[linux-2.6-block.git] / arch / arm / crypto / aesbs-glue.c

/*
 * linux/arch/arm/crypto/aesbs-glue.c - glue code for NEON bit sliced AES
 *
 * Copyright (C) 2013 Linaro Ltd <ard.biesheuvel@linaro.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 */

#include <asm/neon.h>
#include <crypto/aes.h>
#include <crypto/ablk_helper.h>
#include <crypto/algapi.h>
#include <linux/module.h>

#include "aes_glue.h"

#define BIT_SLICED_KEY_MAXSIZE	(128 * (AES_MAXNR - 1) + 2 * AES_BLOCK_SIZE)

struct BS_KEY {
	struct AES_KEY	rk;
	int		converted;
	u8 __aligned(8)	bs[BIT_SLICED_KEY_MAXSIZE];
} __aligned(8);

asmlinkage void bsaes_enc_key_convert(u8 out[], struct AES_KEY const *in);
asmlinkage void bsaes_dec_key_convert(u8 out[], struct AES_KEY const *in);

asmlinkage void bsaes_cbc_encrypt(u8 const in[], u8 out[], u32 bytes,
				  struct BS_KEY *key, u8 iv[]);

asmlinkage void bsaes_ctr32_encrypt_blocks(u8 const in[], u8 out[], u32 blocks,
					   struct BS_KEY *key, u8 const iv[]);

asmlinkage void bsaes_xts_encrypt(u8 const in[], u8 out[], u32 bytes,
				  struct BS_KEY *key, u8 tweak[]);

asmlinkage void bsaes_xts_decrypt(u8 const in[], u8 out[], u32 bytes,
				  struct BS_KEY *key, u8 tweak[]);

struct aesbs_cbc_ctx {
	struct AES_KEY	enc;
	struct BS_KEY	dec;
};

struct aesbs_ctr_ctx {
	struct BS_KEY	enc;
};

struct aesbs_xts_ctx {
	struct BS_KEY	enc;
	struct BS_KEY	dec;
	struct AES_KEY	twkey;
};

static int aesbs_cbc_set_key(struct crypto_tfm *tfm, const u8 *in_key,
			     unsigned int key_len)
{
	struct aesbs_cbc_ctx *ctx = crypto_tfm_ctx(tfm);
	int bits = key_len * 8;

	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc)) {
		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
		return -EINVAL;
	}
	ctx->dec.rk = ctx->enc;
	private_AES_set_decrypt_key(in_key, bits, &ctx->dec.rk);
	ctx->dec.converted = 0;
	return 0;
}

static int aesbs_ctr_set_key(struct crypto_tfm *tfm, const u8 *in_key,
			     unsigned int key_len)
{
	struct aesbs_ctr_ctx *ctx = crypto_tfm_ctx(tfm);
	int bits = key_len * 8;

	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc.rk)) {
		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
		return -EINVAL;
	}
	ctx->enc.converted = 0;
	return 0;
}

static int aesbs_xts_set_key(struct crypto_tfm *tfm, const u8 *in_key,
			     unsigned int key_len)
{
	struct aesbs_xts_ctx *ctx = crypto_tfm_ctx(tfm);
	int bits = key_len * 4;

	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc.rk)) {
		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
		return -EINVAL;
	}
	ctx->dec.rk = ctx->enc.rk;
	private_AES_set_decrypt_key(in_key, bits, &ctx->dec.rk);
	private_AES_set_encrypt_key(in_key + key_len / 2, bits, &ctx->twkey);
	ctx->enc.converted = ctx->dec.converted = 0;
	return 0;
}

static int aesbs_cbc_encrypt(struct blkcipher_desc *desc,
			     struct scatterlist *dst,
			     struct scatterlist *src, unsigned int nbytes)
{
	struct aesbs_cbc_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	struct blkcipher_walk walk;
	int err;

	blkcipher_walk_init(&walk, dst, src, nbytes);
	err = blkcipher_walk_virt(desc, &walk);

	while (walk.nbytes) {
		u32 blocks = walk.nbytes / AES_BLOCK_SIZE;
		u8 *src = walk.src.virt.addr;

		if (walk.dst.virt.addr == walk.src.virt.addr) {
			u8 *iv = walk.iv;

			do {
				crypto_xor(src, iv, AES_BLOCK_SIZE);
				AES_encrypt(src, src, &ctx->enc);
				iv = src;
				src += AES_BLOCK_SIZE;
			} while (--blocks);
			memcpy(walk.iv, iv, AES_BLOCK_SIZE);
		} else {
			u8 *dst = walk.dst.virt.addr;

			do {
				crypto_xor(walk.iv, src, AES_BLOCK_SIZE);
				AES_encrypt(walk.iv, dst, &ctx->enc);
				memcpy(walk.iv, dst, AES_BLOCK_SIZE);
				src += AES_BLOCK_SIZE;
				dst += AES_BLOCK_SIZE;
			} while (--blocks);
		}
		err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
	}
	return err;
}

static int aesbs_cbc_decrypt(struct blkcipher_desc *desc,
			     struct scatterlist *dst,
			     struct scatterlist *src, unsigned int nbytes)
{
	struct aesbs_cbc_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	struct blkcipher_walk walk;
	int err;

	blkcipher_walk_init(&walk, dst, src, nbytes);
	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);

	while ((walk.nbytes / AES_BLOCK_SIZE) >= 8) {
		kernel_neon_begin();
		bsaes_cbc_encrypt(walk.src.virt.addr, walk.dst.virt.addr,
				  walk.nbytes, &ctx->dec, walk.iv);
		kernel_neon_end();
		err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
	}
	while (walk.nbytes) {
		u32 blocks = walk.nbytes / AES_BLOCK_SIZE;
		u8 *dst = walk.dst.virt.addr;
		u8 *src = walk.src.virt.addr;
		u8 bk[2][AES_BLOCK_SIZE];
		u8 *iv = walk.iv;

		do {
			if (walk.dst.virt.addr == walk.src.virt.addr)
				memcpy(bk[blocks & 1], src, AES_BLOCK_SIZE);

			AES_decrypt(src, dst, &ctx->dec.rk);
			crypto_xor(dst, iv, AES_BLOCK_SIZE);

			if (walk.dst.virt.addr == walk.src.virt.addr)
				iv = bk[blocks & 1];
			else
				iv = src;

			dst += AES_BLOCK_SIZE;
			src += AES_BLOCK_SIZE;
		} while (--blocks);
		err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
	}
	return err;
}

static void inc_be128_ctr(__be32 ctr[], u32 addend)
{
	int i;

	for (i = 3; i >= 0; i--, addend = 1) {
		u32 n = be32_to_cpu(ctr[i]) + addend;

		ctr[i] = cpu_to_be32(n);
		if (n >= addend)
			break;
	}
}

static int aesbs_ctr_encrypt(struct blkcipher_desc *desc,
			     struct scatterlist *dst, struct scatterlist *src,
			     unsigned int nbytes)
{
	struct aesbs_ctr_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	struct blkcipher_walk walk;
	u32 blocks;
	int err;

	blkcipher_walk_init(&walk, dst, src, nbytes);
	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);

	while ((blocks = walk.nbytes / AES_BLOCK_SIZE)) {
		u32 tail = walk.nbytes % AES_BLOCK_SIZE;
		__be32 *ctr = (__be32 *)walk.iv;
		u32 headroom = UINT_MAX - be32_to_cpu(ctr[3]);

		/* avoid 32 bit counter overflow in the NEON code */
		if (unlikely(headroom < blocks)) {
			blocks = headroom + 1;
			tail = walk.nbytes - blocks * AES_BLOCK_SIZE;
		}
		kernel_neon_begin();
		bsaes_ctr32_encrypt_blocks(walk.src.virt.addr,
					   walk.dst.virt.addr, blocks,
					   &ctx->enc, walk.iv);
		kernel_neon_end();
		inc_be128_ctr(ctr, blocks);

		nbytes -= blocks * AES_BLOCK_SIZE;
		if (nbytes && nbytes == tail && nbytes <= AES_BLOCK_SIZE)
			break;

		err = blkcipher_walk_done(desc, &walk, tail);
	}
	if (walk.nbytes) {
		u8 *tdst = walk.dst.virt.addr + blocks * AES_BLOCK_SIZE;
		u8 *tsrc = walk.src.virt.addr + blocks * AES_BLOCK_SIZE;
		u8 ks[AES_BLOCK_SIZE];

		AES_encrypt(walk.iv, ks, &ctx->enc.rk);
		if (tdst != tsrc)
			memcpy(tdst, tsrc, nbytes);
		crypto_xor(tdst, ks, nbytes);
		err = blkcipher_walk_done(desc, &walk, 0);
	}
	return err;
}

static int aesbs_xts_encrypt(struct blkcipher_desc *desc,
			     struct scatterlist *dst,
			     struct scatterlist *src, unsigned int nbytes)
{
	struct aesbs_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	struct blkcipher_walk walk;
	int err;

	blkcipher_walk_init(&walk, dst, src, nbytes);
	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);

	/* generate the initial tweak */
	AES_encrypt(walk.iv, walk.iv, &ctx->twkey);

	while (walk.nbytes) {
		kernel_neon_begin();
		bsaes_xts_encrypt(walk.src.virt.addr, walk.dst.virt.addr,
				  walk.nbytes, &ctx->enc, walk.iv);
		kernel_neon_end();
		err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
	}
	return err;
}

static int aesbs_xts_decrypt(struct blkcipher_desc *desc,
			     struct scatterlist *dst,
			     struct scatterlist *src, unsigned int nbytes)
{
	struct aesbs_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	struct blkcipher_walk walk;
	int err;

	blkcipher_walk_init(&walk, dst, src, nbytes);
	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);

	/* generate the initial tweak */
	AES_encrypt(walk.iv, walk.iv, &ctx->twkey);

	while (walk.nbytes) {
		kernel_neon_begin();
		bsaes_xts_decrypt(walk.src.virt.addr, walk.dst.virt.addr,
				  walk.nbytes, &ctx->dec, walk.iv);
		kernel_neon_end();
		err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
	}
	return err;
}

static struct crypto_alg aesbs_algs[] = { {
	.cra_name		= "__cbc-aes-neonbs",
	.cra_driver_name	= "__driver-cbc-aes-neonbs",
	.cra_priority		= 0,
	.cra_flags		= CRYPTO_ALG_TYPE_BLKCIPHER,
	.cra_blocksize		= AES_BLOCK_SIZE,
	.cra_ctxsize		= sizeof(struct aesbs_cbc_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_blkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_blkcipher = {
		.min_keysize	= AES_MIN_KEY_SIZE,
		.max_keysize	= AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= aesbs_cbc_set_key,
		.encrypt	= aesbs_cbc_encrypt,
		.decrypt	= aesbs_cbc_decrypt,
	},
}, {
	.cra_name		= "__ctr-aes-neonbs",
	.cra_driver_name	= "__driver-ctr-aes-neonbs",
	.cra_priority		= 0,
	.cra_flags		= CRYPTO_ALG_TYPE_BLKCIPHER,
	.cra_blocksize		= 1,
	.cra_ctxsize		= sizeof(struct aesbs_ctr_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_blkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_blkcipher = {
		.min_keysize	= AES_MIN_KEY_SIZE,
		.max_keysize	= AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= aesbs_ctr_set_key,
		.encrypt	= aesbs_ctr_encrypt,
		.decrypt	= aesbs_ctr_encrypt,
	},
}, {
	.cra_name		= "__xts-aes-neonbs",
	.cra_driver_name	= "__driver-xts-aes-neonbs",
	.cra_priority		= 0,
	.cra_flags		= CRYPTO_ALG_TYPE_BLKCIPHER,
	.cra_blocksize		= AES_BLOCK_SIZE,
	.cra_ctxsize		= sizeof(struct aesbs_xts_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_blkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_blkcipher = {
		.min_keysize	= 2 * AES_MIN_KEY_SIZE,
		.max_keysize	= 2 * AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= aesbs_xts_set_key,
		.encrypt	= aesbs_xts_encrypt,
		.decrypt	= aesbs_xts_decrypt,
	},
}, {
	.cra_name		= "cbc(aes)",
	.cra_driver_name	= "cbc-aes-neonbs",
	.cra_priority		= 300,
	.cra_flags		= CRYPTO_ALG_TYPE_ABLKCIPHER|CRYPTO_ALG_ASYNC,
	.cra_blocksize		= AES_BLOCK_SIZE,
	.cra_ctxsize		= sizeof(struct async_helper_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_ablkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_init		= ablk_init,
	.cra_exit		= ablk_exit,
	.cra_ablkcipher = {
		.min_keysize	= AES_MIN_KEY_SIZE,
		.max_keysize	= AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= ablk_set_key,
		.encrypt	= __ablk_encrypt,
		.decrypt	= ablk_decrypt,
	}
}, {
	.cra_name		= "ctr(aes)",
	.cra_driver_name	= "ctr-aes-neonbs",
	.cra_priority		= 300,
	.cra_flags		= CRYPTO_ALG_TYPE_ABLKCIPHER|CRYPTO_ALG_ASYNC,
	.cra_blocksize		= 1,
	.cra_ctxsize		= sizeof(struct async_helper_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_ablkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_init		= ablk_init,
	.cra_exit		= ablk_exit,
	.cra_ablkcipher = {
		.min_keysize	= AES_MIN_KEY_SIZE,
		.max_keysize	= AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= ablk_set_key,
		.encrypt	= ablk_encrypt,
		.decrypt	= ablk_decrypt,
	}
}, {
	.cra_name		= "xts(aes)",
	.cra_driver_name	= "xts-aes-neonbs",
	.cra_priority		= 300,
	.cra_flags		= CRYPTO_ALG_TYPE_ABLKCIPHER|CRYPTO_ALG_ASYNC,
	.cra_blocksize		= AES_BLOCK_SIZE,
	.cra_ctxsize		= sizeof(struct async_helper_ctx),
	.cra_alignmask		= 7,
	.cra_type		= &crypto_ablkcipher_type,
	.cra_module		= THIS_MODULE,
	.cra_init		= ablk_init,
	.cra_exit		= ablk_exit,
	.cra_ablkcipher = {
		.min_keysize	= 2 * AES_MIN_KEY_SIZE,
		.max_keysize	= 2 * AES_MAX_KEY_SIZE,
		.ivsize		= AES_BLOCK_SIZE,
		.setkey		= ablk_set_key,
		.encrypt	= ablk_encrypt,
		.decrypt	= ablk_decrypt,
	}
} };

static int __init aesbs_mod_init(void)
{
	if (!cpu_has_neon())
		return -ENODEV;

	return crypto_register_algs(aesbs_algs, ARRAY_SIZE(aesbs_algs));
}

static void __exit aesbs_mod_exit(void)
{
	crypto_unregister_algs(aesbs_algs, ARRAY_SIZE(aesbs_algs));
}

module_init(aesbs_mod_init);
module_exit(aesbs_mod_exit);

MODULE_DESCRIPTION("Bit sliced AES in CBC/CTR/XTS modes using NEON");
MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
MODULE_LICENSE("GPL");
Commit	Line	Data
e4e7f10b AB	1	/*
	2	* linux/arch/arm/crypto/aesbs-glue.c - glue code for NEON bit sliced AES
	3	*
	4	* Copyright (C) 2013 Linaro Ltd <ard.biesheuvel@linaro.org>
	5	*
	6	* This program is free software; you can redistribute it and/or modify
	7	* it under the terms of the GNU General Public License version 2 as
	8	* published by the Free Software Foundation.
	9	*/
	10
	11	#include <asm/neon.h>
	12	#include <crypto/aes.h>
	13	#include <crypto/ablk_helper.h>
	14	#include <crypto/algapi.h>
	15	#include <linux/module.h>
	16
	17	#include "aes_glue.h"
	18
	19	#define BIT_SLICED_KEY_MAXSIZE (128 * (AES_MAXNR - 1) + 2 * AES_BLOCK_SIZE)
	20
	21	struct BS_KEY {
	22	struct AES_KEY rk;
	23	int converted;
	24	u8 __aligned(8) bs[BIT_SLICED_KEY_MAXSIZE];
	25	} __aligned(8);
	26
	27	asmlinkage void bsaes_enc_key_convert(u8 out[], struct AES_KEY const *in);
	28	asmlinkage void bsaes_dec_key_convert(u8 out[], struct AES_KEY const *in);
	29
	30	asmlinkage void bsaes_cbc_encrypt(u8 const in[], u8 out[], u32 bytes,
	31	struct BS_KEY *key, u8 iv[]);
	32
	33	asmlinkage void bsaes_ctr32_encrypt_blocks(u8 const in[], u8 out[], u32 blocks,
	34	struct BS_KEY *key, u8 const iv[]);
	35
	36	asmlinkage void bsaes_xts_encrypt(u8 const in[], u8 out[], u32 bytes,
	37	struct BS_KEY *key, u8 tweak[]);
	38
	39	asmlinkage void bsaes_xts_decrypt(u8 const in[], u8 out[], u32 bytes,
	40	struct BS_KEY *key, u8 tweak[]);
	41
	42	struct aesbs_cbc_ctx {
	43	struct AES_KEY enc;
	44	struct BS_KEY dec;
	45	};
	46
	47	struct aesbs_ctr_ctx {
	48	struct BS_KEY enc;
	49	};
	50
	51	struct aesbs_xts_ctx {
	52	struct BS_KEY enc;
	53	struct BS_KEY dec;
	54	struct AES_KEY twkey;
	55	};
	56
	57	static int aesbs_cbc_set_key(struct crypto_tfm tfm, const u8 in_key,
	58	unsigned int key_len)
	59	{
	60	struct aesbs_cbc_ctx *ctx = crypto_tfm_ctx(tfm);
	61	int bits = key_len * 8;
	62
	63	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc)) {
	64	tfm->crt_flags \|= CRYPTO_TFM_RES_BAD_KEY_LEN;
65	return -EINVAL;
66	}
67	ctx->dec.rk = ctx->enc;
68	private_AES_set_decrypt_key(in_key, bits, &ctx->dec.rk);
69	ctx->dec.converted = 0;
70	return 0;
71	}
72
73	static int aesbs_ctr_set_key(struct crypto_tfm tfm, const u8 in_key,
74	unsigned int key_len)
75	{
76	struct aesbs_ctr_ctx *ctx = crypto_tfm_ctx(tfm);
77	int bits = key_len * 8;
78
79	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc.rk)) {
80	tfm->crt_flags \|= CRYPTO_TFM_RES_BAD_KEY_LEN;
81	return -EINVAL;
82	}
83	ctx->enc.converted = 0;
84	return 0;
85	}
86
87	static int aesbs_xts_set_key(struct crypto_tfm tfm, const u8 in_key,
88	unsigned int key_len)
89	{
90	struct aesbs_xts_ctx *ctx = crypto_tfm_ctx(tfm);
91	int bits = key_len * 4;
92
93	if (private_AES_set_encrypt_key(in_key, bits, &ctx->enc.rk)) {
94	tfm->crt_flags \|= CRYPTO_TFM_RES_BAD_KEY_LEN;
95	return -EINVAL;
96	}
97	ctx->dec.rk = ctx->enc.rk;
98	private_AES_set_decrypt_key(in_key, bits, &ctx->dec.rk);
99	private_AES_set_encrypt_key(in_key + key_len / 2, bits, &ctx->twkey);
100	ctx->enc.converted = ctx->dec.converted = 0;
101	return 0;
102	}
103
104	static int aesbs_cbc_encrypt(struct blkcipher_desc *desc,
105	struct scatterlist *dst,
106	struct scatterlist *src, unsigned int nbytes)
107	{
108	struct aesbs_cbc_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
109	struct blkcipher_walk walk;
110	int err;
111
112	blkcipher_walk_init(&walk, dst, src, nbytes);
113	err = blkcipher_walk_virt(desc, &walk);
114
115	while (walk.nbytes) {
116	u32 blocks = walk.nbytes / AES_BLOCK_SIZE;
117	u8 *src = walk.src.virt.addr;
118
119	if (walk.dst.virt.addr == walk.src.virt.addr) {
120	u8 *iv = walk.iv;
121
122	do {
123	crypto_xor(src, iv, AES_BLOCK_SIZE);
124	AES_encrypt(src, src, &ctx->enc);
125	iv = src;
126	src += AES_BLOCK_SIZE;
127	} while (--blocks);
128	memcpy(walk.iv, iv, AES_BLOCK_SIZE);
129	} else {
130	u8 *dst = walk.dst.virt.addr;
131
132	do {
133	crypto_xor(walk.iv, src, AES_BLOCK_SIZE);
134	AES_encrypt(walk.iv, dst, &ctx->enc);
135	memcpy(walk.iv, dst, AES_BLOCK_SIZE);
136	src += AES_BLOCK_SIZE;
137	dst += AES_BLOCK_SIZE;
138	} while (--blocks);
139	}
f3c400ef	140	err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
e4e7f10b AB	141	}
	142	return err;
	143	}
	144
	145	static int aesbs_cbc_decrypt(struct blkcipher_desc *desc,
	146	struct scatterlist *dst,
	147	struct scatterlist *src, unsigned int nbytes)
	148	{
	149	struct aesbs_cbc_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	150	struct blkcipher_walk walk;
	151	int err;
	152
	153	blkcipher_walk_init(&walk, dst, src, nbytes);
	154	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);
	155
	156	while ((walk.nbytes / AES_BLOCK_SIZE) >= 8) {
	157	kernel_neon_begin();
	158	bsaes_cbc_encrypt(walk.src.virt.addr, walk.dst.virt.addr,
	159	walk.nbytes, &ctx->dec, walk.iv);
	160	kernel_neon_end();
f3c400ef	161	err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
e4e7f10b AB	162	}
	163	while (walk.nbytes) {
	164	u32 blocks = walk.nbytes / AES_BLOCK_SIZE;
	165	u8 *dst = walk.dst.virt.addr;
	166	u8 *src = walk.src.virt.addr;
	167	u8 bk[2][AES_BLOCK_SIZE];
	168	u8 *iv = walk.iv;
	169
	170	do {
	171	if (walk.dst.virt.addr == walk.src.virt.addr)
	172	memcpy(bk[blocks & 1], src, AES_BLOCK_SIZE);
	173
	174	AES_decrypt(src, dst, &ctx->dec.rk);
	175	crypto_xor(dst, iv, AES_BLOCK_SIZE);
	176
	177	if (walk.dst.virt.addr == walk.src.virt.addr)
	178	iv = bk[blocks & 1];
	179	else
	180	iv = src;
	181
	182	dst += AES_BLOCK_SIZE;
	183	src += AES_BLOCK_SIZE;
	184	} while (--blocks);
f3c400ef	185	err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
e4e7f10b AB	186	}
	187	return err;
	188	}
	189
	190	static void inc_be128_ctr(__be32 ctr[], u32 addend)
	191	{
	192	int i;
	193
	194	for (i = 3; i >= 0; i--, addend = 1) {
	195	u32 n = be32_to_cpu(ctr[i]) + addend;
	196
	197	ctr[i] = cpu_to_be32(n);
	198	if (n >= addend)
	199	break;
	200	}
	201	}
	202
	203	static int aesbs_ctr_encrypt(struct blkcipher_desc *desc,
	204	struct scatterlist dst, struct scatterlist src,
	205	unsigned int nbytes)
	206	{
	207	struct aesbs_ctr_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	208	struct blkcipher_walk walk;
	209	u32 blocks;
	210	int err;
	211
	212	blkcipher_walk_init(&walk, dst, src, nbytes);
	213	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);
	214
	215	while ((blocks = walk.nbytes / AES_BLOCK_SIZE)) {
	216	u32 tail = walk.nbytes % AES_BLOCK_SIZE;
	217	__be32 ctr = (__be32 )walk.iv;
	218	u32 headroom = UINT_MAX - be32_to_cpu(ctr[3]);
	219
	220	/* avoid 32 bit counter overflow in the NEON code */
	221	if (unlikely(headroom < blocks)) {
	222	blocks = headroom + 1;
	223	tail = walk.nbytes - blocks * AES_BLOCK_SIZE;
	224	}
	225	kernel_neon_begin();
	226	bsaes_ctr32_encrypt_blocks(walk.src.virt.addr,
	227	walk.dst.virt.addr, blocks,
	228	&ctx->enc, walk.iv);
	229	kernel_neon_end();
	230	inc_be128_ctr(ctr, blocks);
	231
	232	nbytes -= blocks * AES_BLOCK_SIZE;
	233	if (nbytes && nbytes == tail && nbytes <= AES_BLOCK_SIZE)
	234	break;
	235
	236	err = blkcipher_walk_done(desc, &walk, tail);
	237	}
	238	if (walk.nbytes) {
	239	u8 tdst = walk.dst.virt.addr + blocks AES_BLOCK_SIZE;
	240	u8 tsrc = walk.src.virt.addr + blocks AES_BLOCK_SIZE;
	241	u8 ks[AES_BLOCK_SIZE];
	242
	243	AES_encrypt(walk.iv, ks, &ctx->enc.rk);
	244	if (tdst != tsrc)
	245	memcpy(tdst, tsrc, nbytes);
	246	crypto_xor(tdst, ks, nbytes);
	247	err = blkcipher_walk_done(desc, &walk, 0);
	248	}
	249	return err;
250	}
251
252	static int aesbs_xts_encrypt(struct blkcipher_desc *desc,
253	struct scatterlist *dst,
254	struct scatterlist *src, unsigned int nbytes)
255	{
256	struct aesbs_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
257	struct blkcipher_walk walk;
258	int err;
259
260	blkcipher_walk_init(&walk, dst, src, nbytes);
261	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);
262
263	/* generate the initial tweak */
264	AES_encrypt(walk.iv, walk.iv, &ctx->twkey);
265
266	while (walk.nbytes) {
267	kernel_neon_begin();
268	bsaes_xts_encrypt(walk.src.virt.addr, walk.dst.virt.addr,
269	walk.nbytes, &ctx->enc, walk.iv);
270	kernel_neon_end();
f3c400ef	271	err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
e4e7f10b AB	272	}
	273	return err;
	274	}
	275
	276	static int aesbs_xts_decrypt(struct blkcipher_desc *desc,
	277	struct scatterlist *dst,
	278	struct scatterlist *src, unsigned int nbytes)
	279	{
	280	struct aesbs_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm);
	281	struct blkcipher_walk walk;
	282	int err;
	283
	284	blkcipher_walk_init(&walk, dst, src, nbytes);
	285	err = blkcipher_walk_virt_block(desc, &walk, 8 * AES_BLOCK_SIZE);
	286
	287	/* generate the initial tweak */
	288	AES_encrypt(walk.iv, walk.iv, &ctx->twkey);
	289
	290	while (walk.nbytes) {
	291	kernel_neon_begin();
	292	bsaes_xts_decrypt(walk.src.virt.addr, walk.dst.virt.addr,
	293	walk.nbytes, &ctx->dec, walk.iv);
	294	kernel_neon_end();
f3c400ef	295	err = blkcipher_walk_done(desc, &walk, walk.nbytes % AES_BLOCK_SIZE);
e4e7f10b AB	296	}
	297	return err;
	298	}
	299
	300	static struct crypto_alg aesbs_algs[] = { {
	301	.cra_name = "__cbc-aes-neonbs",
	302	.cra_driver_name = "__driver-cbc-aes-neonbs",
	303	.cra_priority = 0,
	304	.cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
	305	.cra_blocksize = AES_BLOCK_SIZE,
	306	.cra_ctxsize = sizeof(struct aesbs_cbc_ctx),
	307	.cra_alignmask = 7,
	308	.cra_type = &crypto_blkcipher_type,
	309	.cra_module = THIS_MODULE,
	310	.cra_blkcipher = {
	311	.min_keysize = AES_MIN_KEY_SIZE,
	312	.max_keysize = AES_MAX_KEY_SIZE,
	313	.ivsize = AES_BLOCK_SIZE,
	314	.setkey = aesbs_cbc_set_key,
	315	.encrypt = aesbs_cbc_encrypt,
	316	.decrypt = aesbs_cbc_decrypt,
	317	},
	318	}, {
	319	.cra_name = "__ctr-aes-neonbs",
	320	.cra_driver_name = "__driver-ctr-aes-neonbs",
	321	.cra_priority = 0,
	322	.cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
	323	.cra_blocksize = 1,
	324	.cra_ctxsize = sizeof(struct aesbs_ctr_ctx),
	325	.cra_alignmask = 7,
	326	.cra_type = &crypto_blkcipher_type,
	327	.cra_module = THIS_MODULE,
	328	.cra_blkcipher = {
	329	.min_keysize = AES_MIN_KEY_SIZE,
	330	.max_keysize = AES_MAX_KEY_SIZE,
	331	.ivsize = AES_BLOCK_SIZE,
	332	.setkey = aesbs_ctr_set_key,
	333	.encrypt = aesbs_ctr_encrypt,
	334	.decrypt = aesbs_ctr_encrypt,
	335	},
	336	}, {
	337	.cra_name = "__xts-aes-neonbs",
	338	.cra_driver_name = "__driver-xts-aes-neonbs",
	339	.cra_priority = 0,
	340	.cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER,
	341	.cra_blocksize = AES_BLOCK_SIZE,
	342	.cra_ctxsize = sizeof(struct aesbs_xts_ctx),
	343	.cra_alignmask = 7,
	344	.cra_type = &crypto_blkcipher_type,
	345	.cra_module = THIS_MODULE,
	346	.cra_blkcipher = {
	347	.min_keysize = 2 * AES_MIN_KEY_SIZE,
	348	.max_keysize = 2 * AES_MAX_KEY_SIZE,
	349	.ivsize = AES_BLOCK_SIZE,
	350	.setkey = aesbs_xts_set_key,
	351	.encrypt = aesbs_xts_encrypt,
	352	.decrypt = aesbs_xts_decrypt,
	353	},
	354	}, {
	355	.cra_name = "cbc(aes)",
	356	.cra_driver_name = "cbc-aes-neonbs",
	357	.cra_priority = 300,
	358	.cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER\|CRYPTO_ALG_ASYNC,
	359	.cra_blocksize = AES_BLOCK_SIZE,
360	.cra_ctxsize = sizeof(struct async_helper_ctx),
361	.cra_alignmask = 7,
362	.cra_type = &crypto_ablkcipher_type,
363	.cra_module = THIS_MODULE,
364	.cra_init = ablk_init,
365	.cra_exit = ablk_exit,
366	.cra_ablkcipher = {
367	.min_keysize = AES_MIN_KEY_SIZE,
368	.max_keysize = AES_MAX_KEY_SIZE,
369	.ivsize = AES_BLOCK_SIZE,
370	.setkey = ablk_set_key,
371	.encrypt = __ablk_encrypt,
372	.decrypt = ablk_decrypt,
373	}
374	}, {
375	.cra_name = "ctr(aes)",
376	.cra_driver_name = "ctr-aes-neonbs",
377	.cra_priority = 300,
378	.cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER\|CRYPTO_ALG_ASYNC,
379	.cra_blocksize = 1,
380	.cra_ctxsize = sizeof(struct async_helper_ctx),
381	.cra_alignmask = 7,
382	.cra_type = &crypto_ablkcipher_type,
383	.cra_module = THIS_MODULE,
384	.cra_init = ablk_init,
385	.cra_exit = ablk_exit,
386	.cra_ablkcipher = {
387	.min_keysize = AES_MIN_KEY_SIZE,
388	.max_keysize = AES_MAX_KEY_SIZE,
389	.ivsize = AES_BLOCK_SIZE,
390	.setkey = ablk_set_key,
391	.encrypt = ablk_encrypt,
392	.decrypt = ablk_decrypt,
393	}
394	}, {
395	.cra_name = "xts(aes)",
396	.cra_driver_name = "xts-aes-neonbs",
397	.cra_priority = 300,
398	.cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER\|CRYPTO_ALG_ASYNC,
399	.cra_blocksize = AES_BLOCK_SIZE,
400	.cra_ctxsize = sizeof(struct async_helper_ctx),
401	.cra_alignmask = 7,
402	.cra_type = &crypto_ablkcipher_type,
403	.cra_module = THIS_MODULE,
404	.cra_init = ablk_init,
405	.cra_exit = ablk_exit,
406	.cra_ablkcipher = {
407	.min_keysize = 2 * AES_MIN_KEY_SIZE,
408	.max_keysize = 2 * AES_MAX_KEY_SIZE,
409	.ivsize = AES_BLOCK_SIZE,
410	.setkey = ablk_set_key,
411	.encrypt = ablk_encrypt,
412	.decrypt = ablk_decrypt,
413	}
414	} };
415
416	static int __init aesbs_mod_init(void)
417	{
418	if (!cpu_has_neon())
419	return -ENODEV;
420
421	return crypto_register_algs(aesbs_algs, ARRAY_SIZE(aesbs_algs));
422	}
423
424	static void __exit aesbs_mod_exit(void)
425	{
426	crypto_unregister_algs(aesbs_algs, ARRAY_SIZE(aesbs_algs));
427	}
428
429	module_init(aesbs_mod_init);
430	module_exit(aesbs_mod_exit);
431
432	MODULE_DESCRIPTION("Bit sliced AES in CBC/CTR/XTS modes using NEON");
433	MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
434	MODULE_LICENSE("GPL");