crypto: des_3des - add x86-64 assembly implementation

Patch adds x86_64 assembly implementation of Triple DES EDE cipher algorithm.
Two assembly implementations are provided. First is regular 'one-block at
time' encrypt/decrypt function. Second is 'three-blocks at time' function that
gains performance increase on out-of-order CPUs.

tcrypt test results:

Intel Core i5-4570:

des3_ede-asm vs des3_ede-generic:
size    ecb-enc ecb-dec cbc-enc cbc-dec ctr-enc ctr-dec
16B     1.21x   1.22x   1.27x   1.36x   1.25x   1.25x
64B     1.98x   1.96x   1.23x   2.04x   2.01x   2.00x
256B    2.34x   2.37x   1.21x   2.40x   2.38x   2.39x
1024B   2.50x   2.47x   1.22x   2.51x   2.52x   2.51x
8192B   2.51x   2.53x   1.21x   2.56x   2.54x   2.55x
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

arch/x86/crypto/Makefile

@@ -14,6 +14,7 @@ obj-$(CONFIG_CRYPTO_SALSA20_586) += salsa20-i586.o
 obj-$(CONFIG_CRYPTO_SERPENT_SSE2_586) += serpent-sse2-i586.o
 
 obj-$(CONFIG_CRYPTO_AES_X86_64) += aes-x86_64.o
+obj-$(CONFIG_CRYPTO_DES3_EDE_X86_64) += des3_ede-x86_64.o
 obj-$(CONFIG_CRYPTO_CAMELLIA_X86_64) += camellia-x86_64.o
 obj-$(CONFIG_CRYPTO_BLOWFISH_X86_64) += blowfish-x86_64.o
 obj-$(CONFIG_CRYPTO_TWOFISH_X86_64) += twofish-x86_64.o
@@ -52,6 +53,7 @@ salsa20-i586-y := salsa20-i586-asm_32.o salsa20_glue.o
 serpent-sse2-i586-y := serpent-sse2-i586-asm_32.o serpent_sse2_glue.o
 
 aes-x86_64-y := aes-x86_64-asm_64.o aes_glue.o
+des3_ede-x86_64-y := des3_ede-asm_64.o des3_ede_glue.o
 camellia-x86_64-y := camellia-x86_64-asm_64.o camellia_glue.o
 blowfish-x86_64-y := blowfish-x86_64-asm_64.o blowfish_glue.o
 twofish-x86_64-y := twofish-x86_64-asm_64.o twofish_glue.o

crypto/Kconfig

@@ -1019,6 +1019,19 @@ config CRYPTO_DES_SPARC64
 	  DES cipher algorithm (FIPS 46-2), and Triple DES EDE (FIPS 46-3),
 	  optimized using SPARC64 crypto opcodes.
 
+config CRYPTO_DES3_EDE_X86_64
+	tristate "Triple DES EDE cipher algorithm (x86-64)"
+	depends on X86 && 64BIT
+	select CRYPTO_ALGAPI
+	select CRYPTO_DES
+	help
+	  Triple DES EDE (FIPS 46-3) algorithm.
+
+	  This module provides implementation of the Triple DES EDE cipher
+	  algorithm that is optimized for x86-64 processors. Two versions of
+	  algorithm are provided; regular processing one input block and
+	  one that processes three blocks parallel.
+
 config CRYPTO_FCRYPT
 	tristate "FCrypt cipher algorithm"
 	select CRYPTO_ALGAPI

crypto/des_generic.c

@@ -859,13 +859,10 @@ static void des_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
  *   property.
  *
  */
-static int des3_ede_setkey(struct crypto_tfm *tfm, const u8 *key,
+int __des3_ede_setkey(u32 *expkey, u32 *flags, const u8 *key,
 		      unsigned int keylen)
 {
 	const u32 *K = (const u32 *)key;
-	struct des3_ede_ctx *dctx = crypto_tfm_ctx(tfm);
-	u32 *expkey = dctx->expkey;
-	u32 *flags = &tfm->crt_flags;
 
 	if (unlikely(!((K[0] ^ K[2]) | (K[1] ^ K[3])) ||
 		     !((K[2] ^ K[4]) | (K[3] ^ K[5]))) &&
@@ -880,6 +877,17 @@ static int des3_ede_setkey(struct crypto_tfm *tfm, const u8 *key,
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(__des3_ede_setkey);
+
+static int des3_ede_setkey(struct crypto_tfm *tfm, const u8 *key,
+			   unsigned int keylen)
+{
+	struct des3_ede_ctx *dctx = crypto_tfm_ctx(tfm);
+	u32 *flags = &tfm->crt_flags;
+	u32 *expkey = dctx->expkey;
+
+	return __des3_ede_setkey(expkey, flags, key, keylen);
+}
 
 static void des3_ede_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
 {
@@ -945,6 +953,8 @@ static void des3_ede_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
 
 static struct crypto_alg des_algs[2] = { {
 	.cra_name		=	"des",
+	.cra_driver_name	=	"des-generic",
+	.cra_priority		=	100,
 	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
 	.cra_blocksize		=	DES_BLOCK_SIZE,
 	.cra_ctxsize		=	sizeof(struct des_ctx),
@@ -958,6 +968,8 @@ static struct crypto_alg des_algs[2] = { {
 	.cia_decrypt		=	des_decrypt } }
 }, {
 	.cra_name		=	"des3_ede",
+	.cra_driver_name	=	"des3_ede-generic",
+	.cra_priority		=	100,
 	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
 	.cra_blocksize		=	DES3_EDE_BLOCK_SIZE,
 	.cra_ctxsize		=	sizeof(struct des3_ede_ctx),

include/crypto/des.h

@@ -16,4 +16,7 @@
 
 extern unsigned long des_ekey(u32 *pe, const u8 *k);
 
+extern int __des3_ede_setkey(u32 *expkey, u32 *flags, const u8 *key,
+			     unsigned int keylen);
+
 #endif /* __CRYPTO_DES_H */