xref: /openbmc/linux/drivers/crypto/padlock-aes.c (revision a1e58bbd)
1 /*
2  * Cryptographic API.
3  *
4  * Support for VIA PadLock hardware crypto engine.
5  *
6  * Copyright (c) 2004  Michal Ludvig <michal@logix.cz>
7  *
8  * Key expansion routine taken from crypto/aes_generic.c
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 
46 #include <crypto/algapi.h>
47 #include <crypto/aes.h>
48 #include <linux/module.h>
49 #include <linux/init.h>
50 #include <linux/types.h>
51 #include <linux/errno.h>
52 #include <linux/interrupt.h>
53 #include <linux/kernel.h>
54 #include <asm/byteorder.h>
55 #include "padlock.h"
56 
57 #define AES_EXTENDED_KEY_SIZE	64	/* in uint32_t units */
58 #define AES_EXTENDED_KEY_SIZE_B	(AES_EXTENDED_KEY_SIZE * sizeof(uint32_t))
59 
60 /* Control word. */
61 struct cword {
62 	unsigned int __attribute__ ((__packed__))
63 		rounds:4,
64 		algo:3,
65 		keygen:1,
66 		interm:1,
67 		encdec:1,
68 		ksize:2;
69 } __attribute__ ((__aligned__(PADLOCK_ALIGNMENT)));
70 
71 /* Whenever making any changes to the following
72  * structure *make sure* you keep E, d_data
73  * and cword aligned on 16 Bytes boundaries!!! */
74 struct aes_ctx {
75 	struct {
76 		struct cword encrypt;
77 		struct cword decrypt;
78 	} cword;
79 	u32 *D;
80 	int key_length;
81 	u32 E[AES_EXTENDED_KEY_SIZE]
82 		__attribute__ ((__aligned__(PADLOCK_ALIGNMENT)));
83 	u32 d_data[AES_EXTENDED_KEY_SIZE]
84 		__attribute__ ((__aligned__(PADLOCK_ALIGNMENT)));
85 };
86 
87 /* ====== Key management routines ====== */
88 
89 static inline uint32_t
90 generic_rotr32 (const uint32_t x, const unsigned bits)
91 {
92 	const unsigned n = bits % 32;
93 	return (x >> n) | (x << (32 - n));
94 }
95 
96 static inline uint32_t
97 generic_rotl32 (const uint32_t x, const unsigned bits)
98 {
99 	const unsigned n = bits % 32;
100 	return (x << n) | (x >> (32 - n));
101 }
102 
103 #define rotl generic_rotl32
104 #define rotr generic_rotr32
105 
106 /*
107  * #define byte(x, nr) ((unsigned char)((x) >> (nr*8)))
108  */
109 static inline uint8_t
110 byte(const uint32_t x, const unsigned n)
111 {
112 	return x >> (n << 3);
113 }
114 
115 #define E_KEY ctx->E
116 #define D_KEY ctx->D
117 
118 static uint8_t pow_tab[256];
119 static uint8_t log_tab[256];
120 static uint8_t sbx_tab[256];
121 static uint8_t isb_tab[256];
122 static uint32_t rco_tab[10];
123 static uint32_t ft_tab[4][256];
124 static uint32_t it_tab[4][256];
125 
126 static uint32_t fl_tab[4][256];
127 static uint32_t il_tab[4][256];
128 
129 static inline uint8_t
130 f_mult (uint8_t a, uint8_t b)
131 {
132 	uint8_t aa = log_tab[a], cc = aa + log_tab[b];
133 
134 	return pow_tab[cc + (cc < aa ? 1 : 0)];
135 }
136 
137 #define ff_mult(a,b)    (a && b ? f_mult(a, b) : 0)
138 
139 #define f_rn(bo, bi, n, k)					\
140     bo[n] =  ft_tab[0][byte(bi[n],0)] ^				\
141              ft_tab[1][byte(bi[(n + 1) & 3],1)] ^		\
142              ft_tab[2][byte(bi[(n + 2) & 3],2)] ^		\
143              ft_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
144 
145 #define i_rn(bo, bi, n, k)					\
146     bo[n] =  it_tab[0][byte(bi[n],0)] ^				\
147              it_tab[1][byte(bi[(n + 3) & 3],1)] ^		\
148              it_tab[2][byte(bi[(n + 2) & 3],2)] ^		\
149              it_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
150 
151 #define ls_box(x)				\
152     ( fl_tab[0][byte(x, 0)] ^			\
153       fl_tab[1][byte(x, 1)] ^			\
154       fl_tab[2][byte(x, 2)] ^			\
155       fl_tab[3][byte(x, 3)] )
156 
157 #define f_rl(bo, bi, n, k)					\
158     bo[n] =  fl_tab[0][byte(bi[n],0)] ^				\
159              fl_tab[1][byte(bi[(n + 1) & 3],1)] ^		\
160              fl_tab[2][byte(bi[(n + 2) & 3],2)] ^		\
161              fl_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
162 
163 #define i_rl(bo, bi, n, k)					\
164     bo[n] =  il_tab[0][byte(bi[n],0)] ^				\
165              il_tab[1][byte(bi[(n + 3) & 3],1)] ^		\
166              il_tab[2][byte(bi[(n + 2) & 3],2)] ^		\
167              il_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
168 
169 static void
170 gen_tabs (void)
171 {
172 	uint32_t i, t;
173 	uint8_t p, q;
174 
175 	/* log and power tables for GF(2**8) finite field with
176 	   0x011b as modular polynomial - the simplest prmitive
177 	   root is 0x03, used here to generate the tables */
178 
179 	for (i = 0, p = 1; i < 256; ++i) {
180 		pow_tab[i] = (uint8_t) p;
181 		log_tab[p] = (uint8_t) i;
182 
183 		p ^= (p << 1) ^ (p & 0x80 ? 0x01b : 0);
184 	}
185 
186 	log_tab[1] = 0;
187 
188 	for (i = 0, p = 1; i < 10; ++i) {
189 		rco_tab[i] = p;
190 
191 		p = (p << 1) ^ (p & 0x80 ? 0x01b : 0);
192 	}
193 
194 	for (i = 0; i < 256; ++i) {
195 		p = (i ? pow_tab[255 - log_tab[i]] : 0);
196 		q = ((p >> 7) | (p << 1)) ^ ((p >> 6) | (p << 2));
197 		p ^= 0x63 ^ q ^ ((q >> 6) | (q << 2));
198 		sbx_tab[i] = p;
199 		isb_tab[p] = (uint8_t) i;
200 	}
201 
202 	for (i = 0; i < 256; ++i) {
203 		p = sbx_tab[i];
204 
205 		t = p;
206 		fl_tab[0][i] = t;
207 		fl_tab[1][i] = rotl (t, 8);
208 		fl_tab[2][i] = rotl (t, 16);
209 		fl_tab[3][i] = rotl (t, 24);
210 
211 		t = ((uint32_t) ff_mult (2, p)) |
212 		    ((uint32_t) p << 8) |
213 		    ((uint32_t) p << 16) | ((uint32_t) ff_mult (3, p) << 24);
214 
215 		ft_tab[0][i] = t;
216 		ft_tab[1][i] = rotl (t, 8);
217 		ft_tab[2][i] = rotl (t, 16);
218 		ft_tab[3][i] = rotl (t, 24);
219 
220 		p = isb_tab[i];
221 
222 		t = p;
223 		il_tab[0][i] = t;
224 		il_tab[1][i] = rotl (t, 8);
225 		il_tab[2][i] = rotl (t, 16);
226 		il_tab[3][i] = rotl (t, 24);
227 
228 		t = ((uint32_t) ff_mult (14, p)) |
229 		    ((uint32_t) ff_mult (9, p) << 8) |
230 		    ((uint32_t) ff_mult (13, p) << 16) |
231 		    ((uint32_t) ff_mult (11, p) << 24);
232 
233 		it_tab[0][i] = t;
234 		it_tab[1][i] = rotl (t, 8);
235 		it_tab[2][i] = rotl (t, 16);
236 		it_tab[3][i] = rotl (t, 24);
237 	}
238 }
239 
240 #define star_x(x) (((x) & 0x7f7f7f7f) << 1) ^ ((((x) & 0x80808080) >> 7) * 0x1b)
241 
242 #define imix_col(y,x)       \
243     u   = star_x(x);        \
244     v   = star_x(u);        \
245     w   = star_x(v);        \
246     t   = w ^ (x);          \
247    (y)  = u ^ v ^ w;        \
248    (y) ^= rotr(u ^ t,  8) ^ \
249           rotr(v ^ t, 16) ^ \
250           rotr(t,24)
251 
252 /* initialise the key schedule from the user supplied key */
253 
254 #define loop4(i)                                    \
255 {   t = rotr(t,  8); t = ls_box(t) ^ rco_tab[i];    \
256     t ^= E_KEY[4 * i];     E_KEY[4 * i + 4] = t;    \
257     t ^= E_KEY[4 * i + 1]; E_KEY[4 * i + 5] = t;    \
258     t ^= E_KEY[4 * i + 2]; E_KEY[4 * i + 6] = t;    \
259     t ^= E_KEY[4 * i + 3]; E_KEY[4 * i + 7] = t;    \
260 }
261 
262 #define loop6(i)                                    \
263 {   t = rotr(t,  8); t = ls_box(t) ^ rco_tab[i];    \
264     t ^= E_KEY[6 * i];     E_KEY[6 * i + 6] = t;    \
265     t ^= E_KEY[6 * i + 1]; E_KEY[6 * i + 7] = t;    \
266     t ^= E_KEY[6 * i + 2]; E_KEY[6 * i + 8] = t;    \
267     t ^= E_KEY[6 * i + 3]; E_KEY[6 * i + 9] = t;    \
268     t ^= E_KEY[6 * i + 4]; E_KEY[6 * i + 10] = t;   \
269     t ^= E_KEY[6 * i + 5]; E_KEY[6 * i + 11] = t;   \
270 }
271 
272 #define loop8(i)                                    \
273 {   t = rotr(t,  8); ; t = ls_box(t) ^ rco_tab[i];  \
274     t ^= E_KEY[8 * i];     E_KEY[8 * i + 8] = t;    \
275     t ^= E_KEY[8 * i + 1]; E_KEY[8 * i + 9] = t;    \
276     t ^= E_KEY[8 * i + 2]; E_KEY[8 * i + 10] = t;   \
277     t ^= E_KEY[8 * i + 3]; E_KEY[8 * i + 11] = t;   \
278     t  = E_KEY[8 * i + 4] ^ ls_box(t);    \
279     E_KEY[8 * i + 12] = t;                \
280     t ^= E_KEY[8 * i + 5]; E_KEY[8 * i + 13] = t;   \
281     t ^= E_KEY[8 * i + 6]; E_KEY[8 * i + 14] = t;   \
282     t ^= E_KEY[8 * i + 7]; E_KEY[8 * i + 15] = t;   \
283 }
284 
285 /* Tells whether the ACE is capable to generate
286    the extended key for a given key_len. */
287 static inline int
288 aes_hw_extkey_available(uint8_t key_len)
289 {
290 	/* TODO: We should check the actual CPU model/stepping
291 	         as it's possible that the capability will be
292 	         added in the next CPU revisions. */
293 	if (key_len == 16)
294 		return 1;
295 	return 0;
296 }
297 
298 static inline struct aes_ctx *aes_ctx_common(void *ctx)
299 {
300 	unsigned long addr = (unsigned long)ctx;
301 	unsigned long align = PADLOCK_ALIGNMENT;
302 
303 	if (align <= crypto_tfm_ctx_alignment())
304 		align = 1;
305 	return (struct aes_ctx *)ALIGN(addr, align);
306 }
307 
308 static inline struct aes_ctx *aes_ctx(struct crypto_tfm *tfm)
309 {
310 	return aes_ctx_common(crypto_tfm_ctx(tfm));
311 }
312 
313 static inline struct aes_ctx *blk_aes_ctx(struct crypto_blkcipher *tfm)
314 {
315 	return aes_ctx_common(crypto_blkcipher_ctx(tfm));
316 }
317 
318 static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key,
319 		       unsigned int key_len)
320 {
321 	struct aes_ctx *ctx = aes_ctx(tfm);
322 	const __le32 *key = (const __le32 *)in_key;
323 	u32 *flags = &tfm->crt_flags;
324 	uint32_t i, t, u, v, w;
325 	uint32_t P[AES_EXTENDED_KEY_SIZE];
326 	uint32_t rounds;
327 
328 	if (key_len % 8) {
329 		*flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
330 		return -EINVAL;
331 	}
332 
333 	ctx->key_length = key_len;
334 
335 	/*
336 	 * If the hardware is capable of generating the extended key
337 	 * itself we must supply the plain key for both encryption
338 	 * and decryption.
339 	 */
340 	ctx->D = ctx->E;
341 
342 	E_KEY[0] = le32_to_cpu(key[0]);
343 	E_KEY[1] = le32_to_cpu(key[1]);
344 	E_KEY[2] = le32_to_cpu(key[2]);
345 	E_KEY[3] = le32_to_cpu(key[3]);
346 
347 	/* Prepare control words. */
348 	memset(&ctx->cword, 0, sizeof(ctx->cword));
349 
350 	ctx->cword.decrypt.encdec = 1;
351 	ctx->cword.encrypt.rounds = 10 + (key_len - 16) / 4;
352 	ctx->cword.decrypt.rounds = ctx->cword.encrypt.rounds;
353 	ctx->cword.encrypt.ksize = (key_len - 16) / 8;
354 	ctx->cword.decrypt.ksize = ctx->cword.encrypt.ksize;
355 
356 	/* Don't generate extended keys if the hardware can do it. */
357 	if (aes_hw_extkey_available(key_len))
358 		return 0;
359 
360 	ctx->D = ctx->d_data;
361 	ctx->cword.encrypt.keygen = 1;
362 	ctx->cword.decrypt.keygen = 1;
363 
364 	switch (key_len) {
365 	case 16:
366 		t = E_KEY[3];
367 		for (i = 0; i < 10; ++i)
368 			loop4 (i);
369 		break;
370 
371 	case 24:
372 		E_KEY[4] = le32_to_cpu(key[4]);
373 		t = E_KEY[5] = le32_to_cpu(key[5]);
374 		for (i = 0; i < 8; ++i)
375 			loop6 (i);
376 		break;
377 
378 	case 32:
379 		E_KEY[4] = le32_to_cpu(key[4]);
380 		E_KEY[5] = le32_to_cpu(key[5]);
381 		E_KEY[6] = le32_to_cpu(key[6]);
382 		t = E_KEY[7] = le32_to_cpu(key[7]);
383 		for (i = 0; i < 7; ++i)
384 			loop8 (i);
385 		break;
386 	}
387 
388 	D_KEY[0] = E_KEY[0];
389 	D_KEY[1] = E_KEY[1];
390 	D_KEY[2] = E_KEY[2];
391 	D_KEY[3] = E_KEY[3];
392 
393 	for (i = 4; i < key_len + 24; ++i) {
394 		imix_col (D_KEY[i], E_KEY[i]);
395 	}
396 
397 	/* PadLock needs a different format of the decryption key. */
398 	rounds = 10 + (key_len - 16) / 4;
399 
400 	for (i = 0; i < rounds; i++) {
401 		P[((i + 1) * 4) + 0] = D_KEY[((rounds - i - 1) * 4) + 0];
402 		P[((i + 1) * 4) + 1] = D_KEY[((rounds - i - 1) * 4) + 1];
403 		P[((i + 1) * 4) + 2] = D_KEY[((rounds - i - 1) * 4) + 2];
404 		P[((i + 1) * 4) + 3] = D_KEY[((rounds - i - 1) * 4) + 3];
405 	}
406 
407 	P[0] = E_KEY[(rounds * 4) + 0];
408 	P[1] = E_KEY[(rounds * 4) + 1];
409 	P[2] = E_KEY[(rounds * 4) + 2];
410 	P[3] = E_KEY[(rounds * 4) + 3];
411 
412 	memcpy(D_KEY, P, AES_EXTENDED_KEY_SIZE_B);
413 
414 	return 0;
415 }
416 
417 /* ====== Encryption/decryption routines ====== */
418 
419 /* These are the real call to PadLock. */
420 static inline void padlock_reset_key(void)
421 {
422 	asm volatile ("pushfl; popfl");
423 }
424 
425 static inline void padlock_xcrypt(const u8 *input, u8 *output, void *key,
426 				  void *control_word)
427 {
428 	asm volatile (".byte 0xf3,0x0f,0xa7,0xc8"	/* rep xcryptecb */
429 		      : "+S"(input), "+D"(output)
430 		      : "d"(control_word), "b"(key), "c"(1));
431 }
432 
433 static void aes_crypt_copy(const u8 *in, u8 *out, u32 *key, struct cword *cword)
434 {
435 	u8 buf[AES_BLOCK_SIZE * 2 + PADLOCK_ALIGNMENT - 1];
436 	u8 *tmp = PTR_ALIGN(&buf[0], PADLOCK_ALIGNMENT);
437 
438 	memcpy(tmp, in, AES_BLOCK_SIZE);
439 	padlock_xcrypt(tmp, out, key, cword);
440 }
441 
442 static inline void aes_crypt(const u8 *in, u8 *out, u32 *key,
443 			     struct cword *cword)
444 {
445 	/* padlock_xcrypt requires at least two blocks of data. */
446 	if (unlikely(!(((unsigned long)in ^ (PAGE_SIZE - AES_BLOCK_SIZE)) &
447 		       (PAGE_SIZE - 1)))) {
448 		aes_crypt_copy(in, out, key, cword);
449 		return;
450 	}
451 
452 	padlock_xcrypt(in, out, key, cword);
453 }
454 
455 static inline void padlock_xcrypt_ecb(const u8 *input, u8 *output, void *key,
456 				      void *control_word, u32 count)
457 {
458 	if (count == 1) {
459 		aes_crypt(input, output, key, control_word);
460 		return;
461 	}
462 
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 */
471 		      : "+S"(input), "+D"(output)
472 		      : "d"(control_word), "b"(key), "c"(count)
473 		      : "ax");
474 }
475 
476 static inline u8 *padlock_xcrypt_cbc(const u8 *input, u8 *output, void *key,
477 				     u8 *iv, void *control_word, u32 count)
478 {
479 	/* rep xcryptcbc */
480 	asm volatile (".byte 0xf3,0x0f,0xa7,0xd0"
481 		      : "+S" (input), "+D" (output), "+a" (iv)
482 		      : "d" (control_word), "b" (key), "c" (count));
483 	return iv;
484 }
485 
486 static void aes_encrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in)
487 {
488 	struct aes_ctx *ctx = aes_ctx(tfm);
489 	padlock_reset_key();
490 	aes_crypt(in, out, ctx->E, &ctx->cword.encrypt);
491 }
492 
493 static void aes_decrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in)
494 {
495 	struct aes_ctx *ctx = aes_ctx(tfm);
496 	padlock_reset_key();
497 	aes_crypt(in, out, ctx->D, &ctx->cword.decrypt);
498 }
499 
500 static struct crypto_alg aes_alg = {
501 	.cra_name		=	"aes",
502 	.cra_driver_name	=	"aes-padlock",
503 	.cra_priority		=	PADLOCK_CRA_PRIORITY,
504 	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
505 	.cra_blocksize		=	AES_BLOCK_SIZE,
506 	.cra_ctxsize		=	sizeof(struct aes_ctx),
507 	.cra_alignmask		=	PADLOCK_ALIGNMENT - 1,
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,
516 			.cia_decrypt	  	=	aes_decrypt,
517 		}
518 	}
519 };
520 
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 	padlock_reset_key();
530 
531 	blkcipher_walk_init(&walk, dst, src, nbytes);
532 	err = blkcipher_walk_virt(desc, &walk);
533 
534 	while ((nbytes = walk.nbytes)) {
535 		padlock_xcrypt_ecb(walk.src.virt.addr, walk.dst.virt.addr,
536 				   ctx->E, &ctx->cword.encrypt,
537 				   nbytes / AES_BLOCK_SIZE);
538 		nbytes &= AES_BLOCK_SIZE - 1;
539 		err = blkcipher_walk_done(desc, &walk, nbytes);
540 	}
541 
542 	return err;
543 }
544 
545 static int ecb_aes_decrypt(struct blkcipher_desc *desc,
546 			   struct scatterlist *dst, struct scatterlist *src,
547 			   unsigned int nbytes)
548 {
549 	struct aes_ctx *ctx = blk_aes_ctx(desc->tfm);
550 	struct blkcipher_walk walk;
551 	int err;
552 
553 	padlock_reset_key();
554 
555 	blkcipher_walk_init(&walk, dst, src, nbytes);
556 	err = blkcipher_walk_virt(desc, &walk);
557 
558 	while ((nbytes = walk.nbytes)) {
559 		padlock_xcrypt_ecb(walk.src.virt.addr, walk.dst.virt.addr,
560 				   ctx->D, &ctx->cword.decrypt,
561 				   nbytes / AES_BLOCK_SIZE);
562 		nbytes &= AES_BLOCK_SIZE - 1;
563 		err = blkcipher_walk_done(desc, &walk, nbytes);
564 	}
565 
566 	return err;
567 }
568 
569 static struct crypto_alg ecb_aes_alg = {
570 	.cra_name		=	"ecb(aes)",
571 	.cra_driver_name	=	"ecb-aes-padlock",
572 	.cra_priority		=	PADLOCK_COMPOSITE_PRIORITY,
573 	.cra_flags		=	CRYPTO_ALG_TYPE_BLKCIPHER,
574 	.cra_blocksize		=	AES_BLOCK_SIZE,
575 	.cra_ctxsize		=	sizeof(struct aes_ctx),
576 	.cra_alignmask		=	PADLOCK_ALIGNMENT - 1,
577 	.cra_type		=	&crypto_blkcipher_type,
578 	.cra_module		=	THIS_MODULE,
579 	.cra_list		=	LIST_HEAD_INIT(ecb_aes_alg.cra_list),
580 	.cra_u			=	{
581 		.blkcipher = {
582 			.min_keysize		=	AES_MIN_KEY_SIZE,
583 			.max_keysize		=	AES_MAX_KEY_SIZE,
584 			.setkey	   		= 	aes_set_key,
585 			.encrypt		=	ecb_aes_encrypt,
586 			.decrypt		=	ecb_aes_decrypt,
587 		}
588 	}
589 };
590 
591 static int cbc_aes_encrypt(struct blkcipher_desc *desc,
592 			   struct scatterlist *dst, struct scatterlist *src,
593 			   unsigned int nbytes)
594 {
595 	struct aes_ctx *ctx = blk_aes_ctx(desc->tfm);
596 	struct blkcipher_walk walk;
597 	int err;
598 
599 	padlock_reset_key();
600 
601 	blkcipher_walk_init(&walk, dst, src, nbytes);
602 	err = blkcipher_walk_virt(desc, &walk);
603 
604 	while ((nbytes = walk.nbytes)) {
605 		u8 *iv = padlock_xcrypt_cbc(walk.src.virt.addr,
606 					    walk.dst.virt.addr, ctx->E,
607 					    walk.iv, &ctx->cword.encrypt,
608 					    nbytes / AES_BLOCK_SIZE);
609 		memcpy(walk.iv, iv, AES_BLOCK_SIZE);
610 		nbytes &= AES_BLOCK_SIZE - 1;
611 		err = blkcipher_walk_done(desc, &walk, nbytes);
612 	}
613 
614 	return err;
615 }
616 
617 static int cbc_aes_decrypt(struct blkcipher_desc *desc,
618 			   struct scatterlist *dst, struct scatterlist *src,
619 			   unsigned int nbytes)
620 {
621 	struct aes_ctx *ctx = blk_aes_ctx(desc->tfm);
622 	struct blkcipher_walk walk;
623 	int err;
624 
625 	padlock_reset_key();
626 
627 	blkcipher_walk_init(&walk, dst, src, nbytes);
628 	err = blkcipher_walk_virt(desc, &walk);
629 
630 	while ((nbytes = walk.nbytes)) {
631 		padlock_xcrypt_cbc(walk.src.virt.addr, walk.dst.virt.addr,
632 				   ctx->D, walk.iv, &ctx->cword.decrypt,
633 				   nbytes / AES_BLOCK_SIZE);
634 		nbytes &= AES_BLOCK_SIZE - 1;
635 		err = blkcipher_walk_done(desc, &walk, nbytes);
636 	}
637 
638 	return err;
639 }
640 
641 static struct crypto_alg cbc_aes_alg = {
642 	.cra_name		=	"cbc(aes)",
643 	.cra_driver_name	=	"cbc-aes-padlock",
644 	.cra_priority		=	PADLOCK_COMPOSITE_PRIORITY,
645 	.cra_flags		=	CRYPTO_ALG_TYPE_BLKCIPHER,
646 	.cra_blocksize		=	AES_BLOCK_SIZE,
647 	.cra_ctxsize		=	sizeof(struct aes_ctx),
648 	.cra_alignmask		=	PADLOCK_ALIGNMENT - 1,
649 	.cra_type		=	&crypto_blkcipher_type,
650 	.cra_module		=	THIS_MODULE,
651 	.cra_list		=	LIST_HEAD_INIT(cbc_aes_alg.cra_list),
652 	.cra_u			=	{
653 		.blkcipher = {
654 			.min_keysize		=	AES_MIN_KEY_SIZE,
655 			.max_keysize		=	AES_MAX_KEY_SIZE,
656 			.ivsize			=	AES_BLOCK_SIZE,
657 			.setkey	   		= 	aes_set_key,
658 			.encrypt		=	cbc_aes_encrypt,
659 			.decrypt		=	cbc_aes_decrypt,
660 		}
661 	}
662 };
663 
664 static int __init padlock_init(void)
665 {
666 	int ret;
667 
668 	if (!cpu_has_xcrypt) {
669 		printk(KERN_ERR PFX "VIA PadLock not detected.\n");
670 		return -ENODEV;
671 	}
672 
673 	if (!cpu_has_xcrypt_enabled) {
674 		printk(KERN_ERR PFX "VIA PadLock detected, but not enabled. Hmm, strange...\n");
675 		return -ENODEV;
676 	}
677 
678 	gen_tabs();
679 	if ((ret = crypto_register_alg(&aes_alg)))
680 		goto aes_err;
681 
682 	if ((ret = crypto_register_alg(&ecb_aes_alg)))
683 		goto ecb_aes_err;
684 
685 	if ((ret = crypto_register_alg(&cbc_aes_alg)))
686 		goto cbc_aes_err;
687 
688 	printk(KERN_NOTICE PFX "Using VIA PadLock ACE for AES algorithm.\n");
689 
690 out:
691 	return ret;
692 
693 cbc_aes_err:
694 	crypto_unregister_alg(&ecb_aes_alg);
695 ecb_aes_err:
696 	crypto_unregister_alg(&aes_alg);
697 aes_err:
698 	printk(KERN_ERR PFX "VIA PadLock AES initialization failed.\n");
699 	goto out;
700 }
701 
702 static void __exit padlock_fini(void)
703 {
704 	crypto_unregister_alg(&cbc_aes_alg);
705 	crypto_unregister_alg(&ecb_aes_alg);
706 	crypto_unregister_alg(&aes_alg);
707 }
708 
709 module_init(padlock_init);
710 module_exit(padlock_fini);
711 
712 MODULE_DESCRIPTION("VIA PadLock AES algorithm support");
713 MODULE_LICENSE("GPL");
714 MODULE_AUTHOR("Michal Ludvig");
715 
716 MODULE_ALIAS("aes");
717