[linux-block.git] / arch / x86 / crypto / cast5_avx_glue.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * Glue Code for the AVX assembler implementation of the Cast5 Cipher
 *
 * Copyright (C) 2012 Johannes Goetzfried
 *     <Johannes.Goetzfried@informatik.stud.uni-erlangen.de>
 */

#include <asm/crypto/glue_helper.h>
#include <crypto/algapi.h>
#include <crypto/cast5.h>
#include <crypto/internal/simd.h>
#include <linux/crypto.h>
#include <linux/err.h>
#include <linux/module.h>
#include <linux/types.h>

#define CAST5_PARALLEL_BLOCKS 16

asmlinkage void cast5_ecb_enc_16way(struct cast5_ctx *ctx, u8 *dst,
				    const u8 *src);
asmlinkage void cast5_ecb_dec_16way(struct cast5_ctx *ctx, u8 *dst,
				    const u8 *src);
asmlinkage void cast5_cbc_dec_16way(struct cast5_ctx *ctx, u8 *dst,
				    const u8 *src);
asmlinkage void cast5_ctr_16way(struct cast5_ctx *ctx, u8 *dst, const u8 *src,
				__be64 *iv);

static int cast5_setkey_skcipher(struct crypto_skcipher *tfm, const u8 *key,
				 unsigned int keylen)
{
	return cast5_setkey(&tfm->base, key, keylen);
}

static inline bool cast5_fpu_begin(bool fpu_enabled, struct skcipher_walk *walk,
				   unsigned int nbytes)
{
	return glue_fpu_begin(CAST5_BLOCK_SIZE, CAST5_PARALLEL_BLOCKS,
			      walk, fpu_enabled, nbytes);
}

static inline void cast5_fpu_end(bool fpu_enabled)
{
	return glue_fpu_end(fpu_enabled);
}

static int ecb_crypt(struct skcipher_request *req, bool enc)
{
	bool fpu_enabled = false;
	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	struct skcipher_walk walk;
	const unsigned int bsize = CAST5_BLOCK_SIZE;
	unsigned int nbytes;
	void (*fn)(struct cast5_ctx *ctx, u8 *dst, const u8 *src);
	int err;

	err = skcipher_walk_virt(&walk, req, false);

	while ((nbytes = walk.nbytes)) {
		u8 *wsrc = walk.src.virt.addr;
		u8 *wdst = walk.dst.virt.addr;

		fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);

		/* Process multi-block batch */
		if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
			fn = (enc) ? cast5_ecb_enc_16way : cast5_ecb_dec_16way;
			do {
				fn(ctx, wdst, wsrc);

				wsrc += bsize * CAST5_PARALLEL_BLOCKS;
				wdst += bsize * CAST5_PARALLEL_BLOCKS;
				nbytes -= bsize * CAST5_PARALLEL_BLOCKS;
			} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);

			if (nbytes < bsize)
				goto done;
		}

		fn = (enc) ? __cast5_encrypt : __cast5_decrypt;

		/* Handle leftovers */
		do {
			fn(ctx, wdst, wsrc);

			wsrc += bsize;
			wdst += bsize;
			nbytes -= bsize;
		} while (nbytes >= bsize);

done:
		err = skcipher_walk_done(&walk, nbytes);
	}

	cast5_fpu_end(fpu_enabled);
	return err;
}

static int ecb_encrypt(struct skcipher_request *req)
{
	return ecb_crypt(req, true);
}

static int ecb_decrypt(struct skcipher_request *req)
{
	return ecb_crypt(req, false);
}

static int cbc_encrypt(struct skcipher_request *req)
{
	const unsigned int bsize = CAST5_BLOCK_SIZE;
	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	struct skcipher_walk walk;
	unsigned int nbytes;
	int err;

	err = skcipher_walk_virt(&walk, req, false);

	while ((nbytes = walk.nbytes)) {
		u64 *src = (u64 *)walk.src.virt.addr;
		u64 *dst = (u64 *)walk.dst.virt.addr;
		u64 *iv = (u64 *)walk.iv;

		do {
			*dst = *src ^ *iv;
			__cast5_encrypt(ctx, (u8 *)dst, (u8 *)dst);
			iv = dst;
			src++;
			dst++;
			nbytes -= bsize;
		} while (nbytes >= bsize);

		*(u64 *)walk.iv = *iv;
		err = skcipher_walk_done(&walk, nbytes);
	}

	return err;
}

static unsigned int __cbc_decrypt(struct cast5_ctx *ctx,
				  struct skcipher_walk *walk)
{
	const unsigned int bsize = CAST5_BLOCK_SIZE;
	unsigned int nbytes = walk->nbytes;
	u64 *src = (u64 *)walk->src.virt.addr;
	u64 *dst = (u64 *)walk->dst.virt.addr;
	u64 last_iv;

	/* Start of the last block. */
	src += nbytes / bsize - 1;
	dst += nbytes / bsize - 1;

	last_iv = *src;

	/* Process multi-block batch */
	if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
		do {
			nbytes -= bsize * (CAST5_PARALLEL_BLOCKS - 1);
			src -= CAST5_PARALLEL_BLOCKS - 1;
			dst -= CAST5_PARALLEL_BLOCKS - 1;

			cast5_cbc_dec_16way(ctx, (u8 *)dst, (u8 *)src);

			nbytes -= bsize;
			if (nbytes < bsize)
				goto done;

			*dst ^= *(src - 1);
			src -= 1;
			dst -= 1;
		} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);
	}

	/* Handle leftovers */
	for (;;) {
		__cast5_decrypt(ctx, (u8 *)dst, (u8 *)src);

		nbytes -= bsize;
		if (nbytes < bsize)
			break;

		*dst ^= *(src - 1);
		src -= 1;
		dst -= 1;
	}

done:
	*dst ^= *(u64 *)walk->iv;
	*(u64 *)walk->iv = last_iv;

	return nbytes;
}

static int cbc_decrypt(struct skcipher_request *req)
{
	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	bool fpu_enabled = false;
	struct skcipher_walk walk;
	unsigned int nbytes;
	int err;

	err = skcipher_walk_virt(&walk, req, false);

	while ((nbytes = walk.nbytes)) {
		fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);
		nbytes = __cbc_decrypt(ctx, &walk);
		err = skcipher_walk_done(&walk, nbytes);
	}

	cast5_fpu_end(fpu_enabled);
	return err;
}

static void ctr_crypt_final(struct skcipher_walk *walk, struct cast5_ctx *ctx)
{
	u8 *ctrblk = walk->iv;
	u8 keystream[CAST5_BLOCK_SIZE];
	u8 *src = walk->src.virt.addr;
	u8 *dst = walk->dst.virt.addr;
	unsigned int nbytes = walk->nbytes;

	__cast5_encrypt(ctx, keystream, ctrblk);
	crypto_xor_cpy(dst, keystream, src, nbytes);

	crypto_inc(ctrblk, CAST5_BLOCK_SIZE);
}

static unsigned int __ctr_crypt(struct skcipher_walk *walk,
				struct cast5_ctx *ctx)
{
	const unsigned int bsize = CAST5_BLOCK_SIZE;
	unsigned int nbytes = walk->nbytes;
	u64 *src = (u64 *)walk->src.virt.addr;
	u64 *dst = (u64 *)walk->dst.virt.addr;

	/* Process multi-block batch */
	if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
		do {
			cast5_ctr_16way(ctx, (u8 *)dst, (u8 *)src,
					(__be64 *)walk->iv);

			src += CAST5_PARALLEL_BLOCKS;
			dst += CAST5_PARALLEL_BLOCKS;
			nbytes -= bsize * CAST5_PARALLEL_BLOCKS;
		} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);

		if (nbytes < bsize)
			goto done;
	}

	/* Handle leftovers */
	do {
		u64 ctrblk;

		if (dst != src)
			*dst = *src;

		ctrblk = *(u64 *)walk->iv;
		be64_add_cpu((__be64 *)walk->iv, 1);

		__cast5_encrypt(ctx, (u8 *)&ctrblk, (u8 *)&ctrblk);
		*dst ^= ctrblk;

		src += 1;
		dst += 1;
		nbytes -= bsize;
	} while (nbytes >= bsize);

done:
	return nbytes;
}

static int ctr_crypt(struct skcipher_request *req)
{
	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	bool fpu_enabled = false;
	struct skcipher_walk walk;
	unsigned int nbytes;
	int err;

	err = skcipher_walk_virt(&walk, req, false);

	while ((nbytes = walk.nbytes) >= CAST5_BLOCK_SIZE) {
		fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);
		nbytes = __ctr_crypt(&walk, ctx);
		err = skcipher_walk_done(&walk, nbytes);
	}

	cast5_fpu_end(fpu_enabled);

	if (walk.nbytes) {
		ctr_crypt_final(&walk, ctx);
		err = skcipher_walk_done(&walk, 0);
	}

	return err;
}

static struct skcipher_alg cast5_algs[] = {
	{
		.base.cra_name		= "__ecb(cast5)",
		.base.cra_driver_name	= "__ecb-cast5-avx",
		.base.cra_priority	= 200,
		.base.cra_flags		= CRYPTO_ALG_INTERNAL,
		.base.cra_blocksize	= CAST5_BLOCK_SIZE,
		.base.cra_ctxsize	= sizeof(struct cast5_ctx),
		.base.cra_module	= THIS_MODULE,
		.min_keysize		= CAST5_MIN_KEY_SIZE,
		.max_keysize		= CAST5_MAX_KEY_SIZE,
		.setkey			= cast5_setkey_skcipher,
		.encrypt		= ecb_encrypt,
		.decrypt		= ecb_decrypt,
	}, {
		.base.cra_name		= "__cbc(cast5)",
		.base.cra_driver_name	= "__cbc-cast5-avx",
		.base.cra_priority	= 200,
		.base.cra_flags		= CRYPTO_ALG_INTERNAL,
		.base.cra_blocksize	= CAST5_BLOCK_SIZE,
		.base.cra_ctxsize	= sizeof(struct cast5_ctx),
		.base.cra_module	= THIS_MODULE,
		.min_keysize		= CAST5_MIN_KEY_SIZE,
		.max_keysize		= CAST5_MAX_KEY_SIZE,
		.ivsize			= CAST5_BLOCK_SIZE,
		.setkey			= cast5_setkey_skcipher,
		.encrypt		= cbc_encrypt,
		.decrypt		= cbc_decrypt,
	}, {
		.base.cra_name		= "__ctr(cast5)",
		.base.cra_driver_name	= "__ctr-cast5-avx",
		.base.cra_priority	= 200,
		.base.cra_flags		= CRYPTO_ALG_INTERNAL,
		.base.cra_blocksize	= 1,
		.base.cra_ctxsize	= sizeof(struct cast5_ctx),
		.base.cra_module	= THIS_MODULE,
		.min_keysize		= CAST5_MIN_KEY_SIZE,
		.max_keysize		= CAST5_MAX_KEY_SIZE,
		.ivsize			= CAST5_BLOCK_SIZE,
		.chunksize		= CAST5_BLOCK_SIZE,
		.setkey			= cast5_setkey_skcipher,
		.encrypt		= ctr_crypt,
		.decrypt		= ctr_crypt,
	}
};

static struct simd_skcipher_alg *cast5_simd_algs[ARRAY_SIZE(cast5_algs)];

static int __init cast5_init(void)
{
	const char *feature_name;

	if (!cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM,
				&feature_name)) {
		pr_info("CPU feature '%s' is not supported.\n", feature_name);
		return -ENODEV;
	}

	return simd_register_skciphers_compat(cast5_algs,
					      ARRAY_SIZE(cast5_algs),
					      cast5_simd_algs);
}

static void __exit cast5_exit(void)
{
	simd_unregister_skciphers(cast5_algs, ARRAY_SIZE(cast5_algs),
				  cast5_simd_algs);
}

module_init(cast5_init);
module_exit(cast5_exit);

MODULE_DESCRIPTION("Cast5 Cipher Algorithm, AVX optimized");
MODULE_LICENSE("GPL");
MODULE_ALIAS_CRYPTO("cast5");
Commit	Line	Data
1a59d1b8	1	// SPDX-License-Identifier: GPL-2.0-or-later
4d6d6a2c	2	/*
a97673a1	3	* Glue Code for the AVX assembler implementation of the Cast5 Cipher
4d6d6a2c JG	4	*
	5	* Copyright (C) 2012 Johannes Goetzfried
	6	* <Johannes.Goetzfried@informatik.stud.uni-erlangen.de>
4d6d6a2c JG	7	*/
4d6d6a2c JG	8
1e63183a	9	#include <asm/crypto/glue_helper.h>
4d6d6a2c JG	10	#include <crypto/algapi.h>
4d6d6a2c JG	11	#include <crypto/cast5.h>
1e63183a EB	12	#include <crypto/internal/simd.h>
	13	#include <linux/crypto.h>
	14	#include <linux/err.h>
	15	#include <linux/module.h>
	16	#include <linux/types.h>
4d6d6a2c JG	17
	18	#define CAST5_PARALLEL_BLOCKS 16
	19
c12ab20b	20	asmlinkage void cast5_ecb_enc_16way(struct cast5_ctx ctx, u8 dst,
4d6d6a2c	21	const u8 *src);
c12ab20b JK	22	asmlinkage void cast5_ecb_dec_16way(struct cast5_ctx ctx, u8 dst,
	23	const u8 *src);
	24	asmlinkage void cast5_cbc_dec_16way(struct cast5_ctx ctx, u8 dst,
	25	const u8 *src);
	26	asmlinkage void cast5_ctr_16way(struct cast5_ctx ctx, u8 dst, const u8 *src,
	27	__be64 *iv);
4d6d6a2c	28
1e63183a EB	29	static int cast5_setkey_skcipher(struct crypto_skcipher tfm, const u8 key,
1e63183a EB	30	unsigned int keylen)
4d6d6a2c	31	{
1e63183a EB	32	return cast5_setkey(&tfm->base, key, keylen);
	33	}
	34
	35	static inline bool cast5_fpu_begin(bool fpu_enabled, struct skcipher_walk *walk,
	36	unsigned int nbytes)
	37	{
75d8a553 EB	38	return glue_fpu_begin(CAST5_BLOCK_SIZE, CAST5_PARALLEL_BLOCKS,
75d8a553 EB	39	walk, fpu_enabled, nbytes);
4d6d6a2c JG	40	}
	41
	42	static inline void cast5_fpu_end(bool fpu_enabled)
	43	{
	44	return glue_fpu_end(fpu_enabled);
	45	}
	46
1e63183a	47	static int ecb_crypt(struct skcipher_request *req, bool enc)
4d6d6a2c JG	48	{
4d6d6a2c JG	49	bool fpu_enabled = false;
1e63183a EB	50	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	51	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	52	struct skcipher_walk walk;
4d6d6a2c JG	53	const unsigned int bsize = CAST5_BLOCK_SIZE;
4d6d6a2c JG	54	unsigned int nbytes;
c12ab20b	55	void (fn)(struct cast5_ctx ctx, u8 dst, const u8 src);
4d6d6a2c JG	56	int err;
4d6d6a2c JG	57
1e63183a	58	err = skcipher_walk_virt(&walk, req, false);
4d6d6a2c	59
1e63183a EB	60	while ((nbytes = walk.nbytes)) {
	61	u8 *wsrc = walk.src.virt.addr;
	62	u8 *wdst = walk.dst.virt.addr;
4d6d6a2c	63
1e63183a	64	fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);
4d6d6a2c JG	65
	66	/* Process multi-block batch */
	67	if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
8f461b1e	68	fn = (enc) ? cast5_ecb_enc_16way : cast5_ecb_dec_16way;
4d6d6a2c	69	do {
c12ab20b	70	fn(ctx, wdst, wsrc);
4d6d6a2c JG	71
	72	wsrc += bsize * CAST5_PARALLEL_BLOCKS;
	73	wdst += bsize * CAST5_PARALLEL_BLOCKS;
	74	nbytes -= bsize * CAST5_PARALLEL_BLOCKS;
	75	} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);
	76
	77	if (nbytes < bsize)
	78	goto done;
	79	}
	80
c12ab20b JK	81	fn = (enc) ? __cast5_encrypt : __cast5_decrypt;
c12ab20b JK	82
4d6d6a2c JG	83	/* Handle leftovers */
4d6d6a2c JG	84	do {
c12ab20b	85	fn(ctx, wdst, wsrc);
4d6d6a2c JG	86
	87	wsrc += bsize;
	88	wdst += bsize;
	89	nbytes -= bsize;
	90	} while (nbytes >= bsize);
	91
	92	done:
1e63183a	93	err = skcipher_walk_done(&walk, nbytes);
4d6d6a2c JG	94	}
	95
	96	cast5_fpu_end(fpu_enabled);
	97	return err;
	98	}
	99
1e63183a	100	static int ecb_encrypt(struct skcipher_request *req)
4d6d6a2c	101	{
1e63183a	102	return ecb_crypt(req, true);
4d6d6a2c JG	103	}
4d6d6a2c JG	104
1e63183a	105	static int ecb_decrypt(struct skcipher_request *req)
4d6d6a2c	106	{
1e63183a	107	return ecb_crypt(req, false);
4d6d6a2c JG	108	}
4d6d6a2c JG	109
1e63183a	110	static int cbc_encrypt(struct skcipher_request *req)
4d6d6a2c	111	{
4d6d6a2c	112	const unsigned int bsize = CAST5_BLOCK_SIZE;
1e63183a EB	113	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
	114	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
	115	struct skcipher_walk walk;
	116	unsigned int nbytes;
4d6d6a2c JG	117	int err;
4d6d6a2c JG	118
1e63183a	119	err = skcipher_walk_virt(&walk, req, false);
4d6d6a2c JG	120
4d6d6a2c JG	121	while ((nbytes = walk.nbytes)) {
1e63183a EB	122	u64 src = (u64 )walk.src.virt.addr;
	123	u64 dst = (u64 )walk.dst.virt.addr;
	124	u64 iv = (u64 )walk.iv;
	125
	126	do {
	127	dst = src ^ *iv;
	128	__cast5_encrypt(ctx, (u8 )dst, (u8 )dst);
	129	iv = dst;
	130	src++;
	131	dst++;
	132	nbytes -= bsize;
	133	} while (nbytes >= bsize);
	134
	135	(u64 )walk.iv = *iv;
	136	err = skcipher_walk_done(&walk, nbytes);
4d6d6a2c JG	137	}
	138
	139	return err;
	140	}
	141
1e63183a EB	142	static unsigned int __cbc_decrypt(struct cast5_ctx *ctx,
1e63183a EB	143	struct skcipher_walk *walk)
4d6d6a2c	144	{
4d6d6a2c JG	145	const unsigned int bsize = CAST5_BLOCK_SIZE;
	146	unsigned int nbytes = walk->nbytes;
	147	u64 src = (u64 )walk->src.virt.addr;
	148	u64 dst = (u64 )walk->dst.virt.addr;
4d6d6a2c	149	u64 last_iv;
4d6d6a2c JG	150
	151	/* Start of the last block. */
	152	src += nbytes / bsize - 1;
	153	dst += nbytes / bsize - 1;
	154
	155	last_iv = *src;
	156
	157	/* Process multi-block batch */
	158	if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
	159	do {
	160	nbytes -= bsize * (CAST5_PARALLEL_BLOCKS - 1);
	161	src -= CAST5_PARALLEL_BLOCKS - 1;
	162	dst -= CAST5_PARALLEL_BLOCKS - 1;
	163
c12ab20b	164	cast5_cbc_dec_16way(ctx, (u8 )dst, (u8 )src);
4d6d6a2c JG	165
	166	nbytes -= bsize;
	167	if (nbytes < bsize)
	168	goto done;
	169
	170	dst ^= (src - 1);
	171	src -= 1;
	172	dst -= 1;
	173	} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);
4d6d6a2c JG	174	}
	175
	176	/* Handle leftovers */
	177	for (;;) {
	178	__cast5_decrypt(ctx, (u8 )dst, (u8 )src);
	179
	180	nbytes -= bsize;
	181	if (nbytes < bsize)
	182	break;
	183
	184	dst ^= (src - 1);
	185	src -= 1;
	186	dst -= 1;
	187	}
	188
	189	done:
	190	dst ^= (u64 *)walk->iv;
	191	(u64 )walk->iv = last_iv;
	192
	193	return nbytes;
	194	}
	195
1e63183a	196	static int cbc_decrypt(struct skcipher_request *req)
4d6d6a2c	197	{
1e63183a EB	198	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
1e63183a EB	199	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
4d6d6a2c	200	bool fpu_enabled = false;
1e63183a EB	201	struct skcipher_walk walk;
1e63183a EB	202	unsigned int nbytes;
4d6d6a2c JG	203	int err;
4d6d6a2c JG	204
1e63183a	205	err = skcipher_walk_virt(&walk, req, false);
4d6d6a2c JG	206
4d6d6a2c JG	207	while ((nbytes = walk.nbytes)) {
1e63183a EB	208	fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);
	209	nbytes = __cbc_decrypt(ctx, &walk);
	210	err = skcipher_walk_done(&walk, nbytes);
4d6d6a2c JG	211	}
	212
	213	cast5_fpu_end(fpu_enabled);
	214	return err;
	215	}
	216
1e63183a	217	static void ctr_crypt_final(struct skcipher_walk walk, struct cast5_ctx ctx)
4d6d6a2c	218	{
4d6d6a2c JG	219	u8 *ctrblk = walk->iv;
	220	u8 keystream[CAST5_BLOCK_SIZE];
	221	u8 *src = walk->src.virt.addr;
	222	u8 *dst = walk->dst.virt.addr;
	223	unsigned int nbytes = walk->nbytes;
	224
	225	__cast5_encrypt(ctx, keystream, ctrblk);
45fe93df	226	crypto_xor_cpy(dst, keystream, src, nbytes);
4d6d6a2c JG	227
	228	crypto_inc(ctrblk, CAST5_BLOCK_SIZE);
	229	}
	230
1e63183a EB	231	static unsigned int __ctr_crypt(struct skcipher_walk *walk,
1e63183a EB	232	struct cast5_ctx *ctx)
4d6d6a2c	233	{
4d6d6a2c JG	234	const unsigned int bsize = CAST5_BLOCK_SIZE;
	235	unsigned int nbytes = walk->nbytes;
	236	u64 src = (u64 )walk->src.virt.addr;
	237	u64 dst = (u64 )walk->dst.virt.addr;
4d6d6a2c JG	238
	239	/* Process multi-block batch */
	240	if (nbytes >= bsize * CAST5_PARALLEL_BLOCKS) {
	241	do {
c12ab20b JK	242	cast5_ctr_16way(ctx, (u8 )dst, (u8 )src,
c12ab20b JK	243	(__be64 *)walk->iv);
4d6d6a2c JG	244
	245	src += CAST5_PARALLEL_BLOCKS;
	246	dst += CAST5_PARALLEL_BLOCKS;
	247	nbytes -= bsize * CAST5_PARALLEL_BLOCKS;
	248	} while (nbytes >= bsize * CAST5_PARALLEL_BLOCKS);
	249
	250	if (nbytes < bsize)
	251	goto done;
	252	}
	253
	254	/* Handle leftovers */
	255	do {
c12ab20b JK	256	u64 ctrblk;
c12ab20b JK	257
4d6d6a2c JG	258	if (dst != src)
	259	dst = src;
	260
c12ab20b JK	261	ctrblk = (u64 )walk->iv;
c12ab20b JK	262	be64_add_cpu((__be64 *)walk->iv, 1);
4d6d6a2c	263
c12ab20b JK	264	__cast5_encrypt(ctx, (u8 )&ctrblk, (u8 )&ctrblk);
c12ab20b JK	265	*dst ^= ctrblk;
4d6d6a2c JG	266
	267	src += 1;
	268	dst += 1;
	269	nbytes -= bsize;
	270	} while (nbytes >= bsize);
	271
	272	done:
4d6d6a2c JG	273	return nbytes;
	274	}
	275
1e63183a	276	static int ctr_crypt(struct skcipher_request *req)
4d6d6a2c	277	{
1e63183a EB	278	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
1e63183a EB	279	struct cast5_ctx *ctx = crypto_skcipher_ctx(tfm);
4d6d6a2c	280	bool fpu_enabled = false;
1e63183a EB	281	struct skcipher_walk walk;
1e63183a EB	282	unsigned int nbytes;
4d6d6a2c JG	283	int err;
4d6d6a2c JG	284
1e63183a	285	err = skcipher_walk_virt(&walk, req, false);
4d6d6a2c JG	286
4d6d6a2c JG	287	while ((nbytes = walk.nbytes) >= CAST5_BLOCK_SIZE) {
1e63183a EB	288	fpu_enabled = cast5_fpu_begin(fpu_enabled, &walk, nbytes);
	289	nbytes = __ctr_crypt(&walk, ctx);
	290	err = skcipher_walk_done(&walk, nbytes);
4d6d6a2c JG	291	}
	292
	293	cast5_fpu_end(fpu_enabled);
	294
	295	if (walk.nbytes) {
1e63183a EB	296	ctr_crypt_final(&walk, ctx);
1e63183a EB	297	err = skcipher_walk_done(&walk, 0);
4d6d6a2c JG	298	}
	299
	300	return err;
	301	}
	302
1e63183a EB	303	static struct skcipher_alg cast5_algs[] = {
	304	{
	305	.base.cra_name = "__ecb(cast5)",
	306	.base.cra_driver_name = "__ecb-cast5-avx",
	307	.base.cra_priority = 200,
	308	.base.cra_flags = CRYPTO_ALG_INTERNAL,
	309	.base.cra_blocksize = CAST5_BLOCK_SIZE,
	310	.base.cra_ctxsize = sizeof(struct cast5_ctx),
	311	.base.cra_module = THIS_MODULE,
	312	.min_keysize = CAST5_MIN_KEY_SIZE,
	313	.max_keysize = CAST5_MAX_KEY_SIZE,
	314	.setkey = cast5_setkey_skcipher,
	315	.encrypt = ecb_encrypt,
	316	.decrypt = ecb_decrypt,
	317	}, {
	318	.base.cra_name = "__cbc(cast5)",
	319	.base.cra_driver_name = "__cbc-cast5-avx",
	320	.base.cra_priority = 200,
	321	.base.cra_flags = CRYPTO_ALG_INTERNAL,
	322	.base.cra_blocksize = CAST5_BLOCK_SIZE,
	323	.base.cra_ctxsize = sizeof(struct cast5_ctx),
	324	.base.cra_module = THIS_MODULE,
	325	.min_keysize = CAST5_MIN_KEY_SIZE,
	326	.max_keysize = CAST5_MAX_KEY_SIZE,
	327	.ivsize = CAST5_BLOCK_SIZE,
	328	.setkey = cast5_setkey_skcipher,
	329	.encrypt = cbc_encrypt,
	330	.decrypt = cbc_decrypt,
	331	}, {
	332	.base.cra_name = "__ctr(cast5)",
	333	.base.cra_driver_name = "__ctr-cast5-avx",
	334	.base.cra_priority = 200,
	335	.base.cra_flags = CRYPTO_ALG_INTERNAL,
	336	.base.cra_blocksize = 1,
	337	.base.cra_ctxsize = sizeof(struct cast5_ctx),
	338	.base.cra_module = THIS_MODULE,
	339	.min_keysize = CAST5_MIN_KEY_SIZE,
	340	.max_keysize = CAST5_MAX_KEY_SIZE,
	341	.ivsize = CAST5_BLOCK_SIZE,
	342	.chunksize = CAST5_BLOCK_SIZE,
	343	.setkey = cast5_setkey_skcipher,
	344	.encrypt = ctr_crypt,
	345	.decrypt = ctr_crypt,
	346	}
	347	};
4d6d6a2c	348
1e63183a	349	static struct simd_skcipher_alg *cast5_simd_algs[ARRAY_SIZE(cast5_algs)];
4d6d6a2c JG	350
	351	static int __init cast5_init(void)
	352	{
d5d34d98	353	const char *feature_name;
4d6d6a2c	354
d91cab78 DH	355	if (!cpu_has_xfeatures(XFEATURE_MASK_SSE \| XFEATURE_MASK_YMM,
d91cab78 DH	356	&feature_name)) {
d5d34d98	357	pr_info("CPU feature '%s' is not supported.\n", feature_name);
4d6d6a2c JG	358	return -ENODEV;
	359	}
	360
1e63183a EB	361	return simd_register_skciphers_compat(cast5_algs,
	362	ARRAY_SIZE(cast5_algs),
	363	cast5_simd_algs);
4d6d6a2c JG	364	}
	365
	366	static void __exit cast5_exit(void)
	367	{
1e63183a EB	368	simd_unregister_skciphers(cast5_algs, ARRAY_SIZE(cast5_algs),
1e63183a EB	369	cast5_simd_algs);
4d6d6a2c JG	370	}
	371
	372	module_init(cast5_init);
	373	module_exit(cast5_exit);
	374
	375	MODULE_DESCRIPTION("Cast5 Cipher Algorithm, AVX optimized");
	376	MODULE_LICENSE("GPL");
5d26a105	377	MODULE_ALIAS_CRYPTO("cast5");