0x03 DexHunter代码分析
DexHunter 实现中，只需要修改一处文件：dalvik\vm\native\dalvik_system_DexFile.cpp
下面是BeyondCompare比对：
首先看一下DexHunter的设计原理：
APP 启动时，通过freature string定位dex在内存中位置，并读取classdef块之前的内存为part1，读取classdef之后的内存为data。遍历class_def_item结构，生成文件classdef，并通过code_item_off判断具体的类方法是否在dex范围内，若不在，则写extra文件。
描述几个问题：
-
从哪里dump出dex文件
dex文件打开时
类加载时
类初始化时
类方法调用时
DexHunter中，我们关注，ClassLoader.loadClass->Dalvik_dalvik_system_DexFile_defineClassNative这个函数，它实现了类的加载，实现过程如下：
选择脱壳的时机应是在APP的第一个类加载的时候，为什么呢？
-
类加载之前，类的内容是在内存当中的
-
当类初始化时，该内存的内容可能会被动态修改
-
在一个类方法被调用前，code_item或指令肯定是可用的
那如何做呢？
我们要主动加载并初始化所有的类；
因此，我们代码的注入点，应该是Dalvik_dalvik_system_DexFile_defineClassNative()函数的clazz = dvmDefineClass(pDvmDex, descriptor, loader);语句之前；即在APP加载第一个类之前完成；通过dvmDefineClass主动遍历class_def_item加载每个类，并调用dvmIsClassInitialized和dvmInitClass函数初始化之。
初始化完成之后，内存中的就是将执行的代码，像梆梆加固针对每个方法进行的加密，会在运行时解密、运行完成后清理内存并再次加密，通过这种方法就可以过掉；因为我们模拟了这样一次调用过程；
下面是我加入注释的代码：
//------------------------added begin----------------------//
#include <asm/siginfo.h>
#include "libdex/DexClass.h"
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/mman.h>
static char dexname[100]={0}; //feature string
static char dumppath[100]={0}; //dump的文件路径
static bool readable=true;
static pthread_mutex_t read_mutex;
static bool flag=true;
static pthread_mutex_t mutex;
static bool timer_flag=true;
static timer_t timerId;
struct arg{
DvmDex* pDvmDex;
Object * loader;
}param;
void timer_thread(sigval_t)
{
timer_flag=false;
timer_delete(timerId);
ALOGI("GOT IT time up");
}
void* ReadThread(void *arg){
FILE *fp = NULL;
while (dexname[0]==0||dumppath[0]==0) {
fp=fopen("/data/dexname", "r");
if (fp==NULL) {
sleep(1);
continue;
}
fgets(dexname,99,fp); //读feature string
dexname[strlen(dexname)-1]=0;
fgets(dumppath,99,fp);
dumppath[strlen(dumppath)-1]=0; //取dump路径
fclose(fp);
fp=NULL;
}
struct sigevent sev;
sev.sigev_notify=SIGEV_THREAD;
sev.sigev_value.sival_ptr=&timerId;
sev.sigev_notify_function=timer_thread;
sev.sigev_notify_attributes = NULL;
timer_create(CLOCK_REALTIME,&sev,&timerId);
struct itimerspec ts;
ts.it_value.tv_sec=5;
ts.it_value.tv_nsec=0;
ts.it_interval.tv_sec=0;
ts.it_interval.tv_nsec=0;
timer_settime(timerId,0,&ts,NULL);
return NULL;
}
/*
这里是class_data_item的前4项，称为ClassDataHeader
Dex File->class_defs->class_def_item(class_data_offset)->class_data_item->ClassDataHeader
*/
void ReadClassDataHeader(const uint8_t** pData, DexClassDataHeader *pHeader)
{
pHeader->staticFieldsSize = readUnsignedLeb128(pData);
pHeader->instanceFieldsSize = readUnsignedLeb128(pData);
pHeader->directMethodsSize = readUnsignedLeb128(pData);
pHeader->virtualMethodsSize = readUnsignedLeb128(pData);
}
/*
下面两个函数，分别读class_data_item Header下的内容，分Field和Method
*/
void ReadClassDataField(const uint8_t** pData, DexField* pField)
{
pField->fieldIdx = readUnsignedLeb128(pData);
pField->accessFlags = readUnsignedLeb128(pData);
}
void ReadClassDataMethod(const uint8_t** pData, DexMethod* pMethod)
{
pMethod->methodIdx = readUnsignedLeb128(pData);
pMethod->accessFlags = readUnsignedLeb128(pData);
pMethod->codeOff = readUnsignedLeb128(pData);
}
/*
解析class_data_item结构，使用到上面3个函数，分别解析，Header、Field和Method部分
*/
DexClassData* ReadClassData(const uint8_t** pData)
{
DexClassDataHeader header;
if (*pData == NULL) {
return NULL;
}
//读取 class_data_item的Header
ReadClassDataHeader(pData, &header);
size_t resultSize = sizeof(DexClassData) + (header.staticFieldsSize * sizeof(DexField)) + (header.instanceFieldsSize * sizeof(DexField)) + (header.directMethodsSize * sizeof(DexMethod)) + (header.virtualMethodsSize * sizeof(DexMethod));
DexClassData* result = (DexClassData*) malloc(resultSize); //result指向class_data_item并返回
if (result == NULL) {
return NULL;
}
uint8_t* ptr = ((uint8_t*) result) + sizeof(DexClassData); //指向class_data_item的staic_fields偏移
result->header = header;
//以下依次读class_data_item的staticFields，instanceFields，directMethods和virtualMethods域大小————————begain
if (header.staticFieldsSize != 0) {
result->staticFields = (DexField*) ptr;
ptr += header.staticFieldsSize * sizeof(DexField);
} else {
result->staticFields = NULL;
}
if (header.instanceFieldsSize != 0) {
result->instanceFields = (DexField*) ptr;
ptr += header.instanceFieldsSize * sizeof(DexField);
} else {
result->instanceFields = NULL;
}
if (header.directMethodsSize != 0) {
result->directMethods = (DexMethod*) ptr;
ptr += header.directMethodsSize * sizeof(DexMethod);
} else {
result->directMethods = NULL;
}
if (header.virtualMethodsSize != 0) {
result->virtualMethods = (DexMethod*) ptr;
} else {
result->virtualMethods = NULL;
}
//以下依次读class_data_item的staticFields，instanceFields，directMethods和virtualMethods域大小————————end
//以下依次读staticFields，instanceFields，directMethods，virtualMethods域内容————————begain
for (uint32_t i = 0; i < header.staticFieldsSize; i++) {
ReadClassDataField(pData, &result->staticFields[i]);
}
for (uint32_t i = 0; i < header.instanceFieldsSize; i++) {
ReadClassDataField(pData, &result->instanceFields[i]);
}
for (uint32_t i = 0; i < header.directMethodsSize; i++) {
ReadClassDataMethod(pData, &result->directMethods[i]);
}
for (uint32_t i = 0; i < header.virtualMethodsSize; i++) {
ReadClassDataMethod(pData, &result->virtualMethods[i]);
}
//以下依次读staticFields，instanceFields，directMethods，virtualMethods域内容————————end
return result;
}
/*
class_data_item中的一些域是用LEB128算法编码的
*/
void writeLeb128(uint8_t ** ptr, uint32_t data)
{
while (true) {
uint8_t out = data & 0x7f;
if (out != data) {
*(*ptr)++ = out | 0x80;
data >>= 7;
} else {
*(*ptr)++ = out;
break;
}
}
}
/*
此函数读取class_data_item，并将内容用writeLeb128转码后返回
*/
uint8_t* EncodeClassData(DexClassData *pData, int& len)
{
len=0;
len+=unsignedLeb128Size(pData->header.staticFieldsSize);
len+=unsignedLeb128Size(pData->header.instanceFieldsSize);
len+=unsignedLeb128Size(pData->header.directMethodsSize);
len+=unsignedLeb128Size(pData->header.virtualMethodsSize);
if (pData->staticFields) {
for (uint32_t i = 0; i < pData->header.staticFieldsSize; i++) {
len+=unsignedLeb128Size(pData->staticFields[i].fieldIdx);
len+=unsignedLeb128Size(pData->staticFields[i].accessFlags);
}
}
if (pData->instanceFields) {
for (uint32_t i = 0; i < pData->header.instanceFieldsSize; i++) {
len+=unsignedLeb128Size(pData->instanceFields[i].fieldIdx);
len+=unsignedLeb128Size(pData->instanceFields[i].accessFlags);
}
}
if (pData->directMethods) {
for (uint32_t i=0; i<pData->header.directMethodsSize; i++) {
len+=unsignedLeb128Size(pData->directMethods[i].methodIdx);
len+=unsignedLeb128Size(pData->directMethods[i].accessFlags);
len+=unsignedLeb128Size(pData->directMethods[i].codeOff);
}
}
if (pData->virtualMethods) {
for (uint32_t i=0; i<pData->header.virtualMethodsSize; i++) {
len+=unsignedLeb128Size(pData->virtualMethods[i].methodIdx);
len+=unsignedLeb128Size(pData->virtualMethods[i].accessFlags);
len+=unsignedLeb128Size(pData->virtualMethods[i].codeOff);
}
}
uint8_t * store = (uint8_t *) malloc(len);
if (!store) {
return NULL;
}
uint8_t * result=store;
writeLeb128(&store,pData->header.staticFieldsSize);
writeLeb128(&store,pData->header.instanceFieldsSize);
writeLeb128(&store,pData->header.directMethodsSize);
writeLeb128(&store,pData->header.virtualMethodsSize);
if (pData->staticFields) {
for (uint32_t i = 0; i < pData->header.staticFieldsSize; i++) {
writeLeb128(&store,pData->staticFields[i].fieldIdx);
writeLeb128(&store,pData->staticFields[i].accessFlags);
}
}
if (pData->instanceFields) {
for (uint32_t i = 0; i < pData->header.instanceFieldsSize; i++) {
writeLeb128(&store,pData->instanceFields[i].fieldIdx);
writeLeb128(&store,pData->instanceFields[i].accessFlags);
}
}
if (pData->directMethods) {
for (uint32_t i=0; i<pData->header.directMethodsSize; i++) {
writeLeb128(&store,pData->directMethods[i].methodIdx);
writeLeb128(&store,pData->directMethods[i].accessFlags);
writeLeb128(&store,pData->directMethods[i].codeOff);
}
}
if (pData->virtualMethods) {
for (uint32_t i=0; i<pData->header.virtualMethodsSize; i++) {
writeLeb128(&store,pData->virtualMethods[i].methodIdx);
writeLeb128(&store,pData->virtualMethods[i].accessFlags);
writeLeb128(&store,pData->virtualMethods[i].codeOff);
}
}
free(pData);
return result;
}
uint8_t* codeitem_end(const u1** pData)
{
uint32_t num_of_list = readUnsignedLeb128(pData);
for (;num_of_list>0;num_of_list--) {
int32_t num_of_handlers=readSignedLeb128(pData);
int num=num_of_handlers;
if (num_of_handlers<=0) {
num=-num_of_handlers;
}
for (; num > 0; num--) {
readUnsignedLeb128(pData);
readUnsignedLeb128(pData);
}
if (num_of_handlers<=0) {
readUnsignedLeb128(pData);
}
}
return (uint8_t*)(*pData);
}
代码未完，下一篇继续；