i386: move crypto

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

6 files changed, 1337 insertions, 0 deletions

diff --git a/arch/x86/crypto/Makefile b/arch/x86/crypto/Makefile
new file mode 100644
index 0000000..b1bcf7c63
--- /dev/null
+++ b/arch/x86/crypto/Makefile
@@ -0,0 +1,5 @@
+ifeq ($(CONFIG_X86_32),y)
+include ${srctree}/arch/x86/crypto/Makefile_32
+else
+include ${srctree}/arch/x86_64/crypto/Makefile_64
+endif
diff --git a/arch/x86/crypto/Makefile_32 b/arch/x86/crypto/Makefile_32
new file mode 100644
index 0000000..2d873a2
--- /dev/null
+++ b/arch/x86/crypto/Makefile_32
@@ -0,0 +1,12 @@
+# 
+# x86/crypto/Makefile 
+# 
+# Arch-specific CryptoAPI modules.
+# 
+
+obj-$(CONFIG_CRYPTO_AES_586) += aes-i586.o
+obj-$(CONFIG_CRYPTO_TWOFISH_586) += twofish-i586.o
+
+aes-i586-y := aes-i586-asm_32.o aes_32.o
+twofish-i586-y := twofish-i586-asm_32.o twofish_32.o
+
diff --git a/arch/x86/crypto/aes-i586-asm_32.S b/arch/x86/crypto/aes-i586-asm_32.S
new file mode 100644
index 0000000..f942f0c
--- /dev/null
+++ b/arch/x86/crypto/aes-i586-asm_32.S
@@ -0,0 +1,373 @@
+// -------------------------------------------------------------------------
+// Copyright (c) 2001, Dr Brian Gladman <                 >, Worcester, UK.
+// All rights reserved.
+//
+// LICENSE TERMS
+//
+// The free distribution and use of this software in both source and binary 
+// form is allowed (with or without changes) provided that:
+//
+//   1. distributions of this source code include the above copyright 
+//      notice, this list of conditions and the following disclaimer//
+//
+//   2. distributions in binary form include the above copyright
+//      notice, this list of conditions and the following disclaimer
+//      in the documentation and/or other associated materials//
+//
+//   3. the copyright holder's name is not used to endorse products 
+//      built using this software without specific written permission.
+//
+//
+// ALTERNATIVELY, provided that this notice is retained in full, this product
+// may be distributed under the terms of the GNU General Public License (GPL),
+// in which case the provisions of the GPL apply INSTEAD OF those given above.
+//
+// Copyright (c) 2004 Linus Torvalds <torvalds@osdl.org>
+// Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
+
+// DISCLAIMER
+//
+// This software is provided 'as is' with no explicit or implied warranties
+// in respect of its properties including, but not limited to, correctness 
+// and fitness for purpose.
+// -------------------------------------------------------------------------
+// Issue Date: 29/07/2002
+
+.file "aes-i586-asm.S"
+.text
+
+#include <asm/asm-offsets.h>
+
+#define tlen 1024   // length of each of 4 'xor' arrays (256 32-bit words)
+
+/* offsets to parameters with one register pushed onto stack */
+#define tfm 8
+#define out_blk 12
+#define in_blk 16
+
+/* offsets in crypto_tfm structure */
+#define ekey (crypto_tfm_ctx_offset + 0)
+#define nrnd (crypto_tfm_ctx_offset + 256)
+#define dkey (crypto_tfm_ctx_offset + 260)
+
+// register mapping for encrypt and decrypt subroutines
+
+#define r0  eax
+#define r1  ebx
+#define r2  ecx
+#define r3  edx
+#define r4  esi
+#define r5  edi
+
+#define eaxl  al
+#define eaxh  ah
+#define ebxl  bl
+#define ebxh  bh
+#define ecxl  cl
+#define ecxh  ch
+#define edxl  dl
+#define edxh  dh
+
+#define _h(reg) reg##h
+#define h(reg) _h(reg)
+
+#define _l(reg) reg##l
+#define l(reg) _l(reg)
+
+// This macro takes a 32-bit word representing a column and uses
+// each of its four bytes to index into four tables of 256 32-bit
+// words to obtain values that are then xored into the appropriate
+// output registers r0, r1, r4 or r5.  
+
+// Parameters:
+// table table base address
+//   %1  out_state[0]
+//   %2  out_state[1]
+//   %3  out_state[2]
+//   %4  out_state[3]
+//   idx input register for the round (destroyed)
+//   tmp scratch register for the round
+// sched key schedule
+
+#define do_col(table, a1,a2,a3,a4, idx, tmp)	\
+	movzx   %l(idx),%tmp;			\
+	xor     table(,%tmp,4),%a1;		\
+	movzx   %h(idx),%tmp;			\
+	shr     $16,%idx;			\
+	xor     table+tlen(,%tmp,4),%a2;	\
+	movzx   %l(idx),%tmp;			\
+	movzx   %h(idx),%idx;			\
+	xor     table+2*tlen(,%tmp,4),%a3;	\
+	xor     table+3*tlen(,%idx,4),%a4;
+
+// initialise output registers from the key schedule
+// NB1: original value of a3 is in idx on exit
+// NB2: original values of a1,a2,a4 aren't used
+#define do_fcol(table, a1,a2,a3,a4, idx, tmp, sched) \
+	mov     0 sched,%a1;			\
+	movzx   %l(idx),%tmp;			\
+	mov     12 sched,%a2;			\
+	xor     table(,%tmp,4),%a1;		\
+	mov     4 sched,%a4;			\
+	movzx   %h(idx),%tmp;			\
+	shr     $16,%idx;			\
+	xor     table+tlen(,%tmp,4),%a2;	\
+	movzx   %l(idx),%tmp;			\
+	movzx   %h(idx),%idx;			\
+	xor     table+3*tlen(,%idx,4),%a4;	\
+	mov     %a3,%idx;			\
+	mov     8 sched,%a3;			\
+	xor     table+2*tlen(,%tmp,4),%a3;
+
+// initialise output registers from the key schedule
+// NB1: original value of a3 is in idx on exit
+// NB2: original values of a1,a2,a4 aren't used
+#define do_icol(table, a1,a2,a3,a4, idx, tmp, sched) \
+	mov     0 sched,%a1;			\
+	movzx   %l(idx),%tmp;			\
+	mov     4 sched,%a2;			\
+	xor     table(,%tmp,4),%a1;		\
+	mov     12 sched,%a4;			\
+	movzx   %h(idx),%tmp;			\
+	shr     $16,%idx;			\
+	xor     table+tlen(,%tmp,4),%a2;	\
+	movzx   %l(idx),%tmp;			\
+	movzx   %h(idx),%idx;			\
+	xor     table+3*tlen(,%idx,4),%a4;	\
+	mov     %a3,%idx;			\
+	mov     8 sched,%a3;			\
+	xor     table+2*tlen(,%tmp,4),%a3;
+
+
+// original Gladman had conditional saves to MMX regs.
+#define save(a1, a2)		\
+	mov     %a2,4*a1(%esp)
+
+#define restore(a1, a2)		\
+	mov     4*a2(%esp),%a1
+
+// These macros perform a forward encryption cycle. They are entered with
+// the first previous round column values in r0,r1,r4,r5 and
+// exit with the final values in the same registers, using stack
+// for temporary storage.
+
+// round column values
+// on entry: r0,r1,r4,r5
+// on exit:  r2,r1,r4,r5
+#define fwd_rnd1(arg, table)						\
+	save   (0,r1);							\
+	save   (1,r5);							\
+									\
+	/* compute new column values */					\
+	do_fcol(table, r2,r5,r4,r1, r0,r3, arg);	/* idx=r0 */	\
+	do_col (table, r4,r1,r2,r5, r0,r3);		/* idx=r4 */	\
+	restore(r0,0);							\
+	do_col (table, r1,r2,r5,r4, r0,r3);		/* idx=r1 */	\
+	restore(r0,1);							\
+	do_col (table, r5,r4,r1,r2, r0,r3);		/* idx=r5 */
+
+// round column values
+// on entry: r2,r1,r4,r5
+// on exit:  r0,r1,r4,r5
+#define fwd_rnd2(arg, table)						\
+	save   (0,r1);							\
+	save   (1,r5);							\
+									\
+	/* compute new column values */					\
+	do_fcol(table, r0,r5,r4,r1, r2,r3, arg);	/* idx=r2 */	\
+	do_col (table, r4,r1,r0,r5, r2,r3);		/* idx=r4 */	\
+	restore(r2,0);							\
+	do_col (table, r1,r0,r5,r4, r2,r3);		/* idx=r1 */	\
+	restore(r2,1);							\
+	do_col (table, r5,r4,r1,r0, r2,r3);		/* idx=r5 */
+
+// These macros performs an inverse encryption cycle. They are entered with
+// the first previous round column values in r0,r1,r4,r5 and
+// exit with the final values in the same registers, using stack
+// for temporary storage
+
+// round column values
+// on entry: r0,r1,r4,r5
+// on exit:  r2,r1,r4,r5
+#define inv_rnd1(arg, table)						\
+	save    (0,r1);							\
+	save    (1,r5);							\
+									\
+	/* compute new column values */					\
+	do_icol(table, r2,r1,r4,r5, r0,r3, arg);	/* idx=r0 */	\
+	do_col (table, r4,r5,r2,r1, r0,r3);		/* idx=r4 */	\
+	restore(r0,0);							\
+	do_col (table, r1,r4,r5,r2, r0,r3);		/* idx=r1 */	\
+	restore(r0,1);							\
+	do_col (table, r5,r2,r1,r4, r0,r3);		/* idx=r5 */
+
+// round column values
+// on entry: r2,r1,r4,r5
+// on exit:  r0,r1,r4,r5
+#define inv_rnd2(arg, table)						\
+	save    (0,r1);							\
+	save    (1,r5);							\
+									\
+	/* compute new column values */					\
+	do_icol(table, r0,r1,r4,r5, r2,r3, arg);	/* idx=r2 */	\
+	do_col (table, r4,r5,r0,r1, r2,r3);		/* idx=r4 */	\
+	restore(r2,0);							\
+	do_col (table, r1,r4,r5,r0, r2,r3);		/* idx=r1 */	\
+	restore(r2,1);							\
+	do_col (table, r5,r0,r1,r4, r2,r3);		/* idx=r5 */
+
+// AES (Rijndael) Encryption Subroutine
+/* void aes_enc_blk(struct crypto_tfm *tfm, u8 *out_blk, const u8 *in_blk) */
+
+.global  aes_enc_blk
+
+.extern  ft_tab
+.extern  fl_tab
+
+.align 4
+
+aes_enc_blk:
+	push    %ebp
+	mov     tfm(%esp),%ebp
+
+// CAUTION: the order and the values used in these assigns 
+// rely on the register mappings
+
+1:	push    %ebx
+	mov     in_blk+4(%esp),%r2
+	push    %esi
+	mov     nrnd(%ebp),%r3   // number of rounds
+	push    %edi
+#if ekey != 0
+	lea     ekey(%ebp),%ebp  // key pointer
+#endif
+
+// input four columns and xor in first round key
+
+	mov     (%r2),%r0
+	mov     4(%r2),%r1
+	mov     8(%r2),%r4
+	mov     12(%r2),%r5
+	xor     (%ebp),%r0
+	xor     4(%ebp),%r1
+	xor     8(%ebp),%r4
+	xor     12(%ebp),%r5
+
+	sub     $8,%esp		// space for register saves on stack
+	add     $16,%ebp	// increment to next round key
+	cmp     $12,%r3
+	jb      4f		// 10 rounds for 128-bit key
+	lea     32(%ebp),%ebp
+	je      3f		// 12 rounds for 192-bit key
+	lea     32(%ebp),%ebp
+
+2:	fwd_rnd1( -64(%ebp) ,ft_tab)	// 14 rounds for 256-bit key
+	fwd_rnd2( -48(%ebp) ,ft_tab)
+3:	fwd_rnd1( -32(%ebp) ,ft_tab)	// 12 rounds for 192-bit key
+	fwd_rnd2( -16(%ebp) ,ft_tab)
+4:	fwd_rnd1(    (%ebp) ,ft_tab)	// 10 rounds for 128-bit key
+	fwd_rnd2( +16(%ebp) ,ft_tab)
+	fwd_rnd1( +32(%ebp) ,ft_tab)
+	fwd_rnd2( +48(%ebp) ,ft_tab)
+	fwd_rnd1( +64(%ebp) ,ft_tab)
+	fwd_rnd2( +80(%ebp) ,ft_tab)
+	fwd_rnd1( +96(%ebp) ,ft_tab)
+	fwd_rnd2(+112(%ebp) ,ft_tab)
+	fwd_rnd1(+128(%ebp) ,ft_tab)
+	fwd_rnd2(+144(%ebp) ,fl_tab)	// last round uses a different table
+
+// move final values to the output array.  CAUTION: the 
+// order of these assigns rely on the register mappings
+
+	add     $8,%esp
+	mov     out_blk+12(%esp),%ebp
+	mov     %r5,12(%ebp)
+	pop     %edi
+	mov     %r4,8(%ebp)
+	pop     %esi
+	mov     %r1,4(%ebp)
+	pop     %ebx
+	mov     %r0,(%ebp)
+	pop     %ebp
+	mov     $1,%eax
+	ret
+
+// AES (Rijndael) Decryption Subroutine
+/* void aes_dec_blk(struct crypto_tfm *tfm, u8 *out_blk, const u8 *in_blk) */
+
+.global  aes_dec_blk
+
+.extern  it_tab
+.extern  il_tab
+
+.align 4
+
+aes_dec_blk:
+	push    %ebp
+	mov     tfm(%esp),%ebp
+
+// CAUTION: the order and the values used in these assigns 
+// rely on the register mappings
+
+1:	push    %ebx
+	mov     in_blk+4(%esp),%r2
+	push    %esi
+	mov     nrnd(%ebp),%r3   // number of rounds
+	push    %edi
+#if dkey != 0
+	lea     dkey(%ebp),%ebp  // key pointer
+#endif
+	mov     %r3,%r0
+	shl     $4,%r0
+	add     %r0,%ebp
+	
+// input four columns and xor in first round key
+
+	mov     (%r2),%r0
+	mov     4(%r2),%r1
+	mov     8(%r2),%r4
+	mov     12(%r2),%r5
+	xor     (%ebp),%r0
+	xor     4(%ebp),%r1
+	xor     8(%ebp),%r4
+	xor     12(%ebp),%r5
+
+	sub     $8,%esp		// space for register saves on stack
+	sub     $16,%ebp	// increment to next round key
+	cmp     $12,%r3
+	jb      4f		// 10 rounds for 128-bit key
+	lea     -32(%ebp),%ebp
+	je      3f		// 12 rounds for 192-bit key
+	lea     -32(%ebp),%ebp
+
+2:	inv_rnd1( +64(%ebp), it_tab)	// 14 rounds for 256-bit key
+	inv_rnd2( +48(%ebp), it_tab)
+3:	inv_rnd1( +32(%ebp), it_tab)	// 12 rounds for 192-bit key
+	inv_rnd2( +16(%ebp), it_tab)
+4:	inv_rnd1(    (%ebp), it_tab)	// 10 rounds for 128-bit key
+	inv_rnd2( -16(%ebp), it_tab)
+	inv_rnd1( -32(%ebp), it_tab)
+	inv_rnd2( -48(%ebp), it_tab)
+	inv_rnd1( -64(%ebp), it_tab)
+	inv_rnd2( -80(%ebp), it_tab)
+	inv_rnd1( -96(%ebp), it_tab)
+	inv_rnd2(-112(%ebp), it_tab)
+	inv_rnd1(-128(%ebp), it_tab)
+	inv_rnd2(-144(%ebp), il_tab)	// last round uses a different table
+
+// move final values to the output array.  CAUTION: the 
+// order of these assigns rely on the register mappings
+
+	add     $8,%esp
+	mov     out_blk+12(%esp),%ebp
+	mov     %r5,12(%ebp)
+	pop     %edi
+	mov     %r4,8(%ebp)
+	pop     %esi
+	mov     %r1,4(%ebp)
+	pop     %ebx
+	mov     %r0,(%ebp)
+	pop     %ebp
+	mov     $1,%eax
+	ret
+
diff --git a/arch/x86/crypto/aes_32.c b/arch/x86/crypto/aes_32.c
new file mode 100644
index 0000000..49aad93
--- /dev/null
+++ b/arch/x86/crypto/aes_32.c
@@ -0,0 +1,515 @@
+/* 
+ * 
+ * Glue Code for optimized 586 assembler version of AES
+ *
+ * Copyright (c) 2002, Dr Brian Gladman <>, Worcester, UK.
+ * All rights reserved.
+ *
+ * LICENSE TERMS
+ *
+ * The free distribution and use of this software in both source and binary
+ * form is allowed (with or without changes) provided that:
+ *
+ *   1. distributions of this source code include the above copyright
+ *      notice, this list of conditions and the following disclaimer;
+ *
+ *   2. distributions in binary form include the above copyright
+ *      notice, this list of conditions and the following disclaimer
+ *      in the documentation and/or other associated materials;
+ *
+ *   3. the copyright holder's name is not used to endorse products
+ *      built using this software without specific written permission.
+ *
+ * ALTERNATIVELY, provided that this notice is retained in full, this product
+ * may be distributed under the terms of the GNU General Public License (GPL),
+ * in which case the provisions of the GPL apply INSTEAD OF those given above.
+ *
+ * DISCLAIMER
+ *
+ * This software is provided 'as is' with no explicit or implied warranties
+ * in respect of its properties, including, but not limited to, correctness
+ * and/or fitness for purpose.
+ *
+ * Copyright (c) 2003, Adam J. Richter <adam@yggdrasil.com> (conversion to
+ * 2.5 API).
+ * Copyright (c) 2003, 2004 Fruhwirth Clemens <clemens@endorphin.org>
+ * Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
+ *
+ */
+
+#include <asm/byteorder.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/crypto.h>
+#include <linux/linkage.h>
+
+asmlinkage void aes_enc_blk(struct crypto_tfm *tfm, u8 *dst, const u8 *src);
+asmlinkage void aes_dec_blk(struct crypto_tfm *tfm, u8 *dst, const u8 *src);
+
+#define AES_MIN_KEY_SIZE	16
+#define AES_MAX_KEY_SIZE	32
+#define AES_BLOCK_SIZE		16
+#define AES_KS_LENGTH		4 * AES_BLOCK_SIZE
+#define RC_LENGTH		29
+
+struct aes_ctx {
+	u32 ekey[AES_KS_LENGTH];
+	u32 rounds;
+	u32 dkey[AES_KS_LENGTH];
+};
+
+#define WPOLY 0x011b
+#define bytes2word(b0, b1, b2, b3)  \
+	(((u32)(b3) << 24) | ((u32)(b2) << 16) | ((u32)(b1) << 8) | (b0))
+
+/* define the finite field multiplies required for Rijndael */
+#define f2(x) ((x) ? pow[log[x] + 0x19] : 0)
+#define f3(x) ((x) ? pow[log[x] + 0x01] : 0)
+#define f9(x) ((x) ? pow[log[x] + 0xc7] : 0)
+#define fb(x) ((x) ? pow[log[x] + 0x68] : 0)
+#define fd(x) ((x) ? pow[log[x] + 0xee] : 0)
+#define fe(x) ((x) ? pow[log[x] + 0xdf] : 0)
+#define fi(x) ((x) ?   pow[255 - log[x]]: 0)
+
+static inline u32 upr(u32 x, int n)
+{
+	return (x << 8 * n) | (x >> (32 - 8 * n));
+}
+
+static inline u8 bval(u32 x, int n)
+{
+	return x >> 8 * n;
+}
+
+/* The forward and inverse affine transformations used in the S-box */
+#define fwd_affine(x) \
+	(w = (u32)x, w ^= (w<<1)^(w<<2)^(w<<3)^(w<<4), 0x63^(u8)(w^(w>>8)))
+
+#define inv_affine(x) \
+	(w = (u32)x, w = (w<<1)^(w<<3)^(w<<6), 0x05^(u8)(w^(w>>8)))
+
+static u32 rcon_tab[RC_LENGTH];
+
+u32 ft_tab[4][256];
+u32 fl_tab[4][256];
+static u32 im_tab[4][256];
+u32 il_tab[4][256];
+u32 it_tab[4][256];
+
+static void gen_tabs(void)
+{
+	u32 i, w;
+	u8 pow[512], log[256];
+
+	/*
+	 * log and power tables for GF(2^8) finite field with
+	 * WPOLY as modular polynomial - the simplest primitive
+	 * root is 0x03, used here to generate the tables.
+	 */
+	i = 0; w = 1; 
+	
+	do {
+		pow[i] = (u8)w;
+		pow[i + 255] = (u8)w;
+		log[w] = (u8)i++;
+		w ^=  (w << 1) ^ (w & 0x80 ? WPOLY : 0);
+	} while (w != 1);
+	
+	for(i = 0, w = 1; i < RC_LENGTH; ++i) {
+		rcon_tab[i] = bytes2word(w, 0, 0, 0);
+		w = f2(w);
+	}
+
+	for(i = 0; i < 256; ++i) {
+		u8 b;
+		
+		b = fwd_affine(fi((u8)i));
+		w = bytes2word(f2(b), b, b, f3(b));
+
+		/* tables for a normal encryption round */
+		ft_tab[0][i] = w;
+		ft_tab[1][i] = upr(w, 1);
+		ft_tab[2][i] = upr(w, 2);
+		ft_tab[3][i] = upr(w, 3);
+		w = bytes2word(b, 0, 0, 0);
+		
+		/*
+		 * tables for last encryption round
+		 * (may also be used in the key schedule)
+		 */
+		fl_tab[0][i] = w;
+		fl_tab[1][i] = upr(w, 1);
+		fl_tab[2][i] = upr(w, 2);
+		fl_tab[3][i] = upr(w, 3);
+		
+		b = fi(inv_affine((u8)i));
+		w = bytes2word(fe(b), f9(b), fd(b), fb(b));
+
+		/* tables for the inverse mix column operation  */
+		im_tab[0][b] = w;
+		im_tab[1][b] = upr(w, 1);
+		im_tab[2][b] = upr(w, 2);
+		im_tab[3][b] = upr(w, 3);
+
+		/* tables for a normal decryption round */
+		it_tab[0][i] = w;
+		it_tab[1][i] = upr(w,1);
+		it_tab[2][i] = upr(w,2);
+		it_tab[3][i] = upr(w,3);
+
+		w = bytes2word(b, 0, 0, 0);
+		
+		/* tables for last decryption round */
+		il_tab[0][i] = w;
+		il_tab[1][i] = upr(w,1);
+		il_tab[2][i] = upr(w,2);
+		il_tab[3][i] = upr(w,3);
+    }
+}
+
+#define four_tables(x,tab,vf,rf,c)		\
+(	tab[0][bval(vf(x,0,c),rf(0,c))]	^	\
+	tab[1][bval(vf(x,1,c),rf(1,c))] ^	\
+	tab[2][bval(vf(x,2,c),rf(2,c))] ^	\
+	tab[3][bval(vf(x,3,c),rf(3,c))]		\
+)
+
+#define vf1(x,r,c)  (x)
+#define rf1(r,c)    (r)
+#define rf2(r,c)    ((r-c)&3)
+
+#define inv_mcol(x) four_tables(x,im_tab,vf1,rf1,0)
+#define ls_box(x,c) four_tables(x,fl_tab,vf1,rf2,c)
+
+#define ff(x) inv_mcol(x)
+
+#define ke4(k,i)							\
+{									\
+	k[4*(i)+4] = ss[0] ^= ls_box(ss[3],3) ^ rcon_tab[i];		\
+	k[4*(i)+5] = ss[1] ^= ss[0];					\
+	k[4*(i)+6] = ss[2] ^= ss[1];					\
+	k[4*(i)+7] = ss[3] ^= ss[2];					\
+}
+
+#define kel4(k,i)							\
+{									\
+	k[4*(i)+4] = ss[0] ^= ls_box(ss[3],3) ^ rcon_tab[i];		\
+	k[4*(i)+5] = ss[1] ^= ss[0];					\
+	k[4*(i)+6] = ss[2] ^= ss[1]; k[4*(i)+7] = ss[3] ^= ss[2];	\
+}
+
+#define ke6(k,i)							\
+{									\
+	k[6*(i)+ 6] = ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i];		\
+	k[6*(i)+ 7] = ss[1] ^= ss[0];					\
+	k[6*(i)+ 8] = ss[2] ^= ss[1];					\
+	k[6*(i)+ 9] = ss[3] ^= ss[2];					\
+	k[6*(i)+10] = ss[4] ^= ss[3];					\
+	k[6*(i)+11] = ss[5] ^= ss[4];					\
+}
+
+#define kel6(k,i)							\
+{									\
+	k[6*(i)+ 6] = ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i];		\
+	k[6*(i)+ 7] = ss[1] ^= ss[0];					\
+	k[6*(i)+ 8] = ss[2] ^= ss[1];					\
+	k[6*(i)+ 9] = ss[3] ^= ss[2];					\
+}
+
+#define ke8(k,i)							\
+{									\
+	k[8*(i)+ 8] = ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i];		\
+	k[8*(i)+ 9] = ss[1] ^= ss[0];					\
+	k[8*(i)+10] = ss[2] ^= ss[1];					\
+	k[8*(i)+11] = ss[3] ^= ss[2];					\
+	k[8*(i)+12] = ss[4] ^= ls_box(ss[3],0);				\
+	k[8*(i)+13] = ss[5] ^= ss[4];					\
+	k[8*(i)+14] = ss[6] ^= ss[5];					\
+	k[8*(i)+15] = ss[7] ^= ss[6];					\
+}
+
+#define kel8(k,i)							\
+{									\
+	k[8*(i)+ 8] = ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i];		\
+	k[8*(i)+ 9] = ss[1] ^= ss[0];					\
+	k[8*(i)+10] = ss[2] ^= ss[1];					\
+	k[8*(i)+11] = ss[3] ^= ss[2];					\
+}
+
+#define kdf4(k,i)							\
+{									\
+	ss[0] = ss[0] ^ ss[2] ^ ss[1] ^ ss[3];				\
+	ss[1] = ss[1] ^ ss[3];						\
+	ss[2] = ss[2] ^ ss[3];						\
+	ss[3] = ss[3];							\
+	ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i];			\
+	ss[i % 4] ^= ss[4];						\
+	ss[4] ^= k[4*(i)];						\
+	k[4*(i)+4] = ff(ss[4]);						\
+	ss[4] ^= k[4*(i)+1];						\
+	k[4*(i)+5] = ff(ss[4]);						\
+	ss[4] ^= k[4*(i)+2];						\
+	k[4*(i)+6] = ff(ss[4]);						\
+	ss[4] ^= k[4*(i)+3];						\
+	k[4*(i)+7] = ff(ss[4]);						\
+}
+
+#define kd4(k,i)							\
+{									\
+	ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i];			\
+	ss[i % 4] ^= ss[4];						\
+	ss[4] = ff(ss[4]);						\
+	k[4*(i)+4] = ss[4] ^= k[4*(i)];					\
+	k[4*(i)+5] = ss[4] ^= k[4*(i)+1];				\
+	k[4*(i)+6] = ss[4] ^= k[4*(i)+2];				\
+	k[4*(i)+7] = ss[4] ^= k[4*(i)+3];				\
+}
+
+#define kdl4(k,i)							\
+{									\
+	ss[4] = ls_box(ss[(i+3) % 4], 3) ^ rcon_tab[i];			\
+	ss[i % 4] ^= ss[4];						\
+	k[4*(i)+4] = (ss[0] ^= ss[1]) ^ ss[2] ^ ss[3];			\
+	k[4*(i)+5] = ss[1] ^ ss[3];					\
+	k[4*(i)+6] = ss[0];						\
+	k[4*(i)+7] = ss[1];						\
+}
+
+#define kdf6(k,i)							\
+{									\
+	ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i];				\
+	k[6*(i)+ 6] = ff(ss[0]);					\
+	ss[1] ^= ss[0];							\
+	k[6*(i)+ 7] = ff(ss[1]);					\
+	ss[2] ^= ss[1];							\
+	k[6*(i)+ 8] = ff(ss[2]);					\
+	ss[3] ^= ss[2];							\
+	k[6*(i)+ 9] = ff(ss[3]);					\
+	ss[4] ^= ss[3];							\
+	k[6*(i)+10] = ff(ss[4]);					\
+	ss[5] ^= ss[4];							\
+	k[6*(i)+11] = ff(ss[5]);					\
+}
+
+#define kd6(k,i)							\
+{									\
+	ss[6] = ls_box(ss[5],3) ^ rcon_tab[i];				\
+	ss[0] ^= ss[6]; ss[6] = ff(ss[6]);				\
+	k[6*(i)+ 6] = ss[6] ^= k[6*(i)];				\
+	ss[1] ^= ss[0];							\
+	k[6*(i)+ 7] = ss[6] ^= k[6*(i)+ 1];				\
+	ss[2] ^= ss[1];							\
+	k[6*(i)+ 8] = ss[6] ^= k[6*(i)+ 2];				\
+	ss[3] ^= ss[2];							\
+	k[6*(i)+ 9] = ss[6] ^= k[6*(i)+ 3];				\
+	ss[4] ^= ss[3];							\
+	k[6*(i)+10] = ss[6] ^= k[6*(i)+ 4];				\
+	ss[5] ^= ss[4];							\
+	k[6*(i)+11] = ss[6] ^= k[6*(i)+ 5];				\
+}
+
+#define kdl6(k,i)							\
+{									\
+	ss[0] ^= ls_box(ss[5],3) ^ rcon_tab[i];				\
+	k[6*(i)+ 6] = ss[0];						\
+	ss[1] ^= ss[0];							\
+	k[6*(i)+ 7] = ss[1];						\
+	ss[2] ^= ss[1];							\
+	k[6*(i)+ 8] = ss[2];						\
+	ss[3] ^= ss[2];							\
+	k[6*(i)+ 9] = ss[3];						\
+}
+
+#define kdf8(k,i)							\
+{									\
+	ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i];				\
+	k[8*(i)+ 8] = ff(ss[0]);					\
+	ss[1] ^= ss[0];							\
+	k[8*(i)+ 9] = ff(ss[1]);					\
+	ss[2] ^= ss[1];							\
+	k[8*(i)+10] = ff(ss[2]);					\
+	ss[3] ^= ss[2];							\
+	k[8*(i)+11] = ff(ss[3]);					\
+	ss[4] ^= ls_box(ss[3],0);					\
+	k[8*(i)+12] = ff(ss[4]);					\
+	ss[5] ^= ss[4];							\
+	k[8*(i)+13] = ff(ss[5]);					\
+	ss[6] ^= ss[5];							\
+	k[8*(i)+14] = ff(ss[6]);					\
+	ss[7] ^= ss[6];							\
+	k[8*(i)+15] = ff(ss[7]);					\
+}
+
+#define kd8(k,i)							\
+{									\
+	u32 __g = ls_box(ss[7],3) ^ rcon_tab[i];			\
+	ss[0] ^= __g;							\
+	__g = ff(__g);							\
+	k[8*(i)+ 8] = __g ^= k[8*(i)];					\
+	ss[1] ^= ss[0];							\
+	k[8*(i)+ 9] = __g ^= k[8*(i)+ 1];				\
+	ss[2] ^= ss[1];							\
+	k[8*(i)+10] = __g ^= k[8*(i)+ 2];				\
+	ss[3] ^= ss[2];							\
+	k[8*(i)+11] = __g ^= k[8*(i)+ 3];				\
+	__g = ls_box(ss[3],0);						\
+	ss[4] ^= __g;							\
+	__g = ff(__g);							\
+	k[8*(i)+12] = __g ^= k[8*(i)+ 4];				\
+	ss[5] ^= ss[4];							\
+	k[8*(i)+13] = __g ^= k[8*(i)+ 5];				\
+	ss[6] ^= ss[5];							\
+	k[8*(i)+14] = __g ^= k[8*(i)+ 6];				\
+	ss[7] ^= ss[6];							\
+	k[8*(i)+15] = __g ^= k[8*(i)+ 7];				\
+}
+
+#define kdl8(k,i)							\
+{									\
+	ss[0] ^= ls_box(ss[7],3) ^ rcon_tab[i];				\
+	k[8*(i)+ 8] = ss[0];						\
+	ss[1] ^= ss[0];							\
+	k[8*(i)+ 9] = ss[1];						\
+	ss[2] ^= ss[1];							\
+	k[8*(i)+10] = ss[2];						\
+	ss[3] ^= ss[2];							\
+	k[8*(i)+11] = ss[3];						\
+}
+
+static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key,
+		       unsigned int key_len)
+{
+	int i;
+	u32 ss[8];
+	struct aes_ctx *ctx = crypto_tfm_ctx(tfm);
+	const __le32 *key = (const __le32 *)in_key;
+	u32 *flags = &tfm->crt_flags;
+
+	/* encryption schedule */
+	
+	ctx->ekey[0] = ss[0] = le32_to_cpu(key[0]);
+	ctx->ekey[1] = ss[1] = le32_to_cpu(key[1]);
+	ctx->ekey[2] = ss[2] = le32_to_cpu(key[2]);
+	ctx->ekey[3] = ss[3] = le32_to_cpu(key[3]);
+
+	switch(key_len) {
+	case 16:
+		for (i = 0; i < 9; i++)
+			ke4(ctx->ekey, i);
+		kel4(ctx->ekey, 9);
+		ctx->rounds = 10;
+		break;
+		
+	case 24:
+		ctx->ekey[4] = ss[4] = le32_to_cpu(key[4]);
+		ctx->ekey[5] = ss[5] = le32_to_cpu(key[5]);
+		for (i = 0; i < 7; i++)
+			ke6(ctx->ekey, i);
+		kel6(ctx->ekey, 7); 
+		ctx->rounds = 12;
+		break;
+
+	case 32:
+		ctx->ekey[4] = ss[4] = le32_to_cpu(key[4]);
+		ctx->ekey[5] = ss[5] = le32_to_cpu(key[5]);
+		ctx->ekey[6] = ss[6] = le32_to_cpu(key[6]);
+		ctx->ekey[7] = ss[7] = le32_to_cpu(key[7]);
+		for (i = 0; i < 6; i++)
+			ke8(ctx->ekey, i);
+		kel8(ctx->ekey, 6);
+		ctx->rounds = 14;
+		break;
+
+	default:
+		*flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
+		return -EINVAL;
+	}
+	
+	/* decryption schedule */
+	
+	ctx->dkey[0] = ss[0] = le32_to_cpu(key[0]);
+	ctx->dkey[1] = ss[1] = le32_to_cpu(key[1]);
+	ctx->dkey[2] = ss[2] = le32_to_cpu(key[2]);
+	ctx->dkey[3] = ss[3] = le32_to_cpu(key[3]);
+
+	switch (key_len) {
+	case 16:
+		kdf4(ctx->dkey, 0);
+		for (i = 1; i < 9; i++)
+			kd4(ctx->dkey, i);
+		kdl4(ctx->dkey, 9);
+		break;
+		
+	case 24:
+		ctx->dkey[4] = ff(ss[4] = le32_to_cpu(key[4]));
+		ctx->dkey[5] = ff(ss[5] = le32_to_cpu(key[5]));
+		kdf6(ctx->dkey, 0);
+		for (i = 1; i < 7; i++)
+			kd6(ctx->dkey, i);
+		kdl6(ctx->dkey, 7);
+		break;
+
+	case 32:
+		ctx->dkey[4] = ff(ss[4] = le32_to_cpu(key[4]));
+		ctx->dkey[5] = ff(ss[5] = le32_to_cpu(key[5]));
+		ctx->dkey[6] = ff(ss[6] = le32_to_cpu(key[6]));
+		ctx->dkey[7] = ff(ss[7] = le32_to_cpu(key[7]));
+		kdf8(ctx->dkey, 0);
+		for (i = 1; i < 6; i++)
+			kd8(ctx->dkey, i);
+		kdl8(ctx->dkey, 6);
+		break;
+	}
+	return 0;
+}
+
+static void aes_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	aes_enc_blk(tfm, dst, src);
+}
+
+static void aes_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	aes_dec_blk(tfm, dst, src);
+}
+
+static struct crypto_alg aes_alg = {
+	.cra_name		=	"aes",
+	.cra_driver_name	=	"aes-i586",
+	.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)
+{
+	gen_tabs();
+	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, i586 asm optimized");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_AUTHOR("Fruhwirth Clemens, James Morris, Brian Gladman, Adam Richter");
+MODULE_ALIAS("aes");
diff --git a/arch/x86/crypto/twofish-i586-asm_32.S b/arch/x86/crypto/twofish-i586-asm_32.S
new file mode 100644
index 0000000..39b98ed
--- /dev/null
+++ b/arch/x86/crypto/twofish-i586-asm_32.S
@@ -0,0 +1,335 @@
+/***************************************************************************
+*   Copyright (C) 2006 by Joachim Fritschi, <jfritschi@freenet.de>        *
+*                                                                         *
+*   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.                                   *
+*                                                                         *
+*   This program is distributed in the hope that it will be useful,       *
+*   but WITHOUT ANY WARRANTY; without even the implied warranty of        *
+*   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         *
+*   GNU General Public License for more details.                          *
+*                                                                         *
+*   You should have received a copy of the GNU General Public License     *
+*   along with this program; if not, write to the                         *
+*   Free Software Foundation, Inc.,                                       *
+*   59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.             *
+***************************************************************************/
+
+.file "twofish-i586-asm.S"
+.text
+
+#include <asm/asm-offsets.h>
+
+/* return adress at 0 */
+
+#define in_blk    12  /* input byte array address parameter*/
+#define out_blk   8  /* output byte array address parameter*/
+#define tfm       4  /* Twofish context structure */
+
+#define a_offset	0
+#define b_offset	4
+#define c_offset	8
+#define d_offset	12
+
+/* Structure of the crypto context struct*/
+
+#define s0	0	/* S0 Array 256 Words each */
+#define s1	1024	/* S1 Array */
+#define s2	2048	/* S2 Array */
+#define s3	3072	/* S3 Array */
+#define w	4096	/* 8 whitening keys (word) */
+#define k	4128	/* key 1-32 ( word ) */
+
+/* define a few register aliases to allow macro substitution */
+
+#define R0D    %eax
+#define R0B    %al
+#define R0H    %ah
+
+#define R1D    %ebx
+#define R1B    %bl
+#define R1H    %bh
+
+#define R2D    %ecx
+#define R2B    %cl
+#define R2H    %ch
+
+#define R3D    %edx
+#define R3B    %dl
+#define R3H    %dh
+
+
+/* performs input whitening */
+#define input_whitening(src,context,offset)\
+	xor	w+offset(context),	src;
+
+/* performs input whitening */
+#define output_whitening(src,context,offset)\
+	xor	w+16+offset(context),	src;
+
+/*
+ * a input register containing a (rotated 16)
+ * b input register containing b
+ * c input register containing c
+ * d input register containing d (already rol $1)
+ * operations on a and b are interleaved to increase performance
+ */
+#define encrypt_round(a,b,c,d,round)\
+	push	d ## D;\
+	movzx	b ## B,		%edi;\
+	mov	s1(%ebp,%edi,4),d ## D;\
+	movzx	a ## B,		%edi;\
+	mov	s2(%ebp,%edi,4),%esi;\
+	movzx	b ## H,		%edi;\
+	ror	$16,		b ## D;\
+	xor	s2(%ebp,%edi,4),d ## D;\
+	movzx	a ## H,		%edi;\
+	ror	$16,		a ## D;\
+	xor	s3(%ebp,%edi,4),%esi;\
+	movzx	b ## B,		%edi;\
+	xor	s3(%ebp,%edi,4),d ## D;\
+	movzx	a ## B,		%edi;\
+	xor	(%ebp,%edi,4),	%esi;\
+	movzx	b ## H,		%edi;\
+	ror	$15,		b ## D;\
+	xor	(%ebp,%edi,4),	d ## D;\
+	movzx	a ## H,		%edi;\
+	xor	s1(%ebp,%edi,4),%esi;\
+	pop	%edi;\
+	add	d ## D,		%esi;\
+	add	%esi,		d ## D;\
+	add	k+round(%ebp),	%esi;\
+	xor	%esi,		c ## D;\
+	rol	$15,		c ## D;\
+	add	k+4+round(%ebp),d ## D;\
+	xor	%edi,		d ## D;
+
+/*
+ * a input register containing a (rotated 16)
+ * b input register containing b
+ * c input register containing c
+ * d input register containing d (already rol $1)
+ * operations on a and b are interleaved to increase performance
+ * last round has different rotations for the output preparation
+ */
+#define encrypt_last_round(a,b,c,d,round)\
+	push	d ## D;\
+	movzx	b ## B,		%edi;\
+	mov	s1(%ebp,%edi,4),d ## D;\
+	movzx	a ## B,		%edi;\
+	mov	s2(%ebp,%edi,4),%esi;\
+	movzx	b ## H,		%edi;\
+	ror	$16,		b ## D;\
+	xor	s2(%ebp,%edi,4),d ## D;\
+	movzx	a ## H,		%edi;\
+	ror	$16,		a ## D;\
+	xor	s3(%ebp,%edi,4),%esi;\
+	movzx	b ## B,		%edi;\
+	xor	s3(%ebp,%edi,4),d ## D;\
+	movzx	a ## B,		%edi;\
+	xor	(%ebp,%edi,4),	%esi;\
+	movzx	b ## H,		%edi;\
+	ror	$16,		b ## D;\
+	xor	(%ebp,%edi,4),	d ## D;\
+	movzx	a ## H,		%edi;\
+	xor	s1(%ebp,%edi,4),%esi;\
+	pop	%edi;\
+	add	d ## D,		%esi;\
+	add	%esi,		d ## D;\
+	add	k+round(%ebp),	%esi;\
+	xor	%esi,		c ## D;\
+	ror	$1,		c ## D;\
+	add	k+4+round(%ebp),d ## D;\
+	xor	%edi,		d ## D;
+
+/*
+ * a input register containing a
+ * b input register containing b (rotated 16)
+ * c input register containing c
+ * d input register containing d (already rol $1)
+ * operations on a and b are interleaved to increase performance
+ */
+#define decrypt_round(a,b,c,d,round)\
+	push	c ## D;\
+	movzx	a ## B,		%edi;\
+	mov	(%ebp,%edi,4),	c ## D;\
+	movzx	b ## B,		%edi;\
+	mov	s3(%ebp,%edi,4),%esi;\
+	movzx	a ## H,		%edi;\
+	ror	$16,		a ## D;\
+	xor	s1(%ebp,%edi,4),c ## D;\
+	movzx	b ## H,		%edi;\
+	ror	$16,		b ## D;\
+	xor	(%ebp,%edi,4),	%esi;\
+	movzx	a ## B,		%edi;\
+	xor	s2(%ebp,%edi,4),c ## D;\
+	movzx	b ## B,		%edi;\
+	xor	s1(%ebp,%edi,4),%esi;\
+	movzx	a ## H,		%edi;\
+	ror	$15,		a ## D;\
+	xor	s3(%ebp,%edi,4),c ## D;\
+	movzx	b ## H,		%edi;\
+	xor	s2(%ebp,%edi,4),%esi;\
+	pop	%edi;\
+	add	%esi,		c ## D;\
+	add	c ## D,		%esi;\
+	add	k+round(%ebp),	c ## D;\
+	xor	%edi,		c ## D;\
+	add	k+4+round(%ebp),%esi;\
+	xor	%esi,		d ## D;\
+	rol	$15,		d ## D;
+
+/*
+ * a input register containing a
+ * b input register containing b (rotated 16)
+ * c input register containing c
+ * d input register containing d (already rol $1)
+ * operations on a and b are interleaved to increase performance
+ * last round has different rotations for the output preparation
+ */
+#define decrypt_last_round(a,b,c,d,round)\
+	push	c ## D;\
+	movzx	a ## B,		%edi;\
+	mov	(%ebp,%edi,4),	c ## D;\
+	movzx	b ## B,		%edi;\
+	mov	s3(%ebp,%edi,4),%esi;\
+	movzx	a ## H,		%edi;\
+	ror	$16,		a ## D;\
+	xor	s1(%ebp,%edi,4),c ## D;\
+	movzx	b ## H,		%edi;\
+	ror	$16,		b ## D;\
+	xor	(%ebp,%edi,4),	%esi;\
+	movzx	a ## B,		%edi;\
+	xor	s2(%ebp,%edi,4),c ## D;\
+	movzx	b ## B,		%edi;\
+	xor	s1(%ebp,%edi,4),%esi;\
+	movzx	a ## H,		%edi;\
+	ror	$16,		a ## D;\
+	xor	s3(%ebp,%edi,4),c ## D;\
+	movzx	b ## H,		%edi;\
+	xor	s2(%ebp,%edi,4),%esi;\
+	pop	%edi;\
+	add	%esi,		c ## D;\
+	add	c ## D,		%esi;\
+	add	k+round(%ebp),	c ## D;\
+	xor	%edi,		c ## D;\
+	add	k+4+round(%ebp),%esi;\
+	xor	%esi,		d ## D;\
+	ror	$1,		d ## D;
+
+.align 4
+.global twofish_enc_blk
+.global twofish_dec_blk
+
+twofish_enc_blk:
+	push	%ebp			/* save registers according to calling convention*/
+	push    %ebx
+	push    %esi
+	push    %edi
+
+	mov	tfm + 16(%esp),	%ebp	/* abuse the base pointer: set new base bointer to the crypto tfm */
+	add	$crypto_tfm_ctx_offset, %ebp	/* ctx adress */
+	mov     in_blk+16(%esp),%edi	/* input adress in edi */
+
+	mov	(%edi),		%eax
+	mov	b_offset(%edi),	%ebx
+	mov	c_offset(%edi),	%ecx
+	mov	d_offset(%edi),	%edx
+	input_whitening(%eax,%ebp,a_offset)
+	ror	$16,	%eax
+	input_whitening(%ebx,%ebp,b_offset)
+	input_whitening(%ecx,%ebp,c_offset)
+	input_whitening(%edx,%ebp,d_offset)
+	rol	$1,	%edx
+
+	encrypt_round(R0,R1,R2,R3,0);
+	encrypt_round(R2,R3,R0,R1,8);
+	encrypt_round(R0,R1,R2,R3,2*8);
+	encrypt_round(R2,R3,R0,R1,3*8);
+	encrypt_round(R0,R1,R2,R3,4*8);
+	encrypt_round(R2,R3,R0,R1,5*8);
+	encrypt_round(R0,R1,R2,R3,6*8);
+	encrypt_round(R2,R3,R0,R1,7*8);
+	encrypt_round(R0,R1,R2,R3,8*8);
+	encrypt_round(R2,R3,R0,R1,9*8);
+	encrypt_round(R0,R1,R2,R3,10*8);
+	encrypt_round(R2,R3,R0,R1,11*8);
+	encrypt_round(R0,R1,R2,R3,12*8);
+	encrypt_round(R2,R3,R0,R1,13*8);
+	encrypt_round(R0,R1,R2,R3,14*8);
+	encrypt_last_round(R2,R3,R0,R1,15*8);
+
+	output_whitening(%eax,%ebp,c_offset)
+	output_whitening(%ebx,%ebp,d_offset)
+	output_whitening(%ecx,%ebp,a_offset)
+	output_whitening(%edx,%ebp,b_offset)
+	mov	out_blk+16(%esp),%edi;
+	mov	%eax,		c_offset(%edi)
+	mov	%ebx,		d_offset(%edi)
+	mov	%ecx,		(%edi)
+	mov	%edx,		b_offset(%edi)
+
+	pop	%edi
+	pop	%esi
+	pop	%ebx
+	pop	%ebp
+	mov	$1,	%eax
+	ret
+
+twofish_dec_blk:
+	push	%ebp			/* save registers according to calling convention*/
+	push    %ebx
+	push    %esi
+	push    %edi
+
+
+	mov	tfm + 16(%esp),	%ebp	/* abuse the base pointer: set new base bointer to the crypto tfm */
+	add	$crypto_tfm_ctx_offset, %ebp	/* ctx adress */
+	mov     in_blk+16(%esp),%edi	/* input adress in edi */
+
+	mov	(%edi),		%eax
+	mov	b_offset(%edi),	%ebx
+	mov	c_offset(%edi),	%ecx
+	mov	d_offset(%edi),	%edx
+	output_whitening(%eax,%ebp,a_offset)
+	output_whitening(%ebx,%ebp,b_offset)
+	ror	$16,	%ebx
+	output_whitening(%ecx,%ebp,c_offset)
+	output_whitening(%edx,%ebp,d_offset)
+	rol	$1,	%ecx
+
+	decrypt_round(R0,R1,R2,R3,15*8);
+	decrypt_round(R2,R3,R0,R1,14*8);
+	decrypt_round(R0,R1,R2,R3,13*8);
+	decrypt_round(R2,R3,R0,R1,12*8);
+	decrypt_round(R0,R1,R2,R3,11*8);
+	decrypt_round(R2,R3,R0,R1,10*8);
+	decrypt_round(R0,R1,R2,R3,9*8);
+	decrypt_round(R2,R3,R0,R1,8*8);
+	decrypt_round(R0,R1,R2,R3,7*8);
+	decrypt_round(R2,R3,R0,R1,6*8);
+	decrypt_round(R0,R1,R2,R3,5*8);
+	decrypt_round(R2,R3,R0,R1,4*8);
+	decrypt_round(R0,R1,R2,R3,3*8);
+	decrypt_round(R2,R3,R0,R1,2*8);
+	decrypt_round(R0,R1,R2,R3,1*8);
+	decrypt_last_round(R2,R3,R0,R1,0);
+
+	input_whitening(%eax,%ebp,c_offset)
+	input_whitening(%ebx,%ebp,d_offset)
+	input_whitening(%ecx,%ebp,a_offset)
+	input_whitening(%edx,%ebp,b_offset)
+	mov	out_blk+16(%esp),%edi;
+	mov	%eax,		c_offset(%edi)
+	mov	%ebx,		d_offset(%edi)
+	mov	%ecx,		(%edi)
+	mov	%edx,		b_offset(%edi)
+
+	pop	%edi
+	pop	%esi
+	pop	%ebx
+	pop	%ebp
+	mov	$1,	%eax
+	ret
diff --git a/arch/x86/crypto/twofish_32.c b/arch/x86/crypto/twofish_32.c
new file mode 100644
index 0000000..e3004df
--- /dev/null
+++ b/arch/x86/crypto/twofish_32.c
@@ -0,0 +1,97 @@
+/*
+ *  Glue Code for optimized 586 assembler version of TWOFISH
+ *
+ * Originally Twofish for GPG
+ * By Matthew Skala <mskala@ansuz.sooke.bc.ca>, July 26, 1998
+ * 256-bit key length added March 20, 1999
+ * Some modifications to reduce the text size by Werner Koch, April, 1998
+ * Ported to the kerneli patch by Marc Mutz <Marc@Mutz.com>
+ * Ported to CryptoAPI by Colin Slater <hoho@tacomeat.net>
+ *
+ * The original author has disclaimed all copyright interest in this
+ * code and thus put it in the public domain. The subsequent authors
+ * have put this under the GNU General Public License.
+ *
+ * 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.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307
+ * USA
+ *
+ * This code is a "clean room" implementation, written from the paper
+ * _Twofish: A 128-Bit Block Cipher_ by Bruce Schneier, John Kelsey,
+ * Doug Whiting, David Wagner, Chris Hall, and Niels Ferguson, available
+ * through http://www.counterpane.com/twofish.html
+ *
+ * For background information on multiplication in finite fields, used for
+ * the matrix operations in the key schedule, see the book _Contemporary
+ * Abstract Algebra_ by Joseph A. Gallian, especially chapter 22 in the
+ * Third Edition.
+ */
+
+#include <crypto/twofish.h>
+#include <linux/crypto.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/types.h>
+
+
+asmlinkage void twofish_enc_blk(struct crypto_tfm *tfm, u8 *dst, const u8 *src);
+asmlinkage void twofish_dec_blk(struct crypto_tfm *tfm, u8 *dst, const u8 *src);
+
+static void twofish_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	twofish_enc_blk(tfm, dst, src);
+}
+
+static void twofish_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src)
+{
+	twofish_dec_blk(tfm, dst, src);
+}
+
+static struct crypto_alg alg = {
+	.cra_name		=	"twofish",
+	.cra_driver_name	=	"twofish-i586",
+	.cra_priority		=	200,
+	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
+	.cra_blocksize		=	TF_BLOCK_SIZE,
+	.cra_ctxsize		=	sizeof(struct twofish_ctx),
+	.cra_alignmask		=	3,
+	.cra_module		=	THIS_MODULE,
+	.cra_list		=	LIST_HEAD_INIT(alg.cra_list),
+	.cra_u			=	{
+		.cipher = {
+			.cia_min_keysize	=	TF_MIN_KEY_SIZE,
+			.cia_max_keysize	=	TF_MAX_KEY_SIZE,
+			.cia_setkey		=	twofish_setkey,
+			.cia_encrypt		=	twofish_encrypt,
+			.cia_decrypt		=	twofish_decrypt
+		}
+	}
+};
+
+static int __init init(void)
+{
+	return crypto_register_alg(&alg);
+}
+
+static void __exit fini(void)
+{
+	crypto_unregister_alg(&alg);
+}
+
+module_init(init);
+module_exit(fini);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION ("Twofish Cipher Algorithm, i586 asm optimized");
+MODULE_ALIAS("twofish");