xref: /linux/arch/arm64/crypto/sm4-ce-glue.c (revision 2697b79a469b68e3ad3640f55284359c1396278d)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * SM4 Cipher Algorithm, using ARMv8 Crypto Extensions
4  * as specified in
5  * https://tools.ietf.org/id/draft-ribose-cfrg-sm4-10.html
6  *
7  * Copyright (C) 2022, Alibaba Group.
8  * Copyright (C) 2022 Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
9  */
10 
11 #include <linux/module.h>
12 #include <linux/crypto.h>
13 #include <linux/kernel.h>
14 #include <linux/cpufeature.h>
15 #include <asm/neon.h>
16 #include <asm/simd.h>
17 #include <crypto/b128ops.h>
18 #include <crypto/internal/simd.h>
19 #include <crypto/internal/skcipher.h>
20 #include <crypto/internal/hash.h>
21 #include <crypto/scatterwalk.h>
22 #include <crypto/xts.h>
23 #include <crypto/sm4.h>
24 
25 #define BYTES2BLKS(nbytes)	((nbytes) >> 4)
26 
27 asmlinkage void sm4_ce_expand_key(const u8 *key, u32 *rkey_enc, u32 *rkey_dec,
28 				  const u32 *fk, const u32 *ck);
29 asmlinkage void sm4_ce_crypt_block(const u32 *rkey, u8 *dst, const u8 *src);
30 asmlinkage void sm4_ce_crypt(const u32 *rkey, u8 *dst, const u8 *src,
31 			     unsigned int nblks);
32 asmlinkage void sm4_ce_cbc_enc(const u32 *rkey, u8 *dst, const u8 *src,
33 			       u8 *iv, unsigned int nblocks);
34 asmlinkage void sm4_ce_cbc_dec(const u32 *rkey, u8 *dst, const u8 *src,
35 			       u8 *iv, unsigned int nblocks);
36 asmlinkage void sm4_ce_cbc_cts_enc(const u32 *rkey, u8 *dst, const u8 *src,
37 				   u8 *iv, unsigned int nbytes);
38 asmlinkage void sm4_ce_cbc_cts_dec(const u32 *rkey, u8 *dst, const u8 *src,
39 				   u8 *iv, unsigned int nbytes);
40 asmlinkage void sm4_ce_ctr_enc(const u32 *rkey, u8 *dst, const u8 *src,
41 			       u8 *iv, unsigned int nblks);
42 asmlinkage void sm4_ce_xts_enc(const u32 *rkey1, u8 *dst, const u8 *src,
43 			       u8 *tweak, unsigned int nbytes,
44 			       const u32 *rkey2_enc);
45 asmlinkage void sm4_ce_xts_dec(const u32 *rkey1, u8 *dst, const u8 *src,
46 			       u8 *tweak, unsigned int nbytes,
47 			       const u32 *rkey2_enc);
48 asmlinkage void sm4_ce_mac_update(const u32 *rkey_enc, u8 *digest,
49 				  const u8 *src, unsigned int nblocks,
50 				  bool enc_before, bool enc_after);
51 
52 EXPORT_SYMBOL(sm4_ce_expand_key);
53 EXPORT_SYMBOL(sm4_ce_crypt_block);
54 EXPORT_SYMBOL(sm4_ce_cbc_enc);
55 
56 struct sm4_xts_ctx {
57 	struct sm4_ctx key1;
58 	struct sm4_ctx key2;
59 };
60 
61 struct sm4_mac_tfm_ctx {
62 	struct sm4_ctx key;
63 	u8 __aligned(8) consts[];
64 };
65 
66 struct sm4_mac_desc_ctx {
67 	unsigned int len;
68 	u8 digest[SM4_BLOCK_SIZE];
69 };
70 
71 static int sm4_setkey(struct crypto_skcipher *tfm, const u8 *key,
72 		      unsigned int key_len)
73 {
74 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
75 
76 	if (key_len != SM4_KEY_SIZE)
77 		return -EINVAL;
78 
79 	kernel_neon_begin();
80 	sm4_ce_expand_key(key, ctx->rkey_enc, ctx->rkey_dec,
81 			  crypto_sm4_fk, crypto_sm4_ck);
82 	kernel_neon_end();
83 	return 0;
84 }
85 
86 static int sm4_xts_setkey(struct crypto_skcipher *tfm, const u8 *key,
87 			  unsigned int key_len)
88 {
89 	struct sm4_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
90 	int ret;
91 
92 	if (key_len != SM4_KEY_SIZE * 2)
93 		return -EINVAL;
94 
95 	ret = xts_verify_key(tfm, key, key_len);
96 	if (ret)
97 		return ret;
98 
99 	kernel_neon_begin();
100 	sm4_ce_expand_key(key, ctx->key1.rkey_enc,
101 			  ctx->key1.rkey_dec, crypto_sm4_fk, crypto_sm4_ck);
102 	sm4_ce_expand_key(&key[SM4_KEY_SIZE], ctx->key2.rkey_enc,
103 			  ctx->key2.rkey_dec, crypto_sm4_fk, crypto_sm4_ck);
104 	kernel_neon_end();
105 
106 	return 0;
107 }
108 
109 static int sm4_ecb_do_crypt(struct skcipher_request *req, const u32 *rkey)
110 {
111 	struct skcipher_walk walk;
112 	unsigned int nbytes;
113 	int err;
114 
115 	err = skcipher_walk_virt(&walk, req, false);
116 
117 	while ((nbytes = walk.nbytes) > 0) {
118 		const u8 *src = walk.src.virt.addr;
119 		u8 *dst = walk.dst.virt.addr;
120 		unsigned int nblks;
121 
122 		kernel_neon_begin();
123 
124 		nblks = BYTES2BLKS(nbytes);
125 		if (nblks) {
126 			sm4_ce_crypt(rkey, dst, src, nblks);
127 			nbytes -= nblks * SM4_BLOCK_SIZE;
128 		}
129 
130 		kernel_neon_end();
131 
132 		err = skcipher_walk_done(&walk, nbytes);
133 	}
134 
135 	return err;
136 }
137 
138 static int sm4_ecb_encrypt(struct skcipher_request *req)
139 {
140 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
141 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
142 
143 	return sm4_ecb_do_crypt(req, ctx->rkey_enc);
144 }
145 
146 static int sm4_ecb_decrypt(struct skcipher_request *req)
147 {
148 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
149 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
150 
151 	return sm4_ecb_do_crypt(req, ctx->rkey_dec);
152 }
153 
154 static int sm4_cbc_crypt(struct skcipher_request *req,
155 			 struct sm4_ctx *ctx, bool encrypt)
156 {
157 	struct skcipher_walk walk;
158 	unsigned int nbytes;
159 	int err;
160 
161 	err = skcipher_walk_virt(&walk, req, false);
162 	if (err)
163 		return err;
164 
165 	while ((nbytes = walk.nbytes) > 0) {
166 		const u8 *src = walk.src.virt.addr;
167 		u8 *dst = walk.dst.virt.addr;
168 		unsigned int nblocks;
169 
170 		nblocks = nbytes / SM4_BLOCK_SIZE;
171 		if (nblocks) {
172 			kernel_neon_begin();
173 
174 			if (encrypt)
175 				sm4_ce_cbc_enc(ctx->rkey_enc, dst, src,
176 					       walk.iv, nblocks);
177 			else
178 				sm4_ce_cbc_dec(ctx->rkey_dec, dst, src,
179 					       walk.iv, nblocks);
180 
181 			kernel_neon_end();
182 		}
183 
184 		err = skcipher_walk_done(&walk, nbytes % SM4_BLOCK_SIZE);
185 	}
186 
187 	return err;
188 }
189 
190 static int sm4_cbc_encrypt(struct skcipher_request *req)
191 {
192 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
193 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
194 
195 	return sm4_cbc_crypt(req, ctx, true);
196 }
197 
198 static int sm4_cbc_decrypt(struct skcipher_request *req)
199 {
200 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
201 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
202 
203 	return sm4_cbc_crypt(req, ctx, false);
204 }
205 
206 static int sm4_cbc_cts_crypt(struct skcipher_request *req, bool encrypt)
207 {
208 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
209 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
210 	struct scatterlist *src = req->src;
211 	struct scatterlist *dst = req->dst;
212 	struct scatterlist sg_src[2], sg_dst[2];
213 	struct skcipher_request subreq;
214 	struct skcipher_walk walk;
215 	int cbc_blocks;
216 	int err;
217 
218 	if (req->cryptlen < SM4_BLOCK_SIZE)
219 		return -EINVAL;
220 
221 	if (req->cryptlen == SM4_BLOCK_SIZE)
222 		return sm4_cbc_crypt(req, ctx, encrypt);
223 
224 	skcipher_request_set_tfm(&subreq, tfm);
225 	skcipher_request_set_callback(&subreq, skcipher_request_flags(req),
226 				      NULL, NULL);
227 
228 	/* handle the CBC cryption part */
229 	cbc_blocks = DIV_ROUND_UP(req->cryptlen, SM4_BLOCK_SIZE) - 2;
230 	if (cbc_blocks) {
231 		skcipher_request_set_crypt(&subreq, src, dst,
232 					   cbc_blocks * SM4_BLOCK_SIZE,
233 					   req->iv);
234 
235 		err = sm4_cbc_crypt(&subreq, ctx, encrypt);
236 		if (err)
237 			return err;
238 
239 		dst = src = scatterwalk_ffwd(sg_src, src, subreq.cryptlen);
240 		if (req->dst != req->src)
241 			dst = scatterwalk_ffwd(sg_dst, req->dst,
242 					       subreq.cryptlen);
243 	}
244 
245 	/* handle ciphertext stealing */
246 	skcipher_request_set_crypt(&subreq, src, dst,
247 				   req->cryptlen - cbc_blocks * SM4_BLOCK_SIZE,
248 				   req->iv);
249 
250 	err = skcipher_walk_virt(&walk, &subreq, false);
251 	if (err)
252 		return err;
253 
254 	kernel_neon_begin();
255 
256 	if (encrypt)
257 		sm4_ce_cbc_cts_enc(ctx->rkey_enc, walk.dst.virt.addr,
258 				   walk.src.virt.addr, walk.iv, walk.nbytes);
259 	else
260 		sm4_ce_cbc_cts_dec(ctx->rkey_dec, walk.dst.virt.addr,
261 				   walk.src.virt.addr, walk.iv, walk.nbytes);
262 
263 	kernel_neon_end();
264 
265 	return skcipher_walk_done(&walk, 0);
266 }
267 
268 static int sm4_cbc_cts_encrypt(struct skcipher_request *req)
269 {
270 	return sm4_cbc_cts_crypt(req, true);
271 }
272 
273 static int sm4_cbc_cts_decrypt(struct skcipher_request *req)
274 {
275 	return sm4_cbc_cts_crypt(req, false);
276 }
277 
278 static int sm4_ctr_crypt(struct skcipher_request *req)
279 {
280 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
281 	struct sm4_ctx *ctx = crypto_skcipher_ctx(tfm);
282 	struct skcipher_walk walk;
283 	unsigned int nbytes;
284 	int err;
285 
286 	err = skcipher_walk_virt(&walk, req, false);
287 
288 	while ((nbytes = walk.nbytes) > 0) {
289 		const u8 *src = walk.src.virt.addr;
290 		u8 *dst = walk.dst.virt.addr;
291 		unsigned int nblks;
292 
293 		kernel_neon_begin();
294 
295 		nblks = BYTES2BLKS(nbytes);
296 		if (nblks) {
297 			sm4_ce_ctr_enc(ctx->rkey_enc, dst, src, walk.iv, nblks);
298 			dst += nblks * SM4_BLOCK_SIZE;
299 			src += nblks * SM4_BLOCK_SIZE;
300 			nbytes -= nblks * SM4_BLOCK_SIZE;
301 		}
302 
303 		/* tail */
304 		if (walk.nbytes == walk.total && nbytes > 0) {
305 			u8 keystream[SM4_BLOCK_SIZE];
306 
307 			sm4_ce_crypt_block(ctx->rkey_enc, keystream, walk.iv);
308 			crypto_inc(walk.iv, SM4_BLOCK_SIZE);
309 			crypto_xor_cpy(dst, src, keystream, nbytes);
310 			nbytes = 0;
311 		}
312 
313 		kernel_neon_end();
314 
315 		err = skcipher_walk_done(&walk, nbytes);
316 	}
317 
318 	return err;
319 }
320 
321 static int sm4_xts_crypt(struct skcipher_request *req, bool encrypt)
322 {
323 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
324 	struct sm4_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
325 	int tail = req->cryptlen % SM4_BLOCK_SIZE;
326 	const u32 *rkey2_enc = ctx->key2.rkey_enc;
327 	struct scatterlist sg_src[2], sg_dst[2];
328 	struct skcipher_request subreq;
329 	struct scatterlist *src, *dst;
330 	struct skcipher_walk walk;
331 	unsigned int nbytes;
332 	int err;
333 
334 	if (req->cryptlen < SM4_BLOCK_SIZE)
335 		return -EINVAL;
336 
337 	err = skcipher_walk_virt(&walk, req, false);
338 	if (err)
339 		return err;
340 
341 	if (unlikely(tail > 0 && walk.nbytes < walk.total)) {
342 		int nblocks = DIV_ROUND_UP(req->cryptlen, SM4_BLOCK_SIZE) - 2;
343 
344 		skcipher_walk_abort(&walk);
345 
346 		skcipher_request_set_tfm(&subreq, tfm);
347 		skcipher_request_set_callback(&subreq,
348 					      skcipher_request_flags(req),
349 					      NULL, NULL);
350 		skcipher_request_set_crypt(&subreq, req->src, req->dst,
351 					   nblocks * SM4_BLOCK_SIZE, req->iv);
352 
353 		err = skcipher_walk_virt(&walk, &subreq, false);
354 		if (err)
355 			return err;
356 	} else {
357 		tail = 0;
358 	}
359 
360 	while ((nbytes = walk.nbytes) >= SM4_BLOCK_SIZE) {
361 		if (nbytes < walk.total)
362 			nbytes &= ~(SM4_BLOCK_SIZE - 1);
363 
364 		kernel_neon_begin();
365 
366 		if (encrypt)
367 			sm4_ce_xts_enc(ctx->key1.rkey_enc, walk.dst.virt.addr,
368 				       walk.src.virt.addr, walk.iv, nbytes,
369 				       rkey2_enc);
370 		else
371 			sm4_ce_xts_dec(ctx->key1.rkey_dec, walk.dst.virt.addr,
372 				       walk.src.virt.addr, walk.iv, nbytes,
373 				       rkey2_enc);
374 
375 		kernel_neon_end();
376 
377 		rkey2_enc = NULL;
378 
379 		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
380 		if (err)
381 			return err;
382 	}
383 
384 	if (likely(tail == 0))
385 		return 0;
386 
387 	/* handle ciphertext stealing */
388 
389 	dst = src = scatterwalk_ffwd(sg_src, req->src, subreq.cryptlen);
390 	if (req->dst != req->src)
391 		dst = scatterwalk_ffwd(sg_dst, req->dst, subreq.cryptlen);
392 
393 	skcipher_request_set_crypt(&subreq, src, dst, SM4_BLOCK_SIZE + tail,
394 				   req->iv);
395 
396 	err = skcipher_walk_virt(&walk, &subreq, false);
397 	if (err)
398 		return err;
399 
400 	kernel_neon_begin();
401 
402 	if (encrypt)
403 		sm4_ce_xts_enc(ctx->key1.rkey_enc, walk.dst.virt.addr,
404 			       walk.src.virt.addr, walk.iv, walk.nbytes,
405 			       rkey2_enc);
406 	else
407 		sm4_ce_xts_dec(ctx->key1.rkey_dec, walk.dst.virt.addr,
408 			       walk.src.virt.addr, walk.iv, walk.nbytes,
409 			       rkey2_enc);
410 
411 	kernel_neon_end();
412 
413 	return skcipher_walk_done(&walk, 0);
414 }
415 
416 static int sm4_xts_encrypt(struct skcipher_request *req)
417 {
418 	return sm4_xts_crypt(req, true);
419 }
420 
421 static int sm4_xts_decrypt(struct skcipher_request *req)
422 {
423 	return sm4_xts_crypt(req, false);
424 }
425 
426 static struct skcipher_alg sm4_algs[] = {
427 	{
428 		.base = {
429 			.cra_name		= "ecb(sm4)",
430 			.cra_driver_name	= "ecb-sm4-ce",
431 			.cra_priority		= 400,
432 			.cra_blocksize		= SM4_BLOCK_SIZE,
433 			.cra_ctxsize		= sizeof(struct sm4_ctx),
434 			.cra_module		= THIS_MODULE,
435 		},
436 		.min_keysize	= SM4_KEY_SIZE,
437 		.max_keysize	= SM4_KEY_SIZE,
438 		.setkey		= sm4_setkey,
439 		.encrypt	= sm4_ecb_encrypt,
440 		.decrypt	= sm4_ecb_decrypt,
441 	}, {
442 		.base = {
443 			.cra_name		= "cbc(sm4)",
444 			.cra_driver_name	= "cbc-sm4-ce",
445 			.cra_priority		= 400,
446 			.cra_blocksize		= SM4_BLOCK_SIZE,
447 			.cra_ctxsize		= sizeof(struct sm4_ctx),
448 			.cra_module		= THIS_MODULE,
449 		},
450 		.min_keysize	= SM4_KEY_SIZE,
451 		.max_keysize	= SM4_KEY_SIZE,
452 		.ivsize		= SM4_BLOCK_SIZE,
453 		.setkey		= sm4_setkey,
454 		.encrypt	= sm4_cbc_encrypt,
455 		.decrypt	= sm4_cbc_decrypt,
456 	}, {
457 		.base = {
458 			.cra_name		= "ctr(sm4)",
459 			.cra_driver_name	= "ctr-sm4-ce",
460 			.cra_priority		= 400,
461 			.cra_blocksize		= 1,
462 			.cra_ctxsize		= sizeof(struct sm4_ctx),
463 			.cra_module		= THIS_MODULE,
464 		},
465 		.min_keysize	= SM4_KEY_SIZE,
466 		.max_keysize	= SM4_KEY_SIZE,
467 		.ivsize		= SM4_BLOCK_SIZE,
468 		.chunksize	= SM4_BLOCK_SIZE,
469 		.setkey		= sm4_setkey,
470 		.encrypt	= sm4_ctr_crypt,
471 		.decrypt	= sm4_ctr_crypt,
472 	}, {
473 		.base = {
474 			.cra_name		= "cts(cbc(sm4))",
475 			.cra_driver_name	= "cts-cbc-sm4-ce",
476 			.cra_priority		= 400,
477 			.cra_blocksize		= SM4_BLOCK_SIZE,
478 			.cra_ctxsize		= sizeof(struct sm4_ctx),
479 			.cra_module		= THIS_MODULE,
480 		},
481 		.min_keysize	= SM4_KEY_SIZE,
482 		.max_keysize	= SM4_KEY_SIZE,
483 		.ivsize		= SM4_BLOCK_SIZE,
484 		.walksize	= SM4_BLOCK_SIZE * 2,
485 		.setkey		= sm4_setkey,
486 		.encrypt	= sm4_cbc_cts_encrypt,
487 		.decrypt	= sm4_cbc_cts_decrypt,
488 	}, {
489 		.base = {
490 			.cra_name		= "xts(sm4)",
491 			.cra_driver_name	= "xts-sm4-ce",
492 			.cra_priority		= 400,
493 			.cra_blocksize		= SM4_BLOCK_SIZE,
494 			.cra_ctxsize		= sizeof(struct sm4_xts_ctx),
495 			.cra_module		= THIS_MODULE,
496 		},
497 		.min_keysize	= SM4_KEY_SIZE * 2,
498 		.max_keysize	= SM4_KEY_SIZE * 2,
499 		.ivsize		= SM4_BLOCK_SIZE,
500 		.walksize	= SM4_BLOCK_SIZE * 2,
501 		.setkey		= sm4_xts_setkey,
502 		.encrypt	= sm4_xts_encrypt,
503 		.decrypt	= sm4_xts_decrypt,
504 	}
505 };
506 
507 static int sm4_cbcmac_setkey(struct crypto_shash *tfm, const u8 *key,
508 			     unsigned int key_len)
509 {
510 	struct sm4_mac_tfm_ctx *ctx = crypto_shash_ctx(tfm);
511 
512 	if (key_len != SM4_KEY_SIZE)
513 		return -EINVAL;
514 
515 	kernel_neon_begin();
516 	sm4_ce_expand_key(key, ctx->key.rkey_enc, ctx->key.rkey_dec,
517 			  crypto_sm4_fk, crypto_sm4_ck);
518 	kernel_neon_end();
519 
520 	return 0;
521 }
522 
523 static int sm4_cmac_setkey(struct crypto_shash *tfm, const u8 *key,
524 			   unsigned int key_len)
525 {
526 	struct sm4_mac_tfm_ctx *ctx = crypto_shash_ctx(tfm);
527 	be128 *consts = (be128 *)ctx->consts;
528 	u64 a, b;
529 
530 	if (key_len != SM4_KEY_SIZE)
531 		return -EINVAL;
532 
533 	memset(consts, 0, SM4_BLOCK_SIZE);
534 
535 	kernel_neon_begin();
536 
537 	sm4_ce_expand_key(key, ctx->key.rkey_enc, ctx->key.rkey_dec,
538 			  crypto_sm4_fk, crypto_sm4_ck);
539 
540 	/* encrypt the zero block */
541 	sm4_ce_crypt_block(ctx->key.rkey_enc, (u8 *)consts, (const u8 *)consts);
542 
543 	kernel_neon_end();
544 
545 	/* gf(2^128) multiply zero-ciphertext with u and u^2 */
546 	a = be64_to_cpu(consts[0].a);
547 	b = be64_to_cpu(consts[0].b);
548 	consts[0].a = cpu_to_be64((a << 1) | (b >> 63));
549 	consts[0].b = cpu_to_be64((b << 1) ^ ((a >> 63) ? 0x87 : 0));
550 
551 	a = be64_to_cpu(consts[0].a);
552 	b = be64_to_cpu(consts[0].b);
553 	consts[1].a = cpu_to_be64((a << 1) | (b >> 63));
554 	consts[1].b = cpu_to_be64((b << 1) ^ ((a >> 63) ? 0x87 : 0));
555 
556 	return 0;
557 }
558 
559 static int sm4_xcbc_setkey(struct crypto_shash *tfm, const u8 *key,
560 			   unsigned int key_len)
561 {
562 	struct sm4_mac_tfm_ctx *ctx = crypto_shash_ctx(tfm);
563 	u8 __aligned(8) key2[SM4_BLOCK_SIZE];
564 	static u8 const ks[3][SM4_BLOCK_SIZE] = {
565 		{ [0 ... SM4_BLOCK_SIZE - 1] = 0x1},
566 		{ [0 ... SM4_BLOCK_SIZE - 1] = 0x2},
567 		{ [0 ... SM4_BLOCK_SIZE - 1] = 0x3},
568 	};
569 
570 	if (key_len != SM4_KEY_SIZE)
571 		return -EINVAL;
572 
573 	kernel_neon_begin();
574 
575 	sm4_ce_expand_key(key, ctx->key.rkey_enc, ctx->key.rkey_dec,
576 			  crypto_sm4_fk, crypto_sm4_ck);
577 
578 	sm4_ce_crypt_block(ctx->key.rkey_enc, key2, ks[0]);
579 	sm4_ce_crypt(ctx->key.rkey_enc, ctx->consts, ks[1], 2);
580 
581 	sm4_ce_expand_key(key2, ctx->key.rkey_enc, ctx->key.rkey_dec,
582 			  crypto_sm4_fk, crypto_sm4_ck);
583 
584 	kernel_neon_end();
585 
586 	return 0;
587 }
588 
589 static int sm4_mac_init(struct shash_desc *desc)
590 {
591 	struct sm4_mac_desc_ctx *ctx = shash_desc_ctx(desc);
592 
593 	memset(ctx->digest, 0, SM4_BLOCK_SIZE);
594 	ctx->len = 0;
595 
596 	return 0;
597 }
598 
599 static int sm4_mac_update(struct shash_desc *desc, const u8 *p,
600 			  unsigned int len)
601 {
602 	struct sm4_mac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
603 	struct sm4_mac_desc_ctx *ctx = shash_desc_ctx(desc);
604 	unsigned int l, nblocks;
605 
606 	if (len == 0)
607 		return 0;
608 
609 	if (ctx->len || ctx->len + len < SM4_BLOCK_SIZE) {
610 		l = min(len, SM4_BLOCK_SIZE - ctx->len);
611 
612 		crypto_xor(ctx->digest + ctx->len, p, l);
613 		ctx->len += l;
614 		len -= l;
615 		p += l;
616 	}
617 
618 	if (len && (ctx->len % SM4_BLOCK_SIZE) == 0) {
619 		kernel_neon_begin();
620 
621 		if (len < SM4_BLOCK_SIZE && ctx->len == SM4_BLOCK_SIZE) {
622 			sm4_ce_crypt_block(tctx->key.rkey_enc,
623 					   ctx->digest, ctx->digest);
624 			ctx->len = 0;
625 		} else {
626 			nblocks = len / SM4_BLOCK_SIZE;
627 			len %= SM4_BLOCK_SIZE;
628 
629 			sm4_ce_mac_update(tctx->key.rkey_enc, ctx->digest, p,
630 					  nblocks, (ctx->len == SM4_BLOCK_SIZE),
631 					  (len != 0));
632 
633 			p += nblocks * SM4_BLOCK_SIZE;
634 
635 			if (len == 0)
636 				ctx->len = SM4_BLOCK_SIZE;
637 		}
638 
639 		kernel_neon_end();
640 
641 		if (len) {
642 			crypto_xor(ctx->digest, p, len);
643 			ctx->len = len;
644 		}
645 	}
646 
647 	return 0;
648 }
649 
650 static int sm4_cmac_final(struct shash_desc *desc, u8 *out)
651 {
652 	struct sm4_mac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
653 	struct sm4_mac_desc_ctx *ctx = shash_desc_ctx(desc);
654 	const u8 *consts = tctx->consts;
655 
656 	if (ctx->len != SM4_BLOCK_SIZE) {
657 		ctx->digest[ctx->len] ^= 0x80;
658 		consts += SM4_BLOCK_SIZE;
659 	}
660 
661 	kernel_neon_begin();
662 	sm4_ce_mac_update(tctx->key.rkey_enc, ctx->digest, consts, 1,
663 			  false, true);
664 	kernel_neon_end();
665 
666 	memcpy(out, ctx->digest, SM4_BLOCK_SIZE);
667 
668 	return 0;
669 }
670 
671 static int sm4_cbcmac_final(struct shash_desc *desc, u8 *out)
672 {
673 	struct sm4_mac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
674 	struct sm4_mac_desc_ctx *ctx = shash_desc_ctx(desc);
675 
676 	if (ctx->len) {
677 		kernel_neon_begin();
678 		sm4_ce_crypt_block(tctx->key.rkey_enc, ctx->digest,
679 				   ctx->digest);
680 		kernel_neon_end();
681 	}
682 
683 	memcpy(out, ctx->digest, SM4_BLOCK_SIZE);
684 
685 	return 0;
686 }
687 
688 static struct shash_alg sm4_mac_algs[] = {
689 	{
690 		.base = {
691 			.cra_name		= "cmac(sm4)",
692 			.cra_driver_name	= "cmac-sm4-ce",
693 			.cra_priority		= 400,
694 			.cra_blocksize		= SM4_BLOCK_SIZE,
695 			.cra_ctxsize		= sizeof(struct sm4_mac_tfm_ctx)
696 							+ SM4_BLOCK_SIZE * 2,
697 			.cra_module		= THIS_MODULE,
698 		},
699 		.digestsize	= SM4_BLOCK_SIZE,
700 		.init		= sm4_mac_init,
701 		.update		= sm4_mac_update,
702 		.final		= sm4_cmac_final,
703 		.setkey		= sm4_cmac_setkey,
704 		.descsize	= sizeof(struct sm4_mac_desc_ctx),
705 	}, {
706 		.base = {
707 			.cra_name		= "xcbc(sm4)",
708 			.cra_driver_name	= "xcbc-sm4-ce",
709 			.cra_priority		= 400,
710 			.cra_blocksize		= SM4_BLOCK_SIZE,
711 			.cra_ctxsize		= sizeof(struct sm4_mac_tfm_ctx)
712 							+ SM4_BLOCK_SIZE * 2,
713 			.cra_module		= THIS_MODULE,
714 		},
715 		.digestsize	= SM4_BLOCK_SIZE,
716 		.init		= sm4_mac_init,
717 		.update		= sm4_mac_update,
718 		.final		= sm4_cmac_final,
719 		.setkey		= sm4_xcbc_setkey,
720 		.descsize	= sizeof(struct sm4_mac_desc_ctx),
721 	}, {
722 		.base = {
723 			.cra_name		= "cbcmac(sm4)",
724 			.cra_driver_name	= "cbcmac-sm4-ce",
725 			.cra_priority		= 400,
726 			.cra_blocksize		= 1,
727 			.cra_ctxsize		= sizeof(struct sm4_mac_tfm_ctx),
728 			.cra_module		= THIS_MODULE,
729 		},
730 		.digestsize	= SM4_BLOCK_SIZE,
731 		.init		= sm4_mac_init,
732 		.update		= sm4_mac_update,
733 		.final		= sm4_cbcmac_final,
734 		.setkey		= sm4_cbcmac_setkey,
735 		.descsize	= sizeof(struct sm4_mac_desc_ctx),
736 	}
737 };
738 
739 static int __init sm4_init(void)
740 {
741 	int err;
742 
743 	err = crypto_register_skciphers(sm4_algs, ARRAY_SIZE(sm4_algs));
744 	if (err)
745 		return err;
746 
747 	err = crypto_register_shashes(sm4_mac_algs, ARRAY_SIZE(sm4_mac_algs));
748 	if (err)
749 		goto out_err;
750 
751 	return 0;
752 
753 out_err:
754 	crypto_unregister_skciphers(sm4_algs, ARRAY_SIZE(sm4_algs));
755 	return err;
756 }
757 
758 static void __exit sm4_exit(void)
759 {
760 	crypto_unregister_shashes(sm4_mac_algs, ARRAY_SIZE(sm4_mac_algs));
761 	crypto_unregister_skciphers(sm4_algs, ARRAY_SIZE(sm4_algs));
762 }
763 
764 module_cpu_feature_match(SM4, sm4_init);
765 module_exit(sm4_exit);
766 
767 MODULE_DESCRIPTION("SM4 ECB/CBC/CTR/XTS using ARMv8 Crypto Extensions");
768 MODULE_ALIAS_CRYPTO("sm4-ce");
769 MODULE_ALIAS_CRYPTO("sm4");
770 MODULE_ALIAS_CRYPTO("ecb(sm4)");
771 MODULE_ALIAS_CRYPTO("cbc(sm4)");
772 MODULE_ALIAS_CRYPTO("ctr(sm4)");
773 MODULE_ALIAS_CRYPTO("cts(cbc(sm4))");
774 MODULE_ALIAS_CRYPTO("xts(sm4)");
775 MODULE_ALIAS_CRYPTO("cmac(sm4)");
776 MODULE_ALIAS_CRYPTO("xcbc(sm4)");
777 MODULE_ALIAS_CRYPTO("cbcmac(sm4)");
778 MODULE_AUTHOR("Tianjia Zhang <tianjia.zhang@linux.alibaba.com>");
779 MODULE_LICENSE("GPL v2");
780