arm/crypto: Add optimized AES and SHA1 routines

Add assembler versions of AES and SHA1 for ARM platforms.  This has provided
up to a 50% improvement in IPsec/TCP throughout for tunnels using AES128/SHA1.

Platform   CPU SPeed    Endian   Before (bps)   After (bps)   Improvement

IXP425      533 MHz      big     11217042        15566294        ~38%
KS8695      166 MHz     little    3828549         5795373        ~51%
Signed-off-by: NDavid McCullough <ucdevel@gmail.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

arch/arm/Makefile

@@ -255,6 +255,7 @@ core-$(CONFIG_VFP)		+= arch/arm/vfp/
 # If we have a machine-specific directory, then include it in the build.
 core-y				+= arch/arm/kernel/ arch/arm/mm/ arch/arm/common/
 core-y				+= arch/arm/net/
+core-y				+= arch/arm/crypto/
 core-y				+= $(machdirs) $(platdirs)
 
 drivers-$(CONFIG_OPROFILE)      += arch/arm/oprofile/

arch/arm/crypto/Makefile

+#
+# Arch-specific CryptoAPI modules.
+#
+
+obj-$(CONFIG_CRYPTO_AES_ARM) += aes-arm.o
+obj-$(CONFIG_CRYPTO_SHA1_ARM) += sha1-arm.o
+
+aes-arm-y  := aes-armv4.o aes_glue.o
+sha1-arm-y := sha1-armv4-large.o sha1_glue.o

arch/arm/crypto/aes_glue.c

+/*
+ * Glue Code for the asm optimized version of the AES Cipher Algorithm
+ */
+
+#include <linux/module.h>
+#include <linux/crypto.h>
+#include <crypto/aes.h>
+
+#define AES_MAXNR 14
+
+typedef struct {
+	unsigned int rd_key[4 *(AES_MAXNR + 1)];
+	int rounds;
+} AES_KEY;
+
+struct AES_CTX {
+	AES_KEY enc_key;
+	AES_KEY dec_key;
+};
+
+asmlinkage void AES_encrypt(const u8 *in, u8 *out, AES_KEY *ctx);
+asmlinkage void AES_decrypt(const u8 *in, u8 *out, AES_KEY *ctx);
+asmlinkage int private_AES_set_decrypt_key(const unsigned char *userKey, const int bits, AES_KEY *key);
+asmlinkage int private_AES_set_encrypt_key(const unsigned char *userKey, const int bits, AES_KEY *key);
+
+static void aes_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	struct AES_CTX *ctx = crypto_tfm_ctx(tfm);
+	AES_encrypt(src, dst, &ctx->enc_key);
+}
+
+static void aes_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	struct AES_CTX *ctx = crypto_tfm_ctx(tfm);
+	AES_decrypt(src, dst, &ctx->dec_key);
+}
+
+static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key,
+		unsigned int key_len)
+{
+	struct AES_CTX *ctx = crypto_tfm_ctx(tfm);
+
+	switch (key_len) {
+	case AES_KEYSIZE_128:
+		key_len = 128;
+		break;
+	case AES_KEYSIZE_192:
+		key_len = 192;
+		break;
+	case AES_KEYSIZE_256:
+		key_len = 256;
+		break;
+	default:
+		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
+		return -EINVAL;
+	}
+
+	if (private_AES_set_encrypt_key(in_key, key_len, &ctx->enc_key) == -1) {
+		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
+		return -EINVAL;
+	}
+	/* private_AES_set_decrypt_key expects an encryption key as input */
+	ctx->dec_key = ctx->enc_key;
+	if (private_AES_set_decrypt_key(in_key, key_len, &ctx->dec_key) == -1) {
+		tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static struct crypto_alg aes_alg = {
+	.cra_name		= "aes",
+	.cra_driver_name	= "aes-asm",
+	.cra_priority		= 200,
+	.cra_flags		= CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize		= AES_BLOCK_SIZE,
+	.cra_ctxsize		= sizeof(struct AES_CTX),
+	.cra_module		= THIS_MODULE,
+	.cra_list		= LIST_HEAD_INIT(aes_alg.cra_list),
+	.cra_u	= {
+		.cipher	= {
+			.cia_min_keysize	= AES_MIN_KEY_SIZE,
+			.cia_max_keysize	= AES_MAX_KEY_SIZE,
+			.cia_setkey			= aes_set_key,
+			.cia_encrypt		= aes_encrypt,
+			.cia_decrypt		= aes_decrypt
+		}
+	}
+};
+
+static int __init aes_init(void)
+{
+	return crypto_register_alg(&aes_alg);
+}
+
+static void __exit aes_fini(void)
+{
+	crypto_unregister_alg(&aes_alg);
+}
+
+module_init(aes_init);
+module_exit(aes_fini);
+
+MODULE_DESCRIPTION("Rijndael (AES) Cipher Algorithm (ASM)");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("aes");
+MODULE_ALIAS("aes-asm");
+MODULE_AUTHOR("David McCullough <ucdevel@gmail.com>");

arch/arm/crypto/sha1-armv4-large.S

+#define __ARM_ARCH__ __LINUX_ARM_ARCH__
+@ ====================================================================
+@ Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+@ project. The module is, however, dual licensed under OpenSSL and
+@ CRYPTOGAMS licenses depending on where you obtain it. For further
+@ details see http://www.openssl.org/~appro/cryptogams/.
+@ ====================================================================
+
+@ sha1_block procedure for ARMv4.
+@
+@ January 2007.
+
+@ Size/performance trade-off
+@ ====================================================================
+@ impl		size in bytes	comp cycles[*]	measured performance
+@ ====================================================================
+@ thumb		304		3212		4420
+@ armv4-small	392/+29%	1958/+64%	2250/+96%
+@ armv4-compact	740/+89%	1552/+26%	1840/+22%
+@ armv4-large	1420/+92%	1307/+19%	1370/+34%[***]
+@ full unroll	~5100/+260%	~1260/+4%	~1300/+5%
+@ ====================================================================
+@ thumb		= same as 'small' but in Thumb instructions[**] and
+@		  with recurring code in two private functions;
+@ small		= detached Xload/update, loops are folded;
+@ compact	= detached Xload/update, 5x unroll;
+@ large		= interleaved Xload/update, 5x unroll;
+@ full unroll	= interleaved Xload/update, full unroll, estimated[!];
+@
+@ [*]	Manually counted instructions in "grand" loop body. Measured
+@	performance is affected by prologue and epilogue overhead,
+@	i-cache availability, branch penalties, etc.
+@ [**]	While each Thumb instruction is twice smaller, they are not as
+@	diverse as ARM ones: e.g., there are only two arithmetic
+@	instructions with 3 arguments, no [fixed] rotate, addressing
+@	modes are limited. As result it takes more instructions to do
+@	the same job in Thumb, therefore the code is never twice as
+@	small and always slower.
+@ [***]	which is also ~35% better than compiler generated code. Dual-
+@	issue Cortex A8 core was measured to process input block in
+@	~990 cycles.
+
+@ August 2010.
+@
+@ Rescheduling for dual-issue pipeline resulted in 13% improvement on
+@ Cortex A8 core and in absolute terms ~870 cycles per input block
+@ [or 13.6 cycles per byte].
+
+@ February 2011.
+@
+@ Profiler-assisted and platform-specific optimization resulted in 10%
+@ improvement on Cortex A8 core and 12.2 cycles per byte.
+
+.text
+
+.global	sha1_block_data_order
+.type	sha1_block_data_order,%function
+
+.align	2
+sha1_block_data_order:
+	stmdb	sp!,{r4-r12,lr}
+	add	r2,r1,r2,lsl#6	@ r2 to point at the end of r1
+	ldmia	r0,{r3,r4,r5,r6,r7}
+.Lloop:
+	ldr	r8,.LK_00_19
+	mov	r14,sp
+	sub	sp,sp,#15*4
+	mov	r5,r5,ror#30
+	mov	r6,r6,ror#30
+	mov	r7,r7,ror#30		@ [6]
+.L_00_15:
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r7,r8,r7,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r5,r6			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r7,r8,r7,ror#2			@ E+=K_00_19
+	eor	r10,r5,r6			@ F_xx_xx
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r4,r10,ror#2
+	add	r7,r7,r9			@ E+=X[i]
+	eor	r10,r10,r6,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r7,r7,r10			@ E+=F_00_19(B,C,D)
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r6,r8,r6,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r4,r5			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r6,r6,r7,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r6,r8,r6,ror#2			@ E+=K_00_19
+	eor	r10,r4,r5			@ F_xx_xx
+	add	r6,r6,r7,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r3,r10,ror#2
+	add	r6,r6,r9			@ E+=X[i]
+	eor	r10,r10,r5,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r6,r6,r10			@ E+=F_00_19(B,C,D)
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r5,r8,r5,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r3,r4			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r5,r5,r6,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r5,r8,r5,ror#2			@ E+=K_00_19
+	eor	r10,r3,r4			@ F_xx_xx
+	add	r5,r5,r6,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r7,r10,ror#2
+	add	r5,r5,r9			@ E+=X[i]
+	eor	r10,r10,r4,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r5,r5,r10			@ E+=F_00_19(B,C,D)
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r4,r8,r4,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r7,r3			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r4,r4,r5,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r4,r8,r4,ror#2			@ E+=K_00_19
+	eor	r10,r7,r3			@ F_xx_xx
+	add	r4,r4,r5,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r6,r10,ror#2
+	add	r4,r4,r9			@ E+=X[i]
+	eor	r10,r10,r3,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r4,r4,r10			@ E+=F_00_19(B,C,D)
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r3,r8,r3,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r6,r7			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r3,r3,r4,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r3,r8,r3,ror#2			@ E+=K_00_19
+	eor	r10,r6,r7			@ F_xx_xx
+	add	r3,r3,r4,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r5,r10,ror#2
+	add	r3,r3,r9			@ E+=X[i]
+	eor	r10,r10,r7,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r3,r3,r10			@ E+=F_00_19(B,C,D)
+	teq	r14,sp
+	bne	.L_00_15		@ [((11+4)*5+2)*3]
+#if __ARM_ARCH__<7
+	ldrb	r10,[r1,#2]
+	ldrb	r9,[r1,#3]
+	ldrb	r11,[r1,#1]
+	add	r7,r8,r7,ror#2			@ E+=K_00_19
+	ldrb	r12,[r1],#4
+	orr	r9,r9,r10,lsl#8
+	eor	r10,r5,r6			@ F_xx_xx
+	orr	r9,r9,r11,lsl#16
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+	orr	r9,r9,r12,lsl#24
+#else
+	ldr	r9,[r1],#4			@ handles unaligned
+	add	r7,r8,r7,ror#2			@ E+=K_00_19
+	eor	r10,r5,r6			@ F_xx_xx
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	r9,r9				@ byte swap
+#endif
+#endif
+	and	r10,r4,r10,ror#2
+	add	r7,r7,r9			@ E+=X[i]
+	eor	r10,r10,r6,ror#2		@ F_00_19(B,C,D)
+	str	r9,[r14,#-4]!
+	add	r7,r7,r10			@ E+=F_00_19(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r6,r8,r6,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r4,r5			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r6,r6,r7,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r3,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r6,r6,r9			@ E+=X[i]
+	eor	r10,r10,r5,ror#2		@ F_00_19(B,C,D)
+	add	r6,r6,r10			@ E+=F_00_19(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r5,r8,r5,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r3,r4			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r5,r5,r6,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r7,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r5,r5,r9			@ E+=X[i]
+	eor	r10,r10,r4,ror#2		@ F_00_19(B,C,D)
+	add	r5,r5,r10			@ E+=F_00_19(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r4,r8,r4,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r7,r3			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r4,r4,r5,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r6,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r4,r4,r9			@ E+=X[i]
+	eor	r10,r10,r3,ror#2		@ F_00_19(B,C,D)
+	add	r4,r4,r10			@ E+=F_00_19(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r3,r8,r3,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r6,r7			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r3,r3,r4,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r5,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r3,r3,r9			@ E+=X[i]
+	eor	r10,r10,r7,ror#2		@ F_00_19(B,C,D)
+	add	r3,r3,r10			@ E+=F_00_19(B,C,D)
+
+	ldr	r8,.LK_20_39		@ [+15+16*4]
+	sub	sp,sp,#25*4
+	cmn	sp,#0			@ [+3], clear carry to denote 20_39
+.L_20_39_or_60_79:
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r7,r8,r7,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r5,r6			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	eor r10,r4,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r7,r7,r9			@ E+=X[i]
+	add	r7,r7,r10			@ E+=F_20_39(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r6,r8,r6,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r4,r5			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r6,r6,r7,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	eor r10,r3,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r6,r6,r9			@ E+=X[i]
+	add	r6,r6,r10			@ E+=F_20_39(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r5,r8,r5,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r3,r4			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r5,r5,r6,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	eor r10,r7,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r5,r5,r9			@ E+=X[i]
+	add	r5,r5,r10			@ E+=F_20_39(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r4,r8,r4,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r7,r3			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r4,r4,r5,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	eor r10,r6,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r4,r4,r9			@ E+=X[i]
+	add	r4,r4,r10			@ E+=F_20_39(B,C,D)
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r3,r8,r3,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r6,r7			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r3,r3,r4,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	eor r10,r5,r10,ror#2					@ F_xx_xx
+						@ F_xx_xx
+	add	r3,r3,r9			@ E+=X[i]
+	add	r3,r3,r10			@ E+=F_20_39(B,C,D)
+	teq	r14,sp			@ preserve carry
+	bne	.L_20_39_or_60_79	@ [+((12+3)*5+2)*4]
+	bcs	.L_done			@ [+((12+3)*5+2)*4], spare 300 bytes
+
+	ldr	r8,.LK_40_59
+	sub	sp,sp,#20*4		@ [+2]
+.L_40_59:
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r7,r8,r7,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r5,r6			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r7,r7,r3,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r4,r10,ror#2					@ F_xx_xx
+	and r11,r5,r6					@ F_xx_xx
+	add	r7,r7,r9			@ E+=X[i]
+	add	r7,r7,r10			@ E+=F_40_59(B,C,D)
+	add	r7,r7,r11,ror#2
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r6,r8,r6,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r4,r5			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r6,r6,r7,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r3,r10,ror#2					@ F_xx_xx
+	and r11,r4,r5					@ F_xx_xx
+	add	r6,r6,r9			@ E+=X[i]
+	add	r6,r6,r10			@ E+=F_40_59(B,C,D)
+	add	r6,r6,r11,ror#2
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r5,r8,r5,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r3,r4			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r5,r5,r6,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r7,r10,ror#2					@ F_xx_xx
+	and r11,r3,r4					@ F_xx_xx
+	add	r5,r5,r9			@ E+=X[i]
+	add	r5,r5,r10			@ E+=F_40_59(B,C,D)
+	add	r5,r5,r11,ror#2
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r4,r8,r4,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r7,r3			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r4,r4,r5,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r6,r10,ror#2					@ F_xx_xx
+	and r11,r7,r3					@ F_xx_xx
+	add	r4,r4,r9			@ E+=X[i]
+	add	r4,r4,r10			@ E+=F_40_59(B,C,D)
+	add	r4,r4,r11,ror#2
+	ldr	r9,[r14,#15*4]
+	ldr	r10,[r14,#13*4]
+	ldr	r11,[r14,#7*4]
+	add	r3,r8,r3,ror#2			@ E+=K_xx_xx
+	ldr	r12,[r14,#2*4]
+	eor	r9,r9,r10
+	eor	r11,r11,r12			@ 1 cycle stall
+	eor	r10,r6,r7			@ F_xx_xx
+	mov	r9,r9,ror#31
+	add	r3,r3,r4,ror#27			@ E+=ROR(A,27)
+	eor	r9,r9,r11,ror#31
+	str	r9,[r14,#-4]!
+	and r10,r5,r10,ror#2					@ F_xx_xx
+	and r11,r6,r7					@ F_xx_xx
+	add	r3,r3,r9			@ E+=X[i]
+	add	r3,r3,r10			@ E+=F_40_59(B,C,D)
+	add	r3,r3,r11,ror#2
+	teq	r14,sp
+	bne	.L_40_59		@ [+((12+5)*5+2)*4]
+
+	ldr	r8,.LK_60_79
+	sub	sp,sp,#20*4
+	cmp	sp,#0			@ set carry to denote 60_79
+	b	.L_20_39_or_60_79	@ [+4], spare 300 bytes
+.L_done:
+	add	sp,sp,#80*4		@ "deallocate" stack frame
+	ldmia	r0,{r8,r9,r10,r11,r12}
+	add	r3,r8,r3
+	add	r4,r9,r4
+	add	r5,r10,r5,ror#2
+	add	r6,r11,r6,ror#2
+	add	r7,r12,r7,ror#2
+	stmia	r0,{r3,r4,r5,r6,r7}
+	teq	r1,r2
+	bne	.Lloop			@ [+18], total 1307
+
+#if __ARM_ARCH__>=5
+	ldmia	sp!,{r4-r12,pc}
+#else
+	ldmia	sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	.word	0xe12fff1e			@ interoperable with Thumb ISA:-)
+#endif
+.align	2
+.LK_00_19:	.word	0x5a827999
+.LK_20_39:	.word	0x6ed9eba1
+.LK_40_59:	.word	0x8f1bbcdc
+.LK_60_79:	.word	0xca62c1d6
+.size	sha1_block_data_order,.-sha1_block_data_order
+.asciz	"SHA1 block transform for ARMv4, CRYPTOGAMS by <appro@openssl.org>"
+.align	2

arch/arm/crypto/sha1_glue.c

+/*
+ * Cryptographic API.
+ * Glue code for the SHA1 Secure Hash Algorithm assembler implementation
+ *
+ * This file is based on sha1_generic.c and sha1_ssse3_glue.c
+ *
+ * Copyright (c) Alan Smithee.
+ * Copyright (c) Andrew McDonald <andrew@mcdonald.org.uk>
+ * Copyright (c) Jean-Francois Dive <jef@linuxbe.org>
+ * Copyright (c) Mathias Krause <minipli@googlemail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ */
+
+#include <crypto/internal/hash.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/cryptohash.h>
+#include <linux/types.h>
+#include <crypto/sha.h>
+#include <asm/byteorder.h>
+
+struct SHA1_CTX {
+	uint32_t h0,h1,h2,h3,h4;
+	u64 count;
+	u8 data[SHA1_BLOCK_SIZE];
+};
+
+asmlinkage void sha1_block_data_order(struct SHA1_CTX *digest,
+		const unsigned char *data, unsigned int rounds);
+
+
+static int sha1_init(struct shash_desc *desc)
+{
+	struct SHA1_CTX *sctx = shash_desc_ctx(desc);
+	memset(sctx, 0, sizeof(*sctx));
+	sctx->h0 = SHA1_H0;
+	sctx->h1 = SHA1_H1;
+	sctx->h2 = SHA1_H2;
+	sctx->h3 = SHA1_H3;
+	sctx->h4 = SHA1_H4;
+	return 0;
+}
+
+
+static int __sha1_update(struct SHA1_CTX *sctx, const u8 *data,
+			       unsigned int len, unsigned int partial)
+{
+	unsigned int done = 0;
+
+	sctx->count += len;
+
+	if (partial) {
+		done = SHA1_BLOCK_SIZE - partial;
+		memcpy(sctx->data + partial, data, done);
+		sha1_block_data_order(sctx, sctx->data, 1);
+	}
+
+	if (len - done >= SHA1_BLOCK_SIZE) {
+		const unsigned int rounds = (len - done) / SHA1_BLOCK_SIZE;
+		sha1_block_data_order(sctx, data + done, rounds);
+		done += rounds * SHA1_BLOCK_SIZE;
+	}
+
+	memcpy(sctx->data, data + done, len - done);
+	return 0;
+}
+
+
+static int sha1_update(struct shash_desc *desc, const u8 *data,
+			     unsigned int len)
+{
+	struct SHA1_CTX *sctx = shash_desc_ctx(desc);
+	unsigned int partial = sctx->count % SHA1_BLOCK_SIZE;
+	int res;
+
+	/* Handle the fast case right here */
+	if (partial + len < SHA1_BLOCK_SIZE) {
+		sctx->count += len;
+		memcpy(sctx->data + partial, data, len);
+		return 0;
+	}
+	res = __sha1_update(sctx, data, len, partial);
+	return res;
+}
+
+
+/* Add padding and return the message digest. */
+static int sha1_final(struct shash_desc *desc, u8 *out)
+{
+	struct SHA1_CTX *sctx = shash_desc_ctx(desc);
+	unsigned int i, index, padlen;
+	__be32 *dst = (__be32 *)out;
+	__be64 bits;
+	static const u8 padding[SHA1_BLOCK_SIZE] = { 0x80, };
+
+	bits = cpu_to_be64(sctx->count << 3);
+
+	/* Pad out to 56 mod 64 and append length */
+	index = sctx->count % SHA1_BLOCK_SIZE;
+	padlen = (index < 56) ? (56 - index) : ((SHA1_BLOCK_SIZE+56) - index);
+	/* We need to fill a whole block for __sha1_update() */
+	if (padlen <= 56) {
+		sctx->count += padlen;
+		memcpy(sctx->data + index, padding, padlen);
+	} else {
+		__sha1_update(sctx, padding, padlen, index);
+	}
+	__sha1_update(sctx, (const u8 *)&bits, sizeof(bits), 56);
+
+	/* Store state in digest */
+	for (i = 0; i < 5; i++)
+		dst[i] = cpu_to_be32(((u32 *)sctx)[i]);
+
+	/* Wipe context */
+	memset(sctx, 0, sizeof(*sctx));
+	return 0;
+}
+
+
+static int sha1_export(struct shash_desc *desc, void *out)
+{
+	struct SHA1_CTX *sctx = shash_desc_ctx(desc);
+	memcpy(out, sctx, sizeof(*sctx));
+	return 0;
+}
+
+
+static int sha1_import(struct shash_desc *desc, const void *in)
+{
+	struct SHA1_CTX *sctx = shash_desc_ctx(desc);
+	memcpy(sctx, in, sizeof(*sctx));
+	return 0;
+}
+
+
+static struct shash_alg alg = {
+	.digestsize	=	SHA1_DIGEST_SIZE,
+	.init		=	sha1_init,
+	.update		=	sha1_update,
+	.final		=	sha1_final,
+	.export		=	sha1_export,
+	.import		=	sha1_import,
+	.descsize	=	sizeof(struct SHA1_CTX),
+	.statesize	=	sizeof(struct SHA1_CTX),
+	.base		=	{
+		.cra_name	=	"sha1",
+		.cra_driver_name=	"sha1-asm",
+		.cra_priority	=	150,
+		.cra_flags	=	CRYPTO_ALG_TYPE_SHASH,
+		.cra_blocksize	=	SHA1_BLOCK_SIZE,
+		.cra_module	=	THIS_MODULE,
+	}
+};
+
+
+static int __init sha1_mod_init(void)
+{
+	return crypto_register_shash(&alg);
+}
+
+
+static void __exit sha1_mod_fini(void)
+{
+	crypto_unregister_shash(&alg);
+}
+
+
+module_init(sha1_mod_init);
+module_exit(sha1_mod_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("SHA1 Secure Hash Algorithm (ARM)");
+MODULE_ALIAS("sha1");
+MODULE_AUTHOR("David McCullough <ucdevel@gmail.com>");

crypto/Kconfig

@@ -433,6 +433,15 @@ config CRYPTO_SHA1_SSSE3
 	  using Supplemental SSE3 (SSSE3) instructions or Advanced Vector
 	  Extensions (AVX), when available.
 
+config CRYPTO_SHA1_ARM
+	tristate "SHA1 digest algorithm (ARM-asm)"
+	depends on ARM
+	select CRYPTO_SHA1
+	select CRYPTO_HASH
+	help
+	  SHA-1 secure hash standard (FIPS 180-1/DFIPS 180-2) implemented
+	  using optimized ARM assembler.
+
 config CRYPTO_SHA256
 	tristate "SHA224 and SHA256 digest algorithm"
 	select CRYPTO_HASH
@@ -590,6 +599,30 @@ config CRYPTO_AES_NI_INTEL
 	  ECB, CBC, LRW, PCBC, XTS. The 64 bit version has additional
 	  acceleration for CTR.
 
+config CRYPTO_AES_ARM
+	tristate "AES cipher algorithms (ARM-asm)"
+	depends on ARM
+	select CRYPTO_ALGAPI
+	select CRYPTO_AES
+	help
+	  Use optimized AES assembler routines for ARM platforms.
+
+	  AES cipher algorithms (FIPS-197). AES uses the Rijndael
+	  algorithm.
+
+	  Rijndael appears to be consistently a very good performer in
+	  both hardware and software across a wide range of computing
+	  environments regardless of its use in feedback or non-feedback
+	  modes. Its key setup time is excellent, and its key agility is
+	  good. Rijndael's very low memory requirements make it very well
+	  suited for restricted-space environments, in which it also
+	  demonstrates excellent performance. Rijndael's operations are
+	  among the easiest to defend against power and timing attacks.
+
+	  The AES specifies three key sizes: 128, 192 and 256 bits
+
+	  See <http://csrc.nist.gov/encryption/aes/> for more information.
+
 config CRYPTO_ANUBIS
 	tristate "Anubis cipher algorithm"
 	select CRYPTO_ALGAPI