Merge rsync://rsync.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6

arch/x86_64/Makefile

@@ -65,7 +65,9 @@ CFLAGS += $(call cc-option,-mno-sse -mno-mmx -mno-sse2 -mno-3dnow,)
 head-y := arch/x86_64/kernel/head.o arch/x86_64/kernel/head64.o arch/x86_64/kernel/init_task.o
 
 libs-y 					+= arch/x86_64/lib/
-core-y					+= arch/x86_64/kernel/ arch/x86_64/mm/
+core-y					+= arch/x86_64/kernel/ \
+					   arch/x86_64/mm/ \
+					   arch/x86_64/crypto/
 core-$(CONFIG_IA32_EMULATION)		+= arch/x86_64/ia32/
 drivers-$(CONFIG_PCI)			+= arch/x86_64/pci/
 drivers-$(CONFIG_OPROFILE)		+= arch/x86_64/oprofile/

arch/x86_64/crypto/Makefile

@@ -0,0 +1,9 @@
+# 
+# x86_64/crypto/Makefile 
+# 
+# Arch-specific CryptoAPI modules.
+# 
+
+obj-$(CONFIG_CRYPTO_AES_X86_64) += aes-x86_64.o
+
+aes-x86_64-y := aes-x86_64-asm.o aes.o

arch/x86_64/crypto/aes-x86_64-asm.S

@@ -0,0 +1,186 @@
+/* AES (Rijndael) implementation (FIPS PUB 197) for x86_64
+ *
+ * Copyright (C) 2005 Andreas Steinmetz, <ast@domdv.de>
+ *
+ * License:
+ * This code can be distributed under the terms of the GNU General Public
+ * License (GPL) Version 2 provided that the above header down to and
+ * including this sentence is retained in full.
+ */
+
+.extern aes_ft_tab
+.extern aes_it_tab
+.extern aes_fl_tab
+.extern aes_il_tab
+
+.text
+
+#define R1	%rax
+#define R1E	%eax
+#define R1X	%ax
+#define R1H	%ah
+#define R1L	%al
+#define R2	%rbx
+#define R2E	%ebx
+#define R2X	%bx
+#define R2H	%bh
+#define R2L	%bl
+#define R3	%rcx
+#define R3E	%ecx
+#define R3X	%cx
+#define R3H	%ch
+#define R3L	%cl
+#define R4	%rdx
+#define R4E	%edx
+#define R4X	%dx
+#define R4H	%dh
+#define R4L	%dl
+#define R5	%rsi
+#define R5E	%esi
+#define R6	%rdi
+#define R6E	%edi
+#define R7	%rbp
+#define R7E	%ebp
+#define R8	%r8
+#define R9	%r9
+#define R10	%r10
+#define R11	%r11
+
+#define prologue(FUNC,BASE,B128,B192,r1,r2,r3,r4,r5,r6,r7,r8,r9,r10,r11) \
+	.global	FUNC;			\
+	.type	FUNC,@function;		\
+	.align	8;			\
+FUNC:	movq	r1,r2;			\
+	movq	r3,r4;			\
+	leaq	BASE+52(r8),r9;		\
+	movq	r10,r11;		\
+	movl	(r7),r5 ## E;		\
+	movl	4(r7),r1 ## E;		\
+	movl	8(r7),r6 ## E;		\
+	movl	12(r7),r7 ## E;		\
+	movl	(r8),r10 ## E;		\
+	xorl	-48(r9),r5 ## E;	\
+	xorl	-44(r9),r1 ## E;	\
+	xorl	-40(r9),r6 ## E;	\
+	xorl	-36(r9),r7 ## E;	\
+	cmpl	$24,r10 ## E;		\
+	jb	B128;			\
+	leaq	32(r9),r9;		\
+	je	B192;			\
+	leaq	32(r9),r9;
+
+#define epilogue(r1,r2,r3,r4,r5,r6,r7,r8,r9) \
+	movq	r1,r2;			\
+	movq	r3,r4;			\
+	movl	r5 ## E,(r9);		\
+	movl	r6 ## E,4(r9);		\
+	movl	r7 ## E,8(r9);		\
+	movl	r8 ## E,12(r9);		\
+	ret;
+
+#define round(TAB,OFFSET,r1,r2,r3,r4,r5,r6,r7,r8,ra,rb,rc,rd) \
+	movzbl	r2 ## H,r5 ## E;	\
+	movzbl	r2 ## L,r6 ## E;	\
+	movl	TAB+1024(,r5,4),r5 ## E;\
+	movw	r4 ## X,r2 ## X;	\
+	movl	TAB(,r6,4),r6 ## E;	\
+	roll	$16,r2 ## E;		\
+	shrl	$16,r4 ## E;		\
+	movzbl	r4 ## H,r7 ## E;	\
+	movzbl	r4 ## L,r4 ## E;	\
+	xorl	OFFSET(r8),ra ## E;	\
+	xorl	OFFSET+4(r8),rb ## E;	\
+	xorl	TAB+3072(,r7,4),r5 ## E;\
+	xorl	TAB+2048(,r4,4),r6 ## E;\
+	movzbl	r1 ## L,r7 ## E;	\
+	movzbl	r1 ## H,r4 ## E;	\
+	movl	TAB+1024(,r4,4),r4 ## E;\
+	movw	r3 ## X,r1 ## X;	\
+	roll	$16,r1 ## E;		\
+	shrl	$16,r3 ## E;		\
+	xorl	TAB(,r7,4),r5 ## E;	\
+	movzbl	r3 ## H,r7 ## E;	\
+	movzbl	r3 ## L,r3 ## E;	\
+	xorl	TAB+3072(,r7,4),r4 ## E;\
+	xorl	TAB+2048(,r3,4),r5 ## E;\
+	movzbl	r1 ## H,r7 ## E;	\
+	movzbl	r1 ## L,r3 ## E;	\
+	shrl	$16,r1 ## E;		\
+	xorl	TAB+3072(,r7,4),r6 ## E;\
+	movl	TAB+2048(,r3,4),r3 ## E;\
+	movzbl	r1 ## H,r7 ## E;	\
+	movzbl	r1 ## L,r1 ## E;	\
+	xorl	TAB+1024(,r7,4),r6 ## E;\
+	xorl	TAB(,r1,4),r3 ## E;	\
+	movzbl	r2 ## H,r1 ## E;	\
+	movzbl	r2 ## L,r7 ## E;	\
+	shrl	$16,r2 ## E;		\
+	xorl	TAB+3072(,r1,4),r3 ## E;\
+	xorl	TAB+2048(,r7,4),r4 ## E;\
+	movzbl	r2 ## H,r1 ## E;	\
+	movzbl	r2 ## L,r2 ## E;	\
+	xorl	OFFSET+8(r8),rc ## E;	\
+	xorl	OFFSET+12(r8),rd ## E;	\
+	xorl	TAB+1024(,r1,4),r3 ## E;\
+	xorl	TAB(,r2,4),r4 ## E;
+
+#define move_regs(r1,r2,r3,r4) \
+	movl	r3 ## E,r1 ## E;	\
+	movl	r4 ## E,r2 ## E;
+
+#define entry(FUNC,BASE,B128,B192) \
+	prologue(FUNC,BASE,B128,B192,R2,R8,R7,R9,R1,R3,R4,R6,R10,R5,R11)
+
+#define return epilogue(R8,R2,R9,R7,R5,R6,R3,R4,R11)
+
+#define encrypt_round(TAB,OFFSET) \
+	round(TAB,OFFSET,R1,R2,R3,R4,R5,R6,R7,R10,R5,R6,R3,R4) \
+	move_regs(R1,R2,R5,R6)
+
+#define encrypt_final(TAB,OFFSET) \
+	round(TAB,OFFSET,R1,R2,R3,R4,R5,R6,R7,R10,R5,R6,R3,R4)
+
+#define decrypt_round(TAB,OFFSET) \
+	round(TAB,OFFSET,R2,R1,R4,R3,R6,R5,R7,R10,R5,R6,R3,R4) \
+	move_regs(R1,R2,R5,R6)
+
+#define decrypt_final(TAB,OFFSET) \
+	round(TAB,OFFSET,R2,R1,R4,R3,R6,R5,R7,R10,R5,R6,R3,R4)
+
+/* void aes_encrypt(void *ctx, u8 *out, const u8 *in) */
+
+	entry(aes_encrypt,0,enc128,enc192)
+	encrypt_round(aes_ft_tab,-96)
+	encrypt_round(aes_ft_tab,-80)
+enc192:	encrypt_round(aes_ft_tab,-64)
+	encrypt_round(aes_ft_tab,-48)
+enc128:	encrypt_round(aes_ft_tab,-32)
+	encrypt_round(aes_ft_tab,-16)
+	encrypt_round(aes_ft_tab,  0)
+	encrypt_round(aes_ft_tab, 16)
+	encrypt_round(aes_ft_tab, 32)
+	encrypt_round(aes_ft_tab, 48)
+	encrypt_round(aes_ft_tab, 64)
+	encrypt_round(aes_ft_tab, 80)
+	encrypt_round(aes_ft_tab, 96)
+	encrypt_final(aes_fl_tab,112)
+	return
+
+/* void aes_decrypt(void *ctx, u8 *out, const u8 *in) */
+
+	entry(aes_decrypt,240,dec128,dec192)
+	decrypt_round(aes_it_tab,-96)
+	decrypt_round(aes_it_tab,-80)
+dec192:	decrypt_round(aes_it_tab,-64)
+	decrypt_round(aes_it_tab,-48)
+dec128:	decrypt_round(aes_it_tab,-32)
+	decrypt_round(aes_it_tab,-16)
+	decrypt_round(aes_it_tab,  0)
+	decrypt_round(aes_it_tab, 16)
+	decrypt_round(aes_it_tab, 32)
+	decrypt_round(aes_it_tab, 48)
+	decrypt_round(aes_it_tab, 64)
+	decrypt_round(aes_it_tab, 80)
+	decrypt_round(aes_it_tab, 96)
+	decrypt_final(aes_il_tab,112)
+	return

arch/x86_64/crypto/aes.c

@@ -0,0 +1,324 @@
+/*
+ * Cryptographic API.
+ *
+ * AES Cipher Algorithm.
+ *
+ * Based on Brian Gladman's code.
+ *
+ * Linux developers:
+ *  Alexander Kjeldaas <astor@fast.no>
+ *  Herbert Valerio Riedel <hvr@hvrlab.org>
+ *  Kyle McMartin <kyle@debian.org>
+ *  Adam J. Richter <adam@yggdrasil.com> (conversion to 2.5 API).
+ *  Andreas Steinmetz <ast@domdv.de> (adapted to x86_64 assembler)
+ *
+ * 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.
+ *
+ * ---------------------------------------------------------------------------
+ * Copyright (c) 2002, Dr Brian Gladman <brg@gladman.me.uk>, 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.
+ * ---------------------------------------------------------------------------
+ */
+
+/* Some changes from the Gladman version:
+    s/RIJNDAEL(e_key)/E_KEY/g
+    s/RIJNDAEL(d_key)/D_KEY/g
+*/
+
+#include <asm/byteorder.h>
+#include <linux/bitops.h>
+#include <linux/crypto.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/types.h>
+
+#define AES_MIN_KEY_SIZE	16
+#define AES_MAX_KEY_SIZE	32
+
+#define AES_BLOCK_SIZE		16
+
+/*
+ * #define byte(x, nr) ((unsigned char)((x) >> (nr*8)))
+ */
+static inline u8 byte(const u32 x, const unsigned n)
+{
+	return x >> (n << 3);
+}
+
+#define u32_in(x) le32_to_cpu(*(const __le32 *)(x))
+
+struct aes_ctx
+{
+	u32 key_length;
+	u32 E[60];
+	u32 D[60];
+};
+
+#define E_KEY ctx->E
+#define D_KEY ctx->D
+
+static u8 pow_tab[256] __initdata;
+static u8 log_tab[256] __initdata;
+static u8 sbx_tab[256] __initdata;
+static u8 isb_tab[256] __initdata;
+static u32 rco_tab[10];
+u32 aes_ft_tab[4][256];
+u32 aes_it_tab[4][256];
+
+u32 aes_fl_tab[4][256];
+u32 aes_il_tab[4][256];
+
+static inline u8 f_mult(u8 a, u8 b)
+{
+	u8 aa = log_tab[a], cc = aa + log_tab[b];
+
+	return pow_tab[cc + (cc < aa ? 1 : 0)];
+}
+
+#define ff_mult(a, b) (a && b ? f_mult(a, b) : 0)
+
+#define ls_box(x)				\
+	(aes_fl_tab[0][byte(x, 0)] ^		\
+	 aes_fl_tab[1][byte(x, 1)] ^		\
+	 aes_fl_tab[2][byte(x, 2)] ^		\
+	 aes_fl_tab[3][byte(x, 3)])
+
+static void __init gen_tabs(void)
+{
+	u32 i, t;
+	u8 p, q;
+
+	/* log and power tables for GF(2**8) finite field with
+	   0x011b as modular polynomial - the simplest primitive
+	   root is 0x03, used here to generate the tables */
+
+	for (i = 0, p = 1; i < 256; ++i) {
+		pow_tab[i] = (u8)p;
+		log_tab[p] = (u8)i;
+
+		p ^= (p << 1) ^ (p & 0x80 ? 0x01b : 0);
+	}
+
+	log_tab[1] = 0;
+
+	for (i = 0, p = 1; i < 10; ++i) {
+		rco_tab[i] = p;
+
+		p = (p << 1) ^ (p & 0x80 ? 0x01b : 0);
+	}
+
+	for (i = 0; i < 256; ++i) {
+		p = (i ? pow_tab[255 - log_tab[i]] : 0);
+		q = ((p >> 7) | (p << 1)) ^ ((p >> 6) | (p << 2));
+		p ^= 0x63 ^ q ^ ((q >> 6) | (q << 2));
+		sbx_tab[i] = p;
+		isb_tab[p] = (u8)i;
+	}
+
+	for (i = 0; i < 256; ++i) {
+		p = sbx_tab[i];
+
+		t = p;
+		aes_fl_tab[0][i] = t;
+		aes_fl_tab[1][i] = rol32(t, 8);
+		aes_fl_tab[2][i] = rol32(t, 16);
+		aes_fl_tab[3][i] = rol32(t, 24);
+
+		t = ((u32)ff_mult(2, p)) |
+		    ((u32)p << 8) |
+		    ((u32)p << 16) | ((u32)ff_mult(3, p) << 24);
+
+		aes_ft_tab[0][i] = t;
+		aes_ft_tab[1][i] = rol32(t, 8);
+		aes_ft_tab[2][i] = rol32(t, 16);
+		aes_ft_tab[3][i] = rol32(t, 24);
+
+		p = isb_tab[i];
+
+		t = p;
+		aes_il_tab[0][i] = t;
+		aes_il_tab[1][i] = rol32(t, 8);
+		aes_il_tab[2][i] = rol32(t, 16);
+		aes_il_tab[3][i] = rol32(t, 24);
+
+		t = ((u32)ff_mult(14, p)) |
+		    ((u32)ff_mult(9, p) << 8) |
+		    ((u32)ff_mult(13, p) << 16) |
+		    ((u32)ff_mult(11, p) << 24);
+
+		aes_it_tab[0][i] = t;
+		aes_it_tab[1][i] = rol32(t, 8);
+		aes_it_tab[2][i] = rol32(t, 16);
+		aes_it_tab[3][i] = rol32(t, 24);
+	}
+}
+
+#define star_x(x) (((x) & 0x7f7f7f7f) << 1) ^ ((((x) & 0x80808080) >> 7) * 0x1b)
+
+#define imix_col(y, x)			\
+	u    = star_x(x);		\
+	v    = star_x(u);		\
+	w    = star_x(v);		\
+	t    = w ^ (x);			\
+	(y)  = u ^ v ^ w;		\
+	(y) ^= ror32(u ^ t,  8) ^	\
+	       ror32(v ^ t, 16) ^	\
+	       ror32(t, 24)
+
+/* initialise the key schedule from the user supplied key */
+
+#define loop4(i)					\
+{							\
+	t = ror32(t,  8); t = ls_box(t) ^ rco_tab[i];	\
+	t ^= E_KEY[4 * i];     E_KEY[4 * i + 4] = t;	\
+	t ^= E_KEY[4 * i + 1]; E_KEY[4 * i + 5] = t;	\
+	t ^= E_KEY[4 * i + 2]; E_KEY[4 * i + 6] = t;	\
+	t ^= E_KEY[4 * i + 3]; E_KEY[4 * i + 7] = t;	\
+}
+
+#define loop6(i)					\
+{							\
+	t = ror32(t,  8); t = ls_box(t) ^ rco_tab[i];	\
+	t ^= E_KEY[6 * i];     E_KEY[6 * i + 6] = t;	\
+	t ^= E_KEY[6 * i + 1]; E_KEY[6 * i + 7] = t;	\
+	t ^= E_KEY[6 * i + 2]; E_KEY[6 * i + 8] = t;	\
+	t ^= E_KEY[6 * i + 3]; E_KEY[6 * i + 9] = t;	\
+	t ^= E_KEY[6 * i + 4]; E_KEY[6 * i + 10] = t;	\
+	t ^= E_KEY[6 * i + 5]; E_KEY[6 * i + 11] = t;	\
+}
+
+#define loop8(i)					\
+{							\
+	t = ror32(t,  8); ; t = ls_box(t) ^ rco_tab[i];	\
+	t ^= E_KEY[8 * i];     E_KEY[8 * i + 8] = t;	\
+	t ^= E_KEY[8 * i + 1]; E_KEY[8 * i + 9] = t;	\
+	t ^= E_KEY[8 * i + 2]; E_KEY[8 * i + 10] = t;	\
+	t ^= E_KEY[8 * i + 3]; E_KEY[8 * i + 11] = t;	\
+	t  = E_KEY[8 * i + 4] ^ ls_box(t);		\
+	E_KEY[8 * i + 12] = t;				\
+	t ^= E_KEY[8 * i + 5]; E_KEY[8 * i + 13] = t;	\
+	t ^= E_KEY[8 * i + 6]; E_KEY[8 * i + 14] = t;	\
+	t ^= E_KEY[8 * i + 7]; E_KEY[8 * i + 15] = t;	\
+}
+
+static int aes_set_key(void *ctx_arg, const u8 *in_key, unsigned int key_len,
+		       u32 *flags)
+{
+	struct aes_ctx *ctx = ctx_arg;
+	u32 i, j, t, u, v, w;
+
+	if (key_len != 16 && key_len != 24 && key_len != 32) {
+		*flags |= CRYPTO_TFM_RES_BAD_KEY_LEN;
+		return -EINVAL;
+	}
+
+	ctx->key_length = key_len;
+
+	D_KEY[key_len + 24] = E_KEY[0] = u32_in(in_key);
+	D_KEY[key_len + 25] = E_KEY[1] = u32_in(in_key + 4);
+	D_KEY[key_len + 26] = E_KEY[2] = u32_in(in_key + 8);
+	D_KEY[key_len + 27] = E_KEY[3] = u32_in(in_key + 12);
+
+	switch (key_len) {
+	case 16:
+		t = E_KEY[3];
+		for (i = 0; i < 10; ++i)
+			loop4(i);
+		break;
+
+	case 24:
+		E_KEY[4] = u32_in(in_key + 16);
+		t = E_KEY[5] = u32_in(in_key + 20);
+		for (i = 0; i < 8; ++i)
+			loop6 (i);
+		break;
+
+	case 32:
+		E_KEY[4] = u32_in(in_key + 16);
+		E_KEY[5] = u32_in(in_key + 20);
+		E_KEY[6] = u32_in(in_key + 24);
+		t = E_KEY[7] = u32_in(in_key + 28);
+		for (i = 0; i < 7; ++i)
+			loop8(i);
+		break;
+	}
+
+	D_KEY[0] = E_KEY[key_len + 24];
+	D_KEY[1] = E_KEY[key_len + 25];
+	D_KEY[2] = E_KEY[key_len + 26];
+	D_KEY[3] = E_KEY[key_len + 27];
+
+	for (i = 4; i < key_len + 24; ++i) {
+		j = key_len + 24 - (i & ~3) + (i & 3);
+		imix_col(D_KEY[j], E_KEY[i]);
+	}
+
+	return 0;
+}
+
+extern void aes_encrypt(void *ctx_arg, u8 *out, const u8 *in);
+extern void aes_decrypt(void *ctx_arg, u8 *out, const u8 *in);
+
+static struct crypto_alg aes_alg = {
+	.cra_name		=	"aes",
+	.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");
+MODULE_LICENSE("GPL");

crypto/Kconfig

@@ -146,7 +146,7 @@ config CRYPTO_SERPENT
 
 config CRYPTO_AES
 	tristate "AES cipher algorithms"
-	depends on CRYPTO && !((X86 || UML_X86) && !64BIT)
+	depends on CRYPTO && !(X86 || UML_X86)
 	help
 	  AES cipher algorithms (FIPS-197). AES uses the Rijndael 
 	  algorithm.
@@ -184,6 +184,26 @@ config CRYPTO_AES_586
 
 	  See <http://csrc.nist.gov/encryption/aes/> for more information.
 
+config CRYPTO_AES_X86_64
+	tristate "AES cipher algorithms (x86_64)"
+	depends on CRYPTO && ((X86 || UML_X86) && 64BIT)
+	help
+	  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_CAST5
 	tristate "CAST5 (CAST-128) cipher algorithm"
 	depends on CRYPTO

crypto/api.c

@@ -13,9 +13,12 @@
  * any later version.
  *
  */
+
+#include <linux/compiler.h>
 #include <linux/init.h>
 #include <linux/crypto.h>
 #include <linux/errno.h>
+#include <linux/kmod.h>
 #include <linux/rwsem.h>
 #include <linux/slab.h>
 #include "internal.h"
@@ -33,7 +36,7 @@ static inline void crypto_alg_put(struct crypto_alg *alg)
 	module_put(alg->cra_module);
 }
 
-struct crypto_alg *crypto_alg_lookup(const char *name)
+static struct crypto_alg *crypto_alg_lookup(const char *name)
 {
 	struct crypto_alg *q, *alg = NULL;
 
@@ -54,6 +57,13 @@ struct crypto_alg *crypto_alg_lookup(const char *name)
 	return alg;
 }
 
+/* A far more intelligent version of this is planned.  For now, just
+ * try an exact match on the name of the algorithm. */
+static inline struct crypto_alg *crypto_alg_mod_lookup(const char *name)
+{
+	return try_then_request_module(crypto_alg_lookup(name), name);
+}
+
 static int crypto_init_flags(struct crypto_tfm *tfm, u32 flags)
 {
 	tfm->crt_flags = 0;
@@ -117,20 +127,46 @@ static void crypto_exit_ops(struct crypto_tfm *tfm)
 	}
 }
 
+static unsigned int crypto_ctxsize(struct crypto_alg *alg, int flags)
+{
+	unsigned int len;
+
+	switch (alg->cra_flags & CRYPTO_ALG_TYPE_MASK) {
+	default:
+		BUG();
+
+	case CRYPTO_ALG_TYPE_CIPHER:
+		len = crypto_cipher_ctxsize(alg, flags);
+		break;
+		
+	case CRYPTO_ALG_TYPE_DIGEST:
+		len = crypto_digest_ctxsize(alg, flags);
+		break;
+		
+	case CRYPTO_ALG_TYPE_COMPRESS:
+		len = crypto_compress_ctxsize(alg, flags);
+		break;
+	}
+
+	return len + alg->cra_alignmask;
+}
+
 struct crypto_tfm *crypto_alloc_tfm(const char *name, u32 flags)
 {
 	struct crypto_tfm *tfm = NULL;
 	struct crypto_alg *alg;
+	unsigned int tfm_size;
 
 	alg = crypto_alg_mod_lookup(name);
 	if (alg == NULL)
 		goto out;
-	
-	tfm = kmalloc(sizeof(*tfm) + alg->cra_ctxsize, GFP_KERNEL);
+
+	tfm_size = sizeof(*tfm) + crypto_ctxsize(alg, flags);
+	tfm = kmalloc(tfm_size, GFP_KERNEL);
 	if (tfm == NULL)
 		goto out_put;
 
-	memset(tfm, 0, sizeof(*tfm) + alg->cra_ctxsize);
+	memset(tfm, 0, tfm_size);
 	
 	tfm->__crt_alg = alg;
 	
@@ -155,8 +191,14 @@ out:
 
 void crypto_free_tfm(struct crypto_tfm *tfm)
 {
-	struct crypto_alg *alg = tfm->__crt_alg;
-	int size = sizeof(*tfm) + alg->cra_ctxsize;
+	struct crypto_alg *alg;
+	int size;
+
+	if (unlikely(!tfm))
+		return;
+
+	alg = tfm->__crt_alg;
+	size = sizeof(*tfm) + alg->cra_ctxsize;
 
 	crypto_exit_ops(tfm);
 	crypto_alg_put(alg);
@@ -168,6 +210,12 @@ int crypto_register_alg(struct crypto_alg *alg)
 {
 	int ret = 0;
 	struct crypto_alg *q;
+
+	if (alg->cra_alignmask & (alg->cra_alignmask + 1))
+		return -EINVAL;
+
+	if (alg->cra_alignmask > PAGE_SIZE)
+		return -EINVAL;
 	
 	down_write(&crypto_alg_sem);
 	

crypto/cipher.c

@@ -4,6 +4,7 @@
  * Cipher operations.
  *
  * Copyright (c) 2002 James Morris <jmorris@intercode.com.au>
+ * Copyright (c) 2005 Herbert Xu <herbert@gondor.apana.org.au>
  *
  * 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
@@ -22,10 +23,6 @@
 #include "internal.h"
 #include "scatterwalk.h"
 
-typedef void (cryptfn_t)(void *, u8 *, const u8 *);
-typedef void (procfn_t)(struct crypto_tfm *, u8 *,
-                        u8*, cryptfn_t, void *);
-
 static inline void xor_64(u8 *a, const u8 *b)
 {
 	((u32 *)a)[0] ^= ((u32 *)b)[0];
@@ -39,63 +36,70 @@ static inline void xor_128(u8 *a, const u8 *b)
 	((u32 *)a)[2] ^= ((u32 *)b)[2];
 	((u32 *)a)[3] ^= ((u32 *)b)[3];
 }
- 
-static inline void *prepare_src(struct scatter_walk *walk, int bsize,
-				void *tmp, int in_place)
+
+static unsigned int crypt_slow(const struct cipher_desc *desc,
+			       struct scatter_walk *in,
+			       struct scatter_walk *out, unsigned int bsize)
 {
-	void *src = walk->data;
-	int n = bsize;
+	unsigned int alignmask = crypto_tfm_alg_alignmask(desc->tfm);
+	u8 buffer[bsize * 2 + alignmask];
+	u8 *src = (u8 *)ALIGN((unsigned long)buffer, alignmask + 1);
+	u8 *dst = src + bsize;
+	unsigned int n;
 
-	if (unlikely(scatterwalk_across_pages(walk, bsize))) {
-		src = tmp;
-		n = scatterwalk_copychunks(src, walk, bsize, 0);
-	}
-	scatterwalk_advance(walk, n);
-	return src;
+	n = scatterwalk_copychunks(src, in, bsize, 0);
+	scatterwalk_advance(in, n);
+
+	desc->prfn(desc, dst, src, bsize);
+
+	n = scatterwalk_copychunks(dst, out, bsize, 1);
+	scatterwalk_advance(out, n);
+
+	return bsize;
 }
 
-static inline void *prepare_dst(struct scatter_walk *walk, int bsize,
-				void *tmp, int in_place)
+static inline unsigned int crypt_fast(const struct cipher_desc *desc,
+				      struct scatter_walk *in,
+				      struct scatter_walk *out,
+				      unsigned int nbytes, u8 *tmp)
 {
-	void *dst = walk->data;
+	u8 *src, *dst;
 
-	if (unlikely(scatterwalk_across_pages(walk, bsize)) || in_place)
+	src = in->data;
+	dst = scatterwalk_samebuf(in, out) ? src : out->data;
+
+	if (tmp) {
+		memcpy(tmp, in->data, nbytes);
+		src = tmp;
 		dst = tmp;
-	return dst;
-}
+	}
 
-static inline void complete_src(struct scatter_walk *walk, int bsize,
-				void *src, int in_place)
-{
-}
+	nbytes = desc->prfn(desc, dst, src, nbytes);
 
-static inline void complete_dst(struct scatter_walk *walk, int bsize,
-				void *dst, int in_place)
-{
-	int n = bsize;
+	if (tmp)
+		memcpy(out->data, tmp, nbytes);
+
+	scatterwalk_advance(in, nbytes);
+	scatterwalk_advance(out, nbytes);
 
-	if (unlikely(scatterwalk_across_pages(walk, bsize)))
-		n = scatterwalk_copychunks(dst, walk, bsize, 1);
-	else if (in_place)
-		memcpy(walk->data, dst, bsize);
-	scatterwalk_advance(walk, n);
+	return nbytes;
 }
 
 /* 
  * Generic encrypt/decrypt wrapper for ciphers, handles operations across
  * multiple page boundaries by using temporary blocks.  In user context,
- * the kernel is given a chance to schedule us once per block.
+ * the kernel is given a chance to schedule us once per page.
  */
-static int crypt(struct crypto_tfm *tfm,
+static int crypt(const struct cipher_desc *desc,
 		 struct scatterlist *dst,
 		 struct scatterlist *src,
-                 unsigned int nbytes, cryptfn_t crfn,
-                 procfn_t prfn, void *info)
+		 unsigned int nbytes)
 {
 	struct scatter_walk walk_in, walk_out;
+	struct crypto_tfm *tfm = desc->tfm;
 	const unsigned int bsize = crypto_tfm_alg_blocksize(tfm);
-	u8 tmp_src[bsize];
-	u8 tmp_dst[bsize];
+	unsigned int alignmask = crypto_tfm_alg_alignmask(tfm);
+	unsigned long buffer = 0;
 
 	if (!nbytes)
 		return 0;
@@ -109,64 +113,144 @@ static int crypt(struct crypto_tfm *tfm,
 	scatterwalk_start(&walk_out, dst);
 
 	for(;;) {
-		u8 *src_p, *dst_p;
-		int in_place;
+		unsigned int n = nbytes;
+		u8 *tmp = NULL;
+
+		if (!scatterwalk_aligned(&walk_in, alignmask) ||
+		    !scatterwalk_aligned(&walk_out, alignmask)) {
+			if (!buffer) {
+				buffer = __get_free_page(GFP_ATOMIC);
+				if (!buffer)
+					n = 0;
+			}
+			tmp = (u8 *)buffer;
+		}
 
 		scatterwalk_map(&walk_in, 0);
 		scatterwalk_map(&walk_out, 1);
 
-		in_place = scatterwalk_samebuf(&walk_in, &walk_out);
-
-		do {
-			src_p = prepare_src(&walk_in, bsize, tmp_src,
-					    in_place);
-			dst_p = prepare_dst(&walk_out, bsize, tmp_dst,
-					    in_place);
-
-			prfn(tfm, dst_p, src_p, crfn, info);
+		n = scatterwalk_clamp(&walk_in, n);
+		n = scatterwalk_clamp(&walk_out, n);
 
-			complete_src(&walk_in, bsize, src_p, in_place);
-			complete_dst(&walk_out, bsize, dst_p, in_place);
+		if (likely(n >= bsize))
+			n = crypt_fast(desc, &walk_in, &walk_out, n, tmp);
+		else
+			n = crypt_slow(desc, &walk_in, &walk_out, bsize);
 
-			nbytes -= bsize;
-		} while (nbytes &&
-			 !scatterwalk_across_pages(&walk_in, bsize) &&
-			 !scatterwalk_across_pages(&walk_out, bsize));
+		nbytes -= n;
 
 		scatterwalk_done(&walk_in, 0, nbytes);
 		scatterwalk_done(&walk_out, 1, nbytes);
 
 		if (!nbytes)
-			return 0;
+			break;
 
 		crypto_yield(tfm);
 	}
+
+	if (buffer)
+		free_page(buffer);
+
+	return 0;
 }
 
-static void cbc_process_encrypt(struct crypto_tfm *tfm, u8 *dst, u8 *src,
-				cryptfn_t fn, void *info)
+static int crypt_iv_unaligned(struct cipher_desc *desc,
+			      struct scatterlist *dst,
+			      struct scatterlist *src,
+			      unsigned int nbytes)
 {
-	u8 *iv = info;
+	struct crypto_tfm *tfm = desc->tfm;
+	unsigned int alignmask = crypto_tfm_alg_alignmask(tfm);
+	u8 *iv = desc->info;
 
-	tfm->crt_u.cipher.cit_xor_block(iv, src);
-	fn(crypto_tfm_ctx(tfm), dst, iv);
-	memcpy(iv, dst, crypto_tfm_alg_blocksize(tfm));
+	if (unlikely(((unsigned long)iv & alignmask))) {
+		unsigned int ivsize = tfm->crt_cipher.cit_ivsize;
+		u8 buffer[ivsize + alignmask];
+		u8 *tmp = (u8 *)ALIGN((unsigned long)buffer, alignmask + 1);
+		int err;
+
+		desc->info = memcpy(tmp, iv, ivsize);
+		err = crypt(desc, dst, src, nbytes);
+		memcpy(iv, tmp, ivsize);
+
+		return err;
+	}
+
+	return crypt(desc, dst, src, nbytes);
 }
 
-static void cbc_process_decrypt(struct crypto_tfm *tfm, u8 *dst, u8 *src,
-				cryptfn_t fn, void *info)
+static unsigned int cbc_process_encrypt(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes)
 {
-	u8 *iv = info;
+	struct crypto_tfm *tfm = desc->tfm;
+	void (*xor)(u8 *, const u8 *) = tfm->crt_u.cipher.cit_xor_block;
+	int bsize = crypto_tfm_alg_blocksize(tfm);
+
+	void (*fn)(void *, u8 *, const u8 *) = desc->crfn;
+	u8 *iv = desc->info;
+	unsigned int done = 0;
+
+	do {
+		xor(iv, src);
+		fn(crypto_tfm_ctx(tfm), dst, iv);
+		memcpy(iv, dst, bsize);
 
-	fn(crypto_tfm_ctx(tfm), dst, src);
-	tfm->crt_u.cipher.cit_xor_block(dst, iv);
-	memcpy(iv, src, crypto_tfm_alg_blocksize(tfm));
+		src += bsize;
+		dst += bsize;
+	} while ((done += bsize) < nbytes);
+
+	return done;
 }
 
-static void ecb_process(struct crypto_tfm *tfm, u8 *dst, u8 *src,
-			cryptfn_t fn, void *info)
+static unsigned int cbc_process_decrypt(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes)
 {
-	fn(crypto_tfm_ctx(tfm), dst, src);
+	struct crypto_tfm *tfm = desc->tfm;
+	void (*xor)(u8 *, const u8 *) = tfm->crt_u.cipher.cit_xor_block;
+	int bsize = crypto_tfm_alg_blocksize(tfm);
+
+	u8 stack[src == dst ? bsize : 0];
+	u8 *buf = stack;
+	u8 **dst_p = src == dst ? &buf : &dst;
+
+	void (*fn)(void *, u8 *, const u8 *) = desc->crfn;
+	u8 *iv = desc->info;
+	unsigned int done = 0;
+
+	do {
+		u8 *tmp_dst = *dst_p;
+
+		fn(crypto_tfm_ctx(tfm), tmp_dst, src);
+		xor(tmp_dst, iv);
+		memcpy(iv, src, bsize);
+		if (tmp_dst != dst)
+			memcpy(dst, tmp_dst, bsize);
+
+		src += bsize;
+		dst += bsize;
+	} while ((done += bsize) < nbytes);
+
+	return done;
+}
+
+static unsigned int ecb_process(const struct cipher_desc *desc, u8 *dst,
+				const u8 *src, unsigned int nbytes)
+{
+	struct crypto_tfm *tfm = desc->tfm;
+	int bsize = crypto_tfm_alg_blocksize(tfm);
+	void (*fn)(void *, u8 *, const u8 *) = desc->crfn;
+	unsigned int done = 0;
+
+	do {
+		fn(crypto_tfm_ctx(tfm), dst, src);
+
+		src += bsize;
+		dst += bsize;
+	} while ((done += bsize) < nbytes);
+
+	return done;
 }
 
 static int setkey(struct crypto_tfm *tfm, const u8 *key, unsigned int keylen)
@@ -185,9 +269,14 @@ static int ecb_encrypt(struct crypto_tfm *tfm,
 		       struct scatterlist *dst,
                        struct scatterlist *src, unsigned int nbytes)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_encrypt,
-	             ecb_process, NULL);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_encrypt;
+	desc.prfn = cipher->cia_encrypt_ecb ?: ecb_process;
+
+	return crypt(&desc, dst, src, nbytes);
 }
 
 static int ecb_decrypt(struct crypto_tfm *tfm,
@@ -195,9 +284,14 @@ static int ecb_decrypt(struct crypto_tfm *tfm,
                        struct scatterlist *src,
 		       unsigned int nbytes)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_decrypt,
-	             ecb_process, NULL);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_decrypt;
+	desc.prfn = cipher->cia_decrypt_ecb ?: ecb_process;
+
+	return crypt(&desc, dst, src, nbytes);
 }
 
 static int cbc_encrypt(struct crypto_tfm *tfm,
@@ -205,9 +299,15 @@ static int cbc_encrypt(struct crypto_tfm *tfm,
                        struct scatterlist *src,
 		       unsigned int nbytes)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_encrypt,
-	             cbc_process_encrypt, tfm->crt_cipher.cit_iv);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_encrypt;
+	desc.prfn = cipher->cia_encrypt_cbc ?: cbc_process_encrypt;
+	desc.info = tfm->crt_cipher.cit_iv;
+
+	return crypt(&desc, dst, src, nbytes);
 }
 
 static int cbc_encrypt_iv(struct crypto_tfm *tfm,
@@ -215,9 +315,15 @@ static int cbc_encrypt_iv(struct crypto_tfm *tfm,
                           struct scatterlist *src,
                           unsigned int nbytes, u8 *iv)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_encrypt,
-	             cbc_process_encrypt, iv);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_encrypt;
+	desc.prfn = cipher->cia_encrypt_cbc ?: cbc_process_encrypt;
+	desc.info = iv;
+
+	return crypt_iv_unaligned(&desc, dst, src, nbytes);
 }
 
 static int cbc_decrypt(struct crypto_tfm *tfm,
@@ -225,9 +331,15 @@ static int cbc_decrypt(struct crypto_tfm *tfm,
                        struct scatterlist *src,
 		       unsigned int nbytes)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_decrypt,
-	             cbc_process_decrypt, tfm->crt_cipher.cit_iv);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_decrypt;
+	desc.prfn = cipher->cia_decrypt_cbc ?: cbc_process_decrypt;
+	desc.info = tfm->crt_cipher.cit_iv;
+
+	return crypt(&desc, dst, src, nbytes);
 }
 
 static int cbc_decrypt_iv(struct crypto_tfm *tfm,
@@ -235,9 +347,15 @@ static int cbc_decrypt_iv(struct crypto_tfm *tfm,
                           struct scatterlist *src,
                           unsigned int nbytes, u8 *iv)
 {
-	return crypt(tfm, dst, src, nbytes,
-	             tfm->__crt_alg->cra_cipher.cia_decrypt,
-	             cbc_process_decrypt, iv);
+	struct cipher_desc desc;
+	struct cipher_alg *cipher = &tfm->__crt_alg->cra_cipher;
+
+	desc.tfm = tfm;
+	desc.crfn = cipher->cia_decrypt;
+	desc.prfn = cipher->cia_decrypt_cbc ?: cbc_process_decrypt;
+	desc.info = iv;
+
+	return crypt_iv_unaligned(&desc, dst, src, nbytes);
 }
 
 static int nocrypt(struct crypto_tfm *tfm,
@@ -306,6 +424,8 @@ int crypto_init_cipher_ops(struct crypto_tfm *tfm)
 	}
 	
 	if (ops->cit_mode == CRYPTO_TFM_MODE_CBC) {
+		unsigned int align;
+		unsigned long addr;
 	    	
 	    	switch (crypto_tfm_alg_blocksize(tfm)) {
 	    	case 8:
@@ -325,9 +445,11 @@ int crypto_init_cipher_ops(struct crypto_tfm *tfm)
 	    	}
 	    	
 		ops->cit_ivsize = crypto_tfm_alg_blocksize(tfm);
-	    	ops->cit_iv = kmalloc(ops->cit_ivsize, GFP_KERNEL);
-		if (ops->cit_iv == NULL)
-			ret = -ENOMEM;
+		align = crypto_tfm_alg_alignmask(tfm) + 1;
+		addr = (unsigned long)crypto_tfm_ctx(tfm);
+		addr = ALIGN(addr, align);
+		addr += ALIGN(tfm->__crt_alg->cra_ctxsize, align);
+		ops->cit_iv = (void *)addr;
 	}
 
 out:	
@@ -336,6 +458,4 @@ out:
 
 void crypto_exit_cipher_ops(struct crypto_tfm *tfm)
 {
-	if (tfm->crt_cipher.cit_iv)
-		kfree(tfm->crt_cipher.cit_iv);
 }

crypto/des.c

@@ -1,18 +1,9 @@
-/* 
+/*
  * Cryptographic API.
  *
  * DES & Triple DES EDE Cipher Algorithms.
  *
- * Originally released as descore by Dana L. How <how@isl.stanford.edu>.
- * Modified by Raimar Falke <rf13@inf.tu-dresden.de> for the Linux-Kernel.
- * Derived from Cryptoapi and Nettle implementations, adapted for in-place
- * scatterlist interface.  Changed LGPL to GPL per section 3 of the LGPL.
- *
- * Copyright (c) 1992 Dana L. How.
- * Copyright (c) Raimar Falke <rf13@inf.tu-dresden.de> 
- * Copyright (c) Gisle Sælensminde <gisle@ii.uib.no>
- * Copyright (C) 2001 Niels Möller.
- * Copyright (c) 2002 James Morris <jmorris@intercode.com.au>
+ * Copyright (c) 2005 Dag Arne Osvik <da@osvik.no>
  *
  * 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
@@ -20,11 +11,11 @@
  * (at your option) any later version.
  *
  */
+
+#include <linux/bitops.h>
 #include <linux/init.h>
 #include <linux/module.h>
-#include <linux/mm.h>
 #include <linux/errno.h>
-#include <asm/scatterlist.h>
 #include <linux/crypto.h>
 
 #define DES_KEY_SIZE		8
@@ -35,1157 +26,826 @@
 #define DES3_EDE_EXPKEY_WORDS	(3 * DES_EXPKEY_WORDS)
 #define DES3_EDE_BLOCK_SIZE	DES_BLOCK_SIZE
 
-#define ROR(d,c,o)	((d) = (d) >> (c) | (d) << (o))
+#define ROL(x, r) ((x) = rol32((x), (r)))
+#define ROR(x, r) ((x) = ror32((x), (r)))
 
 struct des_ctx {
-	u8 iv[DES_BLOCK_SIZE];
 	u32 expkey[DES_EXPKEY_WORDS];
 };
 
 struct des3_ede_ctx {
-	u8 iv[DES_BLOCK_SIZE];
 	u32 expkey[DES3_EDE_EXPKEY_WORDS];
 };
 
-static const u32 des_keymap[] = {
-	0x02080008, 0x02082000, 0x00002008, 0x00000000,
-	0x02002000, 0x00080008, 0x02080000, 0x02082008,
-	0x00000008, 0x02000000, 0x00082000, 0x00002008,
-	0x00082008, 0x02002008, 0x02000008, 0x02080000,
-	0x00002000, 0x00082008, 0x00080008, 0x02002000,
-	0x02082008, 0x02000008, 0x00000000, 0x00082000,
-	0x02000000, 0x00080000, 0x02002008, 0x02080008,
-	0x00080000, 0x00002000, 0x02082000, 0x00000008,
-	0x00080000, 0x00002000, 0x02000008, 0x02082008,
-	0x00002008, 0x02000000, 0x00000000, 0x00082000,
-	0x02080008, 0x02002008, 0x02002000, 0x00080008,
-	0x02082000, 0x00000008, 0x00080008, 0x02002000,
-	0x02082008, 0x00080000, 0x02080000, 0x02000008,
-	0x00082000, 0x00002008, 0x02002008, 0x02080000,
-	0x00000008, 0x02082000, 0x00082008, 0x00000000,
-	0x02000000, 0x02080008, 0x00002000, 0x00082008,
-
-	0x08000004, 0x00020004, 0x00000000, 0x08020200,
-	0x00020004, 0x00000200, 0x08000204, 0x00020000,
-	0x00000204, 0x08020204, 0x00020200, 0x08000000,
-	0x08000200, 0x08000004, 0x08020000, 0x00020204,
-	0x00020000, 0x08000204, 0x08020004, 0x00000000,
-	0x00000200, 0x00000004, 0x08020200, 0x08020004,
-	0x08020204, 0x08020000, 0x08000000, 0x00000204,
-	0x00000004, 0x00020200, 0x00020204, 0x08000200,
-	0x00000204, 0x08000000, 0x08000200, 0x00020204,
-	0x08020200, 0x00020004, 0x00000000, 0x08000200,
-	0x08000000, 0x00000200, 0x08020004, 0x00020000,
-	0x00020004, 0x08020204, 0x00020200, 0x00000004,
-	0x08020204, 0x00020200, 0x00020000, 0x08000204,
-	0x08000004, 0x08020000, 0x00020204, 0x00000000,
-	0x00000200, 0x08000004, 0x08000204, 0x08020200,
-	0x08020000, 0x00000204, 0x00000004, 0x08020004,
-
-	0x80040100, 0x01000100, 0x80000000, 0x81040100,
-	0x00000000, 0x01040000, 0x81000100, 0x80040000,
-	0x01040100, 0x81000000, 0x01000000, 0x80000100,
-	0x81000000, 0x80040100, 0x00040000, 0x01000000,
-	0x81040000, 0x00040100, 0x00000100, 0x80000000,
-	0x00040100, 0x81000100, 0x01040000, 0x00000100,
-	0x80000100, 0x00000000, 0x80040000, 0x01040100,
-	0x01000100, 0x81040000, 0x81040100, 0x00040000,
-	0x81040000, 0x80000100, 0x00040000, 0x81000000,
-	0x00040100, 0x01000100, 0x80000000, 0x01040000,
-	0x81000100, 0x00000000, 0x00000100, 0x80040000,
-	0x00000000, 0x81040000, 0x01040100, 0x00000100,
-	0x01000000, 0x81040100, 0x80040100, 0x00040000,
-	0x81040100, 0x80000000, 0x01000100, 0x80040100,
-	0x80040000, 0x00040100, 0x01040000, 0x81000100,
-	0x80000100, 0x01000000, 0x81000000, 0x01040100,
-
-	0x04010801, 0x00000000, 0x00010800, 0x04010000,
-	0x04000001, 0x00000801, 0x04000800, 0x00010800,
-	0x00000800, 0x04010001, 0x00000001, 0x04000800,
-	0x00010001, 0x04010800, 0x04010000, 0x00000001,
-	0x00010000, 0x04000801, 0x04010001, 0x00000800,
-	0x00010801, 0x04000000, 0x00000000, 0x00010001,
-	0x04000801, 0x00010801, 0x04010800, 0x04000001,
-	0x04000000, 0x00010000, 0x00000801, 0x04010801,
-	0x00010001, 0x04010800, 0x04000800, 0x00010801,
-	0x04010801, 0x00010001, 0x04000001, 0x00000000,
-	0x04000000, 0x00000801, 0x00010000, 0x04010001,
-	0x00000800, 0x04000000, 0x00010801, 0x04000801,
-	0x04010800, 0x00000800, 0x00000000, 0x04000001,
-	0x00000001, 0x04010801, 0x00010800, 0x04010000,
-	0x04010001, 0x00010000, 0x00000801, 0x04000800,
-	0x04000801, 0x00000001, 0x04010000, 0x00010800,
-
-	0x00000400, 0x00000020, 0x00100020, 0x40100000,
-	0x40100420, 0x40000400, 0x00000420, 0x00000000,
-	0x00100000, 0x40100020, 0x40000020, 0x00100400,
-	0x40000000, 0x00100420, 0x00100400, 0x40000020,
-	0x40100020, 0x00000400, 0x40000400, 0x40100420,
-	0x00000000, 0x00100020, 0x40100000, 0x00000420,
-	0x40100400, 0x40000420, 0x00100420, 0x40000000,
-	0x40000420, 0x40100400, 0x00000020, 0x00100000,
-	0x40000420, 0x00100400, 0x40100400, 0x40000020,
-	0x00000400, 0x00000020, 0x00100000, 0x40100400,
-	0x40100020, 0x40000420, 0x00000420, 0x00000000,
-	0x00000020, 0x40100000, 0x40000000, 0x00100020,
-	0x00000000, 0x40100020, 0x00100020, 0x00000420,
-	0x40000020, 0x00000400, 0x40100420, 0x00100000,
-	0x00100420, 0x40000000, 0x40000400, 0x40100420,
-	0x40100000, 0x00100420, 0x00100400, 0x40000400,
-
-	0x00800000, 0x00001000, 0x00000040, 0x00801042,
-	0x00801002, 0x00800040, 0x00001042, 0x00801000,
-	0x00001000, 0x00000002, 0x00800002, 0x00001040,
-	0x00800042, 0x00801002, 0x00801040, 0x00000000,
-	0x00001040, 0x00800000, 0x00001002, 0x00000042,
-	0x00800040, 0x00001042, 0x00000000, 0x00800002,
-	0x00000002, 0x00800042, 0x00801042, 0x00001002,
-	0x00801000, 0x00000040, 0x00000042, 0x00801040,
-	0x00801040, 0x00800042, 0x00001002, 0x00801000,
-	0x00001000, 0x00000002, 0x00800002, 0x00800040,
-	0x00800000, 0x00001040, 0x00801042, 0x00000000,
-	0x00001042, 0x00800000, 0x00000040, 0x00001002,
-	0x00800042, 0x00000040, 0x00000000, 0x00801042,
-	0x00801002, 0x00801040, 0x00000042, 0x00001000,
-	0x00001040, 0x00801002, 0x00800040, 0x00000042,
-	0x00000002, 0x00001042, 0x00801000, 0x00800002,
-
-	0x10400000, 0x00404010, 0x00000010, 0x10400010,
-	0x10004000, 0x00400000, 0x10400010, 0x00004010,
-	0x00400010, 0x00004000, 0x00404000, 0x10000000,
-	0x10404010, 0x10000010, 0x10000000, 0x10404000,
-	0x00000000, 0x10004000, 0x00404010, 0x00000010,
-	0x10000010, 0x10404010, 0x00004000, 0x10400000,
-	0x10404000, 0x00400010, 0x10004010, 0x00404000,
-	0x00004010, 0x00000000, 0x00400000, 0x10004010,
-	0x00404010, 0x00000010, 0x10000000, 0x00004000,
-	0x10000010, 0x10004000, 0x00404000, 0x10400010,
-	0x00000000, 0x00404010, 0x00004010, 0x10404000,
-	0x10004000, 0x00400000, 0x10404010, 0x10000000,
-	0x10004010, 0x10400000, 0x00400000, 0x10404010,
-	0x00004000, 0x00400010, 0x10400010, 0x00004010,
-	0x00400010, 0x00000000, 0x10404000, 0x10000010,
-	0x10400000, 0x10004010, 0x00000010, 0x00404000,
-
-	0x00208080, 0x00008000, 0x20200000, 0x20208080,
-	0x00200000, 0x20008080, 0x20008000, 0x20200000,
-	0x20008080, 0x00208080, 0x00208000, 0x20000080,
-	0x20200080, 0x00200000, 0x00000000, 0x20008000,
-	0x00008000, 0x20000000, 0x00200080, 0x00008080,
-	0x20208080, 0x00208000, 0x20000080, 0x00200080,
-	0x20000000, 0x00000080, 0x00008080, 0x20208000,
-	0x00000080, 0x20200080, 0x20208000, 0x00000000,
-	0x00000000, 0x20208080, 0x00200080, 0x20008000,
-	0x00208080, 0x00008000, 0x20000080, 0x00200080,
-	0x20208000, 0x00000080, 0x00008080, 0x20200000,
-	0x20008080, 0x20000000, 0x20200000, 0x00208000,
-	0x20208080, 0x00008080, 0x00208000, 0x20200080,
-	0x00200000, 0x20000080, 0x20008000, 0x00000000,
-	0x00008000, 0x00200000, 0x20200080, 0x00208080,
-	0x20000000, 0x20208000, 0x00000080, 0x20008080,
+/* Lookup tables for key expansion */
+
+static const u8 pc1[256] = {
+	0x00, 0x00, 0x40, 0x04, 0x10, 0x10, 0x50, 0x14,
+	0x04, 0x40, 0x44, 0x44, 0x14, 0x50, 0x54, 0x54,
+	0x02, 0x02, 0x42, 0x06, 0x12, 0x12, 0x52, 0x16,
+	0x06, 0x42, 0x46, 0x46, 0x16, 0x52, 0x56, 0x56,
+	0x80, 0x08, 0xc0, 0x0c, 0x90, 0x18, 0xd0, 0x1c,
+	0x84, 0x48, 0xc4, 0x4c, 0x94, 0x58, 0xd4, 0x5c,
+	0x82, 0x0a, 0xc2, 0x0e, 0x92, 0x1a, 0xd2, 0x1e,
+	0x86, 0x4a, 0xc6, 0x4e, 0x96, 0x5a, 0xd6, 0x5e,
+	0x20, 0x20, 0x60, 0x24, 0x30, 0x30, 0x70, 0x34,
+	0x24, 0x60, 0x64, 0x64, 0x34, 0x70, 0x74, 0x74,
+	0x22, 0x22, 0x62, 0x26, 0x32, 0x32, 0x72, 0x36,
+	0x26, 0x62, 0x66, 0x66, 0x36, 0x72, 0x76, 0x76,
+	0xa0, 0x28, 0xe0, 0x2c, 0xb0, 0x38, 0xf0, 0x3c,
+	0xa4, 0x68, 0xe4, 0x6c, 0xb4, 0x78, 0xf4, 0x7c,
+	0xa2, 0x2a, 0xe2, 0x2e, 0xb2, 0x3a, 0xf2, 0x3e,
+	0xa6, 0x6a, 0xe6, 0x6e, 0xb6, 0x7a, 0xf6, 0x7e,
+	0x08, 0x80, 0x48, 0x84, 0x18, 0x90, 0x58, 0x94,
+	0x0c, 0xc0, 0x4c, 0xc4, 0x1c, 0xd0, 0x5c, 0xd4,
+	0x0a, 0x82, 0x4a, 0x86, 0x1a, 0x92, 0x5a, 0x96,
+	0x0e, 0xc2, 0x4e, 0xc6, 0x1e, 0xd2, 0x5e, 0xd6,
+	0x88, 0x88, 0xc8, 0x8c, 0x98, 0x98, 0xd8, 0x9c,
+	0x8c, 0xc8, 0xcc, 0xcc, 0x9c, 0xd8, 0xdc, 0xdc,
+	0x8a, 0x8a, 0xca, 0x8e, 0x9a, 0x9a, 0xda, 0x9e,
+	0x8e, 0xca, 0xce, 0xce, 0x9e, 0xda, 0xde, 0xde,
+	0x28, 0xa0, 0x68, 0xa4, 0x38, 0xb0, 0x78, 0xb4,
+	0x2c, 0xe0, 0x6c, 0xe4, 0x3c, 0xf0, 0x7c, 0xf4,
+	0x2a, 0xa2, 0x6a, 0xa6, 0x3a, 0xb2, 0x7a, 0xb6,
+	0x2e, 0xe2, 0x6e, 0xe6, 0x3e, 0xf2, 0x7e, 0xf6,
+	0xa8, 0xa8, 0xe8, 0xac, 0xb8, 0xb8, 0xf8, 0xbc,
+	0xac, 0xe8, 0xec, 0xec, 0xbc, 0xf8, 0xfc, 0xfc,
+	0xaa, 0xaa, 0xea, 0xae, 0xba, 0xba, 0xfa, 0xbe,
+	0xae, 0xea, 0xee, 0xee, 0xbe, 0xfa, 0xfe, 0xfe
 };
 
-static const u8 rotors[] = {
-	34, 13,  5, 46, 47, 18, 32, 41, 11, 53, 33, 20,
-	14, 36, 30, 24, 49,  2, 15, 37, 42, 50,  0, 21,
-	38, 48,  6, 26, 39,  4, 52, 25, 12, 27, 31, 40,
-	1, 17, 28, 29, 23, 51, 35,  7,  3, 22,  9, 43,
-
-	41, 20, 12, 53, 54, 25, 39, 48, 18, 31, 40, 27,
-	21, 43, 37,  0,  1,  9, 22, 44, 49,  2,  7, 28,
-	45, 55, 13, 33, 46, 11,  6, 32, 19, 34, 38, 47,
-	8, 24, 35, 36, 30,  3, 42, 14, 10, 29, 16, 50,
-
-	55, 34, 26, 38, 11, 39, 53,  5, 32, 45, 54, 41,
-	35,  2, 51, 14, 15, 23, 36,  3,  8, 16, 21, 42,
-	6, 12, 27, 47, 31, 25, 20, 46, 33, 48, 52,  4,
-	22,  7, 49, 50, 44, 17,  1, 28, 24, 43, 30,  9,
-
-	12, 48, 40, 52, 25, 53, 38, 19, 46,  6, 11, 55,
-	49, 16, 10, 28, 29, 37, 50, 17, 22, 30, 35,  1,
-	20, 26, 41,  4, 45, 39, 34, 31, 47,  5, 13, 18,
-	36, 21,  8,  9,  3,  0, 15, 42,  7,  2, 44, 23,
-
-	26,  5, 54, 13, 39, 38, 52, 33, 31, 20, 25, 12,
-	8, 30, 24, 42, 43, 51,  9,  0, 36, 44, 49, 15,
-	34, 40, 55, 18,  6, 53, 48, 45,  4, 19, 27, 32,
-	50, 35, 22, 23, 17, 14, 29,  1, 21, 16,  3, 37,
-
-	40, 19, 11, 27, 53, 52, 13, 47, 45, 34, 39, 26,
-	22, 44,  7,  1,  2, 10, 23, 14, 50,  3,  8, 29,
-	48, 54, 12, 32, 20, 38,  5,  6, 18, 33, 41, 46,
-	9, 49, 36, 37,  0, 28, 43, 15, 35, 30, 17, 51,
-
-	54, 33, 25, 41, 38, 13, 27,  4,  6, 48, 53, 40,
-	36,  3, 21, 15, 16, 24, 37, 28,  9, 17, 22, 43,
-	5, 11, 26, 46, 34, 52, 19, 20, 32, 47, 55, 31,
-	23,  8, 50, 51, 14, 42,  2, 29, 49, 44,  0, 10,
-
-	11, 47, 39, 55, 52, 27, 41, 18, 20,  5, 38, 54,
-	50, 17, 35, 29, 30,  7, 51, 42, 23,  0, 36,  2,
-	19, 25, 40, 31, 48, 13, 33, 34, 46,  4, 12, 45,
-	37, 22,  9, 10, 28,  1, 16, 43,  8,  3, 14, 24,
-
-	18, 54, 46,  5,  6, 34, 48, 25, 27, 12, 45,  4,
-	2, 24, 42, 36, 37, 14,  3, 49, 30,  7, 43,  9,
-	26, 32, 47, 38, 55, 20, 40, 41, 53, 11, 19, 52,
-	44, 29, 16, 17, 35,  8, 23, 50, 15, 10, 21,  0,
-
-	32, 11, 31, 19, 20, 48,  5, 39, 41, 26,  6, 18,
-	16,  7,  1, 50, 51, 28, 17,  8, 44, 21,  2, 23,
-	40, 46,  4, 52, 12, 34, 54, 55, 38, 25, 33, 13,
-	3, 43, 30,  0, 49, 22, 37,  9, 29, 24, 35, 14,
-
-	46, 25, 45, 33, 34,  5, 19, 53, 55, 40, 20, 32,
-	30, 21, 15,  9, 10, 42,  0, 22,  3, 35, 16, 37,
-	54, 31, 18, 13, 26, 48, 11, 12, 52, 39, 47, 27,
-	17,  2, 44, 14,  8, 36, 51, 23, 43,  7, 49, 28,
-
-	31, 39,  6, 47, 48, 19, 33, 38, 12, 54, 34, 46,
-	44, 35, 29, 23, 24,  1, 14, 36, 17, 49, 30, 51,
-	11, 45, 32, 27, 40,  5, 25, 26, 13, 53,  4, 41,
-	0, 16,  3, 28, 22, 50, 10, 37,  2, 21,  8, 42,
-
-	45, 53, 20,  4,  5, 33, 47, 52, 26, 11, 48, 31,
-	3, 49, 43, 37,  7, 15, 28, 50,  0,  8, 44, 10,
-	25,  6, 46, 41, 54, 19, 39, 40, 27, 38, 18, 55,
-	14, 30, 17, 42, 36,  9, 24, 51, 16, 35, 22,  1,
-
-	6, 38, 34, 18, 19, 47,  4, 13, 40, 25,  5, 45,
-	17,  8,  2, 51, 21, 29, 42,  9, 14, 22,  3, 24,
-	39, 20, 31, 55, 11, 33, 53, 54, 41, 52, 32, 12,
-	28, 44,  0,  1, 50, 23,  7, 10, 30, 49, 36, 15,
-
-	20, 52, 48, 32, 33,  4, 18, 27, 54, 39, 19,  6,
-	0, 22, 16, 10, 35, 43,  1, 23, 28, 36, 17,  7,
-	53, 34, 45, 12, 25, 47, 38, 11, 55, 13, 46, 26,
-	42,  3, 14, 15,  9, 37, 21, 24, 44,  8, 50, 29,
-
-	27,  6, 55, 39, 40, 11, 25, 34,  4, 46, 26, 13,
-	7, 29, 23, 17, 42, 50,  8, 30, 35, 43, 24, 14,
-	31, 41, 52, 19, 32, 54, 45, 18,  5, 20, 53, 33,
-	49, 10, 21, 22, 16, 44, 28,  0, 51, 15,  2, 36,
+static const u8 rs[256] = {
+	0x00, 0x00, 0x80, 0x80, 0x02, 0x02, 0x82, 0x82,
+	0x04, 0x04, 0x84, 0x84, 0x06, 0x06, 0x86, 0x86,
+	0x08, 0x08, 0x88, 0x88, 0x0a, 0x0a, 0x8a, 0x8a,
+	0x0c, 0x0c, 0x8c, 0x8c, 0x0e, 0x0e, 0x8e, 0x8e,
+	0x10, 0x10, 0x90, 0x90, 0x12, 0x12, 0x92, 0x92,
+	0x14, 0x14, 0x94, 0x94, 0x16, 0x16, 0x96, 0x96,
+	0x18, 0x18, 0x98, 0x98, 0x1a, 0x1a, 0x9a, 0x9a,
+	0x1c, 0x1c, 0x9c, 0x9c, 0x1e, 0x1e, 0x9e, 0x9e,
+	0x20, 0x20, 0xa0, 0xa0, 0x22, 0x22, 0xa2, 0xa2,
+	0x24, 0x24, 0xa4, 0xa4, 0x26, 0x26, 0xa6, 0xa6,
+	0x28, 0x28, 0xa8, 0xa8, 0x2a, 0x2a, 0xaa, 0xaa,
+	0x2c, 0x2c, 0xac, 0xac, 0x2e, 0x2e, 0xae, 0xae,
+	0x30, 0x30, 0xb0, 0xb0, 0x32, 0x32, 0xb2, 0xb2,
+	0x34, 0x34, 0xb4, 0xb4, 0x36, 0x36, 0xb6, 0xb6,
+	0x38, 0x38, 0xb8, 0xb8, 0x3a, 0x3a, 0xba, 0xba,
+	0x3c, 0x3c, 0xbc, 0xbc, 0x3e, 0x3e, 0xbe, 0xbe,
+	0x40, 0x40, 0xc0, 0xc0, 0x42, 0x42, 0xc2, 0xc2,
+	0x44, 0x44, 0xc4, 0xc4, 0x46, 0x46, 0xc6, 0xc6,
+	0x48, 0x48, 0xc8, 0xc8, 0x4a, 0x4a, 0xca, 0xca,
+	0x4c, 0x4c, 0xcc, 0xcc, 0x4e, 0x4e, 0xce, 0xce,
+	0x50, 0x50, 0xd0, 0xd0, 0x52, 0x52, 0xd2, 0xd2,
+	0x54, 0x54, 0xd4, 0xd4, 0x56, 0x56, 0xd6, 0xd6,
+	0x58, 0x58, 0xd8, 0xd8, 0x5a, 0x5a, 0xda, 0xda,
+	0x5c, 0x5c, 0xdc, 0xdc, 0x5e, 0x5e, 0xde, 0xde,
+	0x60, 0x60, 0xe0, 0xe0, 0x62, 0x62, 0xe2, 0xe2,
+	0x64, 0x64, 0xe4, 0xe4, 0x66, 0x66, 0xe6, 0xe6,
+	0x68, 0x68, 0xe8, 0xe8, 0x6a, 0x6a, 0xea, 0xea,
+	0x6c, 0x6c, 0xec, 0xec, 0x6e, 0x6e, 0xee, 0xee,
+	0x70, 0x70, 0xf0, 0xf0, 0x72, 0x72, 0xf2, 0xf2,
+	0x74, 0x74, 0xf4, 0xf4, 0x76, 0x76, 0xf6, 0xf6,
+	0x78, 0x78, 0xf8, 0xf8, 0x7a, 0x7a, 0xfa, 0xfa,
+	0x7c, 0x7c, 0xfc, 0xfc, 0x7e, 0x7e, 0xfe, 0xfe
 };
 
-static const u8 parity[] = {
-	8,1,0,8,0,8,8,0,0,8,8,0,8,0,2,8,0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,3,
-	0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,
-	0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,
-	8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,
-	0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,
-	8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,
-	8,0,0,8,0,8,8,0,0,8,8,0,8,0,0,8,0,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,
-	4,8,8,0,8,0,0,8,8,0,0,8,0,8,8,0,8,5,0,8,0,8,8,0,0,8,8,0,8,0,6,8,
+static const u32 pc2[1024] = {
+	0x00000000, 0x00000000, 0x00000000, 0x00000000,
+	0x00040000, 0x00000000, 0x04000000, 0x00100000,
+	0x00400000, 0x00000008, 0x00000800, 0x40000000,
+	0x00440000, 0x00000008, 0x04000800, 0x40100000,
+	0x00000400, 0x00000020, 0x08000000, 0x00000100,
+	0x00040400, 0x00000020, 0x0c000000, 0x00100100,
+	0x00400400, 0x00000028, 0x08000800, 0x40000100,
+	0x00440400, 0x00000028, 0x0c000800, 0x40100100,
+	0x80000000, 0x00000010, 0x00000000, 0x00800000,
+	0x80040000, 0x00000010, 0x04000000, 0x00900000,
+	0x80400000, 0x00000018, 0x00000800, 0x40800000,
+	0x80440000, 0x00000018, 0x04000800, 0x40900000,
+	0x80000400, 0x00000030, 0x08000000, 0x00800100,
+	0x80040400, 0x00000030, 0x0c000000, 0x00900100,
+	0x80400400, 0x00000038, 0x08000800, 0x40800100,
+	0x80440400, 0x00000038, 0x0c000800, 0x40900100,
+	0x10000000, 0x00000000, 0x00200000, 0x00001000,
+	0x10040000, 0x00000000, 0x04200000, 0x00101000,
+	0x10400000, 0x00000008, 0x00200800, 0x40001000,
+	0x10440000, 0x00000008, 0x04200800, 0x40101000,
+	0x10000400, 0x00000020, 0x08200000, 0x00001100,
+	0x10040400, 0x00000020, 0x0c200000, 0x00101100,
+	0x10400400, 0x00000028, 0x08200800, 0x40001100,
+	0x10440400, 0x00000028, 0x0c200800, 0x40101100,
+	0x90000000, 0x00000010, 0x00200000, 0x00801000,
+	0x90040000, 0x00000010, 0x04200000, 0x00901000,
+	0x90400000, 0x00000018, 0x00200800, 0x40801000,
+	0x90440000, 0x00000018, 0x04200800, 0x40901000,
+	0x90000400, 0x00000030, 0x08200000, 0x00801100,
+	0x90040400, 0x00000030, 0x0c200000, 0x00901100,
+	0x90400400, 0x00000038, 0x08200800, 0x40801100,
+	0x90440400, 0x00000038, 0x0c200800, 0x40901100,
+	0x00000200, 0x00080000, 0x00000000, 0x00000004,
+	0x00040200, 0x00080000, 0x04000000, 0x00100004,
+	0x00400200, 0x00080008, 0x00000800, 0x40000004,
+	0x00440200, 0x00080008, 0x04000800, 0x40100004,
+	0x00000600, 0x00080020, 0x08000000, 0x00000104,
+	0x00040600, 0x00080020, 0x0c000000, 0x00100104,
+	0x00400600, 0x00080028, 0x08000800, 0x40000104,
+	0x00440600, 0x00080028, 0x0c000800, 0x40100104,
+	0x80000200, 0x00080010, 0x00000000, 0x00800004,
+	0x80040200, 0x00080010, 0x04000000, 0x00900004,
+	0x80400200, 0x00080018, 0x00000800, 0x40800004,
+	0x80440200, 0x00080018, 0x04000800, 0x40900004,
+	0x80000600, 0x00080030, 0x08000000, 0x00800104,
+	0x80040600, 0x00080030, 0x0c000000, 0x00900104,
+	0x80400600, 0x00080038, 0x08000800, 0x40800104,
+	0x80440600, 0x00080038, 0x0c000800, 0x40900104,
+	0x10000200, 0x00080000, 0x00200000, 0x00001004,
+	0x10040200, 0x00080000, 0x04200000, 0x00101004,
+	0x10400200, 0x00080008, 0x00200800, 0x40001004,
+	0x10440200, 0x00080008, 0x04200800, 0x40101004,
+	0x10000600, 0x00080020, 0x08200000, 0x00001104,
+	0x10040600, 0x00080020, 0x0c200000, 0x00101104,
+	0x10400600, 0x00080028, 0x08200800, 0x40001104,
+	0x10440600, 0x00080028, 0x0c200800, 0x40101104,
+	0x90000200, 0x00080010, 0x00200000, 0x00801004,
+	0x90040200, 0x00080010, 0x04200000, 0x00901004,
+	0x90400200, 0x00080018, 0x00200800, 0x40801004,
+	0x90440200, 0x00080018, 0x04200800, 0x40901004,
+	0x90000600, 0x00080030, 0x08200000, 0x00801104,
+	0x90040600, 0x00080030, 0x0c200000, 0x00901104,
+	0x90400600, 0x00080038, 0x08200800, 0x40801104,
+	0x90440600, 0x00080038, 0x0c200800, 0x40901104,
+	0x00000002, 0x00002000, 0x20000000, 0x00000001,
+	0x00040002, 0x00002000, 0x24000000, 0x00100001,
+	0x00400002, 0x00002008, 0x20000800, 0x40000001,
+	0x00440002, 0x00002008, 0x24000800, 0x40100001,
+	0x00000402, 0x00002020, 0x28000000, 0x00000101,
+	0x00040402, 0x00002020, 0x2c000000, 0x00100101,
+	0x00400402, 0x00002028, 0x28000800, 0x40000101,
+	0x00440402, 0x00002028, 0x2c000800, 0x40100101,
+	0x80000002, 0x00002010, 0x20000000, 0x00800001,
+	0x80040002, 0x00002010, 0x24000000, 0x00900001,
+	0x80400002, 0x00002018, 0x20000800, 0x40800001,
+	0x80440002, 0x00002018, 0x24000800, 0x40900001,
+	0x80000402, 0x00002030, 0x28000000, 0x00800101,
+	0x80040402, 0x00002030, 0x2c000000, 0x00900101,
+	0x80400402, 0x00002038, 0x28000800, 0x40800101,
+	0x80440402, 0x00002038, 0x2c000800, 0x40900101,
+	0x10000002, 0x00002000, 0x20200000, 0x00001001,
+	0x10040002, 0x00002000, 0x24200000, 0x00101001,
+	0x10400002, 0x00002008, 0x20200800, 0x40001001,
+	0x10440002, 0x00002008, 0x24200800, 0x40101001,
+	0x10000402, 0x00002020, 0x28200000, 0x00001101,
+	0x10040402, 0x00002020, 0x2c200000, 0x00101101,
+	0x10400402, 0x00002028, 0x28200800, 0x40001101,
+	0x10440402, 0x00002028, 0x2c200800, 0x40101101,
+	0x90000002, 0x00002010, 0x20200000, 0x00801001,
+	0x90040002, 0x00002010, 0x24200000, 0x00901001,
+	0x90400002, 0x00002018, 0x20200800, 0x40801001,
+	0x90440002, 0x00002018, 0x24200800, 0x40901001,
+	0x90000402, 0x00002030, 0x28200000, 0x00801101,
+	0x90040402, 0x00002030, 0x2c200000, 0x00901101,
+	0x90400402, 0x00002038, 0x28200800, 0x40801101,
+	0x90440402, 0x00002038, 0x2c200800, 0x40901101,
+	0x00000202, 0x00082000, 0x20000000, 0x00000005,
+	0x00040202, 0x00082000, 0x24000000, 0x00100005,
+	0x00400202, 0x00082008, 0x20000800, 0x40000005,
+	0x00440202, 0x00082008, 0x24000800, 0x40100005,
+	0x00000602, 0x00082020, 0x28000000, 0x00000105,
+	0x00040602, 0x00082020, 0x2c000000, 0x00100105,
+	0x00400602, 0x00082028, 0x28000800, 0x40000105,
+	0x00440602, 0x00082028, 0x2c000800, 0x40100105,
+	0x80000202, 0x00082010, 0x20000000, 0x00800005,
+	0x80040202, 0x00082010, 0x24000000, 0x00900005,
+	0x80400202, 0x00082018, 0x20000800, 0x40800005,
+	0x80440202, 0x00082018, 0x24000800, 0x40900005,
+	0x80000602, 0x00082030, 0x28000000, 0x00800105,
+	0x80040602, 0x00082030, 0x2c000000, 0x00900105,
+	0x80400602, 0x00082038, 0x28000800, 0x40800105,
+	0x80440602, 0x00082038, 0x2c000800, 0x40900105,
+	0x10000202, 0x00082000, 0x20200000, 0x00001005,
+	0x10040202, 0x00082000, 0x24200000, 0x00101005,
+	0x10400202, 0x00082008, 0x20200800, 0x40001005,
+	0x10440202, 0x00082008, 0x24200800, 0x40101005,
+	0x10000602, 0x00082020, 0x28200000, 0x00001105,
+	0x10040602, 0x00082020, 0x2c200000, 0x00101105,
+	0x10400602, 0x00082028, 0x28200800, 0x40001105,
+	0x10440602, 0x00082028, 0x2c200800, 0x40101105,
+	0x90000202, 0x00082010, 0x20200000, 0x00801005,
+	0x90040202, 0x00082010, 0x24200000, 0x00901005,
+	0x90400202, 0x00082018, 0x20200800, 0x40801005,
+	0x90440202, 0x00082018, 0x24200800, 0x40901005,
+	0x90000602, 0x00082030, 0x28200000, 0x00801105,
+	0x90040602, 0x00082030, 0x2c200000, 0x00901105,
+	0x90400602, 0x00082038, 0x28200800, 0x40801105,
+	0x90440602, 0x00082038, 0x2c200800, 0x40901105,
+
+	0x00000000, 0x00000000, 0x00000000, 0x00000000,
+	0x00000000, 0x00000008, 0x00080000, 0x10000000,
+	0x02000000, 0x00000000, 0x00000080, 0x00001000,
+	0x02000000, 0x00000008, 0x00080080, 0x10001000,
+	0x00004000, 0x00000000, 0x00000040, 0x00040000,
+	0x00004000, 0x00000008, 0x00080040, 0x10040000,
+	0x02004000, 0x00000000, 0x000000c0, 0x00041000,
+	0x02004000, 0x00000008, 0x000800c0, 0x10041000,
+	0x00020000, 0x00008000, 0x08000000, 0x00200000,
+	0x00020000, 0x00008008, 0x08080000, 0x10200000,
+	0x02020000, 0x00008000, 0x08000080, 0x00201000,
+	0x02020000, 0x00008008, 0x08080080, 0x10201000,
+	0x00024000, 0x00008000, 0x08000040, 0x00240000,
+	0x00024000, 0x00008008, 0x08080040, 0x10240000,
+	0x02024000, 0x00008000, 0x080000c0, 0x00241000,
+	0x02024000, 0x00008008, 0x080800c0, 0x10241000,
+	0x00000000, 0x01000000, 0x00002000, 0x00000020,
+	0x00000000, 0x01000008, 0x00082000, 0x10000020,
+	0x02000000, 0x01000000, 0x00002080, 0x00001020,
+	0x02000000, 0x01000008, 0x00082080, 0x10001020,
+	0x00004000, 0x01000000, 0x00002040, 0x00040020,
+	0x00004000, 0x01000008, 0x00082040, 0x10040020,
+	0x02004000, 0x01000000, 0x000020c0, 0x00041020,
+	0x02004000, 0x01000008, 0x000820c0, 0x10041020,
+	0x00020000, 0x01008000, 0x08002000, 0x00200020,
+	0x00020000, 0x01008008, 0x08082000, 0x10200020,
+	0x02020000, 0x01008000, 0x08002080, 0x00201020,
+	0x02020000, 0x01008008, 0x08082080, 0x10201020,
+	0x00024000, 0x01008000, 0x08002040, 0x00240020,
+	0x00024000, 0x01008008, 0x08082040, 0x10240020,
+	0x02024000, 0x01008000, 0x080020c0, 0x00241020,
+	0x02024000, 0x01008008, 0x080820c0, 0x10241020,
+	0x00000400, 0x04000000, 0x00100000, 0x00000004,
+	0x00000400, 0x04000008, 0x00180000, 0x10000004,
+	0x02000400, 0x04000000, 0x00100080, 0x00001004,
+	0x02000400, 0x04000008, 0x00180080, 0x10001004,
+	0x00004400, 0x04000000, 0x00100040, 0x00040004,
+	0x00004400, 0x04000008, 0x00180040, 0x10040004,
+	0x02004400, 0x04000000, 0x001000c0, 0x00041004,
+	0x02004400, 0x04000008, 0x001800c0, 0x10041004,
+	0x00020400, 0x04008000, 0x08100000, 0x00200004,
+	0x00020400, 0x04008008, 0x08180000, 0x10200004,
+	0x02020400, 0x04008000, 0x08100080, 0x00201004,
+	0x02020400, 0x04008008, 0x08180080, 0x10201004,
+	0x00024400, 0x04008000, 0x08100040, 0x00240004,
+	0x00024400, 0x04008008, 0x08180040, 0x10240004,
+	0x02024400, 0x04008000, 0x081000c0, 0x00241004,
+	0x02024400, 0x04008008, 0x081800c0, 0x10241004,
+	0x00000400, 0x05000000, 0x00102000, 0x00000024,
+	0x00000400, 0x05000008, 0x00182000, 0x10000024,
+	0x02000400, 0x05000000, 0x00102080, 0x00001024,
+	0x02000400, 0x05000008, 0x00182080, 0x10001024,
+	0x00004400, 0x05000000, 0x00102040, 0x00040024,
+	0x00004400, 0x05000008, 0x00182040, 0x10040024,
+	0x02004400, 0x05000000, 0x001020c0, 0x00041024,
+	0x02004400, 0x05000008, 0x001820c0, 0x10041024,
+	0x00020400, 0x05008000, 0x08102000, 0x00200024,
+	0x00020400, 0x05008008, 0x08182000, 0x10200024,
+	0x02020400, 0x05008000, 0x08102080, 0x00201024,
+	0x02020400, 0x05008008, 0x08182080, 0x10201024,
+	0x00024400, 0x05008000, 0x08102040, 0x00240024,
+	0x00024400, 0x05008008, 0x08182040, 0x10240024,
+	0x02024400, 0x05008000, 0x081020c0, 0x00241024,
+	0x02024400, 0x05008008, 0x081820c0, 0x10241024,
+	0x00000800, 0x00010000, 0x20000000, 0x00000010,
+	0x00000800, 0x00010008, 0x20080000, 0x10000010,
+	0x02000800, 0x00010000, 0x20000080, 0x00001010,
+	0x02000800, 0x00010008, 0x20080080, 0x10001010,
+	0x00004800, 0x00010000, 0x20000040, 0x00040010,
+	0x00004800, 0x00010008, 0x20080040, 0x10040010,
+	0x02004800, 0x00010000, 0x200000c0, 0x00041010,
+	0x02004800, 0x00010008, 0x200800c0, 0x10041010,
+	0x00020800, 0x00018000, 0x28000000, 0x00200010,
+	0x00020800, 0x00018008, 0x28080000, 0x10200010,
+	0x02020800, 0x00018000, 0x28000080, 0x00201010,
+	0x02020800, 0x00018008, 0x28080080, 0x10201010,
+	0x00024800, 0x00018000, 0x28000040, 0x00240010,
+	0x00024800, 0x00018008, 0x28080040, 0x10240010,
+	0x02024800, 0x00018000, 0x280000c0, 0x00241010,
+	0x02024800, 0x00018008, 0x280800c0, 0x10241010,
+	0x00000800, 0x01010000, 0x20002000, 0x00000030,
+	0x00000800, 0x01010008, 0x20082000, 0x10000030,
+	0x02000800, 0x01010000, 0x20002080, 0x00001030,
+	0x02000800, 0x01010008, 0x20082080, 0x10001030,
+	0x00004800, 0x01010000, 0x20002040, 0x00040030,
+	0x00004800, 0x01010008, 0x20082040, 0x10040030,
+	0x02004800, 0x01010000, 0x200020c0, 0x00041030,
+	0x02004800, 0x01010008, 0x200820c0, 0x10041030,
+	0x00020800, 0x01018000, 0x28002000, 0x00200030,
+	0x00020800, 0x01018008, 0x28082000, 0x10200030,
+	0x02020800, 0x01018000, 0x28002080, 0x00201030,
+	0x02020800, 0x01018008, 0x28082080, 0x10201030,
+	0x00024800, 0x01018000, 0x28002040, 0x00240030,
+	0x00024800, 0x01018008, 0x28082040, 0x10240030,
+	0x02024800, 0x01018000, 0x280020c0, 0x00241030,
+	0x02024800, 0x01018008, 0x280820c0, 0x10241030,
+	0x00000c00, 0x04010000, 0x20100000, 0x00000014,
+	0x00000c00, 0x04010008, 0x20180000, 0x10000014,
+	0x02000c00, 0x04010000, 0x20100080, 0x00001014,
+	0x02000c00, 0x04010008, 0x20180080, 0x10001014,
+	0x00004c00, 0x04010000, 0x20100040, 0x00040014,
+	0x00004c00, 0x04010008, 0x20180040, 0x10040014,
+	0x02004c00, 0x04010000, 0x201000c0, 0x00041014,
+	0x02004c00, 0x04010008, 0x201800c0, 0x10041014,
+	0x00020c00, 0x04018000, 0x28100000, 0x00200014,
+	0x00020c00, 0x04018008, 0x28180000, 0x10200014,
+	0x02020c00, 0x04018000, 0x28100080, 0x00201014,
+	0x02020c00, 0x04018008, 0x28180080, 0x10201014,
+	0x00024c00, 0x04018000, 0x28100040, 0x00240014,
+	0x00024c00, 0x04018008, 0x28180040, 0x10240014,
+	0x02024c00, 0x04018000, 0x281000c0, 0x00241014,
+	0x02024c00, 0x04018008, 0x281800c0, 0x10241014,
+	0x00000c00, 0x05010000, 0x20102000, 0x00000034,
+	0x00000c00, 0x05010008, 0x20182000, 0x10000034,
+	0x02000c00, 0x05010000, 0x20102080, 0x00001034,
+	0x02000c00, 0x05010008, 0x20182080, 0x10001034,
+	0x00004c00, 0x05010000, 0x20102040, 0x00040034,
+	0x00004c00, 0x05010008, 0x20182040, 0x10040034,
+	0x02004c00, 0x05010000, 0x201020c0, 0x00041034,
+	0x02004c00, 0x05010008, 0x201820c0, 0x10041034,
+	0x00020c00, 0x05018000, 0x28102000, 0x00200034,
+	0x00020c00, 0x05018008, 0x28182000, 0x10200034,
+	0x02020c00, 0x05018000, 0x28102080, 0x00201034,
+	0x02020c00, 0x05018008, 0x28182080, 0x10201034,
+	0x00024c00, 0x05018000, 0x28102040, 0x00240034,
+	0x00024c00, 0x05018008, 0x28182040, 0x10240034,
+	0x02024c00, 0x05018000, 0x281020c0, 0x00241034,
+	0x02024c00, 0x05018008, 0x281820c0, 0x10241034
 };
 
+/* S-box lookup tables */
+
+static const u32 S1[64] = {
+	0x01010400, 0x00000000, 0x00010000, 0x01010404,
+	0x01010004, 0x00010404, 0x00000004, 0x00010000,
+	0x00000400, 0x01010400, 0x01010404, 0x00000400,
+	0x01000404, 0x01010004, 0x01000000, 0x00000004,
+	0x00000404, 0x01000400, 0x01000400, 0x00010400,
+	0x00010400, 0x01010000, 0x01010000, 0x01000404,
+	0x00010004, 0x01000004, 0x01000004, 0x00010004,
+	0x00000000, 0x00000404, 0x00010404, 0x01000000,
+	0x00010000, 0x01010404, 0x00000004, 0x01010000,
+	0x01010400, 0x01000000, 0x01000000, 0x00000400,
+	0x01010004, 0x00010000, 0x00010400, 0x01000004,
+	0x00000400, 0x00000004, 0x01000404, 0x00010404,
+	0x01010404, 0x00010004, 0x01010000, 0x01000404,
+	0x01000004, 0x00000404, 0x00010404, 0x01010400,
+	0x00000404, 0x01000400, 0x01000400, 0x00000000,
+	0x00010004, 0x00010400, 0x00000000, 0x01010004
+};
 
-static void des_small_fips_encrypt(u32 *expkey, u8 *dst, const u8 *src)
-{
-	u32 x, y, z;
-	
-	x  = src[7];
-	x <<= 8;
-	x |= src[6];
-	x <<= 8;
-	x |= src[5];
-	x <<= 8;
-	x |= src[4];
-	y  = src[3];
-	y <<= 8;
-	y |= src[2];
-	y <<= 8;
-	y |= src[1];
-	y <<= 8;
-	y |= src[0];
-	z  = ((x >> 004) ^ y) & 0x0F0F0F0FL;
-	x ^= z << 004;
-	y ^= z;
-	z  = ((y >> 020) ^ x) & 0x0000FFFFL;
-	y ^= z << 020;
-	x ^= z;
-	z  = ((x >> 002) ^ y) & 0x33333333L;
-	x ^= z << 002;
-	y ^= z;
-	z  = ((y >> 010) ^ x) & 0x00FF00FFL;
-	y ^= z << 010;
-	x ^= z;
-	x  = x >> 1 | x << 31;
-	z  = (x ^ y) & 0x55555555L;
-	y ^= z;
-	x ^= z;
-	y  = y >> 1 | y << 31;
-	z  = expkey[0];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[1];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[2];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[3];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[4];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[5];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[6];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[7];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[8];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[9];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[10];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[11];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[12];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[13];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[14];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[15];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[16];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[17];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[18];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[19];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[20];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[21];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[22];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[23];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[24];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[25];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[26];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[27];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[28];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[29];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[30];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[31];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	x  = x << 1 | x >> 31;
-	z  = (x ^ y) & 0x55555555L;
-	y ^= z;
-	x ^= z;
-	y  = y << 1 | y >> 31;
-	z  = ((x >> 010) ^ y) & 0x00FF00FFL;
-	x ^= z << 010;
-	y ^= z;
-	z  = ((y >> 002) ^ x) & 0x33333333L;
-	y ^= z << 002;
-	x ^= z;
-	z  = ((x >> 020) ^ y) & 0x0000FFFFL;
-	x ^= z << 020;
-	y ^= z;
-	z  = ((y >> 004) ^ x) & 0x0F0F0F0FL;
-	y ^= z << 004;
-	x ^= z;
-	dst[0] = x;
-	x >>= 8;
-	dst[1] = x;
-	x >>= 8;
-	dst[2] = x;
-	x >>= 8;
-	dst[3] = x;
-	dst[4] = y;
-	y >>= 8;
-	dst[5] = y;
-	y >>= 8;
-	dst[6] = y;
-	y >>= 8;
-	dst[7] = y;
-}
+static const u32 S2[64] = {
+	0x80108020, 0x80008000, 0x00008000, 0x00108020,
+	0x00100000, 0x00000020, 0x80100020, 0x80008020,
+	0x80000020, 0x80108020, 0x80108000, 0x80000000,
+	0x80008000, 0x00100000, 0x00000020, 0x80100020,
+	0x00108000, 0x00100020, 0x80008020, 0x00000000,
+	0x80000000, 0x00008000, 0x00108020, 0x80100000,
+	0x00100020, 0x80000020, 0x00000000, 0x00108000,
+	0x00008020, 0x80108000, 0x80100000, 0x00008020,
+	0x00000000, 0x00108020, 0x80100020, 0x00100000,
+	0x80008020, 0x80100000, 0x80108000, 0x00008000,
+	0x80100000, 0x80008000, 0x00000020, 0x80108020,
+	0x00108020, 0x00000020, 0x00008000, 0x80000000,
+	0x00008020, 0x80108000, 0x00100000, 0x80000020,
+	0x00100020, 0x80008020, 0x80000020, 0x00100020,
+	0x00108000, 0x00000000, 0x80008000, 0x00008020,
+	0x80000000, 0x80100020, 0x80108020, 0x00108000
+};
 
-static void des_small_fips_decrypt(u32 *expkey, u8 *dst, const u8 *src)
-{
-	u32 x, y, z;
-	
-	x  = src[7];
-	x <<= 8;
-	x |= src[6];
-	x <<= 8;
-	x |= src[5];
-	x <<= 8;
-	x |= src[4];
-	y  = src[3];
-	y <<= 8;
-	y |= src[2];
-	y <<= 8;
-	y |= src[1];
-	y <<= 8;
-	y |= src[0];
-	z  = ((x >> 004) ^ y) & 0x0F0F0F0FL;
-	x ^= z << 004;
-	y ^= z;
-	z  = ((y >> 020) ^ x) & 0x0000FFFFL;
-	y ^= z << 020;
-	x ^= z;
-	z  = ((x >> 002) ^ y) & 0x33333333L;
-	x ^= z << 002;
-	y ^= z;
-	z  = ((y >> 010) ^ x) & 0x00FF00FFL;
-	y ^= z << 010;
-	x ^= z;
-	x  = x >> 1 | x << 31;
-	z  = (x ^ y) & 0x55555555L;
-	y ^= z;
-	x ^= z;
-	y  = y >> 1 | y << 31;
-	z  = expkey[31];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[30];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[29];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[28];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[27];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[26];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[25];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[24];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[23];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[22];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[21];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[20];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[19];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[18];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[17];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[16];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[15];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[14];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[13];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[12];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[11];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[10];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[9];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[8];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[7];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[6];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[5];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[4];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[3];
-	z ^= y;
-	z  = z << 4 | z >> 28;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[2];
-	z ^= y;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	x ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	z  = expkey[1];
-	z ^= x;
-	z  = z << 4 | z >> 28;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 448) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 384) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 320) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 256) + (0xFC & z));
-	z  = expkey[0];
-	z ^= x;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 192) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 128) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) (des_keymap + 64) + (0xFC & z));
-	z >>= 8;
-	y ^= * (u32 *) ((u8 *) des_keymap + (0xFC & z));
-	x  = x << 1 | x >> 31;
-	z  = (x ^ y) & 0x55555555L;
-	y ^= z;
-	x ^= z;
-	y  = y << 1 | y >> 31;
-	z  = ((x >> 010) ^ y) & 0x00FF00FFL;
-	x ^= z << 010;
-	y ^= z;
-	z  = ((y >> 002) ^ x) & 0x33333333L;
-	y ^= z << 002;
-	x ^= z;
-	z  = ((x >> 020) ^ y) & 0x0000FFFFL;
-	x ^= z << 020;
-	y ^= z;
-	z  = ((y >> 004) ^ x) & 0x0F0F0F0FL;
-	y ^= z << 004;
-	x ^= z;
-	dst[0] = x;
-	x >>= 8;
-	dst[1] = x;
-	x >>= 8;
-	dst[2] = x;
-	x >>= 8;
-	dst[3] = x;
-	dst[4] = y;
-	y >>= 8;
-	dst[5] = y;
-	y >>= 8;
-	dst[6] = y;
-	y >>= 8;
-	dst[7] = y;
-}
+static const u32 S3[64] = {
+	0x00000208, 0x08020200, 0x00000000, 0x08020008,
+	0x08000200, 0x00000000, 0x00020208, 0x08000200,
+	0x00020008, 0x08000008, 0x08000008, 0x00020000,
+	0x08020208, 0x00020008, 0x08020000, 0x00000208,
+	0x08000000, 0x00000008, 0x08020200, 0x00000200,
+	0x00020200, 0x08020000, 0x08020008, 0x00020208,
+	0x08000208, 0x00020200, 0x00020000, 0x08000208,
+	0x00000008, 0x08020208, 0x00000200, 0x08000000,
+	0x08020200, 0x08000000, 0x00020008, 0x00000208,
+	0x00020000, 0x08020200, 0x08000200, 0x00000000,
+	0x00000200, 0x00020008, 0x08020208, 0x08000200,
+	0x08000008, 0x00000200, 0x00000000, 0x08020008,
+	0x08000208, 0x00020000, 0x08000000, 0x08020208,
+	0x00000008, 0x00020208, 0x00020200, 0x08000008,
+	0x08020000, 0x08000208, 0x00000208, 0x08020000,
+	0x00020208, 0x00000008, 0x08020008, 0x00020200
+};
+
+static const u32 S4[64] = {
+	0x00802001, 0x00002081, 0x00002081, 0x00000080,
+	0x00802080, 0x00800081, 0x00800001, 0x00002001,
+	0x00000000, 0x00802000, 0x00802000, 0x00802081,
+	0x00000081, 0x00000000, 0x00800080, 0x00800001,
+	0x00000001, 0x00002000, 0x00800000, 0x00802001,
+	0x00000080, 0x00800000, 0x00002001, 0x00002080,
+	0x00800081, 0x00000001, 0x00002080, 0x00800080,
+	0x00002000, 0x00802080, 0x00802081, 0x00000081,
+	0x00800080, 0x00800001, 0x00802000, 0x00802081,
+	0x00000081, 0x00000000, 0x00000000, 0x00802000,
+	0x00002080, 0x00800080, 0x00800081, 0x00000001,
+	0x00802001, 0x00002081, 0x00002081, 0x00000080,
+	0x00802081, 0x00000081, 0x00000001, 0x00002000,
+	0x00800001, 0x00002001, 0x00802080, 0x00800081,
+	0x00002001, 0x00002080, 0x00800000, 0x00802001,
+	0x00000080, 0x00800000, 0x00002000, 0x00802080
+};
+
+static const u32 S5[64] = {
+	0x00000100, 0x02080100, 0x02080000, 0x42000100,
+	0x00080000, 0x00000100, 0x40000000, 0x02080000,
+	0x40080100, 0x00080000, 0x02000100, 0x40080100,
+	0x42000100, 0x42080000, 0x00080100, 0x40000000,
+	0x02000000, 0x40080000, 0x40080000, 0x00000000,
+	0x40000100, 0x42080100, 0x42080100, 0x02000100,
+	0x42080000, 0x40000100, 0x00000000, 0x42000000,
+	0x02080100, 0x02000000, 0x42000000, 0x00080100,
+	0x00080000, 0x42000100, 0x00000100, 0x02000000,
+	0x40000000, 0x02080000, 0x42000100, 0x40080100,
+	0x02000100, 0x40000000, 0x42080000, 0x02080100,
+	0x40080100, 0x00000100, 0x02000000, 0x42080000,
+	0x42080100, 0x00080100, 0x42000000, 0x42080100,
+	0x02080000, 0x00000000, 0x40080000, 0x42000000,
+	0x00080100, 0x02000100, 0x40000100, 0x00080000,
+	0x00000000, 0x40080000, 0x02080100, 0x40000100
+};
+
+static const u32 S6[64] = {
+	0x20000010, 0x20400000, 0x00004000, 0x20404010,
+	0x20400000, 0x00000010, 0x20404010, 0x00400000,
+	0x20004000, 0x00404010, 0x00400000, 0x20000010,
+	0x00400010, 0x20004000, 0x20000000, 0x00004010,
+	0x00000000, 0x00400010, 0x20004010, 0x00004000,
+	0x00404000, 0x20004010, 0x00000010, 0x20400010,
+	0x20400010, 0x00000000, 0x00404010, 0x20404000,
+	0x00004010, 0x00404000, 0x20404000, 0x20000000,
+	0x20004000, 0x00000010, 0x20400010, 0x00404000,
+	0x20404010, 0x00400000, 0x00004010, 0x20000010,
+	0x00400000, 0x20004000, 0x20000000, 0x00004010,
+	0x20000010, 0x20404010, 0x00404000, 0x20400000,
+	0x00404010, 0x20404000, 0x00000000, 0x20400010,
+	0x00000010, 0x00004000, 0x20400000, 0x00404010,
+	0x00004000, 0x00400010, 0x20004010, 0x00000000,
+	0x20404000, 0x20000000, 0x00400010, 0x20004010
+};
+
+static const u32 S7[64] = {
+	0x00200000, 0x04200002, 0x04000802, 0x00000000,
+	0x00000800, 0x04000802, 0x00200802, 0x04200800,
+	0x04200802, 0x00200000, 0x00000000, 0x04000002,
+	0x00000002, 0x04000000, 0x04200002, 0x00000802,
+	0x04000800, 0x00200802, 0x00200002, 0x04000800,
+	0x04000002, 0x04200000, 0x04200800, 0x00200002,
+	0x04200000, 0x00000800, 0x00000802, 0x04200802,
+	0x00200800, 0x00000002, 0x04000000, 0x00200800,
+	0x04000000, 0x00200800, 0x00200000, 0x04000802,
+	0x04000802, 0x04200002, 0x04200002, 0x00000002,
+	0x00200002, 0x04000000, 0x04000800, 0x00200000,
+	0x04200800, 0x00000802, 0x00200802, 0x04200800,
+	0x00000802, 0x04000002, 0x04200802, 0x04200000,
+	0x00200800, 0x00000000, 0x00000002, 0x04200802,
+	0x00000000, 0x00200802, 0x04200000, 0x00000800,
+	0x04000002, 0x04000800, 0x00000800, 0x00200002
+};
+
+static const u32 S8[64] = {
+	0x10001040, 0x00001000, 0x00040000, 0x10041040,
+	0x10000000, 0x10001040, 0x00000040, 0x10000000,
+	0x00040040, 0x10040000, 0x10041040, 0x00041000,
+	0x10041000, 0x00041040, 0x00001000, 0x00000040,
+	0x10040000, 0x10000040, 0x10001000, 0x00001040,
+	0x00041000, 0x00040040, 0x10040040, 0x10041000,
+	0x00001040, 0x00000000, 0x00000000, 0x10040040,
+	0x10000040, 0x10001000, 0x00041040, 0x00040000,
+	0x00041040, 0x00040000, 0x10041000, 0x00001000,
+	0x00000040, 0x10040040, 0x00001000, 0x00041040,
+	0x10001000, 0x00000040, 0x10000040, 0x10040000,
+	0x10040040, 0x10000000, 0x00040000, 0x10001040,
+	0x00000000, 0x10041040, 0x00040040, 0x10000040,
+	0x10040000, 0x10001000, 0x10001040, 0x00000000,
+	0x10041040, 0x00041000, 0x00041000, 0x00001040,
+	0x00001040, 0x00040040, 0x10000000, 0x10041000
+};
+
+/* Encryption components: IP, FP, and round function */
+
+#define IP(L, R, T)		\
+	ROL(R, 4);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xf0f0f0f0;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROL(R, 12);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xffff0000;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROR(R, 14);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xcccccccc;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROL(R, 6);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xff00ff00;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROR(R, 7);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xaaaaaaaa;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROL(L, 1);
+
+#define FP(L, R, T)		\
+	ROR(L, 1);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xaaaaaaaa;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROL(R, 7);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xff00ff00;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROR(R, 6);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xcccccccc;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROL(R, 14);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xffff0000;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROR(R, 12);		\
+	T  = L;			\
+	L ^= R;			\
+	L &= 0xf0f0f0f0;	\
+	R ^= L;			\
+	L ^= T;			\
+	ROR(R, 4);
+
+#define ROUND(L, R, A, B, K, d)					\
+	B = K[0];			A = K[1];	K += d;	\
+	B ^= R;				A ^= R;			\
+	B &= 0x3f3f3f3f;		ROR(A, 4);		\
+	L ^= S8[0xff & B];		A &= 0x3f3f3f3f;	\
+	L ^= S6[0xff & (B >> 8)];	B >>= 16;		\
+	L ^= S7[0xff & A];					\
+	L ^= S5[0xff & (A >> 8)];	A >>= 16;		\
+	L ^= S4[0xff & B];					\
+	L ^= S2[0xff & (B >> 8)];				\
+	L ^= S3[0xff & A];					\
+	L ^= S1[0xff & (A >> 8)];
+
+/*
+ * PC2 lookup tables are organized as 2 consecutive sets of 4 interleaved
+ * tables of 128 elements.  One set is for C_i and the other for D_i, while
+ * the 4 interleaved tables correspond to four 7-bit subsets of C_i or D_i.
+ *
+ * After PC1 each of the variables a,b,c,d contains a 7 bit subset of C_i
+ * or D_i in bits 7-1 (bit 0 being the least significant).
+ */
+
+#define T1(x) pt[2 * (x) + 0]
+#define T2(x) pt[2 * (x) + 1]
+#define T3(x) pt[2 * (x) + 2]
+#define T4(x) pt[2 * (x) + 3]
+
+#define PC2(a, b, c, d) (T4(d) | T3(c) | T2(b) | T1(a))
 
 /*
+ * Encryption key expansion
+ *
  * RFC2451: Weak key checks SHOULD be performed.
+ *
+ * FIPS 74:
+ *
+ *   Keys having duals are keys which produce all zeros, all ones, or
+ *   alternating zero-one patterns in the C and D registers after Permuted
+ *   Choice 1 has operated on the key.
+ *
  */
-static int setkey(u32 *expkey, const u8 *key, unsigned int keylen, u32 *flags)
+static unsigned long ekey(u32 *pe, const u8 *k)
 {
-	const u8 *k;
-	u8 *b0, *b1;
-	u32 n, w;
-	u8 bits0[56], bits1[56];
-
-	n  = parity[key[0]]; n <<= 4;
-	n |= parity[key[1]]; n <<= 4;
-	n |= parity[key[2]]; n <<= 4;
-	n |= parity[key[3]]; n <<= 4;
-	n |= parity[key[4]]; n <<= 4;
-	n |= parity[key[5]]; n <<= 4;
-	n |= parity[key[6]]; n <<= 4;
-	n |= parity[key[7]];
-	w = 0x88888888L;
-	
-	if ((*flags & CRYPTO_TFM_REQ_WEAK_KEY)
-	    && !((n - (w >> 3)) & w)) {  /* 1 in 10^10 keys passes this test */
-		if (n < 0x41415151) {
-			if (n < 0x31312121) {
-				if (n < 0x14141515) {
-					/* 01 01 01 01 01 01 01 01 */
-					if (n == 0x11111111) goto weak;
-					/* 01 1F 01 1F 01 0E 01 0E */
-					if (n == 0x13131212) goto weak;
-				} else {
-					/* 01 E0 01 E0 01 F1 01 F1 */
-					if (n == 0x14141515) goto weak;
-					/* 01 FE 01 FE 01 FE 01 FE */
-					if (n == 0x16161616) goto weak;
-				}
-			} else {
-				if (n < 0x34342525) {
-					/* 1F 01 1F 01 0E 01 0E 01 */
-					if (n == 0x31312121) goto weak;
-					/* 1F 1F 1F 1F 0E 0E 0E 0E (?) */
-					if (n == 0x33332222) goto weak;
-				} else {
-					/* 1F E0 1F E0 0E F1 0E F1 */
-					if (n == 0x34342525) goto weak;
-					/* 1F FE 1F FE 0E FE 0E FE */
-					if (n == 0x36362626) goto weak;
-				}
-			}
-		} else {
-			if (n < 0x61616161) {
-				if (n < 0x44445555) {
-					/* E0 01 E0 01 F1 01 F1 01 */
-					if (n == 0x41415151) goto weak;
-					/* E0 1F E0 1F F1 0E F1 0E */
-					if (n == 0x43435252) goto weak;
-				} else {
-					/* E0 E0 E0 E0 F1 F1 F1 F1 (?) */
-					if (n == 0x44445555) goto weak;
-					/* E0 FE E0 FE F1 FE F1 FE */
-					if (n == 0x46465656) goto weak;
-				}
-			} else {
-				if (n < 0x64646565) {
-					/* FE 01 FE 01 FE 01 FE 01 */
-					if (n == 0x61616161) goto weak;
-					/* FE 1F FE 1F FE 0E FE 0E */
-					if (n == 0x63636262) goto weak;
-				} else {
-					/* FE E0 FE E0 FE F1 FE F1 */
-					if (n == 0x64646565) goto weak;
-					/* FE FE FE FE FE FE FE FE */
-					if (n == 0x66666666) goto weak;
-				}
-			}
-		}
-	
-		goto not_weak;
-weak:
-		*flags |= CRYPTO_TFM_RES_WEAK_KEY;
-		return -EINVAL;
+	/* K&R: long is at least 32 bits */
+	unsigned long a, b, c, d, w;
+	const u32 *pt = pc2;
+
+	d = k[4]; d &= 0x0e; d <<= 4; d |= k[0] & 0x1e; d = pc1[d];
+	c = k[5]; c &= 0x0e; c <<= 4; c |= k[1] & 0x1e; c = pc1[c];
+	b = k[6]; b &= 0x0e; b <<= 4; b |= k[2] & 0x1e; b = pc1[b];
+	a = k[7]; a &= 0x0e; a <<= 4; a |= k[3] & 0x1e; a = pc1[a];
+
+	pe[15 * 2 + 0] = PC2(a, b, c, d); d = rs[d];
+	pe[14 * 2 + 0] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[13 * 2 + 0] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[12 * 2 + 0] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[11 * 2 + 0] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[10 * 2 + 0] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 9 * 2 + 0] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 8 * 2 + 0] = PC2(d, a, b, c); c = rs[c];
+	pe[ 7 * 2 + 0] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 6 * 2 + 0] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 5 * 2 + 0] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 4 * 2 + 0] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 3 * 2 + 0] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 2 * 2 + 0] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 1 * 2 + 0] = PC2(c, d, a, b); b = rs[b];
+	pe[ 0 * 2 + 0] = PC2(b, c, d, a);
+
+	/* Check if first half is weak */
+	w  = (a ^ c) | (b ^ d) | (rs[a] ^ c) | (b ^ rs[d]);
+
+	/* Skip to next table set */
+	pt += 512;
+
+	d = k[0]; d &= 0xe0; d >>= 4; d |= k[4] & 0xf0; d = pc1[d + 1];
+	c = k[1]; c &= 0xe0; c >>= 4; c |= k[5] & 0xf0; c = pc1[c + 1];
+	b = k[2]; b &= 0xe0; b >>= 4; b |= k[6] & 0xf0; b = pc1[b + 1];
+	a = k[3]; a &= 0xe0; a >>= 4; a |= k[7] & 0xf0; a = pc1[a + 1];
+
+	/* Check if second half is weak */
+	w |= (a ^ c) | (b ^ d) | (rs[a] ^ c) | (b ^ rs[d]);
+
+	pe[15 * 2 + 1] = PC2(a, b, c, d); d = rs[d];
+	pe[14 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[13 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[12 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[11 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[10 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 9 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 8 * 2 + 1] = PC2(d, a, b, c); c = rs[c];
+	pe[ 7 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 6 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 5 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 4 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 3 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 2 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[ 1 * 2 + 1] = PC2(c, d, a, b); b = rs[b];
+	pe[ 0 * 2 + 1] = PC2(b, c, d, a);
+
+	/* Fixup: 2413 5768 -> 1357 2468 */
+	for (d = 0; d < 16; ++d) {
+		a = pe[2 * d];
+		b = pe[2 * d + 1];
+		c = a ^ b;
+		c &= 0xffff0000;
+		a ^= c;
+		b ^= c;
+		ROL(b, 18);
+		pe[2 * d] = a;
+		pe[2 * d + 1] = b;
 	}
 
-not_weak:
-
-	/* explode the bits */
-	n = 56;
-	b0 = bits0;
-	b1 = bits1;
-	
-	do {
-		w = (256 | *key++) << 2;
-		do {
-			--n;
-			b1[n] = 8 & w;
-			w >>= 1;
-			b0[n] = 4 & w;
-		} while ( w >= 16 );
-	} while ( n );
-	
-	/* put the bits in the correct places */
-	n = 16;
-	k = rotors;
-	
-	do {
-		w   = (b1[k[ 0   ]] | b0[k[ 1   ]]) << 4;
-		w  |= (b1[k[ 2   ]] | b0[k[ 3   ]]) << 2;
-		w  |=  b1[k[ 4   ]] | b0[k[ 5   ]];
-		w <<= 8;
-		w  |= (b1[k[ 6   ]] | b0[k[ 7   ]]) << 4;
-		w  |= (b1[k[ 8   ]] | b0[k[ 9   ]]) << 2;
-		w  |=  b1[k[10   ]] | b0[k[11   ]];
-		w <<= 8;
-		w  |= (b1[k[12   ]] | b0[k[13   ]]) << 4;
-		w  |= (b1[k[14   ]] | b0[k[15   ]]) << 2;
-		w  |=  b1[k[16   ]] | b0[k[17   ]];
-		w <<= 8;
-		w  |= (b1[k[18   ]] | b0[k[19   ]]) << 4;
-		w  |= (b1[k[20   ]] | b0[k[21   ]]) << 2;
-		w  |=  b1[k[22   ]] | b0[k[23   ]];
-		expkey[0] = w;
-		
-		w   = (b1[k[ 0+24]] | b0[k[ 1+24]]) << 4;
-		w  |= (b1[k[ 2+24]] | b0[k[ 3+24]]) << 2;
-		w  |=  b1[k[ 4+24]] | b0[k[ 5+24]];
-		w <<= 8;
-		w  |= (b1[k[ 6+24]] | b0[k[ 7+24]]) << 4;
-		w  |= (b1[k[ 8+24]] | b0[k[ 9+24]]) << 2;
-		w  |=  b1[k[10+24]] | b0[k[11+24]];
-		w <<= 8;
-		w  |= (b1[k[12+24]] | b0[k[13+24]]) << 4;
-		w  |= (b1[k[14+24]] | b0[k[15+24]]) << 2;
-		w  |=  b1[k[16+24]] | b0[k[17+24]];
-		w <<= 8;
-		w  |= (b1[k[18+24]] | b0[k[19+24]]) << 4;
-		w  |= (b1[k[20+24]] | b0[k[21+24]]) << 2;
-		w  |=  b1[k[22+24]] | b0[k[23+24]];
-		
-		ROR(w, 4, 28);      /* could be eliminated */
-		expkey[1] = w;
-
-		k += 48;
-		expkey += 2;
-	} while (--n);
+	/* Zero if weak key */
+	return w;
+}
 
-	return 0;
+/*
+ * Decryption key expansion
+ *
+ * No weak key checking is performed, as this is only used by triple DES
+ *
+ */
+static void dkey(u32 *pe, const u8 *k)
+{
+	/* K&R: long is at least 32 bits */
+	unsigned long a, b, c, d;
+	const u32 *pt = pc2;
+
+	d = k[4]; d &= 0x0e; d <<= 4; d |= k[0] & 0x1e; d = pc1[d];
+	c = k[5]; c &= 0x0e; c <<= 4; c |= k[1] & 0x1e; c = pc1[c];
+	b = k[6]; b &= 0x0e; b <<= 4; b |= k[2] & 0x1e; b = pc1[b];
+	a = k[7]; a &= 0x0e; a <<= 4; a |= k[3] & 0x1e; a = pc1[a];
+
+	pe[ 0 * 2] = PC2(a, b, c, d); d = rs[d];
+	pe[ 1 * 2] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 2 * 2] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 3 * 2] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 4 * 2] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 5 * 2] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 6 * 2] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 7 * 2] = PC2(d, a, b, c); c = rs[c];
+	pe[ 8 * 2] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 9 * 2] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[10 * 2] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[11 * 2] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[12 * 2] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[13 * 2] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[14 * 2] = PC2(c, d, a, b); b = rs[b];
+	pe[15 * 2] = PC2(b, c, d, a);
+
+	/* Skip to next table set */
+	pt += 512;
+
+	d = k[0]; d &= 0xe0; d >>= 4; d |= k[4] & 0xf0; d = pc1[d + 1];
+	c = k[1]; c &= 0xe0; c >>= 4; c |= k[5] & 0xf0; c = pc1[c + 1];
+	b = k[2]; b &= 0xe0; b >>= 4; b |= k[6] & 0xf0; b = pc1[b + 1];
+	a = k[3]; a &= 0xe0; a >>= 4; a |= k[7] & 0xf0; a = pc1[a + 1];
+
+	pe[ 0 * 2 + 1] = PC2(a, b, c, d); d = rs[d];
+	pe[ 1 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 2 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 3 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 4 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 5 * 2 + 1] = PC2(d, a, b, c); c = rs[c]; b = rs[b];
+	pe[ 6 * 2 + 1] = PC2(b, c, d, a); a = rs[a]; d = rs[d];
+	pe[ 7 * 2 + 1] = PC2(d, a, b, c); c = rs[c];
+	pe[ 8 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[ 9 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[10 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[11 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[12 * 2 + 1] = PC2(c, d, a, b); b = rs[b]; a = rs[a];
+	pe[13 * 2 + 1] = PC2(a, b, c, d); d = rs[d]; c = rs[c];
+	pe[14 * 2 + 1] = PC2(c, d, a, b); b = rs[b];
+	pe[15 * 2 + 1] = PC2(b, c, d, a);
+
+	/* Fixup: 2413 5768 -> 1357 2468 */
+	for (d = 0; d < 16; ++d) {
+		a = pe[2 * d];
+		b = pe[2 * d + 1];
+		c = a ^ b;
+		c &= 0xffff0000;
+		a ^= c;
+		b ^= c;
+		ROL(b, 18);
+		pe[2 * d] = a;
+		pe[2 * d + 1] = b;
+	}
 }
 
 static int des_setkey(void *ctx, const u8 *key, unsigned int keylen, u32 *flags)
 {
-	return setkey(((struct des_ctx *)ctx)->expkey, key, keylen, flags);
+	struct des_ctx *dctx = ctx;
+	u32 tmp[DES_EXPKEY_WORDS];
+	int ret;
+
+	/* Expand to tmp */
+	ret = ekey(tmp, key);
+
+	if (unlikely(ret == 0) && (*flags & CRYPTO_TFM_REQ_WEAK_KEY)) {
+		*flags |= CRYPTO_TFM_RES_WEAK_KEY;
+		return -EINVAL;
+	}
+
+	/* Copy to output */
+	memcpy(dctx->expkey, tmp, sizeof(dctx->expkey));
+
+	return 0;
 }
 
 static void des_encrypt(void *ctx, u8 *dst, const u8 *src)
 {
-	des_small_fips_encrypt(((struct des_ctx *)ctx)->expkey, dst, src);
+	const u32 *K = ((struct des_ctx *)ctx)->expkey;
+	const __le32 *s = (const __le32 *)src;
+	__le32 *d = (__le32 *)dst;
+	u32 L, R, A, B;
+	int i;
+
+	L = le32_to_cpu(s[0]);
+	R = le32_to_cpu(s[1]);
+
+	IP(L, R, A);
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, 2);
+		ROUND(R, L, A, B, K, 2);
+	}
+	FP(R, L, A);
+
+	d[0] = cpu_to_le32(R);
+	d[1] = cpu_to_le32(L);
 }
 
 static void des_decrypt(void *ctx, u8 *dst, const u8 *src)
 {
-	des_small_fips_decrypt(((struct des_ctx *)ctx)->expkey, dst, src);
+	const u32 *K = ((struct des_ctx *)ctx)->expkey + DES_EXPKEY_WORDS - 2;
+	const __le32 *s = (const __le32 *)src;
+	__le32 *d = (__le32 *)dst;
+	u32 L, R, A, B;
+	int i;
+
+	L = le32_to_cpu(s[0]);
+	R = le32_to_cpu(s[1]);
+
+	IP(L, R, A);
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, -2);
+		ROUND(R, L, A, B, K, -2);
+	}
+	FP(R, L, A);
+
+	d[0] = cpu_to_le32(R);
+	d[1] = cpu_to_le32(L);
 }
 
-/* 
+/*
  * RFC2451:
  *
  *   For DES-EDE3, there is no known need to reject weak or
@@ -1199,44 +859,86 @@ static void des_decrypt(void *ctx, u8 *dst, const u8 *src)
  *
  */
 static int des3_ede_setkey(void *ctx, const u8 *key,
-                           unsigned int keylen, u32 *flags)
+			   unsigned int keylen, u32 *flags)
 {
-	unsigned int i, off;
+	const u32 *K = (const u32 *)key;
 	struct des3_ede_ctx *dctx = ctx;
+	u32 *expkey = dctx->expkey;
 
-	if (!(memcmp(key, &key[DES_KEY_SIZE], DES_KEY_SIZE) && 
-	    memcmp(&key[DES_KEY_SIZE], &key[DES_KEY_SIZE * 2],
-	    					DES_KEY_SIZE))) {
-
+	if (unlikely(!((K[0] ^ K[2]) | (K[1] ^ K[3])) ||
+		     !((K[2] ^ K[4]) | (K[3] ^ K[5]))))
+	{
 		*flags |= CRYPTO_TFM_RES_BAD_KEY_SCHED;
 		return -EINVAL;
 	}
-	
-	for (i = 0, off = 0; i < 3; i++, off += DES_EXPKEY_WORDS,
-							key += DES_KEY_SIZE) {
-		int ret = setkey(&dctx->expkey[off], key, DES_KEY_SIZE, flags);
-		if (ret < 0)
-			return ret;
-	}	
+
+	ekey(expkey, key); expkey += DES_EXPKEY_WORDS; key += DES_KEY_SIZE;
+	dkey(expkey, key); expkey += DES_EXPKEY_WORDS; key += DES_KEY_SIZE;
+	ekey(expkey, key);
+
 	return 0;
 }
 
 static void des3_ede_encrypt(void *ctx, u8 *dst, const u8 *src)
 {
 	struct des3_ede_ctx *dctx = ctx;
-	
-	des_small_fips_encrypt(dctx->expkey, dst, src);
-	des_small_fips_decrypt(&dctx->expkey[DES_EXPKEY_WORDS], dst, dst);
-	des_small_fips_encrypt(&dctx->expkey[DES_EXPKEY_WORDS * 2], dst, dst);
+	const u32 *K = dctx->expkey;
+	const __le32 *s = (const __le32 *)src;
+	__le32 *d = (__le32 *)dst;
+	u32 L, R, A, B;
+	int i;
+
+	L = le32_to_cpu(s[0]);
+	R = le32_to_cpu(s[1]);
+
+	IP(L, R, A);
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, 2);
+		ROUND(R, L, A, B, K, 2);
+	}
+	for (i = 0; i < 8; i++) {
+		ROUND(R, L, A, B, K, 2);
+		ROUND(L, R, A, B, K, 2);
+	}
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, 2);
+		ROUND(R, L, A, B, K, 2);
+	}
+	FP(R, L, A);
+
+	d[0] = cpu_to_le32(R);
+	d[1] = cpu_to_le32(L);
 }
 
 static void des3_ede_decrypt(void *ctx, u8 *dst, const u8 *src)
 {
 	struct des3_ede_ctx *dctx = ctx;
+	const u32 *K = dctx->expkey + DES3_EDE_EXPKEY_WORDS - 2;
+	const __le32 *s = (const __le32 *)src;
+	__le32 *d = (__le32 *)dst;
+	u32 L, R, A, B;
+	int i;
+
+	L = le32_to_cpu(s[0]);
+	R = le32_to_cpu(s[1]);
+
+	IP(L, R, A);
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, -2);
+		ROUND(R, L, A, B, K, -2);
+	}
+	for (i = 0; i < 8; i++) {
+		ROUND(R, L, A, B, K, -2);
+		ROUND(L, R, A, B, K, -2);
+	}
+	for (i = 0; i < 8; i++) {
+		ROUND(L, R, A, B, K, -2);
+		ROUND(R, L, A, B, K, -2);
+	}
+	FP(R, L, A);
 
-	des_small_fips_decrypt(&dctx->expkey[DES_EXPKEY_WORDS * 2], dst, src);
-	des_small_fips_encrypt(&dctx->expkey[DES_EXPKEY_WORDS], dst, dst);
-	des_small_fips_decrypt(dctx->expkey, dst, dst);
+	d[0] = cpu_to_le32(R);
+	d[1] = cpu_to_le32(L);
 }
 
 static struct crypto_alg des_alg = {
@@ -1249,7 +951,7 @@ static struct crypto_alg des_alg = {
 	.cra_u			=	{ .cipher = {
 	.cia_min_keysize	=	DES_KEY_SIZE,
 	.cia_max_keysize	=	DES_KEY_SIZE,
-	.cia_setkey		= 	des_setkey,
+	.cia_setkey		=	des_setkey,
 	.cia_encrypt		=	des_encrypt,
 	.cia_decrypt		=	des_decrypt } }
 };
@@ -1264,9 +966,9 @@ static struct crypto_alg des3_ede_alg = {
 	.cra_u			=	{ .cipher = {
 	.cia_min_keysize	=	DES3_EDE_KEY_SIZE,
 	.cia_max_keysize	=	DES3_EDE_KEY_SIZE,
-	.cia_setkey	   	= 	des3_ede_setkey,
-	.cia_encrypt	 	=	des3_ede_encrypt,
-	.cia_decrypt	  	=	des3_ede_decrypt } }
+	.cia_setkey		=	des3_ede_setkey,
+	.cia_encrypt		=	des3_ede_encrypt,
+	.cia_decrypt		=	des3_ede_decrypt } }
 };
 
 MODULE_ALIAS("des3_ede");
@@ -1274,7 +976,7 @@ MODULE_ALIAS("des3_ede");
 static int __init init(void)
 {
 	int ret = 0;
-	
+
 	ret = crypto_register_alg(&des_alg);
 	if (ret < 0)
 		goto out;
@@ -1282,7 +984,7 @@ static int __init init(void)
 	ret = crypto_register_alg(&des3_ede_alg);
 	if (ret < 0)
 		crypto_unregister_alg(&des_alg);
-out:	
+out:
 	return ret;
 }
 
@@ -1297,3 +999,4 @@ module_exit(fini);
 
 MODULE_LICENSE("GPL");
 MODULE_DESCRIPTION("DES & Triple DES EDE Cipher Algorithms");
+MODULE_AUTHOR("Dag Arne Osvik <da@osvik.no>");

crypto/hmac.c

@@ -49,8 +49,7 @@ int crypto_alloc_hmac_block(struct crypto_tfm *tfm)
 
 void crypto_free_hmac_block(struct crypto_tfm *tfm)
 {
-	if (tfm->crt_digest.dit_hmac_block)
-		kfree(tfm->crt_digest.dit_hmac_block);
+	kfree(tfm->crt_digest.dit_hmac_block);
 }
 
 void crypto_hmac_init(struct crypto_tfm *tfm, u8 *key, unsigned int *keylen)

crypto/internal.h

@@ -16,7 +16,7 @@
 #include <linux/highmem.h>
 #include <linux/interrupt.h>
 #include <linux/init.h>
-#include <linux/kmod.h>
+#include <linux/kernel.h>
 #include <asm/kmap_types.h>
 
 extern enum km_type crypto_km_types[];
@@ -42,20 +42,6 @@ static inline void crypto_yield(struct crypto_tfm *tfm)
 		cond_resched();
 }
 
-static inline void *crypto_tfm_ctx(struct crypto_tfm *tfm)
-{
-	return (void *)&tfm[1];
-}
-
-struct crypto_alg *crypto_alg_lookup(const char *name);
-
-/* A far more intelligent version of this is planned.  For now, just
- * try an exact match on the name of the algorithm. */
-static inline struct crypto_alg *crypto_alg_mod_lookup(const char *name)
-{
-	return try_then_request_module(crypto_alg_lookup(name), name);
-}
-
 #ifdef CONFIG_CRYPTO_HMAC
 int crypto_alloc_hmac_block(struct crypto_tfm *tfm);
 void crypto_free_hmac_block(struct crypto_tfm *tfm);
@@ -76,6 +62,33 @@ static inline void crypto_init_proc(void)
 { }
 #endif
 
+static inline unsigned int crypto_digest_ctxsize(struct crypto_alg *alg,
+						 int flags)
+{
+	return alg->cra_ctxsize;
+}
+
+static inline unsigned int crypto_cipher_ctxsize(struct crypto_alg *alg,
+						 int flags)
+{
+	unsigned int len = alg->cra_ctxsize;
+	
+	switch (flags & CRYPTO_TFM_MODE_MASK) {
+	case CRYPTO_TFM_MODE_CBC:
+		len = ALIGN(len, alg->cra_alignmask + 1);
+		len += alg->cra_blocksize;
+		break;
+	}
+
+	return len;
+}
+
+static inline unsigned int crypto_compress_ctxsize(struct crypto_alg *alg,
+						   int flags)
+{
+	return alg->cra_ctxsize;
+}
+
 int crypto_init_digest_flags(struct crypto_tfm *tfm, u32 flags);
 int crypto_init_cipher_flags(struct crypto_tfm *tfm, u32 flags);
 int crypto_init_compress_flags(struct crypto_tfm *tfm, u32 flags);

crypto/scatterwalk.c

@@ -100,7 +100,7 @@ void scatterwalk_done(struct scatter_walk *walk, int out, int more)
 int scatterwalk_copychunks(void *buf, struct scatter_walk *walk,
 			   size_t nbytes, int out)
 {
-	do {
+	while (nbytes > walk->len_this_page) {
 		memcpy_dir(buf, walk->data, walk->len_this_page, out);
 		buf += walk->len_this_page;
 		nbytes -= walk->len_this_page;
@@ -108,7 +108,7 @@ int scatterwalk_copychunks(void *buf, struct scatter_walk *walk,
 		scatterwalk_unmap(walk, out);
 		scatterwalk_pagedone(walk, out, 1);
 		scatterwalk_map(walk, out);
-	} while (nbytes > walk->len_this_page);
+	}
 
 	memcpy_dir(buf, walk->data, nbytes, out);
 	return nbytes;

crypto/scatterwalk.h

@@ -40,10 +40,10 @@ static inline int scatterwalk_samebuf(struct scatter_walk *walk_in,
 	       walk_in->offset == walk_out->offset;
 }
 
-static inline int scatterwalk_across_pages(struct scatter_walk *walk,
-					   unsigned int nbytes)
+static inline unsigned int scatterwalk_clamp(struct scatter_walk *walk,
+					     unsigned int nbytes)
 {
-	return nbytes > walk->len_this_page;
+	return nbytes > walk->len_this_page ? walk->len_this_page : nbytes;
 }
 
 static inline void scatterwalk_advance(struct scatter_walk *walk,
@@ -55,6 +55,12 @@ static inline void scatterwalk_advance(struct scatter_walk *walk,
 	walk->len_this_segment -= nbytes;
 }
 
+static inline unsigned int scatterwalk_aligned(struct scatter_walk *walk,
+					       unsigned int alignmask)
+{
+	return !(walk->offset & alignmask);
+}
+
 void scatterwalk_start(struct scatter_walk *walk, struct scatterlist *sg);
 int scatterwalk_copychunks(void *buf, struct scatter_walk *walk, size_t nbytes, int out);
 void scatterwalk_map(struct scatter_walk *walk, int out);

crypto/serpent.c

@@ -210,7 +210,6 @@
 	x4 ^= x2;
 
 struct serpent_ctx {
-	u8 iv[SERPENT_BLOCK_SIZE];
 	u32 expkey[SERPENT_EXPKEY_WORDS];
 };
 

drivers/crypto/padlock-aes.c

@@ -49,6 +49,7 @@
 #include <linux/errno.h>
 #include <linux/crypto.h>
 #include <linux/interrupt.h>
+#include <linux/kernel.h>
 #include <asm/byteorder.h>
 #include "padlock.h"
 
@@ -59,8 +60,12 @@
 #define AES_EXTENDED_KEY_SIZE_B	(AES_EXTENDED_KEY_SIZE * sizeof(uint32_t))
 
 struct aes_ctx {
-	uint32_t e_data[AES_EXTENDED_KEY_SIZE+4];
-	uint32_t d_data[AES_EXTENDED_KEY_SIZE+4];
+	uint32_t e_data[AES_EXTENDED_KEY_SIZE];
+	uint32_t d_data[AES_EXTENDED_KEY_SIZE];
+	struct {
+		struct cword encrypt;
+		struct cword decrypt;
+	} cword;
 	uint32_t *E;
 	uint32_t *D;
 	int key_length;
@@ -280,10 +285,15 @@ aes_hw_extkey_available(uint8_t key_len)
 	return 0;
 }
 
+static inline struct aes_ctx *aes_ctx(void *ctx)
+{
+	return (struct aes_ctx *)ALIGN((unsigned long)ctx, PADLOCK_ALIGNMENT);
+}
+
 static int
 aes_set_key(void *ctx_arg, const uint8_t *in_key, unsigned int key_len, uint32_t *flags)
 {
-	struct aes_ctx *ctx = ctx_arg;
+	struct aes_ctx *ctx = aes_ctx(ctx_arg);
 	uint32_t i, t, u, v, w;
 	uint32_t P[AES_EXTENDED_KEY_SIZE];
 	uint32_t rounds;
@@ -295,25 +305,36 @@ aes_set_key(void *ctx_arg, const uint8_t *in_key, unsigned int key_len, uint32_t
 
 	ctx->key_length = key_len;
 
+	/*
+	 * If the hardware is capable of generating the extended key
+	 * itself we must supply the plain key for both encryption
+	 * and decryption.
+	 */
 	ctx->E = ctx->e_data;
-	ctx->D = ctx->d_data;
-
-	/* Ensure 16-Bytes alignmentation of keys for VIA PadLock. */
-	if ((int)(ctx->e_data) & 0x0F)
-		ctx->E += 4 - (((int)(ctx->e_data) & 0x0F) / sizeof (ctx->e_data[0]));
-
-	if ((int)(ctx->d_data) & 0x0F)
-		ctx->D += 4 - (((int)(ctx->d_data) & 0x0F) / sizeof (ctx->d_data[0]));
+	ctx->D = ctx->e_data;
 
 	E_KEY[0] = uint32_t_in (in_key);
 	E_KEY[1] = uint32_t_in (in_key + 4);
 	E_KEY[2] = uint32_t_in (in_key + 8);
 	E_KEY[3] = uint32_t_in (in_key + 12);
 
+	/* Prepare control words. */
+	memset(&ctx->cword, 0, sizeof(ctx->cword));
+
+	ctx->cword.decrypt.encdec = 1;
+	ctx->cword.encrypt.rounds = 10 + (key_len - 16) / 4;
+	ctx->cword.decrypt.rounds = ctx->cword.encrypt.rounds;
+	ctx->cword.encrypt.ksize = (key_len - 16) / 8;
+	ctx->cword.decrypt.ksize = ctx->cword.encrypt.ksize;
+
 	/* Don't generate extended keys if the hardware can do it. */
 	if (aes_hw_extkey_available(key_len))
 		return 0;
 
+	ctx->D = ctx->d_data;
+	ctx->cword.encrypt.keygen = 1;
+	ctx->cword.decrypt.keygen = 1;
+
 	switch (key_len) {
 	case 16:
 		t = E_KEY[3];
@@ -369,10 +390,9 @@ aes_set_key(void *ctx_arg, const uint8_t *in_key, unsigned int key_len, uint32_t
 
 /* ====== Encryption/decryption routines ====== */
 
-/* This is the real call to PadLock. */
-static inline void
-padlock_xcrypt_ecb(uint8_t *input, uint8_t *output, uint8_t *key,
-		   void *control_word, uint32_t count)
+/* These are the real call to PadLock. */
+static inline void padlock_xcrypt_ecb(const u8 *input, u8 *output, void *key,
+				      void *control_word, u32 count)
 {
 	asm volatile ("pushfl; popfl");		/* enforce key reload. */
 	asm volatile (".byte 0xf3,0x0f,0xa7,0xc8"	/* rep xcryptecb */
@@ -380,60 +400,70 @@ padlock_xcrypt_ecb(uint8_t *input, uint8_t *output, uint8_t *key,
 		      : "d"(control_word), "b"(key), "c"(count));
 }
 
-static void
-aes_padlock(void *ctx_arg, uint8_t *out_arg, const uint8_t *in_arg, int encdec)
+static inline u8 *padlock_xcrypt_cbc(const u8 *input, u8 *output, void *key,
+				     u8 *iv, void *control_word, u32 count)
 {
-	/* Don't blindly modify this structure - the items must 
-	   fit on 16-Bytes boundaries! */
-	struct padlock_xcrypt_data {
-		uint8_t buf[AES_BLOCK_SIZE];
-		union cword cword;
-	};
-
-	struct aes_ctx *ctx = ctx_arg;
-	char bigbuf[sizeof(struct padlock_xcrypt_data) + 16];
-	struct padlock_xcrypt_data *data;
-	void *key;
-
-	/* Place 'data' at the first 16-Bytes aligned address in 'bigbuf'. */
-	if (((long)bigbuf) & 0x0F)
-		data = (void*)(bigbuf + 16 - ((long)bigbuf & 0x0F));
-	else
-		data = (void*)bigbuf;
-
-	/* Prepare Control word. */
-	memset (data, 0, sizeof(struct padlock_xcrypt_data));
-	data->cword.b.encdec = !encdec;	/* in the rest of cryptoapi ENC=1/DEC=0 */
-	data->cword.b.rounds = 10 + (ctx->key_length - 16) / 4;
-	data->cword.b.ksize = (ctx->key_length - 16) / 8;
-
-	/* Is the hardware capable to generate the extended key? */
-	if (!aes_hw_extkey_available(ctx->key_length))
-		data->cword.b.keygen = 1;
-
-	/* ctx->E starts with a plain key - if the hardware is capable
-	   to generate the extended key itself we must supply
-	   the plain key for both Encryption and Decryption. */
-	if (encdec == CRYPTO_DIR_ENCRYPT || data->cword.b.keygen == 0)
-		key = ctx->E;
-	else
-		key = ctx->D;
-	
-	memcpy(data->buf, in_arg, AES_BLOCK_SIZE);
-	padlock_xcrypt_ecb(data->buf, data->buf, key, &data->cword, 1);
-	memcpy(out_arg, data->buf, AES_BLOCK_SIZE);
+	/* Enforce key reload. */
+	asm volatile ("pushfl; popfl");
+	/* rep xcryptcbc */
+	asm volatile (".byte 0xf3,0x0f,0xa7,0xd0"
+		      : "+S" (input), "+D" (output), "+a" (iv)
+		      : "d" (control_word), "b" (key), "c" (count));
+	return iv;
 }
 
 static void
 aes_encrypt(void *ctx_arg, uint8_t *out, const uint8_t *in)
 {
-	aes_padlock(ctx_arg, out, in, CRYPTO_DIR_ENCRYPT);
+	struct aes_ctx *ctx = aes_ctx(ctx_arg);
+	padlock_xcrypt_ecb(in, out, ctx->E, &ctx->cword.encrypt, 1);
 }
 
 static void
 aes_decrypt(void *ctx_arg, uint8_t *out, const uint8_t *in)
 {
-	aes_padlock(ctx_arg, out, in, CRYPTO_DIR_DECRYPT);
+	struct aes_ctx *ctx = aes_ctx(ctx_arg);
+	padlock_xcrypt_ecb(in, out, ctx->D, &ctx->cword.decrypt, 1);
+}
+
+static unsigned int aes_encrypt_ecb(const struct cipher_desc *desc, u8 *out,
+				    const u8 *in, unsigned int nbytes)
+{
+	struct aes_ctx *ctx = aes_ctx(crypto_tfm_ctx(desc->tfm));
+	padlock_xcrypt_ecb(in, out, ctx->E, &ctx->cword.encrypt,
+			   nbytes / AES_BLOCK_SIZE);
+	return nbytes & ~(AES_BLOCK_SIZE - 1);
+}
+
+static unsigned int aes_decrypt_ecb(const struct cipher_desc *desc, u8 *out,
+				    const u8 *in, unsigned int nbytes)
+{
+	struct aes_ctx *ctx = aes_ctx(crypto_tfm_ctx(desc->tfm));
+	padlock_xcrypt_ecb(in, out, ctx->D, &ctx->cword.decrypt,
+			   nbytes / AES_BLOCK_SIZE);
+	return nbytes & ~(AES_BLOCK_SIZE - 1);
+}
+
+static unsigned int aes_encrypt_cbc(const struct cipher_desc *desc, u8 *out,
+				    const u8 *in, unsigned int nbytes)
+{
+	struct aes_ctx *ctx = aes_ctx(crypto_tfm_ctx(desc->tfm));
+	u8 *iv;
+
+	iv = padlock_xcrypt_cbc(in, out, ctx->E, desc->info,
+				&ctx->cword.encrypt, nbytes / AES_BLOCK_SIZE);
+	memcpy(desc->info, iv, AES_BLOCK_SIZE);
+
+	return nbytes & ~(AES_BLOCK_SIZE - 1);
+}
+
+static unsigned int aes_decrypt_cbc(const struct cipher_desc *desc, u8 *out,
+				    const u8 *in, unsigned int nbytes)
+{
+	struct aes_ctx *ctx = aes_ctx(crypto_tfm_ctx(desc->tfm));
+	padlock_xcrypt_cbc(in, out, ctx->D, desc->info, &ctx->cword.decrypt,
+			   nbytes / AES_BLOCK_SIZE);
+	return nbytes & ~(AES_BLOCK_SIZE - 1);
 }
 
 static struct crypto_alg aes_alg = {
@@ -441,6 +471,7 @@ static struct crypto_alg aes_alg = {
 	.cra_flags		=	CRYPTO_ALG_TYPE_CIPHER,
 	.cra_blocksize		=	AES_BLOCK_SIZE,
 	.cra_ctxsize		=	sizeof(struct aes_ctx),
+	.cra_alignmask		=	PADLOCK_ALIGNMENT - 1,
 	.cra_module		=	THIS_MODULE,
 	.cra_list		=	LIST_HEAD_INIT(aes_alg.cra_list),
 	.cra_u			=	{
@@ -449,7 +480,11 @@ static struct crypto_alg aes_alg = {
 			.cia_max_keysize	=	AES_MAX_KEY_SIZE,
 			.cia_setkey	   	= 	aes_set_key,
 			.cia_encrypt	 	=	aes_encrypt,
-			.cia_decrypt	  	=	aes_decrypt
+			.cia_decrypt	  	=	aes_decrypt,
+			.cia_encrypt_ecb 	=	aes_encrypt_ecb,
+			.cia_decrypt_ecb  	=	aes_decrypt_ecb,
+			.cia_encrypt_cbc 	=	aes_encrypt_cbc,
+			.cia_decrypt_cbc  	=	aes_decrypt_cbc,
 		}
 	}
 };

drivers/crypto/padlock.h

@@ -13,18 +13,18 @@
 #ifndef _CRYPTO_PADLOCK_H
 #define _CRYPTO_PADLOCK_H
 
+#define PADLOCK_ALIGNMENT 16
+
 /* Control word. */
-union cword {
-	uint32_t cword[4];
-	struct {
-		int rounds:4;
-		int algo:3;
-		int keygen:1;
-		int interm:1;
-		int encdec:1;
-		int ksize:2;
-	} b;
-};
+struct cword {
+	int __attribute__ ((__packed__))
+		rounds:4,
+		algo:3,
+		keygen:1,
+		interm:1,
+		encdec:1,
+		ksize:2;
+} __attribute__ ((__aligned__(PADLOCK_ALIGNMENT)));
 
 #define PFX	"padlock: "
 

include/linux/crypto.h

@@ -61,6 +61,15 @@
 #define CRYPTO_DIR_DECRYPT		0
 
 struct scatterlist;
+struct crypto_tfm;
+
+struct cipher_desc {
+	struct crypto_tfm *tfm;
+	void (*crfn)(void *ctx, u8 *dst, const u8 *src);
+	unsigned int (*prfn)(const struct cipher_desc *desc, u8 *dst,
+			     const u8 *src, unsigned int nbytes);
+	void *info;
+};
 
 /*
  * Algorithms: modular crypto algorithm implementations, managed
@@ -73,6 +82,19 @@ struct cipher_alg {
 	                  unsigned int keylen, u32 *flags);
 	void (*cia_encrypt)(void *ctx, u8 *dst, const u8 *src);
 	void (*cia_decrypt)(void *ctx, u8 *dst, const u8 *src);
+
+	unsigned int (*cia_encrypt_ecb)(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes);
+	unsigned int (*cia_decrypt_ecb)(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes);
+	unsigned int (*cia_encrypt_cbc)(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes);
+	unsigned int (*cia_decrypt_cbc)(const struct cipher_desc *desc,
+					u8 *dst, const u8 *src,
+					unsigned int nbytes);
 };
 
 struct digest_alg {
@@ -102,6 +124,7 @@ struct crypto_alg {
 	u32 cra_flags;
 	unsigned int cra_blocksize;
 	unsigned int cra_ctxsize;
+	unsigned int cra_alignmask;
 	const char cra_name[CRYPTO_MAX_ALG_NAME];
 
 	union {
@@ -136,7 +159,6 @@ static inline int crypto_alg_available(const char *name, u32 flags)
  * and core processing logic.  Managed via crypto_alloc_tfm() and
  * crypto_free_tfm(), as well as the various helpers below.
  */
-struct crypto_tfm;
 
 struct cipher_tfm {
 	void *cit_iv;
@@ -266,6 +288,16 @@ static inline unsigned int crypto_tfm_alg_digestsize(struct crypto_tfm *tfm)
 	return tfm->__crt_alg->cra_digest.dia_digestsize;
 }
 
+static inline unsigned int crypto_tfm_alg_alignmask(struct crypto_tfm *tfm)
+{
+	return tfm->__crt_alg->cra_alignmask;
+}
+
+static inline void *crypto_tfm_ctx(struct crypto_tfm *tfm)
+{
+	return (void *)&tfm[1];
+}
+
 /*
  * API wrappers.
  */