]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Cryptographic API. | |
3 | * | |
4 | * Support for VIA PadLock hardware crypto engine. | |
5 | * | |
6 | * Copyright (c) 2004 Michal Ludvig <michal@logix.cz> | |
7 | * | |
f8246af0 | 8 | * Key expansion routine taken from crypto/aes_generic.c |
1da177e4 LT |
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 | * --------------------------------------------------------------------------- | |
16 | * Copyright (c) 2002, Dr Brian Gladman <brg@gladman.me.uk>, Worcester, UK. | |
17 | * All rights reserved. | |
18 | * | |
19 | * LICENSE TERMS | |
20 | * | |
21 | * The free distribution and use of this software in both source and binary | |
22 | * form is allowed (with or without changes) provided that: | |
23 | * | |
24 | * 1. distributions of this source code include the above copyright | |
25 | * notice, this list of conditions and the following disclaimer; | |
26 | * | |
27 | * 2. distributions in binary form include the above copyright | |
28 | * notice, this list of conditions and the following disclaimer | |
29 | * in the documentation and/or other associated materials; | |
30 | * | |
31 | * 3. the copyright holder's name is not used to endorse products | |
32 | * built using this software without specific written permission. | |
33 | * | |
34 | * ALTERNATIVELY, provided that this notice is retained in full, this product | |
35 | * may be distributed under the terms of the GNU General Public License (GPL), | |
36 | * in which case the provisions of the GPL apply INSTEAD OF those given above. | |
37 | * | |
38 | * DISCLAIMER | |
39 | * | |
40 | * This software is provided 'as is' with no explicit or implied warranties | |
41 | * in respect of its properties, including, but not limited to, correctness | |
42 | * and/or fitness for purpose. | |
43 | * --------------------------------------------------------------------------- | |
44 | */ | |
45 | ||
28ce728a | 46 | #include <crypto/algapi.h> |
1da177e4 LT |
47 | #include <linux/module.h> |
48 | #include <linux/init.h> | |
49 | #include <linux/types.h> | |
50 | #include <linux/errno.h> | |
1da177e4 | 51 | #include <linux/interrupt.h> |
6789b2dc | 52 | #include <linux/kernel.h> |
1da177e4 LT |
53 | #include <asm/byteorder.h> |
54 | #include "padlock.h" | |
55 | ||
56 | #define AES_MIN_KEY_SIZE 16 /* in uint8_t units */ | |
57 | #define AES_MAX_KEY_SIZE 32 /* ditto */ | |
58 | #define AES_BLOCK_SIZE 16 /* ditto */ | |
59 | #define AES_EXTENDED_KEY_SIZE 64 /* in uint32_t units */ | |
60 | #define AES_EXTENDED_KEY_SIZE_B (AES_EXTENDED_KEY_SIZE * sizeof(uint32_t)) | |
61 | ||
ccc17c34 ML |
62 | /* Control word. */ |
63 | struct cword { | |
64 | unsigned int __attribute__ ((__packed__)) | |
65 | rounds:4, | |
66 | algo:3, | |
67 | keygen:1, | |
68 | interm:1, | |
69 | encdec:1, | |
70 | ksize:2; | |
71 | } __attribute__ ((__aligned__(PADLOCK_ALIGNMENT))); | |
72 | ||
cc08632f ML |
73 | /* Whenever making any changes to the following |
74 | * structure *make sure* you keep E, d_data | |
75 | * and cword aligned on 16 Bytes boundaries!!! */ | |
1da177e4 | 76 | struct aes_ctx { |
6789b2dc HX |
77 | struct { |
78 | struct cword encrypt; | |
79 | struct cword decrypt; | |
80 | } cword; | |
82062c72 | 81 | u32 *D; |
1da177e4 | 82 | int key_length; |
cc08632f ML |
83 | u32 E[AES_EXTENDED_KEY_SIZE] |
84 | __attribute__ ((__aligned__(PADLOCK_ALIGNMENT))); | |
85 | u32 d_data[AES_EXTENDED_KEY_SIZE] | |
86 | __attribute__ ((__aligned__(PADLOCK_ALIGNMENT))); | |
1da177e4 LT |
87 | }; |
88 | ||
89 | /* ====== Key management routines ====== */ | |
90 | ||
91 | static inline uint32_t | |
92 | generic_rotr32 (const uint32_t x, const unsigned bits) | |
93 | { | |
94 | const unsigned n = bits % 32; | |
95 | return (x >> n) | (x << (32 - n)); | |
96 | } | |
97 | ||
98 | static inline uint32_t | |
99 | generic_rotl32 (const uint32_t x, const unsigned bits) | |
100 | { | |
101 | const unsigned n = bits % 32; | |
102 | return (x << n) | (x >> (32 - n)); | |
103 | } | |
104 | ||
105 | #define rotl generic_rotl32 | |
106 | #define rotr generic_rotr32 | |
107 | ||
108 | /* | |
109 | * #define byte(x, nr) ((unsigned char)((x) >> (nr*8))) | |
110 | */ | |
111 | static inline uint8_t | |
112 | byte(const uint32_t x, const unsigned n) | |
113 | { | |
114 | return x >> (n << 3); | |
115 | } | |
116 | ||
1da177e4 LT |
117 | #define E_KEY ctx->E |
118 | #define D_KEY ctx->D | |
119 | ||
120 | static uint8_t pow_tab[256]; | |
121 | static uint8_t log_tab[256]; | |
122 | static uint8_t sbx_tab[256]; | |
123 | static uint8_t isb_tab[256]; | |
124 | static uint32_t rco_tab[10]; | |
125 | static uint32_t ft_tab[4][256]; | |
126 | static uint32_t it_tab[4][256]; | |
127 | ||
128 | static uint32_t fl_tab[4][256]; | |
129 | static uint32_t il_tab[4][256]; | |
130 | ||
131 | static inline uint8_t | |
132 | f_mult (uint8_t a, uint8_t b) | |
133 | { | |
134 | uint8_t aa = log_tab[a], cc = aa + log_tab[b]; | |
135 | ||
136 | return pow_tab[cc + (cc < aa ? 1 : 0)]; | |
137 | } | |
138 | ||
139 | #define ff_mult(a,b) (a && b ? f_mult(a, b) : 0) | |
140 | ||
141 | #define f_rn(bo, bi, n, k) \ | |
142 | bo[n] = ft_tab[0][byte(bi[n],0)] ^ \ | |
143 | ft_tab[1][byte(bi[(n + 1) & 3],1)] ^ \ | |
144 | ft_tab[2][byte(bi[(n + 2) & 3],2)] ^ \ | |
145 | ft_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n) | |
146 | ||
147 | #define i_rn(bo, bi, n, k) \ | |
148 | bo[n] = it_tab[0][byte(bi[n],0)] ^ \ | |
149 | it_tab[1][byte(bi[(n + 3) & 3],1)] ^ \ | |
150 | it_tab[2][byte(bi[(n + 2) & 3],2)] ^ \ | |
151 | it_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n) | |
152 | ||
153 | #define ls_box(x) \ | |
154 | ( fl_tab[0][byte(x, 0)] ^ \ | |
155 | fl_tab[1][byte(x, 1)] ^ \ | |
156 | fl_tab[2][byte(x, 2)] ^ \ | |
157 | fl_tab[3][byte(x, 3)] ) | |
158 | ||
159 | #define f_rl(bo, bi, n, k) \ | |
160 | bo[n] = fl_tab[0][byte(bi[n],0)] ^ \ | |
161 | fl_tab[1][byte(bi[(n + 1) & 3],1)] ^ \ | |
162 | fl_tab[2][byte(bi[(n + 2) & 3],2)] ^ \ | |
163 | fl_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n) | |
164 | ||
165 | #define i_rl(bo, bi, n, k) \ | |
166 | bo[n] = il_tab[0][byte(bi[n],0)] ^ \ | |
167 | il_tab[1][byte(bi[(n + 3) & 3],1)] ^ \ | |
168 | il_tab[2][byte(bi[(n + 2) & 3],2)] ^ \ | |
169 | il_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n) | |
170 | ||
171 | static void | |
172 | gen_tabs (void) | |
173 | { | |
174 | uint32_t i, t; | |
175 | uint8_t p, q; | |
176 | ||
177 | /* log and power tables for GF(2**8) finite field with | |
178 | 0x011b as modular polynomial - the simplest prmitive | |
179 | root is 0x03, used here to generate the tables */ | |
180 | ||
181 | for (i = 0, p = 1; i < 256; ++i) { | |
182 | pow_tab[i] = (uint8_t) p; | |
183 | log_tab[p] = (uint8_t) i; | |
184 | ||
185 | p ^= (p << 1) ^ (p & 0x80 ? 0x01b : 0); | |
186 | } | |
187 | ||
188 | log_tab[1] = 0; | |
189 | ||
190 | for (i = 0, p = 1; i < 10; ++i) { | |
191 | rco_tab[i] = p; | |
192 | ||
193 | p = (p << 1) ^ (p & 0x80 ? 0x01b : 0); | |
194 | } | |
195 | ||
196 | for (i = 0; i < 256; ++i) { | |
197 | p = (i ? pow_tab[255 - log_tab[i]] : 0); | |
198 | q = ((p >> 7) | (p << 1)) ^ ((p >> 6) | (p << 2)); | |
199 | p ^= 0x63 ^ q ^ ((q >> 6) | (q << 2)); | |
200 | sbx_tab[i] = p; | |
201 | isb_tab[p] = (uint8_t) i; | |
202 | } | |
203 | ||
204 | for (i = 0; i < 256; ++i) { | |
205 | p = sbx_tab[i]; | |
206 | ||
207 | t = p; | |
208 | fl_tab[0][i] = t; | |
209 | fl_tab[1][i] = rotl (t, 8); | |
210 | fl_tab[2][i] = rotl (t, 16); | |
211 | fl_tab[3][i] = rotl (t, 24); | |
212 | ||
213 | t = ((uint32_t) ff_mult (2, p)) | | |
214 | ((uint32_t) p << 8) | | |
215 | ((uint32_t) p << 16) | ((uint32_t) ff_mult (3, p) << 24); | |
216 | ||
217 | ft_tab[0][i] = t; | |
218 | ft_tab[1][i] = rotl (t, 8); | |
219 | ft_tab[2][i] = rotl (t, 16); | |
220 | ft_tab[3][i] = rotl (t, 24); | |
221 | ||
222 | p = isb_tab[i]; | |
223 | ||
224 | t = p; | |
225 | il_tab[0][i] = t; | |
226 | il_tab[1][i] = rotl (t, 8); | |
227 | il_tab[2][i] = rotl (t, 16); | |
228 | il_tab[3][i] = rotl (t, 24); | |
229 | ||
230 | t = ((uint32_t) ff_mult (14, p)) | | |
231 | ((uint32_t) ff_mult (9, p) << 8) | | |
232 | ((uint32_t) ff_mult (13, p) << 16) | | |
233 | ((uint32_t) ff_mult (11, p) << 24); | |
234 | ||
235 | it_tab[0][i] = t; | |
236 | it_tab[1][i] = rotl (t, 8); | |
237 | it_tab[2][i] = rotl (t, 16); | |
238 | it_tab[3][i] = rotl (t, 24); | |
239 | } | |
240 | } | |
241 | ||
242 | #define star_x(x) (((x) & 0x7f7f7f7f) << 1) ^ ((((x) & 0x80808080) >> 7) * 0x1b) | |
243 | ||
244 | #define imix_col(y,x) \ | |
245 | u = star_x(x); \ | |
246 | v = star_x(u); \ | |
247 | w = star_x(v); \ | |
248 | t = w ^ (x); \ | |
249 | (y) = u ^ v ^ w; \ | |
250 | (y) ^= rotr(u ^ t, 8) ^ \ | |
251 | rotr(v ^ t, 16) ^ \ | |
252 | rotr(t,24) | |
253 | ||
254 | /* initialise the key schedule from the user supplied key */ | |
255 | ||
256 | #define loop4(i) \ | |
257 | { t = rotr(t, 8); t = ls_box(t) ^ rco_tab[i]; \ | |
258 | t ^= E_KEY[4 * i]; E_KEY[4 * i + 4] = t; \ | |
259 | t ^= E_KEY[4 * i + 1]; E_KEY[4 * i + 5] = t; \ | |
260 | t ^= E_KEY[4 * i + 2]; E_KEY[4 * i + 6] = t; \ | |
261 | t ^= E_KEY[4 * i + 3]; E_KEY[4 * i + 7] = t; \ | |
262 | } | |
263 | ||
264 | #define loop6(i) \ | |
265 | { t = rotr(t, 8); t = ls_box(t) ^ rco_tab[i]; \ | |
266 | t ^= E_KEY[6 * i]; E_KEY[6 * i + 6] = t; \ | |
267 | t ^= E_KEY[6 * i + 1]; E_KEY[6 * i + 7] = t; \ | |
268 | t ^= E_KEY[6 * i + 2]; E_KEY[6 * i + 8] = t; \ | |
269 | t ^= E_KEY[6 * i + 3]; E_KEY[6 * i + 9] = t; \ | |
270 | t ^= E_KEY[6 * i + 4]; E_KEY[6 * i + 10] = t; \ | |
271 | t ^= E_KEY[6 * i + 5]; E_KEY[6 * i + 11] = t; \ | |
272 | } | |
273 | ||
274 | #define loop8(i) \ | |
275 | { t = rotr(t, 8); ; t = ls_box(t) ^ rco_tab[i]; \ | |
276 | t ^= E_KEY[8 * i]; E_KEY[8 * i + 8] = t; \ | |
277 | t ^= E_KEY[8 * i + 1]; E_KEY[8 * i + 9] = t; \ | |
278 | t ^= E_KEY[8 * i + 2]; E_KEY[8 * i + 10] = t; \ | |
279 | t ^= E_KEY[8 * i + 3]; E_KEY[8 * i + 11] = t; \ | |
280 | t = E_KEY[8 * i + 4] ^ ls_box(t); \ | |
281 | E_KEY[8 * i + 12] = t; \ | |
282 | t ^= E_KEY[8 * i + 5]; E_KEY[8 * i + 13] = t; \ | |
283 | t ^= E_KEY[8 * i + 6]; E_KEY[8 * i + 14] = t; \ | |
284 | t ^= E_KEY[8 * i + 7]; E_KEY[8 * i + 15] = t; \ | |
285 | } | |
286 | ||
287 | /* Tells whether the ACE is capable to generate | |
288 | the extended key for a given key_len. */ | |
289 | static inline int | |
290 | aes_hw_extkey_available(uint8_t key_len) | |
291 | { | |
292 | /* TODO: We should check the actual CPU model/stepping | |
293 | as it's possible that the capability will be | |
294 | added in the next CPU revisions. */ | |
295 | if (key_len == 16) | |
296 | return 1; | |
297 | return 0; | |
298 | } | |
299 | ||
28ce728a | 300 | static inline struct aes_ctx *aes_ctx_common(void *ctx) |
6789b2dc | 301 | { |
28ce728a | 302 | unsigned long addr = (unsigned long)ctx; |
f10b7897 HX |
303 | unsigned long align = PADLOCK_ALIGNMENT; |
304 | ||
305 | if (align <= crypto_tfm_ctx_alignment()) | |
306 | align = 1; | |
6c2bb98b | 307 | return (struct aes_ctx *)ALIGN(addr, align); |
6789b2dc HX |
308 | } |
309 | ||
28ce728a HX |
310 | static inline struct aes_ctx *aes_ctx(struct crypto_tfm *tfm) |
311 | { | |
312 | return aes_ctx_common(crypto_tfm_ctx(tfm)); | |
313 | } | |
314 | ||
315 | static inline struct aes_ctx *blk_aes_ctx(struct crypto_blkcipher *tfm) | |
316 | { | |
317 | return aes_ctx_common(crypto_blkcipher_ctx(tfm)); | |
318 | } | |
319 | ||
6c2bb98b | 320 | static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key, |
560c06ae | 321 | unsigned int key_len) |
1da177e4 | 322 | { |
6c2bb98b | 323 | struct aes_ctx *ctx = aes_ctx(tfm); |
06ace7a9 | 324 | const __le32 *key = (const __le32 *)in_key; |
560c06ae | 325 | u32 *flags = &tfm->crt_flags; |
1da177e4 LT |
326 | uint32_t i, t, u, v, w; |
327 | uint32_t P[AES_EXTENDED_KEY_SIZE]; | |
328 | uint32_t rounds; | |
329 | ||
560c06ae | 330 | if (key_len % 8) { |
1da177e4 LT |
331 | *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN; |
332 | return -EINVAL; | |
333 | } | |
334 | ||
335 | ctx->key_length = key_len; | |
336 | ||
6789b2dc HX |
337 | /* |
338 | * If the hardware is capable of generating the extended key | |
339 | * itself we must supply the plain key for both encryption | |
340 | * and decryption. | |
341 | */ | |
82062c72 | 342 | ctx->D = ctx->E; |
1da177e4 | 343 | |
06ace7a9 HX |
344 | E_KEY[0] = le32_to_cpu(key[0]); |
345 | E_KEY[1] = le32_to_cpu(key[1]); | |
346 | E_KEY[2] = le32_to_cpu(key[2]); | |
347 | E_KEY[3] = le32_to_cpu(key[3]); | |
1da177e4 | 348 | |
6789b2dc HX |
349 | /* Prepare control words. */ |
350 | memset(&ctx->cword, 0, sizeof(ctx->cword)); | |
351 | ||
352 | ctx->cword.decrypt.encdec = 1; | |
353 | ctx->cword.encrypt.rounds = 10 + (key_len - 16) / 4; | |
354 | ctx->cword.decrypt.rounds = ctx->cword.encrypt.rounds; | |
355 | ctx->cword.encrypt.ksize = (key_len - 16) / 8; | |
356 | ctx->cword.decrypt.ksize = ctx->cword.encrypt.ksize; | |
357 | ||
1da177e4 LT |
358 | /* Don't generate extended keys if the hardware can do it. */ |
359 | if (aes_hw_extkey_available(key_len)) | |
360 | return 0; | |
361 | ||
6789b2dc HX |
362 | ctx->D = ctx->d_data; |
363 | ctx->cword.encrypt.keygen = 1; | |
364 | ctx->cword.decrypt.keygen = 1; | |
365 | ||
1da177e4 LT |
366 | switch (key_len) { |
367 | case 16: | |
368 | t = E_KEY[3]; | |
369 | for (i = 0; i < 10; ++i) | |
370 | loop4 (i); | |
371 | break; | |
372 | ||
373 | case 24: | |
06ace7a9 HX |
374 | E_KEY[4] = le32_to_cpu(key[4]); |
375 | t = E_KEY[5] = le32_to_cpu(key[5]); | |
1da177e4 LT |
376 | for (i = 0; i < 8; ++i) |
377 | loop6 (i); | |
378 | break; | |
379 | ||
380 | case 32: | |
102d60a2 HX |
381 | E_KEY[4] = le32_to_cpu(key[4]); |
382 | E_KEY[5] = le32_to_cpu(key[5]); | |
383 | E_KEY[6] = le32_to_cpu(key[6]); | |
384 | t = E_KEY[7] = le32_to_cpu(key[7]); | |
1da177e4 LT |
385 | for (i = 0; i < 7; ++i) |
386 | loop8 (i); | |
387 | break; | |
388 | } | |
389 | ||
390 | D_KEY[0] = E_KEY[0]; | |
391 | D_KEY[1] = E_KEY[1]; | |
392 | D_KEY[2] = E_KEY[2]; | |
393 | D_KEY[3] = E_KEY[3]; | |
394 | ||
395 | for (i = 4; i < key_len + 24; ++i) { | |
396 | imix_col (D_KEY[i], E_KEY[i]); | |
397 | } | |
398 | ||
399 | /* PadLock needs a different format of the decryption key. */ | |
400 | rounds = 10 + (key_len - 16) / 4; | |
401 | ||
402 | for (i = 0; i < rounds; i++) { | |
403 | P[((i + 1) * 4) + 0] = D_KEY[((rounds - i - 1) * 4) + 0]; | |
404 | P[((i + 1) * 4) + 1] = D_KEY[((rounds - i - 1) * 4) + 1]; | |
405 | P[((i + 1) * 4) + 2] = D_KEY[((rounds - i - 1) * 4) + 2]; | |
406 | P[((i + 1) * 4) + 3] = D_KEY[((rounds - i - 1) * 4) + 3]; | |
407 | } | |
408 | ||
409 | P[0] = E_KEY[(rounds * 4) + 0]; | |
410 | P[1] = E_KEY[(rounds * 4) + 1]; | |
411 | P[2] = E_KEY[(rounds * 4) + 2]; | |
412 | P[3] = E_KEY[(rounds * 4) + 3]; | |
413 | ||
414 | memcpy(D_KEY, P, AES_EXTENDED_KEY_SIZE_B); | |
415 | ||
416 | return 0; | |
417 | } | |
418 | ||
419 | /* ====== Encryption/decryption routines ====== */ | |
420 | ||
28e8c3ad | 421 | /* These are the real call to PadLock. */ |
d4a7dd8e HX |
422 | static inline void padlock_xcrypt(const u8 *input, u8 *output, void *key, |
423 | void *control_word) | |
424 | { | |
425 | asm volatile (".byte 0xf3,0x0f,0xa7,0xc8" /* rep xcryptecb */ | |
426 | : "+S"(input), "+D"(output) | |
427 | : "d"(control_word), "b"(key), "c"(1)); | |
428 | } | |
429 | ||
430 | static void aes_crypt_copy(const u8 *in, u8 *out, u32 *key, struct cword *cword) | |
431 | { | |
490fe3f0 HX |
432 | u8 buf[AES_BLOCK_SIZE * 2 + PADLOCK_ALIGNMENT - 1]; |
433 | u8 *tmp = PTR_ALIGN(&buf[0], PADLOCK_ALIGNMENT); | |
d4a7dd8e HX |
434 | |
435 | memcpy(tmp, in, AES_BLOCK_SIZE); | |
436 | padlock_xcrypt(tmp, out, key, cword); | |
437 | } | |
438 | ||
439 | static inline void aes_crypt(const u8 *in, u8 *out, u32 *key, | |
440 | struct cword *cword) | |
441 | { | |
442 | asm volatile ("pushfl; popfl"); | |
443 | ||
444 | /* padlock_xcrypt requires at least two blocks of data. */ | |
445 | if (unlikely(!(((unsigned long)in ^ (PAGE_SIZE - AES_BLOCK_SIZE)) & | |
446 | (PAGE_SIZE - 1)))) { | |
447 | aes_crypt_copy(in, out, key, cword); | |
448 | return; | |
449 | } | |
450 | ||
451 | padlock_xcrypt(in, out, key, cword); | |
452 | } | |
453 | ||
6789b2dc HX |
454 | static inline void padlock_xcrypt_ecb(const u8 *input, u8 *output, void *key, |
455 | void *control_word, u32 count) | |
1da177e4 | 456 | { |
d4a7dd8e HX |
457 | if (count == 1) { |
458 | aes_crypt(input, output, key, control_word); | |
459 | return; | |
460 | } | |
461 | ||
1da177e4 | 462 | asm volatile ("pushfl; popfl"); /* enforce key reload. */ |
d4a7dd8e HX |
463 | asm volatile ("test $1, %%cl;" |
464 | "je 1f;" | |
465 | "lea -1(%%ecx), %%eax;" | |
466 | "mov $1, %%ecx;" | |
467 | ".byte 0xf3,0x0f,0xa7,0xc8;" /* rep xcryptecb */ | |
468 | "mov %%eax, %%ecx;" | |
469 | "1:" | |
470 | ".byte 0xf3,0x0f,0xa7,0xc8" /* rep xcryptecb */ | |
1da177e4 | 471 | : "+S"(input), "+D"(output) |
d4a7dd8e HX |
472 | : "d"(control_word), "b"(key), "c"(count) |
473 | : "ax"); | |
1da177e4 LT |
474 | } |
475 | ||
476df259 HX |
476 | static inline u8 *padlock_xcrypt_cbc(const u8 *input, u8 *output, void *key, |
477 | u8 *iv, void *control_word, u32 count) | |
28e8c3ad HX |
478 | { |
479 | /* Enforce key reload. */ | |
480 | asm volatile ("pushfl; popfl"); | |
481 | /* rep xcryptcbc */ | |
482 | asm volatile (".byte 0xf3,0x0f,0xa7,0xd0" | |
483 | : "+S" (input), "+D" (output), "+a" (iv) | |
484 | : "d" (control_word), "b" (key), "c" (count)); | |
476df259 | 485 | return iv; |
28e8c3ad HX |
486 | } |
487 | ||
6c2bb98b | 488 | static void aes_encrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) |
1da177e4 | 489 | { |
6c2bb98b | 490 | struct aes_ctx *ctx = aes_ctx(tfm); |
d4a7dd8e | 491 | aes_crypt(in, out, ctx->E, &ctx->cword.encrypt); |
1da177e4 LT |
492 | } |
493 | ||
6c2bb98b | 494 | static void aes_decrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) |
1da177e4 | 495 | { |
6c2bb98b | 496 | struct aes_ctx *ctx = aes_ctx(tfm); |
d4a7dd8e | 497 | aes_crypt(in, out, ctx->D, &ctx->cword.decrypt); |
1da177e4 LT |
498 | } |
499 | ||
500 | static struct crypto_alg aes_alg = { | |
501 | .cra_name = "aes", | |
c8a19c91 | 502 | .cra_driver_name = "aes-padlock", |
ccc17c34 | 503 | .cra_priority = PADLOCK_CRA_PRIORITY, |
1da177e4 LT |
504 | .cra_flags = CRYPTO_ALG_TYPE_CIPHER, |
505 | .cra_blocksize = AES_BLOCK_SIZE, | |
fbdae9f3 | 506 | .cra_ctxsize = sizeof(struct aes_ctx), |
6789b2dc | 507 | .cra_alignmask = PADLOCK_ALIGNMENT - 1, |
1da177e4 LT |
508 | .cra_module = THIS_MODULE, |
509 | .cra_list = LIST_HEAD_INIT(aes_alg.cra_list), | |
510 | .cra_u = { | |
511 | .cipher = { | |
512 | .cia_min_keysize = AES_MIN_KEY_SIZE, | |
513 | .cia_max_keysize = AES_MAX_KEY_SIZE, | |
514 | .cia_setkey = aes_set_key, | |
515 | .cia_encrypt = aes_encrypt, | |
28e8c3ad | 516 | .cia_decrypt = aes_decrypt, |
1da177e4 LT |
517 | } |
518 | } | |
519 | }; | |
520 | ||
28ce728a HX |
521 | static int ecb_aes_encrypt(struct blkcipher_desc *desc, |
522 | struct scatterlist *dst, struct scatterlist *src, | |
523 | unsigned int nbytes) | |
524 | { | |
525 | struct aes_ctx *ctx = blk_aes_ctx(desc->tfm); | |
526 | struct blkcipher_walk walk; | |
527 | int err; | |
528 | ||
529 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
530 | err = blkcipher_walk_virt(desc, &walk); | |
531 | ||
532 | while ((nbytes = walk.nbytes)) { | |
533 | padlock_xcrypt_ecb(walk.src.virt.addr, walk.dst.virt.addr, | |
534 | ctx->E, &ctx->cword.encrypt, | |
535 | nbytes / AES_BLOCK_SIZE); | |
536 | nbytes &= AES_BLOCK_SIZE - 1; | |
537 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
538 | } | |
539 | ||
540 | return err; | |
541 | } | |
542 | ||
543 | static int ecb_aes_decrypt(struct blkcipher_desc *desc, | |
544 | struct scatterlist *dst, struct scatterlist *src, | |
545 | unsigned int nbytes) | |
546 | { | |
547 | struct aes_ctx *ctx = blk_aes_ctx(desc->tfm); | |
548 | struct blkcipher_walk walk; | |
549 | int err; | |
550 | ||
551 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
552 | err = blkcipher_walk_virt(desc, &walk); | |
553 | ||
554 | while ((nbytes = walk.nbytes)) { | |
555 | padlock_xcrypt_ecb(walk.src.virt.addr, walk.dst.virt.addr, | |
556 | ctx->D, &ctx->cword.decrypt, | |
557 | nbytes / AES_BLOCK_SIZE); | |
558 | nbytes &= AES_BLOCK_SIZE - 1; | |
559 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
560 | } | |
561 | ||
562 | return err; | |
563 | } | |
564 | ||
565 | static struct crypto_alg ecb_aes_alg = { | |
566 | .cra_name = "ecb(aes)", | |
567 | .cra_driver_name = "ecb-aes-padlock", | |
568 | .cra_priority = PADLOCK_COMPOSITE_PRIORITY, | |
569 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
570 | .cra_blocksize = AES_BLOCK_SIZE, | |
571 | .cra_ctxsize = sizeof(struct aes_ctx), | |
572 | .cra_alignmask = PADLOCK_ALIGNMENT - 1, | |
573 | .cra_type = &crypto_blkcipher_type, | |
574 | .cra_module = THIS_MODULE, | |
575 | .cra_list = LIST_HEAD_INIT(ecb_aes_alg.cra_list), | |
576 | .cra_u = { | |
577 | .blkcipher = { | |
578 | .min_keysize = AES_MIN_KEY_SIZE, | |
579 | .max_keysize = AES_MAX_KEY_SIZE, | |
580 | .setkey = aes_set_key, | |
581 | .encrypt = ecb_aes_encrypt, | |
582 | .decrypt = ecb_aes_decrypt, | |
583 | } | |
584 | } | |
585 | }; | |
586 | ||
587 | static int cbc_aes_encrypt(struct blkcipher_desc *desc, | |
588 | struct scatterlist *dst, struct scatterlist *src, | |
589 | unsigned int nbytes) | |
590 | { | |
591 | struct aes_ctx *ctx = blk_aes_ctx(desc->tfm); | |
592 | struct blkcipher_walk walk; | |
593 | int err; | |
594 | ||
595 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
596 | err = blkcipher_walk_virt(desc, &walk); | |
597 | ||
598 | while ((nbytes = walk.nbytes)) { | |
599 | u8 *iv = padlock_xcrypt_cbc(walk.src.virt.addr, | |
600 | walk.dst.virt.addr, ctx->E, | |
601 | walk.iv, &ctx->cword.encrypt, | |
602 | nbytes / AES_BLOCK_SIZE); | |
603 | memcpy(walk.iv, iv, AES_BLOCK_SIZE); | |
604 | nbytes &= AES_BLOCK_SIZE - 1; | |
605 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
606 | } | |
607 | ||
608 | return err; | |
609 | } | |
610 | ||
611 | static int cbc_aes_decrypt(struct blkcipher_desc *desc, | |
612 | struct scatterlist *dst, struct scatterlist *src, | |
613 | unsigned int nbytes) | |
614 | { | |
615 | struct aes_ctx *ctx = blk_aes_ctx(desc->tfm); | |
616 | struct blkcipher_walk walk; | |
617 | int err; | |
618 | ||
619 | blkcipher_walk_init(&walk, dst, src, nbytes); | |
620 | err = blkcipher_walk_virt(desc, &walk); | |
621 | ||
622 | while ((nbytes = walk.nbytes)) { | |
623 | padlock_xcrypt_cbc(walk.src.virt.addr, walk.dst.virt.addr, | |
624 | ctx->D, walk.iv, &ctx->cword.decrypt, | |
625 | nbytes / AES_BLOCK_SIZE); | |
626 | nbytes &= AES_BLOCK_SIZE - 1; | |
627 | err = blkcipher_walk_done(desc, &walk, nbytes); | |
628 | } | |
629 | ||
630 | return err; | |
631 | } | |
632 | ||
633 | static struct crypto_alg cbc_aes_alg = { | |
634 | .cra_name = "cbc(aes)", | |
635 | .cra_driver_name = "cbc-aes-padlock", | |
636 | .cra_priority = PADLOCK_COMPOSITE_PRIORITY, | |
637 | .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, | |
638 | .cra_blocksize = AES_BLOCK_SIZE, | |
639 | .cra_ctxsize = sizeof(struct aes_ctx), | |
640 | .cra_alignmask = PADLOCK_ALIGNMENT - 1, | |
641 | .cra_type = &crypto_blkcipher_type, | |
642 | .cra_module = THIS_MODULE, | |
643 | .cra_list = LIST_HEAD_INIT(cbc_aes_alg.cra_list), | |
644 | .cra_u = { | |
645 | .blkcipher = { | |
646 | .min_keysize = AES_MIN_KEY_SIZE, | |
647 | .max_keysize = AES_MAX_KEY_SIZE, | |
648 | .ivsize = AES_BLOCK_SIZE, | |
649 | .setkey = aes_set_key, | |
650 | .encrypt = cbc_aes_encrypt, | |
651 | .decrypt = cbc_aes_decrypt, | |
652 | } | |
653 | } | |
654 | }; | |
655 | ||
1191f0a4 | 656 | static int __init padlock_init(void) |
1da177e4 | 657 | { |
1191f0a4 ML |
658 | int ret; |
659 | ||
660 | if (!cpu_has_xcrypt) { | |
661 | printk(KERN_ERR PFX "VIA PadLock not detected.\n"); | |
662 | return -ENODEV; | |
663 | } | |
664 | ||
665 | if (!cpu_has_xcrypt_enabled) { | |
666 | printk(KERN_ERR PFX "VIA PadLock detected, but not enabled. Hmm, strange...\n"); | |
667 | return -ENODEV; | |
668 | } | |
1da177e4 LT |
669 | |
670 | gen_tabs(); | |
28ce728a HX |
671 | if ((ret = crypto_register_alg(&aes_alg))) |
672 | goto aes_err; | |
673 | ||
674 | if ((ret = crypto_register_alg(&ecb_aes_alg))) | |
675 | goto ecb_aes_err; | |
676 | ||
677 | if ((ret = crypto_register_alg(&cbc_aes_alg))) | |
678 | goto cbc_aes_err; | |
1191f0a4 ML |
679 | |
680 | printk(KERN_NOTICE PFX "Using VIA PadLock ACE for AES algorithm.\n"); | |
681 | ||
28ce728a | 682 | out: |
1191f0a4 | 683 | return ret; |
28ce728a HX |
684 | |
685 | cbc_aes_err: | |
686 | crypto_unregister_alg(&ecb_aes_alg); | |
687 | ecb_aes_err: | |
688 | crypto_unregister_alg(&aes_alg); | |
689 | aes_err: | |
690 | printk(KERN_ERR PFX "VIA PadLock AES initialization failed.\n"); | |
691 | goto out; | |
1da177e4 LT |
692 | } |
693 | ||
1191f0a4 | 694 | static void __exit padlock_fini(void) |
1da177e4 | 695 | { |
28ce728a HX |
696 | crypto_unregister_alg(&cbc_aes_alg); |
697 | crypto_unregister_alg(&ecb_aes_alg); | |
1da177e4 LT |
698 | crypto_unregister_alg(&aes_alg); |
699 | } | |
1191f0a4 ML |
700 | |
701 | module_init(padlock_init); | |
702 | module_exit(padlock_fini); | |
703 | ||
704 | MODULE_DESCRIPTION("VIA PadLock AES algorithm support"); | |
705 | MODULE_LICENSE("GPL"); | |
706 | MODULE_AUTHOR("Michal Ludvig"); | |
707 | ||
f8246af0 | 708 | MODULE_ALIAS("aes"); |