本文转自http://www.9hao.info/pages/2014/08/android-soke-ru-kou-q
前言
开年来开始接触一些加固样本，基本都对了so进行了处理，拖入ida一看，要么没有 JNI_OnLoad ,要么 JNI_OnLoad 汇编代码羞涩难懂，让人无法下手。 JNI_OnLoad 是真正入口么？
先看看几个文档
1 摘自属性服务一节（《深入理解Android卷1》）
利用gcc的constructor属性，这个属性指明了一个__libc_prenit函数(这个函数内部就将完成共享内存到本地进程的映射工作)。用法：当bionic libc库被加载时，将自动调用__libc_prenit函数。这样在bionic libc动态库被装载时，系统属性缓冲区地址就被确定了，后续的API调用就能找对位置了。 /* We flag the __libc_preinit function as a constructor to ensure * that its address is listed in libc.so's .init_array section. * This ensures that the function is called by the dynamic linker * as soon as the shared library is loaded. */ //constructor属性指示加载器加载该库之后，首先调用__libc_prenit函数。这一点和windows上的动态库的DllMain函数类似
void __attribute__((constructor)) __libc_prenit(void);
从英文说明里面提到到.init_array section，我们可以搜索一下这一节的说明
2 .init_array section
.init_array contains pointers to blocks of code that need to be executed when an application is being initialized (before main() is called). It is used for a number of things, but the primary use is in C++ for running static constructors; a secondary use that is sometimes used is to initialize IO systems in the C library.
If you are not using C++ you may (depending on your C library) be able to live without it entirely; but you’d need to hack your startup code to deal with this.
.init_array probably ends up in ram because its marked read/write — that happens because in a dynamic linking environment the dynamic linker has to fix up all the pointers it contains before it can be used. In a static environment you might be able to get away with forcing it into a read-only section.
来源： <http://blog.sina.com.cn/s/blog_a9303fd901019kvq.html>
3 摘自dlopen小结(《程序员的自我修养》)
动态连接器在加载模块时，会执行".init"段的代码,用以完成模块的初始化工作，dlopen的加载过程基本跟动态连接器一致，在完成装载、映射和重定向以后，就会执行".init"段的代码然后返回
看完这个3段资料，我们可以知道在系统加载so，在完成装载、映射和重定向以后，就首先执行.init和.init_array段的代码.
探本溯源，在源码中追踪
我们先从System.loadLibrary ->Runtime.loadLibrary
public void loadLibrary(String libName) {
loadLibrary(libName, VMStack.getCallingClassLoader());
}
/*
* Loads and links a library without security checks.
*/
void loadLibrary(String libraryName, ClassLoader loader) {
代码略...
String error = nativeLoad(filename, loader);
代码略...
}
-> nativeLoad
static void Dalvik_java_lang_Runtime_nativeLoad(const u4* args,
JValue* pResult)
{
代码略...
StringObject* fileNameObj = (StringObject*) args[];
success = dvmLoadNativeCode(fileName, classLoader, &reason);
代码略...
}
来源： <http://androidxref.com/4.1.2/xref/dalvik/vm/native/java_lang_Runtime.cpp#72>
bool dvmLoadNativeCode(const char* pathName, Object* classLoader,
char** detail)
{
代码略...
handle = dlopen(pathName, RTLD_LAZY); 代码略...
vonLoad = dlsym(handle, "JNI_OnLoad");
if (vonLoad == NULL) {
ALOGD("No JNI_OnLoad found in %s %p, skipping init",
pathName, classLoader);
} else {
/*
* Call JNI_OnLoad. We have to override the current class
* loader, which will always be "null" since the stuff at the
* top of the stack is around Runtime.loadLibrary(). (See
* the comments in the JNI FindClass function.)
*/
OnLoadFunc func = (OnLoadFunc)vonLoad;
Object* prevOverride = self->classLoaderOverride; self->classLoaderOverride = classLoader;
oldStatus = dvmChangeStatus(self, THREAD_NATIVE);
if (gDvm.verboseJni) {
ALOGI("[Calling JNI_OnLoad for \"%s\"]", pathName);
}
version = (*func)(gDvmJni.jniVm, NULL);
dvmChangeStatus(self, oldStatus);
self->classLoaderOverride = prevOverride;
}
代码略...
}
来源： <http://androidxref.com/4.1.2/xref/dalvik/vm/Native.cpp#318>
通过dvmLoadNativeCode函数我们知道系统用dlopen加载so完成后，会查看有没有JNI_OnLoad函数，有的话就调用.
我们再到dlopen函数探个究竟:
void *dlopen(const char *filename, int flag)
{
soinfo *ret; pthread_mutex_lock(&dl_lock);
/*find_library 会判断so是否已经加载，如果没有加载，对so进行加载，完成一些初始化工作,有兴趣的读者可自行分析 */
ret = find_library(filename);
if (unlikely(ret == NULL)) {
set_dlerror(DL_ERR_CANNOT_LOAD_LIBRARY);
} else {
call_constructors_recursive(ret);
ret->refcount++;
}
pthread_mutex_unlock(&dl_lock);
return ret;
}
void call_constructors_recursive(soinfo *si)
{
if (si->constructors_called)
return; // Set this before actually calling the constructors, otherwise it doesn't
// protect against recursive constructor calls. One simple example of
// constructor recursion is the libc debug malloc, which is implemented in
// libc_malloc_debug_leak.so:
// 1. The program depends on libc, so libc's constructor is called here.
// 2. The libc constructor calls dlopen() to load libc_malloc_debug_leak.so.
// 3. dlopen() calls call_constructors_recursive() with the newly created
// soinfo for libc_malloc_debug_leak.so.
// 4. The debug so depends on libc, so call_constructors_recursive() is
// called again with the libc soinfo. If it doesn't trigger the early-
// out above, the libc constructor will be called again (recursively!).
si->constructors_called = ; if (si->flags & FLAG_EXE) {
TRACE("[ %5d Calling preinit_array @ 0x%08x [%d] for '%s' ]\n",
pid, (unsigned)si->preinit_array, si->preinit_array_count,
si->name);
call_array(si->preinit_array, si->preinit_array_count, );
TRACE("[ %5d Done calling preinit_array for '%s' ]\n", pid, si->name);
} else {
if (si->preinit_array) {
DL_ERR("%5d Shared library '%s' has a preinit_array table @ 0x%08x."
" This is INVALID.", pid, si->name,
(unsigned)si->preinit_array);
}
} 代码略... if (si->init_func) {
TRACE("[ %5d Calling init_func @ 0x%08x for '%s' ]\n", pid,
(unsigned)si->init_func, si->name);
si->init_func();
TRACE("[ %5d Done calling init_func for '%s' ]\n", pid, si->name);
}
if (si->init_array) {
TRACE("[ %5d Calling init_array @ 0x%08x [%d] for '%s' ]\n", pid,
(unsigned)si->init_array, si->init_array_count, si->name);
//遍历函数数组并执行
call_array(si->init_array, si->init_array_count, );
TRACE("[ %5d Done calling init_array for '%s' ]\n", pid, si->name);
//ps:看到这么多TRACE这么多调试信息，我们把调试开关打开，是不是能拿到诸多信息？
}
}
来源： <http://androidxref.com/4.1.2/xref/bionic/linker/linker.c#1519>
通过可以函数我们知道si->init_func和si->init_array存在的时候，会执行指向的函数
(不知道大家注意到么si->flags & FLAG_EXE时，还有si->preinit_array？ 以后会不会有这方面的东西？)
再找下 si->init_func和si->init_array 的赋值
case DT_INIT:
si->init_func = (void (*)(void))(si->base + *d);
DEBUG("%5d %s constructors (init func) found at %p\n",
pid, si->name, si->init_func); case DT_INIT_ARRAY:
si->init_array = (unsigned *)(si->base + *d);
DEBUG("%5d %s constructors (init_array) found at %p\n",
pid, si->name, si->init_array);
break;
DEBUG里面说明了constructors (init func)和constructors (init_array)。
我们再看看一份文档Android Dynamic Linker Design Notes
DT_INIT
Points to the address of an initialization function
that must be called when the file is loaded.
DT_INIT_ARRAY
Points to an array of function addresses that must be
called, in-order, to perform initialization. Some of
the entries in the array can be or -, and should
be ignored. Note: this is generally stored in a .init_array section
通过层层分析，我们很清楚知道了系统加载so，在完成装载、映射和重定向以后，就首先执行.init和.init_array段的代码.
前面有一篇文章我已经对so加壳进行简单说明
把源码的dlopen复制出来修改，在把自己so加载起来的时候 ,把自己内存里面某部分地址解密后,用自己的dlopen打开返回一个soinfo结构体 然后把当前soinfo结构体替换原来的soinfo结构体
小结
系统加载so，在完成装载、映射和重定向以后，就首先执行.init和.init_array段的代码，之后如果存在JNI_OnLoad 就调用该函数.我们要对一个so进行分析，需要先看看有没有.init_array section和.init section，so加壳一般会在初始化函数进行脱壳操作。
如何在.init和.init_array段添加我们的函数
[] 共享构造函数，在函数声明时加上"__attribute__((constructor))"属性
void __attribute__((constructor)) init_function(void);
对应有共享虚构函数，在程序exit()或者dlclose()返回前执行
void __attribute__((destructor)) fini_function(void); []c++ 静态构造函数
在.init和.init_array下断点
init_array 用ida可以看到， 可以对里面的函数数组下断点
init ida有时没识别出来,可用readelf查看入口点 xxx@xx:~$ readelf -a '/home/xxx/桌面/libsecexe.so' 0x00000010 (SYMBOLIC) 0x0
0x0000000c (INIT) 0x11401
0x00000019 (INIT_ARRAY) 0x28ca4
0x0000001b (INIT_ARRAYSZ) (bytes)
0x0000001a (FINI_ARRAY) 0x28cac
0x0000001c (FINI_ARRAYSZ) (bytes)
0x00000004 (HASH) 0xf4 我们看到 INIT 入口为 0x11401
(ps:有时你在0x11401是数据，你需要make code，由于对齐关系，要从0x11401+1开始）
样本：梆梆 爱加密
参考：
[1] 《深入理解Android卷1》
[2] 《程序员的自我修养-链接、装载与库》
[3] android linker 浅析
[4] Android Dynamic Linker Design Notes