ohroy / blog

A super blog lite -- just one page. use vue with github api !

Home Page:https://blog.6h.work

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

关于ios下dlopen函数的探究

ohroy opened this issue · comments

commented

内容摘要

近日我在开发一个小工具的时候,用到了一个加载动态库的函数dlopen,基于我的windows平台的开发背景,我想当然的认为其返回值是当前moduleload address,即模块基地址。然后经验证后发现,其和windows下的LoadLibrary还是有本质的差别。那么,它返回的究竟是一个什么东西呢?又该如何通过这样的返回值,来顺利获取到我们需要的load address甚至其它更重要的信息呢?带着这样的疑惑,我与几位资深经验的开发者对这个问题进行了深入的探究。本文是探究思路和研究过程的记录。在读文本文后,你应当会对此方面了解有所加深,并且上述疑惑应当能够释疑。

前知储备

本文面向有一定技术水准的读者,所以不再赘述模块基地址等相关定义。但相对地,下文出现的一些术语或者俗语,在此做一个简短的说明,如果需要进一步了解,请参考附录或者自行检索相关资料。

  • aslr(Address space layout randomization),加载地址随机化,通俗来讲,就是同一个模块在每个进程每次被加载后的地址(可能)不一致。
  • load address,模块加载地址,即模块基地址,也等同于header address
  • patch 补丁,这里名次动用做打补丁

为什么需要load address

在通常开发情境下,一般地代码片段如下:

int main(int argc, char **argv) {  
        void *handle;  
        bool (*fuck)(girl);  
        char *error;  
      
        handle = dlopen ("libfuck.dylib", RTLD_LAZY);  
        if (!handle) {  
            fprintf (stderr, "%s ", dlerror());  
            exit(1);  
        }  
      
        fuck = dlsym(handle, "fuck");  
        if ((error = dlerror()) != NULL)  {  
            fprintf (stderr, "%s ", error);  
            exit(1);  
        }
        girl* xx= make_girl_byname("xx");
        fuck(girl);
        dlclose(handle);  
        return 0;  
    }  

但特殊情况下,我们无法通过dlsym来完成符号的寻找。如该模块根本没有导出符号,或者我们想要显式的调用某个未导出符号的函数,或者我们需要patchheader中的某个字段等情形时,我们就需要知道该模块的load address

如何获取load address

由于aslr的原因,现行通用的解决方案是通过遍历模块来获取。典型的代码片段如下:

int32_t nModNums= _dyld_image_count();
const char *pszModName = NULL;
intptr_t pModSlide = 0;
const struct mach_header* pModHeader = NULL;
    
for (uint32_t i = 0; i < nModNums; i++)
{
	
	pszModName = _dyld_get_image_name(i);
	if(!strcmp(pszModName,"/path/to/xx.dylib")){
    	pModSlide  = _dyld_get_image_vmaddr_slide(i);
    	pModHeader = _dyld_get_image_header(i);
    	NSLog(@"[FLWB]:vm_slide:%lx,load addr:%p",pModSlide,pModHeader);
	}

}

这样,我们就通过对比名字的方式,来获取到了header address
然而我们思考一下,这样做真的没有问题吗???
我认为,起码有两三个问题。

  1. 遍历导致的性能损失。
  2. 此处至少需要调用三个api来完成操作,而此三个api有较明显的特征,非常容易被hook针对。
  3. 也是最重要的,此处使用的几个api是线程安全的。这里是一般情况下我们不会注意到的,实际上,这个是一个常见的安全漏洞。而文档上也对此有说明
/*
 * The following functions allow you to iterate through all loaded images.  
 * This is not a thread safe operation.  Another thread can add or remove
 * an image during the iteration.  
 *
 * Many uses of these routines can be replace by a call to dladdr() which 
 * will return the mach_header and name of an image, given an address in 
 * the image. dladdr() is thread safe.
 */
extern uint32_t                    _dyld_image_count(void)                              __OSX_AVAILABLE_STARTING(__MAC_10_1, __IPHONE_2_0);
extern const struct mach_header*   _dyld_get_image_header(uint32_t image_index)         __OSX_AVAILABLE_STARTING(__MAC_10_1, __IPHONE_2_0);
extern intptr_t                    _dyld_get_image_vmaddr_slide(uint32_t image_index)   __OSX_AVAILABLE_STARTING(__MAC_10_1, __IPHONE_2_0);
extern const char*                 _dyld_get_image_name(uint32_t image_index)           __OSX_AVAILABLE_STARTING(__MAC_10_1, __IPHONE_2_0);

大意是要警惕在遍历时其它进程的添加或者修改操作。那么。显然地,假使我们在遍历的过程中,其它线程删除了一个模块,那么将导致我们的索引移位,得到一个完全不匹配的结果,同时也有可能造成内存访问错误。

那么,要想同时解决以上三个问题,(我认为)最优秀的解决应当是通过dlopen返回的结果来获取。从推理上来讲,dlsym既然能通过这个参数来获取到符号信息,而符号表是需要通过header来计算的。换言之,通过这个dlopen的返回值一定可以获取到header。搜索了一番,网上对此没有任何资料。
但理论上行得通的事情,实际上应该是可以的。于是我开始了探寻。

探寻过程

有了上文的资料,显然,dlopen的返回值必定是一个结构。我们由此来做发散思维。与此同时,我与几位大佬交流后,他们也都给出了非常重要的建议。

头文件及注释

这个是一个最简单的方案,我们去看一下dlopen的头文件声明,其位于/usr/include/dlfcn.h

extern int dlclose(void * __handle);
extern char * dlerror(void);
extern void * dlopen(const char * __path, int __mode);
extern void * dlsym(void * __handle, const char * __symbol);

然而这里面定义它为void* 类型,且变量名字为handle
我们幻想中此处应为一个结构体指针的美梦破碎了。

借用linux头文件

此方案由咸🐟提出。

On Linux, dlopen doesn't return the address where the ELF binary was loaded. It returns struct link_map instead, which has .l_addr member. So you'll want something like:

struct link_map *lm = (struct link_map*) dlopen(0, RTLD_NOW);
printf("%p\n", lm->l_addr);

However, despite what comment in /usr/include/link.h says, .l_addr is actually not a load address either. Instead, it's the difference between where ELF image was linked to load, and where it was actually loaded.

看到这份资料的时候我还是比较兴奋的,因为这无疑证明了我们的推断是正确的。然而我一番寻找,发现在macos下根本没有link.h这个头文件。一番搜索没有答案之下,我们安装windows subsystem然后顺利找到这个头文件。
其中关键部分如下:

struct link_map
  {
    /* These first few members are part of the protocol with the debugger.
       This is the same format used in SVR4.  */

    ElfW(Addr) l_addr;		/* Difference between the address in the ELF
				   file and the addresses in memory.  */
    char *l_name;		/* Absolute file name object was found in.  */
    ElfW(Dyn) *l_ld;		/* Dynamic section of the shared object.  */
    struct link_map *l_next, *l_prev; /* Chain of loaded objects.  */
  };

这里我们简单的复制此关键部分,并修改为

struct link_map
  {
    /* These first few members are part of the protocol with the debugger.
       This is the same format used in SVR4.  */

    void l_addr;		/* Difference between the address in the ELF
				   file and the addresses in memory.  */
    char *l_name;		/* Absolute file name object was found in.  */
    void*l_ld;		/* Dynamic section of the shared object.  */
    struct link_map *l_next, *l_prev; /* Chain of loaded objects.  */
  };

我们的代码如下:

	void* xxHandle=dlopen("xx.dylib",RTLD_LAZY);
	NSLog(@"[FLWB]:xxHandle:%p",xxHandle);
	struct link_map *lm = (struct link_map*)xxHandle;
	NSLog(@"[FLWB]:%s = %p",lm->l_name, lm->l_addr);

运行之后神奇的事情发生了,模块名字竟然正确的输出了!虽然基地址返回的明显不对。但至少证明了它确实是有组织的。但可能因为苹果做了修改导致了不同。
我们继续尝试,发现除了名字奇迹般的正确输出以外,其它的无一相符。
这里提一句题外话,我们显然可以看到linux下的设计还是比较精巧的,它使用了链表来存在模块信息,方便插入和遍历。这一点上和windows是一致的。在此情形下,我们需要隐藏模块的话,就需要修改相邻的两个链表左右节点,也就是所谓的“断链”。有兴趣的话,可以自行研究。

借用anroid头文件

linux的头文件无法适用于ios,是否是因为版本的问题呢,android是否可行?一番搜索之下我们找到来安卓的对应头文件

struct soinfo {
 public:
  char name[SOINFO_NAME_LEN];
  const Elf32_Phdr* phdr;
  size_t phnum;
  Elf32_Addr entry;
  Elf32_Addr base;
  unsigned size;

  uint32_t unused1;  // DO NOT USE, maintained for compatibility.

  Elf32_Dyn* dynamic;

  uint32_t unused2; // DO NOT USE, maintained for compatibility
  uint32_t unused3; // DO NOT USE, maintained for compatibility

  soinfo* next;
  unsigned flags;

  const char* strtab;
  Elf32_Sym* symtab;

  size_t nbucket;
  size_t nchain;
  unsigned* bucket;
  unsigned* chain;

  unsigned* plt_got;

  Elf32_Rel* plt_rel;
  size_t plt_rel_count;

  Elf32_Rel* rel;
  size_t rel_count;

  linker_function_t* preinit_array;
  size_t preinit_array_count;

  linker_function_t* init_array;
  size_t init_array_count;
  linker_function_t* fini_array;
  size_t fini_array_count;

  linker_function_t init_func;
  linker_function_t fini_func;

#if defined(ANDROID_ARM_LINKER)
  // ARM EABI section used for stack unwinding.
  unsigned* ARM_exidx;
  size_t ARM_exidx_count;
#elif defined(ANDROID_MIPS_LINKER)
  unsigned mips_symtabno;
  unsigned mips_local_gotno;
  unsigned mips_gotsym;
#endif

  size_t ref_count;
  link_map_t link_map;

  bool constructors_called;

  // When you read a virtual address from the ELF file, add this
  // value to get the corresponding address in the process' address space.
  Elf32_Addr load_bias;

  bool has_text_relocations;
  bool has_DT_SYMBOLIC;

  void CallConstructors();
  void CallDestructors();
  void CallPreInitConstructors();

 private:
  void CallArray(const char* array_name, linker_function_t* functions, size_t count, bool reverse);
  void CallFunction(const char* function_name, linker_function_t function);
};

然而一番尝试后同样无果,这彻底说明了,至少这部分apple是有自己的实现。

特殊符号定位法

注意,这里的特殊符号不是指的标点符号,而是特殊的symbol
此方案同样由咸🐟提出,并由Zhang扩充。
原理是通过macho里导出的特殊符号来直接获取。具体的符号如

/*
 * The value of the link editor defined symbol _MH_DYLIB_SYM is the address
 * of the mach header in a Mach-O dylib file type.  It does not appear in
 * any file type other than a MH_DYLIB file type.  The type of the symbol is
 * an N_SECT symbol even thought the header is not part of any section.  This
 * symbol is private to the code in the library it is a part of.
 */
#define _MH_DYLIB_SYM	"__mh_dylib_header"
#define MH_DYLIB_SYM	"_mh_dylib_header"

我们修改代码如下:

	void* xxHandle=dlopen("fuck.dylib",RTLD_LAZY);
	void* fuck1 = dlsym(xxHandle, "__mh_dylib_header");  
    void* fuck2 = dlsym(xxHandle, "_mh_dylib_header"); 

结果均获取不到这两个符号,推测为我这边测试环境的问题,或者在链接的过程中被去掉了。

特定符号偏移修正法

此方案由柳下惠提出,原理是通过查找给定的某个符号,然后减去对应的偏移来获取。这种是理论上肯定能行得通的解决方案,但实际上如若是遇到某个模块根本没有导出符号的话,此方案也是束手无策。此处有jmpzws基于此而来的优化方案。即是查找dyld_stub_binder这个特定的符号,align然后减去0x4000。
但同样地,我这边也是无法获取到这个符号。

回调获取法

此方案由Zhang提出,原理是先通过_dyld_register_func_for_add_image来注册一个回调,然后再dlopen即可。因为此回调回调用void (*func)(const mach_header *mh, intptr_t vmaddr_slide)这样的callback,在参数里面刚好就有mach_header,以此来获取模块基地址。此方案从理论上来讲,是完全可行的。但由于操作过于复杂,且(我)没有找到对应的取消注册钩子的办法导致使用代价过大,所以没有深入探寻。

暴力偏移法

多番尝试无果后,现行简单粗暴的方法。即dump出疑似结构的handle的内存,观察load address的位置,然后计算偏移,往后通过偏移来获取。
这种思路类似于游戏外挂的思路,即不管缘由,不考虑理论,先行实现。
于是我们修改代码如下

	void* xxHandle=dlopen("xx.dylib",RTLD_LAZY);
	NSLog(@"[FLWB]:xxHandle:%p",xxHandle);
	NSData *dataData = [NSData dataWithBytes:xxHandle length:0x80];
	NSLog(@"[FLWB]:%p = %@",xxHandle, dataData);

果然,我们在第0x58个成员的位置找到了疑似load address的位置。但这在实际使用上会有很大的问题。其中最为麻烦的是32位和64位的问题,由于类型长度可能不一致,所以在32位下未必是0x58/2,而且在执行时要定义宏或者手动判断系统位数,显然是一个比较繁琐的事情。
但这进一步说明了我们的思路的正确的,一定存在这样的一个结构体。正是由于种种方案给定我们信心,我们才能坚定的走下去。
所以我们下一步就是要找到这个头文件,如果找不到,那我们就需要猜测,然后自行做一个这样的头文件。(反正windows下都是这么做的。)

源码追溯法

正在此时,突然想起来jmpzws曾发的一个截图。

ImageLoader* image = (ImageLoader*)(((uintptr_t)handle) & (-4));	// clear mode bits

那么,这个ImageLoader可不就是我们苦苦寻找的结构体吗?于是我们找到对应的dyld的源码。
我们一番寻找后,发现其原来是个类,头文件如下

/* -*- mode: C++; c-basic-offset: 4; tab-width: 4 -*-
 *
 * Copyright (c) 2004-2010 Apple Inc. All rights reserved.
 *
 * @APPLE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this
 * file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */


#ifndef __IMAGELOADER__
#define __IMAGELOADER__

#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <mach/mach_time.h> // struct mach_timebase_info
#include <mach/mach_init.h> // struct mach_thread_self
#include <mach/shared_region.h>
#include <mach-o/loader.h> 
#include <mach-o/nlist.h> 
#include <stdint.h>
#include <stdlib.h>
#include <TargetConditionals.h>
#include <vector>
#include <new>
#include <uuid/uuid.h>

#if __arm__
 #include <mach/vm_page_size.h>
#endif

#if __x86_64__ || __i386__
	#include <CrashReporterClient.h>
#else
	// work around until iOS has CrashReporterClient.h
	#define CRSetCrashLogMessage(x)
	#define CRSetCrashLogMessage2(x)
#endif

#ifndef SHARED_REGION_BASE_ARM64
	#define SHARED_REGION_BASE_ARM64 0x7FFF80000000LL
#endif

#ifndef SHARED_REGION_SIZE_ARM64
	#define SHARED_REGION_SIZE_ARM64 0x10000000LL
#endif


#define LOG_BINDINGS 0

#include "mach-o/dyld_images.h"
#include "mach-o/dyld_priv.h"
#include "DyldSharedCache.h"

#if __i386__
	#define SHARED_REGION_BASE SHARED_REGION_BASE_I386
	#define SHARED_REGION_SIZE SHARED_REGION_SIZE_I386
#elif __x86_64__
	#define SHARED_REGION_BASE SHARED_REGION_BASE_X86_64
	#define SHARED_REGION_SIZE SHARED_REGION_SIZE_X86_64
#elif __arm__
	#define SHARED_REGION_BASE SHARED_REGION_BASE_ARM
	#define SHARED_REGION_SIZE SHARED_REGION_SIZE_ARM
#elif __arm64__
	#define SHARED_REGION_BASE SHARED_REGION_BASE_ARM64
	#define SHARED_REGION_SIZE SHARED_REGION_SIZE_ARM64
#endif

#ifndef EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER
	#define EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER 0x10
#endif
#ifndef EXPORT_SYMBOL_FLAGS_REEXPORT
	#define EXPORT_SYMBOL_FLAGS_REEXPORT 0x08
#endif

#ifndef LC_MAIN
	#define LC_MAIN (0x28|LC_REQ_DYLD) /* replacement for LC_UNIXTHREAD */
	struct entry_point_command {
		uint32_t  cmd;	/* LC_MAIN only used in MH_EXECUTE filetypes */
		uint32_t  cmdsize;	/* 24 */
		uint64_t  entryoff;	/* file (__TEXT) offset of main() */
		uint64_t  stacksize;/* if not zero, initial stack size */
	};
#endif

#if __IPHONE_OS_VERSION_MIN_REQUIRED          
	#define SPLIT_SEG_SHARED_REGION_SUPPORT 0
	#define SPLIT_SEG_DYLIB_SUPPORT			0
	#define PREBOUND_IMAGE_SUPPORT			__arm__
	#define TEXT_RELOC_SUPPORT				__i386__
	#define SUPPORT_OLD_CRT_INITIALIZATION	0
	#define SUPPORT_LC_DYLD_ENVIRONMENT		1
	#define SUPPORT_VERSIONED_PATHS			1
	#define SUPPORT_CLASSIC_MACHO			__arm__
	#define SUPPORT_ZERO_COST_EXCEPTIONS	(!__USING_SJLJ_EXCEPTIONS__)
	#define INITIAL_IMAGE_COUNT				150
	#define SUPPORT_ACCELERATE_TABLES		!TARGET_IPHONE_SIMULATOR
	#define SUPPORT_ROOT_PATH				TARGET_IPHONE_SIMULATOR
	#define USES_CHAINED_BINDS				(__arm64e__)
#else
	#define SPLIT_SEG_SHARED_REGION_SUPPORT 0
	#define SPLIT_SEG_DYLIB_SUPPORT			__i386__
	#define PREBOUND_IMAGE_SUPPORT			__i386__
	#define TEXT_RELOC_SUPPORT				__i386__
	#define SUPPORT_OLD_CRT_INITIALIZATION	__i386__
	#define SUPPORT_LC_DYLD_ENVIRONMENT		(__i386__ || __x86_64__)
	#define SUPPORT_VERSIONED_PATHS			1
	#define SUPPORT_CLASSIC_MACHO			1
	#define SUPPORT_ZERO_COST_EXCEPTIONS	1
	#define INITIAL_IMAGE_COUNT				200
	#define SUPPORT_ACCELERATE_TABLES		0
	#define SUPPORT_ROOT_PATH				1
	#define USES_CHAINED_BINDS				0
#endif

#define MAX_MACH_O_HEADER_AND_LOAD_COMMANDS_SIZE (32*1024)

#define MH_HAS_OBJC			0x40000000


// <rdar://problem/13590567> optimize away dyld's initializers
#define VECTOR_NEVER_DESTRUCTED(type) \
	namespace std { \
		template <> \
		__vector_base<type, std::allocator<type> >::~__vector_base() { } \
	}
#define VECTOR_NEVER_DESTRUCTED_EXTERN(type) \
       namespace std { \
               template <> \
               __vector_base<type, std::allocator<type> >::~__vector_base(); \
       }
#define VECTOR_NEVER_DESTRUCTED_IMPL(type) \
       namespace std { \
               template <> \
               __vector_base<type, std::allocator<type> >::~__vector_base() { } \
       }

// utilities
namespace dyld {
	extern __attribute__((noreturn)) void throwf(const char* format, ...)  __attribute__((format(printf, 1, 2)));
	extern void log(const char* format, ...)  __attribute__((format(printf, 1, 2)));
	extern void warn(const char* format, ...)  __attribute__((format(printf, 1, 2)));
	extern const char* mkstringf(const char* format, ...)  __attribute__((format(printf, 1, 2)));
#if LOG_BINDINGS
	extern void logBindings(const char* format, ...)  __attribute__((format(printf, 1, 2)));
#endif
}
extern "C" 	int   vm_alloc(vm_address_t* addr, vm_size_t size, uint32_t flags);
extern "C" 	void* xmmap(void* addr, size_t len, int prot, int flags, int fd, off_t offset);


#if __LP64__
	struct macho_header				: public mach_header_64  {};
	struct macho_nlist				: public nlist_64  {};	
#else
	struct macho_header				: public mach_header  {};
	struct macho_nlist				: public nlist  {};	
#endif


#if __arm64__
	#define dyld_page_trunc(__addr)     (__addr & (-16384))
	#define dyld_page_round(__addr)     ((__addr + 16383) & (-16384))
	#define dyld_page_size              16384
#elif __arm__
	#define dyld_page_trunc(__addr)     trunc_page_kernel(__addr)
	#define dyld_page_round(__addr)     round_page_kernel(__addr)
	#define dyld_page_size              vm_kernel_page_size
#else
	#define dyld_page_trunc(__addr)     (__addr & (-4096))
	#define dyld_page_round(__addr)     ((__addr + 4095) & (-4096))
	#define dyld_page_size              4096
#endif



struct ProgramVars
{
	const void*		mh;
	int*			NXArgcPtr;
	const char***	NXArgvPtr;
	const char***	environPtr;
	const char**	__prognamePtr;
};



//
// ImageLoader is an abstract base class.  To support loading a particular executable
// file format, you make a concrete subclass of ImageLoader.
//
// For each executable file (dynamic shared object) in use, an ImageLoader is instantiated.
//
// The ImageLoader base class does the work of linking together images, but it knows nothing
// about any particular file format.
//
//
class ImageLoader {
public:

	typedef uint32_t DefinitionFlags;
	static const DefinitionFlags kNoDefinitionOptions = 0;
	static const DefinitionFlags kWeakDefinition = 1;
	
	typedef uint32_t ReferenceFlags;
	static const ReferenceFlags kNoReferenceOptions = 0;
	static const ReferenceFlags kWeakReference = 1;
	static const ReferenceFlags kTentativeDefinition = 2;
	
	enum PrebindMode { kUseAllPrebinding, kUseSplitSegPrebinding, kUseAllButAppPredbinding, kUseNoPrebinding };
	enum BindingOptions { kBindingNone, kBindingLazyPointers, kBindingNeverSetLazyPointers };
	enum SharedRegionMode { kUseSharedRegion, kUsePrivateSharedRegion, kDontUseSharedRegion, kSharedRegionIsSharedCache };
	
	struct Symbol;  // abstact symbol

	struct MappedRegion {
		uintptr_t	address;
		size_t		size;
	};

	struct RPathChain {
		RPathChain(const RPathChain* n, std::vector<const char*>* p) : next(n), paths(p) {};
		const RPathChain*			next;
		std::vector<const char*>*	paths;
	};

	struct DOFInfo {
		void*				dof;
		const mach_header*	imageHeader;
		const char*			imageShortName;
	};

	struct DynamicReference {
		ImageLoader* from;
		ImageLoader* to;
	};
	
	struct InitializerTimingList
	{
		uintptr_t	count;
		struct {
			const char*		shortName;
			uint64_t		initTime;
		}			images[1];

		void addTime(const char* name, uint64_t time);
	};

	typedef void (^CoalesceNotifier)(const Symbol* implSym, const ImageLoader* implIn, const mach_header* implMh);
	
	struct LinkContext {
		ImageLoader*	(*loadLibrary)(const char* libraryName, bool search, const char* origin, const RPathChain* rpaths, bool enforceIOSMac, unsigned& cacheIndex);
		void			(*terminationRecorder)(ImageLoader* image);
		bool			(*flatExportFinder)(const char* name, const Symbol** sym, const ImageLoader** image);
		bool			(*coalescedExportFinder)(const char* name, const Symbol** sym, const ImageLoader** image, CoalesceNotifier);
		unsigned int	(*getCoalescedImages)(ImageLoader* images[], unsigned imageIndex[]);
		void			(*undefinedHandler)(const char* name);
		MappedRegion*	(*getAllMappedRegions)(MappedRegion*);
		void *			(*bindingHandler)(const char *, const char *, void *);
		void			(*notifySingle)(dyld_image_states, const ImageLoader* image, InitializerTimingList*);
		void			(*notifyBatch)(dyld_image_states state, bool preflightOnly);
		void			(*removeImage)(ImageLoader* image);
		void			(*registerDOFs)(const std::vector<DOFInfo>& dofs);
		void			(*clearAllDepths)();
		void			(*printAllDepths)();
		unsigned int	(*imageCount)();
		void			(*setNewProgramVars)(const ProgramVars&);
		bool			(*inSharedCache)(const char* path);
		void			(*setErrorStrings)(unsigned errorCode, const char* errorClientOfDylibPath,
										const char* errorTargetDylibPath, const char* errorSymbol);
		ImageLoader*	(*findImageContainingAddress)(const void* addr);
		void			(*addDynamicReference)(ImageLoader* from, ImageLoader* to);
#if SUPPORT_ACCELERATE_TABLES
		void			(*notifySingleFromCache)(dyld_image_states, const mach_header* mh, const char* path);
		dyld_image_state_change_handler (*getPreInitNotifyHandler)(unsigned index);
		dyld_image_state_change_handler (*getBoundBatchHandler)(unsigned index);
#endif
		
#if SUPPORT_OLD_CRT_INITIALIZATION
		void			(*setRunInitialzersOldWay)();
#endif
		BindingOptions	bindingOptions;
		int				argc;
		const char**	argv;
		const char**	envp;
		const char**	apple;
		const char*		progname;
		ProgramVars		programVars;
		ImageLoader*	mainExecutable;
		const char* const * imageSuffix;
#if SUPPORT_ROOT_PATH
		const char**	rootPaths;
#endif
		const DyldSharedCache*  dyldCache;
		const dyld_interpose_tuple*	dynamicInterposeArray;
		size_t			dynamicInterposeCount;
		PrebindMode		prebindUsage;
		SharedRegionMode sharedRegionMode;
		bool			dyldLoadedAtSameAddressNeededBySharedCache;
		bool			strictMachORequired;
		bool			allowAtPaths;
		bool			allowEnvVarsPrint;
		bool			allowEnvVarsPath;
		bool			allowEnvVarsSharedCache;
		bool			allowClassicFallbackPaths;
		bool			allowInsertFailures;
		bool			mainExecutableCodeSigned;
		bool			preFetchDisabled;
		bool			prebinding;
		bool			bindFlat;
		bool			linkingMainExecutable;
		bool			startedInitializingMainExecutable;
#if __MAC_OS_X_VERSION_MIN_REQUIRED
		bool			marzipan;
#endif
		bool			verboseOpts;
		bool			verboseEnv;
		bool			verboseLoading;
		bool			verboseMapping;
		bool			verboseRebase;
		bool			verboseBind;
		bool			verboseWeakBind;
		bool			verboseInit;
		bool			verboseDOF;
		bool			verbosePrebinding;
		bool			verboseCoreSymbolication;
		bool			verboseWarnings;
		bool			verboseRPaths;
		bool			verboseInterposing;
		bool			verboseCodeSignatures;
	};
	
	struct CoalIterator
	{
		ImageLoader*	image;
		const char*		symbolName;
		unsigned int	loadOrder;
		bool			weakSymbol;
		bool			symbolMatches;
		bool			done;
		// the following are private to the ImageLoader subclass
		uintptr_t		curIndex;
		uintptr_t		endIndex;
		uintptr_t		address;
		uintptr_t		type;
		uintptr_t		addend;
		uintptr_t		imageIndex;
	};
	
	virtual	void			initializeCoalIterator(CoalIterator&, unsigned int loadOrder, unsigned imageIndex) = 0;
	virtual	bool			incrementCoalIterator(CoalIterator&) = 0;
	virtual	uintptr_t		getAddressCoalIterator(CoalIterator&, const LinkContext& context) = 0;
	virtual	void			updateUsesCoalIterator(CoalIterator&, uintptr_t newAddr, ImageLoader* target, unsigned targetIndex, const LinkContext& context) = 0;
	
	struct UninitedUpwards
	{
		uintptr_t	 count;
		ImageLoader* images[1];
	};

	
										// constructor is protected, but anyone can delete an image
	virtual								~ImageLoader();
	
										// link() takes a newly instantiated ImageLoader and does all 
										// fixups needed to make it usable by the process
	void								link(const LinkContext& context, bool forceLazysBound, bool preflight, bool neverUnload, const RPathChain& loaderRPaths, const char* imagePath);
	
										// runInitializers() is normally called in link() but the main executable must 
										// run crt code before initializers
	void								runInitializers(const LinkContext& context, InitializerTimingList& timingInfo);
	
										// called after link() forces all lazy pointers to be bound
	void								bindAllLazyPointers(const LinkContext& context, bool recursive);
	
										// used by dyld to see if a requested library is already loaded (might be symlink)
	bool								statMatch(const struct stat& stat_buf) const;

										// get short name of this image
	const char*							getShortName() const;

										// returns leaf name
	static const char*					shortName(const char* fullName);

										// get path used to load this image, not necessarily the "real" path
	const char*							getPath() const { return fPath; }

	uint32_t							getPathHash() const { return fPathHash; }

										// get the "real" path for this image (e.g. no @rpath)
	const char*							getRealPath() const;

										// get path this image is intended to be placed on disk or NULL if no preferred install location
	virtual const char*					getInstallPath() const = 0;

										// image was loaded with NSADDIMAGE_OPTION_MATCH_FILENAME_BY_INSTALLNAME and all clients are looking for install path
	bool								matchInstallPath() const;
	void								setMatchInstallPath(bool);
	
										// mark that this image's exported symbols should be ignored when linking other images (e.g. RTLD_LOCAL)
	void								setHideExports(bool hide = true);
	
										// check if this image's exported symbols should be ignored when linking other images 
	bool								hasHiddenExports() const;
	
										// checks if this image is already linked into the process
	bool								isLinked() const;
	
										// even if image is deleted, leave segments mapped in
	void								setLeaveMapped();
	
										// even if image is deleted, leave segments mapped in
	bool								leaveMapped() { return fLeaveMapped; }

										// image resides in dyld shared cache
	virtual bool						inSharedCache() const { return false; };

										// checks if the specifed address is within one of this image's segments
	virtual bool						containsAddress(const void* addr) const;

										// checks if the specifed symbol is within this image's symbol table
	virtual bool						containsSymbol(const void* addr) const = 0;

										// checks if the specifed address range overlaps any of this image's segments
	virtual bool						overlapsWithAddressRange(const void* start, const void* end) const;

										// adds to list of ranges of memory mapped in
	void								getMappedRegions(MappedRegion*& region) const;

										// st_mtime from stat() on file
	time_t								lastModified() const;

										// only valid for main executables, returns a pointer its entry point from LC_MAIN
	virtual void*						getEntryFromLC_MAIN() const = 0;
	
										// only valid for main executables, returns a pointer its main from LC_UNIXTHREAD
	virtual void*						getEntryFromLC_UNIXTHREAD() const = 0;
	
										// dyld API's require each image to have an associated mach_header
	virtual const struct mach_header*   machHeader() const = 0;
	
										// dyld API's require each image to have a slide (actual load address minus preferred load address)
	virtual uintptr_t					getSlide() const = 0;
	
										// last address mapped by image
	virtual const void*					getEnd() const = 0;
	
										// image has exports that participate in runtime coalescing
	virtual bool						hasCoalescedExports() const = 0;

										// search symbol table of definitions in this image for requested name
	virtual bool						findExportedSymbolAddress(const LinkContext& context, const char* symbolName,
																const ImageLoader* requestorImage, int requestorOrdinalOfDef,
																bool runResolver, const ImageLoader** foundIn, uintptr_t* address) const;

										// search symbol table of definitions in this image for requested name
	virtual const Symbol*				findExportedSymbol(const char* name, bool searchReExports, const char* thisPath, const ImageLoader** foundIn) const = 0;
	
										// search symbol table of definitions in this image for requested name
	virtual const Symbol*				findExportedSymbol(const char* name, bool searchReExports, const ImageLoader** foundIn) const {
											return findExportedSymbol(name, searchReExports, this->getPath(), foundIn);
										}
	
										// gets address of implementation (code) of the specified exported symbol
	virtual uintptr_t					getExportedSymbolAddress(const Symbol* sym, const LinkContext& context, 
													const ImageLoader* requestor=NULL, bool runResolver=false, const char* symbolName=NULL) const = 0;
	
										// gets attributes of the specified exported symbol
	virtual DefinitionFlags				getExportedSymbolInfo(const Symbol* sym) const = 0;
	
										// gets name of the specified exported symbol
	virtual const char*					getExportedSymbolName(const Symbol* sym) const = 0;
	
										// gets how many symbols are exported by this image
	virtual uint32_t					getExportedSymbolCount() const = 0;
			
										// gets the i'th exported symbol
	virtual const Symbol*				getIndexedExportedSymbol(uint32_t index) const = 0;
			
										// find exported symbol as if imported by this image
										// used by RTLD_NEXT
	virtual const Symbol*				findExportedSymbolInDependentImages(const char* name, const LinkContext& context, const ImageLoader** foundIn) const;
	
										// find exported symbol as if imported by this image
										// used by RTLD_SELF
	virtual const Symbol*				findExportedSymbolInImageOrDependentImages(const char* name, const LinkContext& context, const ImageLoader** foundIn) const;
	
										// gets how many symbols are imported by this image
	virtual uint32_t					getImportedSymbolCount() const = 0;
	
										// gets the i'th imported symbol
	virtual const Symbol*				getIndexedImportedSymbol(uint32_t index) const = 0;
			
										// gets attributes of the specified imported symbol
	virtual ReferenceFlags				getImportedSymbolInfo(const Symbol* sym) const = 0;
			
										// gets name of the specified imported symbol
	virtual const char*					getImportedSymbolName(const Symbol* sym) const = 0;
			
										// find the closest symbol before addr
	virtual const char*					findClosestSymbol(const void* addr, const void** closestAddr) const = 0;
	
										// for use with accelerator tables
	virtual const char*					getIndexedPath(unsigned) const { return getPath(); }
	virtual const char*					getIndexedShortName(unsigned) const { return getShortName(); }

										// checks if this image is a bundle and can be loaded but not linked
	virtual bool						isBundle() const = 0;
	
										// checks if this image is a dylib 
	virtual bool						isDylib() const = 0;
	
										// checks if this image is a main executable 
	virtual bool						isExecutable() const = 0;
	
										// checks if this image is a main executable 
	virtual bool						isPositionIndependentExecutable() const = 0;
	
										// only for main executable
	virtual bool						forceFlat() const = 0;
	
										// called at runtime when a lazily bound function is first called
	virtual uintptr_t					doBindLazySymbol(uintptr_t* lazyPointer, const LinkContext& context) = 0;
	
										// called at runtime when a fast lazily bound function is first called
	virtual uintptr_t					doBindFastLazySymbol(uint32_t lazyBindingInfoOffset, const LinkContext& context,
															void (*lock)(), void (*unlock)()) = 0;

										// calls termination routines (e.g. C++ static destructors for image)
	virtual void						doTermination(const LinkContext& context) = 0;
					
										// return if this image has initialization routines
	virtual bool						needsInitialization() = 0;
			
										// return if this image has specified section and set start and length
	virtual bool						getSectionContent(const char* segmentName, const char* sectionName, void** start, size_t* length) = 0;

										// fills in info about __eh_frame and __unwind_info sections
	virtual void						getUnwindInfo(dyld_unwind_sections* info) = 0;

										// given a pointer into an image, find which segment and section it is in
	virtual const struct macho_section* findSection(const void* imageInterior) const = 0;

										// given a pointer into an image, find which segment and section it is in
	virtual bool						findSection(const void* imageInterior, const char** segmentName, const char** sectionName, size_t* sectionOffset) = 0;
	
										// the image supports being prebound
	virtual bool						isPrebindable() const = 0;
	
										// the image is prebindable and its prebinding is valid
	virtual bool						usablePrebinding(const LinkContext& context) const = 0;
	
										// add all RPATH paths this image contains
	virtual	void						getRPaths(const LinkContext& context, std::vector<const char*>&) const = 0;
	
										// image has or uses weak definitions that need runtime coalescing
	virtual bool						participatesInCoalescing() const = 0;
		
										// if image has a UUID, copy into parameter and return true
	virtual	bool						getUUID(uuid_t) const = 0;

										// dynamic interpose values onto this image
	virtual void						dynamicInterpose(const LinkContext& context) = 0;

										// record interposing for any late binding
	void								addDynamicInterposingTuples(const struct dyld_interpose_tuple array[], size_t count);
		
	virtual const char*					libPath(unsigned int) const = 0;

										// Image has objc sections, so information objc about when it comes and goes
	virtual	bool						notifyObjC() const { return false; }

	virtual bool						overridesCachedDylib(uint32_t& num) const { return false; }
	virtual void						setOverridesCachedDylib(uint32_t num) { }


//
// A segment is a chunk of an executable file that is mapped into memory.  
//
	virtual unsigned int				segmentCount() const = 0;
	virtual const char*					segName(unsigned int) const = 0;
	virtual uintptr_t					segSize(unsigned int) const = 0;
	virtual uintptr_t					segFileSize(unsigned int) const = 0;
	virtual bool						segHasTrailingZeroFill(unsigned int) = 0;
	virtual uintptr_t					segFileOffset(unsigned int) const = 0;
	virtual bool						segReadable(unsigned int) const = 0;
	virtual bool						segWriteable(unsigned int) const = 0;
	virtual bool						segExecutable(unsigned int) const = 0;
	virtual bool						segUnaccessible(unsigned int) const = 0;
	virtual bool						segHasPreferredLoadAddress(unsigned int) const = 0;
	virtual uintptr_t					segPreferredLoadAddress(unsigned int) const = 0;
	virtual uintptr_t					segActualLoadAddress(unsigned int) const = 0;
	virtual uintptr_t					segActualEndAddress(unsigned int) const = 0;

	
										// info from LC_VERSION_MIN_MACOSX or LC_VERSION_MIN_IPHONEOS
	virtual uint32_t					sdkVersion() const = 0;
	virtual uint32_t					minOSVersion() const = 0;
	
										// if the image contains interposing functions, register them
	virtual void						registerInterposing(const LinkContext& context) = 0;

	virtual bool						usesChainedFixups() const { return false; }


										// when resolving symbols look in subImage if symbol can't be found
	void								reExport(ImageLoader* subImage);

	virtual void						recursiveBind(const LinkContext& context, bool forceLazysBound, bool neverUnload);
	virtual void						recursiveBindWithAccounting(const LinkContext& context, bool forceLazysBound, bool neverUnload);
	void								weakBind(const LinkContext& context);

	void								applyInterposing(const LinkContext& context);

	dyld_image_states					getState() { return (dyld_image_states)fState; }

	ino_t								getInode() const { return fInode; }
    dev_t                               getDevice() const { return fDevice; }

										// used to sort images bottom-up
	int									compare(const ImageLoader* right) const;
	
	void								incrementDlopenReferenceCount() { ++fDlopenReferenceCount; }

	bool								decrementDlopenReferenceCount();
	
	void								printReferenceCounts();

	uint32_t							dlopenCount() const { return fDlopenReferenceCount; }

	void								setCanUnload() { fNeverUnload = false; fLeaveMapped = false; }

	bool								neverUnload() const { return fNeverUnload; }

	void								setNeverUnload() { fNeverUnload = true; fLeaveMapped = true; }
	void								setNeverUnloadRecursive();
	
	bool								isReferencedDownward() { return fIsReferencedDownward; }

	virtual uintptr_t					resolveWeak(const LinkContext& context, const char* symbolName, bool weak_import, bool runResolver,
													const ImageLoader** foundIn) { return 0; } 

										// triggered by DYLD_PRINT_STATISTICS to write info on work done and how fast
	static void							printStatistics(unsigned int imageCount, const InitializerTimingList& timingInfo);
	static void							printStatisticsDetails(unsigned int imageCount, const InitializerTimingList& timingInfo);

										// used with DYLD_IMAGE_SUFFIX
	static void							addSuffix(const char* path, const char* suffix, char* result);
	
	static uint32_t						hash(const char*);
	
	static const uint8_t*				trieWalk(const uint8_t* start, const uint8_t* end, const char* stringToFind);

										// used instead of directly deleting image
	static void							deleteImage(ImageLoader*);

	static bool							haveInterposingTuples() { return !fgInterposingTuples.empty(); }
	static void							clearInterposingTuples() { fgInterposingTuples.clear(); }

	static void							applyInterposingToDyldCache(const LinkContext& context);

			bool						dependsOn(ImageLoader* image);
			
 			void						setPath(const char* path);
			void						setPaths(const char* path, const char* realPath);
			void						setPathUnowned(const char* path);
						
			void						clearDepth() { fDepth = 0; }
			int							getDepth() { return fDepth; }
			
			void						setBeingRemoved() { fBeingRemoved = true; }
			bool						isBeingRemoved() const { return fBeingRemoved; }
			
			void						markNotUsed() { fMarkedInUse = false; }
			void						markedUsedRecursive(const std::vector<DynamicReference>&);
			bool						isMarkedInUse() const	{ return fMarkedInUse; }
		
			void						setAddFuncNotified() { fAddFuncNotified = true; }
			bool						addFuncNotified() const { return fAddFuncNotified; }
	
	struct InterposeTuple { 
		uintptr_t		replacement; 
		ImageLoader*	neverImage;			// don't apply replacement to this image
		ImageLoader*	onlyImage;			// only apply replacement to this image
		uintptr_t		replacee; 
	};

	static uintptr_t read_uleb128(const uint8_t*& p, const uint8_t* end);
	static intptr_t read_sleb128(const uint8_t*& p, const uint8_t* end);

	void			vmAccountingSetSuspended(const LinkContext& context, bool suspend);

protected:			
	// abstract base class so all constructors protected
					ImageLoader(const char* path, unsigned int libCount); 
					ImageLoader(const ImageLoader&);
	void			operator=(const ImageLoader&);
	void			operator delete(void* image) throw() { ::free(image); }
	

	struct LibraryInfo {
		uint32_t		checksum;
		uint32_t		minVersion;
		uint32_t		maxVersion;
	};

	struct DependentLibrary {
		ImageLoader*	image;
		uint32_t		required : 1,
						checksumMatches : 1,
						isReExported : 1,
						isSubFramework : 1;
	};
	
	struct DependentLibraryInfo {
		const char*			name;
		LibraryInfo			info;
		bool				required;
		bool				reExported;
		bool				upward;
	};


	typedef void (*Initializer)(int argc, const char* argv[], const char* envp[], const char* apple[], const ProgramVars* vars);
	typedef void (*Terminator)(void);
	


	unsigned int			libraryCount() const { return fLibraryCount; }
	virtual ImageLoader*	libImage(unsigned int) const = 0;
	virtual bool			libReExported(unsigned int) const = 0;
	virtual bool			libIsUpward(unsigned int) const = 0;
	virtual void			setLibImage(unsigned int, ImageLoader*, bool, bool) = 0;

						// To link() an image, its dependent libraries are loaded, it is rebased, bound, and initialized.
						// These methods do the above, exactly once, and it the right order
	virtual void		recursiveLoadLibraries(const LinkContext& context, bool preflightOnly, const RPathChain& loaderRPaths, const char* loadPath);
	virtual unsigned 	recursiveUpdateDepth(unsigned int maxDepth);
	virtual void		recursiveRebase(const LinkContext& context);
	virtual void		recursiveApplyInterposing(const LinkContext& context);
	virtual void		recursiveGetDOFSections(const LinkContext& context, std::vector<DOFInfo>& dofs);
	virtual void		recursiveInitialization(const LinkContext& context, mach_port_t this_thread, const char* pathToInitialize,
												ImageLoader::InitializerTimingList&, ImageLoader::UninitedUpwards&);

								// fill in information about dependent libraries (array length is fLibraryCount)
	virtual void				doGetDependentLibraries(DependentLibraryInfo libs[]) = 0;
	
								// called on images that are libraries, returns info about itself
	virtual LibraryInfo			doGetLibraryInfo(const LibraryInfo& requestorInfo) = 0;
	
								// do any fix ups in this image that depend only on the load address of the image
	virtual void				doRebase(const LinkContext& context) = 0;
	
								// do any symbolic fix ups in this image
	virtual void				doBind(const LinkContext& context, bool forceLazysBound) = 0;
	
								// called later via API to force all lazy pointer to be bound
	virtual void				doBindJustLazies(const LinkContext& context) = 0;
	
								// if image has any dtrace DOF sections, append them to list to be registered
	virtual void				doGetDOFSections(const LinkContext& context, std::vector<DOFInfo>& dofs) = 0;
	
								// do interpose
	virtual void				doInterpose(const LinkContext& context) = 0;

								// run any initialization routines in this image
	virtual bool				doInitialization(const LinkContext& context) = 0;
	
								// return if this image has termination routines
	virtual bool				needsTermination() = 0;
	
								// support for runtimes in which segments don't have to maintain their relative positions
	virtual bool				segmentsMustSlideTogether() const = 0;	
	
								// built with PIC code and can load at any address
	virtual bool				segmentsCanSlide() const = 0;		

								// set how much all segments slide
	virtual void				setSlide(intptr_t slide) = 0;		

								// returns if all dependent libraries checksum's were as expected and none slide
			bool				allDependentLibrariesAsWhenPreBound() const;

								// in mach-o a child tells it parent to re-export, instead of the other way around...
	virtual	bool				isSubframeworkOf(const LinkContext& context, const ImageLoader* image) const = 0;

								// in mach-o a parent library knows name of sub libraries it re-exports..
	virtual	bool				hasSubLibrary(const LinkContext& context, const ImageLoader* child) const  = 0;

	virtual bool				weakSymbolsBound(unsigned index) { return fWeakSymbolsBound; }
	virtual void				setWeakSymbolsBound(unsigned index) { fWeakSymbolsBound = true; }

								// set fState to dyld_image_state_memory_mapped
	void						setMapped(const LinkContext& context);
		
	void						setFileInfo(dev_t device, ino_t inode, time_t modDate);

	void						setDepth(uint16_t depth) { fDepth = depth; }

	static uintptr_t			interposedAddress(const LinkContext& context, uintptr_t address, const ImageLoader* notInImage, const ImageLoader* onlyInImage=NULL);
	
	static uintptr_t			fgNextPIEDylibAddress;
	static uint32_t				fgImagesWithUsedPrebinding;
	static uint32_t				fgImagesUsedFromSharedCache;
	static uint32_t				fgImagesHasWeakDefinitions;
	static uint32_t				fgImagesRequiringCoalescing;
	static uint32_t				fgTotalRebaseFixups;
	static uint32_t				fgTotalBindFixups;
	static uint32_t				fgTotalBindSymbolsResolved;
	static uint32_t				fgTotalBindImageSearches;
	static uint32_t				fgTotalLazyBindFixups;
	static uint32_t				fgTotalPossibleLazyBindFixups;
	static uint32_t				fgTotalSegmentsMapped;
	static uint32_t				fgSymbolTrieSearchs;
	static uint64_t				fgTotalBytesMapped;
	static uint64_t				fgTotalBytesPreFetched;
	static uint64_t				fgTotalLoadLibrariesTime;
public:
	static uint64_t				fgTotalObjCSetupTime;
	static uint64_t				fgTotalDebuggerPausedTime;
	static uint64_t				fgTotalRebindCacheTime;
	static uint64_t				fgTotalRebaseTime;
	static uint64_t				fgTotalBindTime;
	static uint64_t				fgTotalWeakBindTime;
	static uint64_t				fgTotalDOF;
	static uint64_t				fgTotalInitTime;

protected:
	static std::vector<InterposeTuple>	fgInterposingTuples;
	
	const char*					fPath;
	const char*					fRealPath;
	dev_t						fDevice;
	ino_t						fInode;
	time_t						fLastModified;
	uint32_t					fPathHash;
	uint32_t					fDlopenReferenceCount;	// count of how many dlopens have been done on this image

	struct recursive_lock {
						recursive_lock(mach_port_t t) : thread(t), count(0) {}
		mach_port_t		thread;
		int				count;
	};
	void						recursiveSpinLock(recursive_lock&);
	void						recursiveSpinUnLock();

private:
	const ImageLoader::Symbol*	findExportedSymbolInDependentImagesExcept(const char* name, const ImageLoader** dsiStart, 
										const ImageLoader**& dsiCur, const ImageLoader** dsiEnd, const ImageLoader** foundIn) const;

	void						processInitializers(const LinkContext& context, mach_port_t this_thread,
													InitializerTimingList& timingInfo, ImageLoader::UninitedUpwards& ups);


	recursive_lock*				fInitializerRecursiveLock;
	uint16_t					fDepth;
	uint16_t					fLoadOrder;
	uint32_t					fState : 8,
								fLibraryCount : 10,
								fAllLibraryChecksumsAndLoadAddressesMatch : 1,
								fLeaveMapped : 1,		// when unloaded, leave image mapped in cause some other code may have pointers into it
								fNeverUnload : 1,		// image was statically loaded by main executable
								fHideSymbols : 1,		// ignore this image's exported symbols when linking other images
								fMatchByInstallName : 1,// look at image's install-path not its load path
								fInterposed : 1,
								fRegisteredDOF : 1,
								fAllLazyPointersBound : 1,
								fMarkedInUse : 1,
								fBeingRemoved : 1,
								fAddFuncNotified : 1,
								fPathOwnedByImage : 1,
								fIsReferencedDownward : 1,
								fWeakSymbolsBound : 1;

	static uint16_t				fgLoadOrdinal;
};


VECTOR_NEVER_DESTRUCTED_EXTERN(ImageLoader::InterposeTuple);


#endif

如此之长的头文件,我们去掉方法和静态成员,整理以后,只有以下成员

    const char*					fPath;
	const char*					fRealPath;
	dev_t						fDevice;
	ino_t						fInode;
	time_t						fLastModified;
	uint32_t					fPathHash;
	uint32_t					fDlopenReferenceCount;	// count of how many dlopens have been done on this image
	recursive_lock*				fInitializerRecursiveLock;
	uint16_t					fDepth;
	uint16_t					fLoadOrder;
	uint32_t					fState : 8,
								fLibraryCount : 10,
								fAllLibraryChecksumsAndLoadAddressesMatch : 1,
								fLeaveMapped : 1,		// when unloaded, leave image mapped in cause some other code may have pointers into it
								fNeverUnload : 1,		// image was statically loaded by main executable
								fHideSymbols : 1,		// ignore this image's exported symbols when linking other images
								fMatchByInstallName : 1,// look at image's install-path not its load path
								fInterposed : 1,
								fRegisteredDOF : 1,
								fAllLazyPointersBound : 1,
								fMarkedInUse : 1,
								fBeingRemoved : 1,
								fAddFuncNotified : 1,
								fPathOwnedByImage : 1,
								fIsReferencedDownward : 1,
								fWeakSymbolsBound : 1;

根据我们上次dump的结果来看,然后是不够的,至少这里面根本没有我们需要的load address
那么,这是什么原因呢?
由于c++继承的特性,在c++里,类指针的第一个成员是虚表地址,往后依次是父类的成员(包括私有成员)和自己类成员。显然这里真正的类型应是它的某个子类。
结合我们上面的dump结果和apple的源码,一番尝试后,我们可以总结出它的结构如下

typedef struct {
    void*                       vTable;
    const char*					fPath;
	const char*					fRealPath;
	dev_t						fDevice;
	ino_t						fInode;
	time_t						fLastModified;
	uint32_t					fPathHash;
	uint32_t					fDlopenReferenceCount;	// count of how many dlopens have been done on this image
	void*				fInitializerRecursiveLock;
	uint16_t					fDepth;
	uint16_t					fLoadOrder;
	uint32_t					fState : 8,
								fLibraryCount : 10,
								fAllLibraryChecksumsAndLoadAddressesMatch : 1,
								fLeaveMapped : 1,		// when unloaded, leave image mapped in cause some other code may have pointers into it
								fNeverUnload : 1,		// image was statically loaded by main executable
								fHideSymbols : 1,		// ignore this image's exported symbols when linking other images
								fMatchByInstallName : 1,// look at image's install-path not its load path
								fInterposed : 1,
								fRegisteredDOF : 1,
								fAllLazyPointersBound : 1,
								fMarkedInUse : 1,
								fBeingRemoved : 1,
								fAddFuncNotified : 1,
								fPathOwnedByImage : 1,
								fIsReferencedDownward : 1,
								fWeakSymbolsBound : 1;
	uint64_t								fCoveredCodeLength;
	const uint8_t*							fMachOData;
	const uint8_t*							fLinkEditBase; // add any internal "offset" to this to get mapped address
	uintptr_t								fSlide;
	uint32_t								fEHFrameSectionOffset;
	uint32_t								fUnwindInfoSectionOffset;
	uint32_t								fDylibIDOffset;
	uint32_t								fSegmentsCount : 8,
											fIsSplitSeg : 1,
											fInSharedCache : 1,
#if TEXT_RELOC_SUPPORT
											fTextSegmentRebases : 1, 
											fTextSegmentBinds : 1, 
#endif
#if __i386__
											fReadOnlyImportSegment : 1,
#endif
											fHasSubLibraries : 1,
											fHasSubUmbrella : 1,
											fInUmbrella : 1,
											fHasDOFSections : 1,
											fHasDashInit : 1,
											fHasInitializers : 1,
											fHasTerminators : 1,
											fNotifyObjC : 1,
											fRetainForObjC : 1,
											fRegisteredAsRequiresCoalescing : 1, 	// <rdar://problem/7886402> Loading MH_DYLIB_STUB causing coalescable miscount
											fOverrideOfCacheImageNum : 12;

											
	static uint32_t					fgSymbolTableBinarySearchs;
} MachoImage;

由此,我们可以顺利的通过以下代码来获取到他基地址。

	MachoImage* image=(MachoImage*)dlopen("xx.dylib",RTLD_LAZY);
	NSLog(@"[FLWB]:xxHandle:%p",image->fLinkEditBase);

当然fMachOData也是可以的,但考虑到后面的注释

add any internal "offset" to this to get mapped address

我们还是优先使用fLinkEditBase
此外,我们也可以通过此头文件来获取其它未公开的成员。
到此为止,我们不但达到了我们获取基地址的需求,还能够额外的获取甚至关键的信息。

免责声明

本文所公开的技术细节仅做技术交流,任何权利责任人需自行承担其用途产生的后果。另外,转载本文需注明来源及作者信息。

鸣谢

排名不分先后

资料文献

  1. https://stackoverflow.com/questions/19451791/get-loaded-address-of-a-elf-binary-dlopen-is-not-working-as-expected
  2. 从dlsym()源码看android 动态链接过程
  3. dyld source

某些系统下发现获取出来的不对,fCoveredCodeLength 是 fMachOData的地址,这是为什么