?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux-2.6.12-rc2
[linux-2.6/kvm.git] / arch / i386 / crypto / aes.c
blob1019430fc1f11015128f2f22fdf44f7d24330ab3
1 /*
2 *
3 * Glue Code for optimized 586 assembler version of AES
4 *
5 * Copyright (c) 2002, Dr Brian Gladman <>, Worcester, UK.
6 * All rights reserved.
7 *
8 * LICENSE TERMS
9 *
10 * The free distribution and use of this software in both source and binary
11 * form is allowed (with or without changes) provided that:
12 *
13 * 1. distributions of this source code include the above copyright
14 * notice, this list of conditions and the following disclaimer;
15 *
16 * 2. distributions in binary form include the above copyright
17 * notice, this list of conditions and the following disclaimer
18 * in the documentation and/or other associated materials;
19 *
20 * 3. the copyright holder's name is not used to endorse products
21 * built using this software without specific written permission.
22 *
23 * ALTERNATIVELY, provided that this notice is retained in full, this product
24 * may be distributed under the terms of the GNU General Public License (GPL),
25 * in which case the provisions of the GPL apply INSTEAD OF those given above.
26 *
27 * DISCLAIMER
28 *
29 * This software is provided 'as is' with no explicit or implied warranties
30 * in respect of its properties, including, but not limited to, correctness
31 * and/or fitness for purpose.
32 *
33 * Copyright (c) 2003, Adam J. Richter <adam@yggdrasil.com> (conversion to
34 * 2.5 API).
35 * Copyright (c) 2003, 2004 Fruhwirth Clemens <clemens@endorphin.org>
36 * Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
37 *
38 */
39 #include <linux/kernel.h>
40 #include <linux/module.h>
41 #include <linux/init.h>
42 #include <linux/types.h>
43 #include <linux/crypto.h>
44 #include <linux/linkage.h>
45
46 asmlinkage void aes_enc_blk(const u8 *src, u8 *dst, void *ctx);
47 asmlinkage void aes_dec_blk(const u8 *src, u8 *dst, void *ctx);
48
49 #define AES_MIN_KEY_SIZE 16
50 #define AES_MAX_KEY_SIZE 32
51 #define AES_BLOCK_SIZE 16
52 #define AES_KS_LENGTH 4 * AES_BLOCK_SIZE
53 #define RC_LENGTH 29
54
55 struct aes_ctx {
56 u32 ekey[AES_KS_LENGTH];
57 u32 rounds;
58 u32 dkey[AES_KS_LENGTH];
59 };
60
61 #define WPOLY 0x011b
62 #define u32_in(x) le32_to_cpu(*(const u32 *)(x))
63 #define bytes2word(b0, b1, b2, b3) \
64 (((u32)(b3) << 24) | ((u32)(b2) << 16) | ((u32)(b1) << 8) | (b0))
65
66 /* define the finite field multiplies required for Rijndael */
67 #define f2(x) ((x) ? pow[log[x] + 0x19] : 0)
68 #define f3(x) ((x) ? pow[log[x] + 0x01] : 0)
69 #define f9(x) ((x) ? pow[log[x] + 0xc7] : 0)
70 #define fb(x) ((x) ? pow[log[x] + 0x68] : 0)
71 #define fd(x) ((x) ? pow[log[x] + 0xee] : 0)
72 #define fe(x) ((x) ? pow[log[x] + 0xdf] : 0)
73 #define fi(x) ((x) ? pow[255 - log[x]]: 0)
74
75 static inline u32 upr(u32 x, int n)
76 {
77 return (x << 8 * n) | (x >> (32 - 8 * n));
78 }
79
80 static inline u8 bval(u32 x, int n)
81 {
82 return x >> 8 * n;
83 }
84
85 /* The forward and inverse affine transformations used in the S-box */
86 #define fwd_affine(x) \
87 (w = (u32)x, w ^= (w<<1)^(w<<2)^(w<<3)^(w<<4), 0x63^(u8)(w^(w>>8)))
88
89 #define inv_affine(x) \
90 (w = (u32)x, w = (w<<1)^(w<<3)^(w<<6), 0x05^(u8)(w^(w>>8)))
91
92 static u32 rcon_tab[RC_LENGTH];
93
94 u32 ft_tab[4][256];
95 u32 fl_tab[4][256];
96 static u32 ls_tab[4][256];
97 static u32 im_tab[4][256];
98 u32 il_tab[4][256];
99 u32 it_tab[4][256];
100
101 static void gen_tabs(void)
102 {
103 u32 i, w;
104 u8 pow[512], log[256];
105
106 /*
107 * log and power tables for GF(2^8) finite field with
108 * WPOLY as modular polynomial - the simplest primitive
109 * root is 0x03, used here to generate the tables.
110 */
111 i = 0; w = 1;
112
113 do {
114 pow[i] = (u8)w;
115 pow[i + 255] = (u8)w;
116 log[w] = (u8)i++;
117 w ^= (w << 1) ^ (w & 0x80 ? WPOLY : 0);
118 } while (w != 1);
119
120 for(i = 0, w = 1; i < RC_LENGTH; ++i) {
121 rcon_tab[i] = bytes2word(w, 0, 0, 0);
122 w = f2(w);
123 }
124
125 for(i = 0; i < 256; ++i) {
126 u8 b;
127
128 b = fwd_affine(fi((u8)i));
129 w = bytes2word(f2(b), b, b, f3(b));
130
131 /* tables for a normal encryption round */
132 ft_tab[0][i] = w;
133 ft_tab[1][i] = upr(w, 1);
134 ft_tab[2][i] = upr(w, 2);
135 ft_tab[3][i] = upr(w, 3);
136 w = bytes2word(b, 0, 0, 0);
137
138 /*
139 * tables for last encryption round
140 * (may also be used in the key schedule)
141 */
142 fl_tab[0][i] = w;
143 fl_tab[1][i] = upr(w, 1);
144 fl_tab[2][i] = upr(w, 2);
145 fl_tab[3][i] = upr(w, 3);
146
147 /*
148 * table for key schedule if fl_tab above is
149 * not of the required form
150 */
151 ls_tab[0][i] = w;
152 ls_tab[1][i] = upr(w, 1);
153 ls_tab[2][i] = upr(w, 2);
154 ls_tab[3][i] = upr(w, 3);
155
156 b = fi(inv_affine((u8)i));
157 w = bytes2word(fe(b), f9(b), fd(b), fb(b));
158
159 /* tables for the inverse mix column operation */
160 im_tab[0][b] = w;
161 im_tab[1][b] = upr(w, 1);
162 im_tab[2][b] = upr(w, 2);
163 im_tab[3][b] = upr(w, 3);
164
165 /* tables for a normal decryption round */
166 it_tab[0][i] = w;
167 it_tab[1][i] = upr(w,1);
168 it_tab[2][i] = upr(w,2);
169 it_tab[3][i] = upr(w,3);
170
171 w = bytes2word(b, 0, 0, 0);
172
173 /* tables for last decryption round */
174 il_tab[0][i] = w;
175 il_tab[1][i] = upr(w,1);
176 il_tab[2][i] = upr(w,2);
177 il_tab[3][i] = upr(w,3);
178 }
179 }
180
181 #define four_tables(x,tab,vf,rf,c) \
182 ( tab[0][bval(vf(x,0,c),rf(0,c))] ^ \
183 tab[1][bval(vf(x,1,c),rf(1,c))] ^ \
184 tab[2][bval(vf(x,2,c),rf(2,c))] ^ \
185 tab[3][bval(vf(x,3,c),rf(3,c))] \
186 )
187
188 #define vf1(x,r,c) (x)
189 #define rf1(r,c) (r)
190 #define rf2(r,c) ((r-c)&3)
191
192 #define inv_mcol(x) four_tables(x,im_tab,vf1,rf1,0)
193 #define ls_box(x,c) four_tables(x,fl_tab,vf1,rf2,c)
194
195 #define ff(x) inv_mcol(x)
196
197 #define ke4(k,i) \
198 { \
199 k[4*(i)+4] = ss[0] ^= ls_box(ss[3],3) ^ rcon_tab[i]; \
200 k[4*(i)+5] = ss[1] ^= ss[0]; \
201 k[4*(i)+6] = ss[2] ^= ss[1]; \
202 k[4*(i)+7] = ss[3] ^= ss[2]; \
203 }
204
205 #define kel4(k,i) \
206 { \
207 k[4*(i)+4] = ss[0] ^= ls_box(ss[3],3) ^ rcon_tab[i]; \
208 k[4*(i)+5] = ss[1] ^= ss[0]; \
209 k[4*(i)+6] = ss[2] ^= ss[1]; k[4*(i)+7] = ss[3] ^= ss[2]; \
210 }
211
212 #define ke6(k,i) \
213 { \
214 k[6*(i)+ 6] = ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i]; \
215 k[6*(i)+ 7] = ss[1] ^= ss[0]; \
216 k[6*(i)+ 8] = ss[2] ^= ss[1]; \
217 k[6*(i)+ 9] = ss[3] ^= ss[2]; \
218 k[6*(i)+10] = ss[4] ^= ss[3]; \
219 k[6*(i)+11] = ss[5] ^= ss[4]; \
220 }
221
222 #define kel6(k,i) \
223 { \
224 k[6*(i)+ 6] = ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i]; \
225 k[6*(i)+ 7] = ss[1] ^= ss[0]; \
226 k[6*(i)+ 8] = ss[2] ^= ss[1]; \
227 k[6*(i)+ 9] = ss[3] ^= ss[2]; \
228 }
229
230 #define ke8(k,i) \
231 { \
232 k[8*(i)+ 8] = ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i]; \
233 k[8*(i)+ 9] = ss[1] ^= ss[0]; \
234 k[8*(i)+10] = ss[2] ^= ss[1]; \
235 k[8*(i)+11] = ss[3] ^= ss[2]; \
236 k[8*(i)+12] = ss[4] ^= ls_box(ss[3],0); \
237 k[8*(i)+13] = ss[5] ^= ss[4]; \
238 k[8*(i)+14] = ss[6] ^= ss[5]; \
239 k[8*(i)+15] = ss[7] ^= ss[6]; \
240 }
241
242 #define kel8(k,i) \
243 { \
244 k[8*(i)+ 8] = ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i]; \
245 k[8*(i)+ 9] = ss[1] ^= ss[0]; \
246 k[8*(i)+10] = ss[2] ^= ss[1]; \
247 k[8*(i)+11] = ss[3] ^= ss[2]; \
248 }
249
250 #define kdf4(k,i) \
251 { \
252 ss[0] = ss[0] ^ ss[2] ^ ss[1] ^ ss[3]; \
253 ss[1] = ss[1] ^ ss[3]; \
254 ss[2] = ss[2] ^ ss[3]; \
255 ss[3] = ss[3]; \
256 ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i]; \
257 ss[i % 4] ^= ss[4]; \
258 ss[4] ^= k[4*(i)]; \
259 k[4*(i)+4] = ff(ss[4]); \
260 ss[4] ^= k[4*(i)+1]; \
261 k[4*(i)+5] = ff(ss[4]); \
262 ss[4] ^= k[4*(i)+2]; \
263 k[4*(i)+6] = ff(ss[4]); \
264 ss[4] ^= k[4*(i)+3]; \
265 k[4*(i)+7] = ff(ss[4]); \
266 }
267
268 #define kd4(k,i) \
269 { \
270 ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i]; \
271 ss[i % 4] ^= ss[4]; \
272 ss[4] = ff(ss[4]); \
273 k[4*(i)+4] = ss[4] ^= k[4*(i)]; \
274 k[4*(i)+5] = ss[4] ^= k[4*(i)+1]; \
275 k[4*(i)+6] = ss[4] ^= k[4*(i)+2]; \
276 k[4*(i)+7] = ss[4] ^= k[4*(i)+3]; \
277 }
278
279 #define kdl4(k,i) \
280 { \
281 ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i]; \
282 ss[i % 4] ^= ss[4]; \
283 k[4*(i)+4] = (ss[0] ^= ss[1]) ^ ss[2] ^ ss[3]; \
284 k[4*(i)+5] = ss[1] ^ ss[3]; \
285 k[4*(i)+6] = ss[0]; \
286 k[4*(i)+7] = ss[1]; \
287 }
288
289 #define kdf6(k,i) \
290 { \
291 ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i]; \
292 k[6*(i)+ 6] = ff(ss[0]); \
293 ss[1] ^= ss[0]; \
294 k[6*(i)+ 7] = ff(ss[1]); \
295 ss[2] ^= ss[1]; \
296 k[6*(i)+ 8] = ff(ss[2]); \
297 ss[3] ^= ss[2]; \
298 k[6*(i)+ 9] = ff(ss[3]); \
299 ss[4] ^= ss[3]; \
300 k[6*(i)+10] = ff(ss[4]); \
301 ss[5] ^= ss[4]; \
302 k[6*(i)+11] = ff(ss[5]); \
303 }
304
305 #define kd6(k,i) \
306 { \
307 ss[6] = ls_box(ss[5],3) ^ rcon_tab[i]; \
308 ss[0] ^= ss[6]; ss[6] = ff(ss[6]); \
309 k[6*(i)+ 6] = ss[6] ^= k[6*(i)]; \
310 ss[1] ^= ss[0]; \
311 k[6*(i)+ 7] = ss[6] ^= k[6*(i)+ 1]; \
312 ss[2] ^= ss[1]; \
313 k[6*(i)+ 8] = ss[6] ^= k[6*(i)+ 2]; \
314 ss[3] ^= ss[2]; \
315 k[6*(i)+ 9] = ss[6] ^= k[6*(i)+ 3]; \
316 ss[4] ^= ss[3]; \
317 k[6*(i)+10] = ss[6] ^= k[6*(i)+ 4]; \
318 ss[5] ^= ss[4]; \
319 k[6*(i)+11] = ss[6] ^= k[6*(i)+ 5]; \
320 }
321
322 #define kdl6(k,i) \
323 { \
324 ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i]; \
325 k[6*(i)+ 6] = ss[0]; \
326 ss[1] ^= ss[0]; \
327 k[6*(i)+ 7] = ss[1]; \
328 ss[2] ^= ss[1]; \
329 k[6*(i)+ 8] = ss[2]; \
330 ss[3] ^= ss[2]; \
331 k[6*(i)+ 9] = ss[3]; \
332 }
333
334 #define kdf8(k,i) \
335 { \
336 ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i]; \
337 k[8*(i)+ 8] = ff(ss[0]); \
338 ss[1] ^= ss[0]; \
339 k[8*(i)+ 9] = ff(ss[1]); \
340 ss[2] ^= ss[1]; \
341 k[8*(i)+10] = ff(ss[2]); \
342 ss[3] ^= ss[2]; \
343 k[8*(i)+11] = ff(ss[3]); \
344 ss[4] ^= ls_box(ss[3],0); \
345 k[8*(i)+12] = ff(ss[4]); \
346 ss[5] ^= ss[4]; \
347 k[8*(i)+13] = ff(ss[5]); \
348 ss[6] ^= ss[5]; \
349 k[8*(i)+14] = ff(ss[6]); \
350 ss[7] ^= ss[6]; \
351 k[8*(i)+15] = ff(ss[7]); \
352 }
353
354 #define kd8(k,i) \
355 { \
356 u32 __g = ls_box(ss[7],3) ^ rcon_tab[i]; \
357 ss[0] ^= __g; \
358 __g = ff(__g); \
359 k[8*(i)+ 8] = __g ^= k[8*(i)]; \
360 ss[1] ^= ss[0]; \
361 k[8*(i)+ 9] = __g ^= k[8*(i)+ 1]; \
362 ss[2] ^= ss[1]; \
363 k[8*(i)+10] = __g ^= k[8*(i)+ 2]; \
364 ss[3] ^= ss[2]; \
365 k[8*(i)+11] = __g ^= k[8*(i)+ 3]; \
366 __g = ls_box(ss[3],0); \
367 ss[4] ^= __g; \
368 __g = ff(__g); \
369 k[8*(i)+12] = __g ^= k[8*(i)+ 4]; \
370 ss[5] ^= ss[4]; \
371 k[8*(i)+13] = __g ^= k[8*(i)+ 5]; \
372 ss[6] ^= ss[5]; \
373 k[8*(i)+14] = __g ^= k[8*(i)+ 6]; \
374 ss[7] ^= ss[6]; \
375 k[8*(i)+15] = __g ^= k[8*(i)+ 7]; \
376 }
377
378 #define kdl8(k,i) \
379 { \
380 ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i]; \
381 k[8*(i)+ 8] = ss[0]; \
382 ss[1] ^= ss[0]; \
383 k[8*(i)+ 9] = ss[1]; \
384 ss[2] ^= ss[1]; \
385 k[8*(i)+10] = ss[2]; \
386 ss[3] ^= ss[2]; \
387 k[8*(i)+11] = ss[3]; \
388 }
389
390 static int
391 aes_set_key(void *ctx_arg, const u8 *in_key, unsigned int key_len, u32 *flags)
392 {
393 int i;
394 u32 ss[8];
395 struct aes_ctx *ctx = ctx_arg;
396
397 /* encryption schedule */
398
399 ctx->ekey[0] = ss[0] = u32_in(in_key);
400 ctx->ekey[1] = ss[1] = u32_in(in_key + 4);
401 ctx->ekey[2] = ss[2] = u32_in(in_key + 8);
402 ctx->ekey[3] = ss[3] = u32_in(in_key + 12);
403
404 switch(key_len) {
405 case 16:
406 for (i = 0; i < 9; i++)
407 ke4(ctx->ekey, i);
408 kel4(ctx->ekey, 9);
409 ctx->rounds = 10;
410 break;
411
412 case 24:
413 ctx->ekey[4] = ss[4] = u32_in(in_key + 16);
414 ctx->ekey[5] = ss[5] = u32_in(in_key + 20);
415 for (i = 0; i < 7; i++)
416 ke6(ctx->ekey, i);
417 kel6(ctx->ekey, 7);
418 ctx->rounds = 12;
419 break;
420
421 case 32:
422 ctx->ekey[4] = ss[4] = u32_in(in_key + 16);
423 ctx->ekey[5] = ss[5] = u32_in(in_key + 20);
424 ctx->ekey[6] = ss[6] = u32_in(in_key + 24);
425 ctx->ekey[7] = ss[7] = u32_in(in_key + 28);
426 for (i = 0; i < 6; i++)
427 ke8(ctx->ekey, i);
428 kel8(ctx->ekey, 6);
429 ctx->rounds = 14;
430 break;
431
432 default:
433 *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
434 return -EINVAL;
435 }
436
437 /* decryption schedule */
438
439 ctx->dkey[0] = ss[0] = u32_in(in_key);
440 ctx->dkey[1] = ss[1] = u32_in(in_key + 4);
441 ctx->dkey[2] = ss[2] = u32_in(in_key + 8);
442 ctx->dkey[3] = ss[3] = u32_in(in_key + 12);
443
444 switch (key_len) {
445 case 16:
446 kdf4(ctx->dkey, 0);
447 for (i = 1; i < 9; i++)
448 kd4(ctx->dkey, i);
449 kdl4(ctx->dkey, 9);
450 break;
451
452 case 24:
453 ctx->dkey[4] = ff(ss[4] = u32_in(in_key + 16));
454 ctx->dkey[5] = ff(ss[5] = u32_in(in_key + 20));
455 kdf6(ctx->dkey, 0);
456 for (i = 1; i < 7; i++)
457 kd6(ctx->dkey, i);
458 kdl6(ctx->dkey, 7);
459 break;
460
461 case 32:
462 ctx->dkey[4] = ff(ss[4] = u32_in(in_key + 16));
463 ctx->dkey[5] = ff(ss[5] = u32_in(in_key + 20));
464 ctx->dkey[6] = ff(ss[6] = u32_in(in_key + 24));
465 ctx->dkey[7] = ff(ss[7] = u32_in(in_key + 28));
466 kdf8(ctx->dkey, 0);
467 for (i = 1; i < 6; i++)
468 kd8(ctx->dkey, i);
469 kdl8(ctx->dkey, 6);
470 break;
471 }
472 return 0;
473 }
474
475 static inline void aes_encrypt(void *ctx, u8 *dst, const u8 *src)
476 {
477 aes_enc_blk(src, dst, ctx);
478 }
479 static inline void aes_decrypt(void *ctx, u8 *dst, const u8 *src)
480 {
481 aes_dec_blk(src, dst, ctx);
482 }
483
484
485 static struct crypto_alg aes_alg = {
486 .cra_name = "aes",
487 .cra_flags = CRYPTO_ALG_TYPE_CIPHER,
488 .cra_blocksize = AES_BLOCK_SIZE,
489 .cra_ctxsize = sizeof(struct aes_ctx),
490 .cra_module = THIS_MODULE,
491 .cra_list = LIST_HEAD_INIT(aes_alg.cra_list),
492 .cra_u = {
493 .cipher = {
494 .cia_min_keysize = AES_MIN_KEY_SIZE,
495 .cia_max_keysize = AES_MAX_KEY_SIZE,
496 .cia_setkey = aes_set_key,
497 .cia_encrypt = aes_encrypt,
498 .cia_decrypt = aes_decrypt
499 }
500 }
501 };
502
503 static int __init aes_init(void)
504 {
505 gen_tabs();
506 return crypto_register_alg(&aes_alg);
507 }
508
509 static void __exit aes_fini(void)
510 {
511 crypto_unregister_alg(&aes_alg);
512 }
513
514 module_init(aes_init);
515 module_exit(aes_fini);
516
517 MODULE_DESCRIPTION("Rijndael (AES) Cipher Algorithm, i586 asm optimized");
518 MODULE_LICENSE("Dual BSD/GPL");
519 MODULE_AUTHOR("Fruhwirth Clemens, James Morris, Brian Gladman, Adam Richter");
520 MODULE_ALIAS("aes");
kernel-based virtual machine