_fn = XTS_TWEAK_CAST(twofish_enc_blk),
328 .crypt_ctx = &ctx->crypt_ctx,
329 .crypt_fn = decrypt_callback,
330 };
331
332 return xts_crypt(desc, dst, src, nbytes, &req);
333 }
334
335 static struct crypto_alg tf_algs[5] = { {
336 .cra_name = "ecb(twofish)",
337 .cra_driver_name = "ecb-twofish-3way",
338 .cra_priority = 300,
339 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
340 .cra_blocksize = TF_BLOCK_SIZE,
341 .cra_ctxsize = sizeof(struct twofish_ctx),
342 .cra_alignmask = 0,
343 .cra_type = &crypto_blkcipher_type,
344 .cra_module = THIS_MODULE,
345 .cra_u = {
346 .blkcipher = {
347 .min_keysize = TF_MIN_KEY_SIZE,
348 .max_keysize = TF_MAX_KEY_SIZE,
349 .setkey = twofish_setkey,
350 .encrypt = ecb_encrypt,
351 .decrypt = ecb_decrypt,
352 },
353 },
354 }, {
355 .cra_name = "cbc(twofish)",
356 .cra_driver_name = "cbc-twofish-3way",
357 .cra_priority = 300,
358 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
359 .cra_blocksize = TF_BLOCK_SIZE,
360 .cra_ctxsize = sizeof(struct twofish_ctx),
361 .cra_alignmask = 0,
362 .cra_type = &crypto_blkcipher_type,
363 .cra_module = THIS_MODULE,
364 .cra_u = {
365 .blkcipher = {
366 .min_keysize = TF_MIN_KEY_SIZE,
367 .max_keysize = TF_MAX_KEY_SIZE,
368 .ivsize = TF_BLOCK_SIZE,
369 .setkey = twofish_setkey,
370 .encrypt = cbc_encrypt,
371 .decrypt = cbc_decrypt,
372 },
373 },
374 }, {
375 .cra_name = "ctr(twofish)",
376 .cra_driver_name = "ctr-twofish-3way",
377 .cra_priority = 300,
378 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
379 .cra_blocksize = 1,
380 .cra_ctxsize = sizeof(struct twofish_ctx),
381 .cra_alignmask = 0,
382 .cra_type = &crypto_blkcipher_type,
383 .cra_module = THIS_MODULE,
384 .cra_u = {
385 .blkcipher = {
386 .min_keysize = TF_MIN_KEY_SIZE,
387 .max_keysize = TF_MAX_KEY_SIZE,
388 .ivsize = TF_BLOCK_SIZE,
389 .setkey = twofish_setkey,
390 .encrypt = ctr_crypt,
391 .decrypt = ctr_crypt,
392 },
393 },
394 }, {
395 .cra_name = "lrw(twofish)",
396 .cra_driver_name = "lrw-twofish-3way",
397 .cra_priority = 300,
398 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
399 .cra_blocksize = TF_BLOCK_SIZE,
400 .cra_ctxsize = sizeof(struct twofish_lrw_ctx),
401 .cra_alignmask = 0,
402 .cra_type = &crypto_blkcipher_type,
403 .cra_module = THIS_MODULE,
404 .cra_exit = lrw_twofish_exit_tfm,
405 .cra_u = {
406 .blkcipher = {
407 .min_keysize = TF_MIN_KEY_SIZE + TF_BLOCK_SIZE,
408 .max_keysize = TF_MAX_KEY_SIZE + TF_BLOCK_SIZE,
409 .ivsize = TF_BLOCK_SIZE,
410 .setkey = lrw_twofish_setkey,
411 .encrypt = lrw_encrypt,
412 .decrypt = lrw_decrypt,
413 },
414 },
415 }, {
416 .cra_name = "xts(twofish)",
417 .cra_driver_name = "xts-twofish-3way",
418 .cra_priority = 300,
419 .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
420 .cra_blocksize = TF_BLOCK_SIZE,
421 .cra_ctxsize = sizeof(struct twofish_xts_ctx),
422 .cra_alignmask = 0,
423 .cra_type = &crypto_blkcipher_type,
424 .cra_module = THIS_MODULE,
425 .cra_u = {
426 .blkcipher = {
427 .min_keysize = TF_MIN_KEY_SIZE * 2,
428 .max_keysize = TF_MAX_KEY_SIZE * 2,
429 .ivsize = TF_BLOCK_SIZE,
430 .setkey = xts_twofish_setkey,
431 .encrypt = xts_encrypt,
432 .decrypt = xts_decrypt,
433 },
434 },
435 } };
436
is_blacklisted_cpu(void)437 static bool is_blacklisted_cpu(void)
438 {
439 if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
440 return false;
441
442 if (boot_cpu_data.x86 == 0x06 &&
443 (boot_cpu_data.x86_model == 0x1c ||
444 boot_cpu_data.x86_model == 0x26 ||
445 boot_cpu_data.x86_model == 0x36)) {
446 /*
447 * On Atom, twofish-3way is slower than original assembler
448 * implementation. Twofish-3way trades off some performance in
449 * storing blocks in 64bit registers to allow three blocks to
450 * be processed parallel. Parallel operation then allows gaining
451 * more performance than was trade off, on out-of-order CPUs.
452 * However Atom does not benefit from this parallellism and
453 * should be blacklisted.
454 */
455 return true;
456 }
457
458 if (boot_cpu_data.x86 == 0x0f) {
459 /*
460 * On Pentium 4, twofish-3way is slower than original assembler
461 * implementation because excessive uses of 64bit rotate and
462 * left-shifts (which are really slow on P4) needed to store and
463 * handle 128bit block in two 64bit registers.
464 */
465 return true;
466 }
467
468 return false;
469 }
470
471 static int force;
472 module_param(force, int, 0);
473 MODULE_PARM_DESC(force, "Force module load, ignore CPU blacklist");
474
init(void)475 static int __init init(void)
476 {
477 if (!force && is_blacklisted_cpu()) {
478 printk(KERN_INFO
479 "twofish-x86_64-3way: performance on this CPU "
480 "would be suboptimal: disabling "
481 "twofish-x86_64-3way.\n");
482 return -ENODEV;
483 }
484
485 return crypto_register_algs(tf_algs, ARRAY_SIZE(tf_algs));
486 }
487
fini(void)488 static void __exit fini(void)
489 {
490 crypto_unregister_algs(tf_algs, ARRAY_SIZE(tf_algs));
491 }
492
493 module_init(init);
494 module_exit(fini);
495
496 MODULE_LICENSE("GPL");
497 MODULE_DESCRIPTION("Twofish Cipher Algorithm, 3-way parallel asm optimized");
498 MODULE_ALIAS("twofish");
499 MODULE_ALIAS("twofish-asm");
500