x86.s

UNIX、linux密码的破密程序源代码实现

/*
 * This file is part of John the Ripper password cracker,
 * Copyright (c) 1996-2003 by Solar Designer
 */

/*
 * x86 assembly routines.
 */

#include "arch.h"

#ifdef UNDERSCORES
#define DES_IV				_DES_IV
#define DES_count			_DES_count
#define DES_KS_current			_DES_KS_current
#define DES_KS_copy			_DES_KS_copy
#define DES_KS_table			_DES_KS_table
#define DES_SPE_L			_DES_SPE_L
#define DES_SPE_H			_DES_SPE_H
#define DES_SPE_F			_DES_SPE_F
#define DES_std_crypt			_DES_std_crypt
#define DES_xor_key1			_DES_xor_key1
#define DES_xor_key2			_DES_xor_key2
#define MD5_body			_MD5_body
#define BF_body				_BF_body
#define BF_current			_BF_current
#define CPU_detect			_CPU_detect
#endif

/*
 * Some broken systems don't offer section alignments larger than 4 bytes,
 * while for the MMX code we need at least an 8 byte alignment. ALIGN_FIX
 * is here to work around this issue when we happen to get bad addresses.
 */
#ifndef ALIGN_FIX
#ifdef ALIGN_LOG
#define DO_ALIGN(log)			.align (log)
#elif defined(DUMBAS)
#define DO_ALIGN(log)			.align 1 << log
#else
#define DO_ALIGN(log)			.align (1 << (log))
#endif
#else
#ifdef ALIGN_LOG
#define DO_ALIGN(log)			.align (log); .space 4
#else
#define DO_ALIGN(log)			.align (1 << (log)); .space 4
#endif
#endif

/*
 * DES stuff.
 *
 * Included are versions for:
 * 1. MMX (Pentium MMX and newer Intel x86 CPUs, some clones);
 * 2. Intel Pentium without MMX;
 * 3. The rest (good for Pentium Pro, 486, clones).
 *
 * MMX has to be enabled at compile time (via DES_X2 in arch.h), while #2
 * or #3 is chosen based on CPUID info at runtime.
 *
 * Note: MMX code for the bitslice DES implementation is in x86-mmx.S, so
 * you probably want to look in there instead.
 */

.text

#if DES_X2

/*
 * DES MMX routines.
 */

#define R				%mm0
#define L				%mm1
#define tmp1				%mm2
#define tmp2				%mm3
#define K1				%mm4
#define K2				%mm5
#define K3				%mm6
#define K4				%mm7

#define DES_copy(ofs1, ofs2) \
	movq ofs1(%edx),%mm0; \
	movq ofs2(%edx),%mm1; \
	movq %mm0,DES_KS_copy+ofs1; \
	movq %mm1,DES_KS_copy+ofs2

#define DES_2_ROUNDS_START(K) \
	pxor R,tmp1; \
	movd tmp1,%eax; \
	movb %al,%bl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	psrlq $32,tmp1; \
	pxor DES_SPE_F(%ebx,%ebx),L; \
	pxor DES_SPE_F+0x200(%edx,%edx),L; \
	movb %al,%bl; \
	movb %ah,%dl; \
	movd tmp1,%eax; \
	pxor DES_SPE_F+0x400(%ebx,%ebx),L; \
	pxor DES_SPE_F+0x600(%edx,%edx),L; \
	movb %al,%bl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	pxor DES_SPE_F+0x800(%ebx,%ebx),L; \
	pxor DES_SPE_F+0xA00(%edx,%edx),L; \
	movb %al,%bl; \
	movb %ah,%dl; \
	movq K,tmp1; \
	pxor DES_SPE_F+0xC00(%ebx,%ebx),L; \
	pxor DES_SPE_F+0xE00(%edx,%edx),L; \
	pxor L,tmp1; \
	movd tmp1,%eax; \
	movb %al,%bl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	psrlq $32,tmp1; \
	pxor DES_SPE_F(%ebx,%ebx),R; \
	pxor DES_SPE_F+0x200(%edx,%edx),R; \
	movb %al,%bl; \
	movb %ah,%dl; \
	movd tmp1,%eax; \
	pxor DES_SPE_F+0x400(%ebx,%ebx),R; \
	pxor DES_SPE_F+0x600(%edx,%edx),R; \
	movb %al,%bl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	pxor DES_SPE_F+0x800(%ebx,%ebx),R; \
	pxor DES_SPE_F+0xA00(%edx,%edx),R; \
	movb %al,%bl; \
	movb %ah,%dl

#define DES_2_ROUNDS(K1, K2) \
	DES_2_ROUNDS_START(K1); \
	movq K2,tmp1; \
	pxor DES_SPE_F+0xC00(%ebx,%ebx),R; \
	pxor DES_SPE_F+0xE00(%edx,%edx),R

DO_ALIGN(12)
.globl DES_std_crypt
DES_std_crypt:
	movl 4(%esp),%edx
	pushl %ebx
	movl DES_count,%ecx
	xorl %ebx,%ebx
	movq (%edx),K1
	movq 32(%edx),K2
	movq K1,tmp1
	movq 8(%edx),K3
	movq 16(%edx),K4
	DES_copy(24, 40)
	DES_copy(48, 56)
	DES_copy(64, 96)
	DES_copy(72, 80)
	DES_copy(88, 104)
	DES_copy(112, 120)
	movq DES_IV,R
	xorl %edx,%edx
	movq DES_IV+8,L
DES_loop:
	DES_2_ROUNDS(K3, K4)
	DES_2_ROUNDS(DES_KS_copy+24, K2)
	DES_2_ROUNDS(DES_KS_copy+40, DES_KS_copy+48)
	DES_2_ROUNDS(DES_KS_copy+56, DES_KS_copy+64)
	DES_2_ROUNDS(DES_KS_copy+72, DES_KS_copy+80)
	DES_2_ROUNDS(DES_KS_copy+88, DES_KS_copy+96)
	DES_2_ROUNDS(DES_KS_copy+104, DES_KS_copy+112)
	DES_2_ROUNDS_START(DES_KS_copy+120)
	movq K1,tmp1
	pxor DES_SPE_F+0xC00(%ebx,%ebx),R
	movq L,tmp2
	pxor DES_SPE_F+0xE00(%edx,%edx),R
	decl %ecx
	movq R,L
	movq tmp2,R
	jnz DES_loop
	movl 12(%esp),%edx
	popl %ebx
	movq R,(%edx)
	movq L,8(%edx)
#ifdef EMMS
	emms
#endif
	ret

#define DES_xor1(ofs1, ofs2) \
	movq DES_KS_current+ofs1,%mm0; \
	movq DES_KS_current+ofs2,%mm1; \
	pxor ofs1(%edx),%mm0; \
	pxor ofs2(%edx),%mm1; \
	movq %mm0,DES_KS_current+ofs1; \
	movq %mm1,DES_KS_current+ofs2

DO_ALIGN(4)
.globl DES_xor_key1
DES_xor_key1:
	movl 4(%esp),%edx
	DES_xor1(0, 32)
	DES_xor1(8, 16)
	DES_xor1(24, 40)
	DES_xor1(48, 56)
	DES_xor1(64, 96)
	DES_xor1(72, 80)
	DES_xor1(88, 104)
	DES_xor1(112, 120)
#ifdef EMMS
	emms
#endif
	ret

#define DES_xor2(ofs1, ofs2) \
	movq DES_KS_current+ofs1,%mm0; \
	movq DES_KS_current+ofs2,%mm1; \
	pxor ofs1(%edx),%mm0; \
	pxor ofs2(%edx),%mm1; \
	pxor ofs1(%ecx),%mm0; \
	pxor ofs2(%ecx),%mm1; \
	movq %mm0,DES_KS_current+ofs1; \
	movq %mm1,DES_KS_current+ofs2

DO_ALIGN(4)
.globl DES_xor_key2
DES_xor_key2:
	movl 4(%esp),%edx
	movl 8(%esp),%ecx
	DES_xor2(0, 32)
	DES_xor2(8, 16)
	DES_xor2(24, 40)
	DES_xor2(48, 56)
	DES_xor2(64, 96)
	DES_xor2(72, 80)
	DES_xor2(88, 104)
	DES_xor2(112, 120)
#ifdef EMMS
	emms
#endif
	ret

#else

/*
 * DES non-MMX routines.
 */

#define Rl				%ecx
#define Rh				%ebp
#define Ll				%esi
#define Lh				%edi

#define DES_copy(ofs) \
	movl ofs(%edx),%esi; \
	movl ofs+4(%edx),%edi; \
	movl %esi,DES_KS_copy+ofs; \
	movl %edi,DES_KS_copy+ofs+4

#define DES_CRYPT_START \
	movl 4(%esp),%edx; \
	pushl %ebp; \
	pushl %esi; \
	pushl %edi; \
	pushl %ebx; \
	movl (%edx),%eax; \
	movl 4(%edx),%edi; \
	movl %eax,DES_KS_copy; \
	movl %edi,DES_KS_copy+4; \
	DES_copy(32); \
	DES_copy(8); \
	DES_copy(16); \
	DES_copy(24); \
	DES_copy(40); \
	DES_copy(48); \
	DES_copy(56); \
	DES_copy(64); \
	DES_copy(96); \
	DES_copy(72); \
	DES_copy(80); \
	DES_copy(88); \
	DES_copy(104); \
	DES_copy(112); \
	DES_copy(120); \
	movl DES_IV,Rl; \
	movl DES_IV+4,Rh; \
	movl DES_IV+8,Ll; \
	movl DES_IV+12,Lh; \
	movl DES_count,%edx; \
	xorl %ebx,%ebx; \
	movl %edx,DES_count_tmp; \
	xorl %edx,%edx

#define DES_CRYPT_END \
	movl 24(%esp),%edx; \
	popl %ebx; \
	movl Rl,(%edx); \
	movl Rh,4(%edx); \
	movl Ll,8(%edx); \
	movl Lh,12(%edx); \
	popl %edi; \
	popl %esi; \
	popl %ebp; \
	ret

/*
 * Intel Pentium optimized version, extra operations are used to avoid
 * imperfect pairing.
 */

#define DES_2_ROUNDS_START_P5(K) \
	xorl Rl,%eax; \
	movb %al,%bl; \
	movb %ah,%dl; \
	movl Rl,DES_SavedL; \
	shrl $16,%eax; \
	movl DES_SPE_L(%ebx),Rl; \
	xorl Rl,Ll; \
	movl DES_SPE_H(%ebx),Rl; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x100(%edx),Rl; \
	xorl Rl,Ll; \
	movl DES_SPE_H+0x100(%edx),Rl; \
	movb %al,%bl; \
	xorl Rl,Lh; \
	movb %ah,%dl; \
	movl K+4,%eax; \
	movl DES_SPE_L+0x200(%ebx),Rl; \
	xorl Rh,%eax; \
	xorl Rl,Ll; \
	movl DES_SPE_H+0x200(%ebx),Rl; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x300(%edx),Rl; \
	xorl Rl,Ll; \
	movb %al,%bl; \
	movl DES_SPE_H+0x300(%edx),Rl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x400(%ebx),Rl; \
	xorl Rl,Ll; \
	movl DES_SPE_H+0x400(%ebx),Rl; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x500(%edx),Rl; \
	xorl Rl,Ll; \
	movb %al,%bl; \
	movl DES_SPE_H+0x500(%edx),Rl; \
	movb %ah,%dl; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x600(%ebx),Rl; \
	xorl Rl,Ll; \
	movl DES_SPE_H+0x600(%ebx),Rl; \
	xorl Rl,Lh; \
	movl DES_SPE_L+0x700(%edx),Rl; \
	xorl Rl,Ll; \
	movl K+8,%eax; \
	movl DES_SPE_H+0x700(%edx),Rl; \
	xorl Ll,%eax; \
	xorl Rl,Lh; \
	movb %al,%bl; \
	movl DES_SavedL,Rl; \
	movb %ah,%dl; \
	movl Ll,DES_SavedL; \
	shrl $16,%eax; \
	movl DES_SPE_L(%ebx),Ll; \
	xorl Ll,Rl; \
	movl DES_SPE_H(%ebx),Ll; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x100(%edx),Ll; \
	xorl Ll,Rl; \
	movl DES_SPE_H+0x100(%edx),Ll; \
	movb %al,%bl; \
	xorl Ll,Rh; \
	movb %ah,%dl; \
	movl K+12,%eax; \
	movl DES_SPE_L+0x200(%ebx),Ll; \
	xorl Lh,%eax; \
	xorl Ll,Rl; \
	movl DES_SPE_H+0x200(%ebx),Ll; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x300(%edx),Ll; \
	xorl Ll,Rl; \
	movb %al,%bl; \
	movl DES_SPE_H+0x300(%edx),Ll; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x400(%ebx),Ll; \
	xorl Ll,Rl; \
	movl DES_SPE_H+0x400(%ebx),Ll; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x500(%edx),Ll; \
	xorl Ll,Rl; \
	movb %al,%bl; \
	movl DES_SPE_H+0x500(%edx),Ll; \
	movb %ah,%dl; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x600(%ebx),Ll; \
	xorl Ll,Rl; \
	movl DES_SPE_H+0x600(%ebx),Ll; \
	xorl Ll,Rh; \
	movl DES_SPE_L+0x700(%edx),Ll

#define DES_2_ROUNDS_P5(K) \
	DES_2_ROUNDS_START_P5(K); \
	xorl Ll,Rl; \
	movl DES_SPE_H+0x700(%edx),Ll; \
	xorl Ll,Rh; \
	movl K+16,%eax; \
	movl DES_SavedL,Ll

DO_ALIGN(12)
DES_std_crypt_P5:
	DES_CRYPT_START
DES_loop_P5:
	DES_2_ROUNDS_P5(DES_KS_copy)
	DES_2_ROUNDS_P5(DES_KS_copy+16)
	DES_2_ROUNDS_P5(DES_KS_copy+32)
	DES_2_ROUNDS_P5(DES_KS_copy+48)
	DES_2_ROUNDS_P5(DES_KS_copy+64)
	DES_2_ROUNDS_P5(DES_KS_copy+80)
	DES_2_ROUNDS_P5(DES_KS_copy+96)
	DES_2_ROUNDS_START_P5(DES_KS_copy+112)
	xorl Rl,Ll
	movl DES_SPE_H+0x700(%edx),%eax
	xorl Rh,%eax
	movl Lh,Rh
	movl %eax,Lh
	movl DES_count_tmp,%eax
	movl DES_SavedL,Rl
	decl %eax
	movl %eax,DES_count_tmp
	movl DES_KS_copy,%eax
	jnz DES_loop_P5
	DES_CRYPT_END

/*
 * Generic x86 version.
 */

#define DES_2_ROUNDS_START_ANY(K) \
	xorl Rl,%eax; \
	movb %al,%bl; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	xorl DES_SPE_L(%ebx),Ll; \
	xorl DES_SPE_H(%ebx),Lh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x100(%edx),Ll; \
	xorl DES_SPE_H+0x100(%edx),Lh; \
	movb %ah,%dl; \
	movl K+4,%eax; \
	xorl DES_SPE_L+0x200(%ebx),Ll; \
	xorl Rh,%eax; \
	xorl DES_SPE_H+0x200(%ebx),Lh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x300(%edx),Ll; \
	xorl DES_SPE_H+0x300(%edx),Lh; \
	movb %ah,%dl; \
	xorl DES_SPE_L+0x400(%ebx),Ll; \
	shrl $16,%eax; \
	xorl DES_SPE_H+0x400(%ebx),Lh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x500(%edx),Ll; \
	xorl DES_SPE_H+0x500(%edx),Lh; \
	movb %ah,%dl; \
	xorl DES_SPE_L+0x600(%ebx),Ll; \
	xorl DES_SPE_H+0x600(%ebx),Lh; \
	xorl DES_SPE_L+0x700(%edx),Ll; \
	movl K+8,%eax; \
	xorl Ll,%eax; \
	movb %al,%bl; \
	xorl DES_SPE_H+0x700(%edx),Lh; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	xorl DES_SPE_L(%ebx),Rl; \
	xorl DES_SPE_H(%ebx),Rh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x100(%edx),Rl; \
	xorl DES_SPE_H+0x100(%edx),Rh; \
	movb %ah,%dl; \
	movl K+12,%eax; \
	xorl DES_SPE_L+0x200(%ebx),Rl; \
	xorl Lh,%eax; \
	xorl DES_SPE_H+0x200(%ebx),Rh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x300(%edx),Rl; \
	xorl DES_SPE_H+0x300(%edx),Rh; \
	movb %ah,%dl; \
	shrl $16,%eax; \
	xorl DES_SPE_L+0x400(%ebx),Rl; \
	xorl DES_SPE_H+0x400(%ebx),Rh; \
	movb %al,%bl; \
	xorl DES_SPE_L+0x500(%edx),Rl; \
	xorl DES_SPE_H+0x500(%edx),Rh; \
	movb %ah,%dl

#define DES_2_ROUNDS_ANY(K) \
	DES_2_ROUNDS_START_ANY(K); \
	xorl DES_SPE_L+0x600(%ebx),Rl; \
	xorl DES_SPE_H+0x600(%ebx),Rh; \
	movl K+16,%eax; \
	xorl DES_SPE_L+0x700(%edx),Rl; \
	xorl DES_SPE_H+0x700(%edx),Rh

DO_ALIGN(12)
DES_std_crypt_generic:
	DES_CRYPT_START
DES_loop_generic:
	DES_2_ROUNDS_ANY(DES_KS_copy)
	DES_2_ROUNDS_ANY(DES_KS_copy+16)
	DES_2_ROUNDS_ANY(DES_KS_copy+32)
	DES_2_ROUNDS_ANY(DES_KS_copy+48)
	DES_2_ROUNDS_ANY(DES_KS_copy+64)
	DES_2_ROUNDS_ANY(DES_KS_copy+80)
	DES_2_ROUNDS_ANY(DES_KS_copy+96)
	DES_2_ROUNDS_START_ANY(DES_KS_copy+112)
	movl Ll,%eax
	xorl DES_SPE_L+0x600(%ebx),Rl
	xorl DES_SPE_H+0x600(%ebx),Rh
	movl DES_SPE_L+0x700(%edx),Ll
	xorl Rl,Ll
	movl %eax,Rl
	movl Lh,%eax
	movl DES_SPE_H+0x700(%edx),Lh
	xorl Rh,Lh
	decl DES_count_tmp
	movl %eax,Rh
	movl DES_KS_copy,%eax
	jnz DES_loop_generic
	DES_CRYPT_END

#define DES_xor1(ofs) \
	movl DES_KS_current+ofs,%esi; \
	movl DES_KS_current+ofs+4,%edi; \
	movl ofs(%edx),%eax; \
	movl ofs+4(%edx),%ecx; \
	xorl %esi,%eax; \
	xorl %edi,%ecx; \
	movl %eax,DES_KS_current+ofs; \
	movl %ecx,DES_KS_current+ofs+4

DO_ALIGN(4)
.globl DES_xor_key1
DES_xor_key1:
	movl 4(%esp),%edx
	pushl %esi
	pushl %edi
	DES_xor1(0)
	DES_xor1(32)
	DES_xor1(8)
	DES_xor1(16)
	DES_xor1(24)
	DES_xor1(40)
	DES_xor1(48)
	DES_xor1(56)
	DES_xor1(64)
	DES_xor1(96)
	DES_xor1(72)
	DES_xor1(80)
	DES_xor1(88)
	DES_xor1(104)
	DES_xor1(112)
	DES_xor1(120)
	popl %edi
	popl %esi
	ret

#define DES_xor2(ofs) \
	movl ofs(%edx),%eax; \
	movl ofs+4(%edx),%ecx; \
	movl ofs(%ebx),%esi; \
	movl ofs+4(%ebx),%edi; \
	xorl %esi,%eax; \
	xorl %edi,%ecx; \
	movl DES_KS_current+ofs,%esi; \
	movl DES_KS_current+ofs+4,%edi; \
	xorl %esi,%eax; \
	xorl %edi,%ecx; \
	movl %eax,DES_KS_current+ofs; \
	movl %ecx,DES_KS_current+ofs+4

DO_ALIGN(4)
.globl DES_xor_key2
DES_xor_key2:
	pushl %ebx
	movl 8(%esp),%edx
	movl 12(%esp),%ebx
	pushl %esi
	pushl %edi
	DES_xor2(0)
	DES_xor2(32)
	DES_xor2(8)
	DES_xor2(16)
	DES_xor2(24)
	DES_xor2(40)
	DES_xor2(48)
	DES_xor2(56)
	DES_xor2(64)
	DES_xor2(96)
	DES_xor2(72)
	DES_xor2(80)
	DES_xor2(88)
	DES_xor2(104)
	DES_xor2(112)
	DES_xor2(120)
	popl %edi
	popl %esi
	popl %ebx
	ret

#endif

.data

/*
 * Weird alignments to make sure KS address's bits 5-11 never match current
 * instruction pointer's bits 5-11.
 */
#ifdef DUMBAS
DO_ALIGN(12)
.zero 0x1000 - 32 - 128
#elif defined(__DJGPP__)
.text
DO_ALIGN(12)
.space (0x1000 - 32 - 128 - 0xE0)
#else
DO_ALIGN(12)
.space (0x1000 - 32 - 128)
#endif

#if !DES_X2
/*
 * The function pointer, set by CPU_detect().
 */
.globl DES_std_crypt
DES_std_crypt:
.long DES_std_crypt_generic

DES_SavedL:
.long 0
#endif

/*
 * These are in .data, not .bss, to get them in a cache line that has to be
 * already loaded at DES_std_crypt() startup.
 */
.globl DES_IV
DES_IV:
#ifdef DUMBAS
.zero 16
#else
.space 16
#endif

.globl DES_count
DES_count:
.long 0

DES_count_tmp:
.long 0

DO_ALIGN(5)
.globl DES_KS_copy
DES_KS_copy:
#ifdef DUMBAS
.zero 128
#else
.space 128
#endif

#ifdef __DJGPP__
.space 32
#endif

#ifndef BSD
.bss
#endif

#if DES_X2

.globl DES_SPE_F