armeabi-v7a 在手机上运行
x86 在夜神模拟器上运行
编译elf需要添加pie，编译so不需要
LOCAL_CFLAGS += -fPIE
LOCAL_LDFLAGS += -fPIE -pie
在夜神模拟上面通过so方式ptrace(PTRACE_ATTACH会失败，没找到原因（如果将程序放在virtualxposed中运行，PTRACE_ATTACH是成功的，应该还是权限的问题）
在夜神模拟上面通过eof方式ptrace(PTRACE_ATTACH成功
ninja是一个小型的关注速度的构建系统，不需要关心ninja的脚本，知道怎么配置CMake就可以了
/data/local/tmp目录下面的elf可以运行（chmod 777），程序入口函数是main
安卓上层调用so里面的本地方法
socket都是用c++实现
char* io_buffer = "W A8F79F30 1222";
int data;
char *ptr;
char *ptr1;
char* cursor = io_buffer;
cursor += 1;
void * addr = (void *)strtoll(cursor, &ptr, 16);
printf("%0x", addr);//打印 A8F79F30
//printf("%s", addr);会崩溃
int value = strtol(ptr, &ptr, 10);https://github.com/android/ndk-samples
ndk-samples\other-builds\ndkbuild\two-libs
这个例子是ndk还是使用android.mk来编译，但是c++代码是无法调试的，也看不了c++代码
Android Studio只能写CMakeList.txt，才能调试c++代码？？？
externalNativeBuild {
ndkBuild {
path 'src/main/cpp/Android.mk'
}
}
使用夜神模拟器只能跑程序，无法调试程序???
static jint init_native(JNIEnv *env, jobject thiz)
{
... ...
//return 0;
}
真机可以进入java和c++断点 安卓自带的模拟器可以进入java和c++断点 夜神模拟器可以进入java断点，无法进入c++断点，没有找到原因 (夜神模拟器应该需要使用x86的so，使用了还是无法进入c++断点)
ndk {
abiFilters 'armeabi-v7a', 'arm64-v8a', 'x86', 'x86_64'
}
使用elf，/data/local/tmp目下调用ptrace：
root的手机 uid = 2000，用adb shell ps -ef查看，uid是shell
夜神模拟 uid = 0
uid为0时ptrace(PTRACE_DETACH,才会成功
在Virtual Xposed中使用so调用ptrace，是可以成功的，但是无法看到日志。
text、rodata、data、bss、stack、heap
https://www.cnblogs.com/DataArt/p/9879192.html
.text (r-x), .rodata (r--), .data (rw-) and .bss (rw-)
/proc/${pid}/maps，这个文件有内存所有的详细信息
查找的内存段：
heap, stack, .data, .bss
/*
* get the load address for regions of the same ELF file
*
* When the ELF loader loads an executable or a library into
* memory, there is one region per ELF segment created:
* .text (r-x), .rodata (r--), .data (rw-) and .bss (rw-). The
* 'x' permission of .text is used to detect the load address
* (region start) and the end of the ELF file in memory. All
* these regions have the same filename. The only exception
* is the .bss region. Its filename is empty and it is
* consecutive with the .data region. But the regions .bss and
* .rodata may not be present with some ELF files. This is why
* we can't rely on other regions to be consecutive in memory.
* There should never be more than these four regions.
* The data regions use their variables relative to the load
* address. So determining it makes sense as we can get the
* variable address used within the ELF file with it.
* But for the executable there is the special case that there
* is a gap between .text and .rodata. Other regions might be
* loaded via mmap() to it. So we have to count the number of
* regions belonging to the exe separately to handle that.
* References:
* http://en.wikipedia.org/wiki/Executable_and_Linkable_Format
* http://wiki.osdev.org/ELF
* http://lwn.net/Articles/531148/
*/
##表示连接符
#define DEFINE_INTEGER_MATCHANY_ROUTINE(DATATYPENAME, DATAWIDTH) \
int scan_routine_##DATATYPENAME##_ANY ()//DATATYPENAME的后面跟的是字符串,所以需要加##
{
}
DEFINE_INTEGER_MATCHANY_ROUTINE(INTEGER8, 8)
展开之后就是：
int scan_routine_INTEGER8_ANY ()
{
}
注意下面的区别
#define REGISTER_NATIVE(env, module) { \
extern int register_##module(JNIEnv *); \
if (register_##module(env) < 0) \
return JNI_FALSE; \
}
module后面不是跟的字符串，所以不需要加##
REGISTER_NATIVE(env, Emulator);会替换成
//##表示连接符
{
extern int register_Emulator(JNIEnv *);//register_Emulator(JNIEnv *)在emulator.cpp中
if (register_Emulator(env) < 0)
return JNI_FALSE;
}
#define FLAG_MACRO_NEW(stringType) true ? (stringType " ") : ""
printf("%s\n", FLAG_MACRO_NEW("I32"));
打印：I32 (注意有一个空格)
#define FLAG_MACRO_NEW(stringType) true ? (stringType " aditext") : ""
printf("%s\n", FLAG_MACRO_NEW("I32"));
打印：I32 aditext
先看一段代码，old_value_and_match_info data[0];没有暂用内存空间
typedef struct {
uint8_t old_value;
match_flags match_info;
} old_value_and_match_info;//8字节
typedef struct {
void *first_byte_in_child;
unsigned long number_of_bytes;
old_value_and_match_info data[0];
} matches_and_old_values_swath;//8字节
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
typedef struct _Info
{
int i;
char data[0];
}Info;
int main(int argc, char* argv[])
{
char buf[10] = "123456789";
void* p = NULL;
printf("%d/n",sizeof(Info));
Info* info = (Info*)malloc(sizeof(Info) + 10);
p = (void*)info->data;
printf("addr of info is %p. addr of data is %p ./n", info, p);
strcpy((char*)p, buf);
printf("%s/n", (char*)p);
return 0;
}
data是一个数组名；该数组没有元素；该数组的真实地址紧随结构体Info之后；这种声明方法可以巧妙的实现C语言里的数组扩展。
输出结果：
4
addr of info is 0072B1D0. addr of data is 0072B1D4 .
123456789
/proc/$pid/mem以与进程中相同的方式显示$pid的内存映射,即伪文件中偏移x处的字节与进程中地址x处的字节相同.
如果在该过程中未映射地址,则从文件中的相应偏移量读取将返回EIO(输入/输出错误).
例如,由于进程中的第一个页面从未映射(因此取消引用NULL指针会干净地失败而不是无意中访问实际内存),读取/proc/$pid/mem的第一个字节总会产生I/O错误.
找出进程内存映射的哪些部分的方法是读取/proc/$pid/maps.此文件包含每个映射区域一行,如下所示：
08048000-08054000 r-xp 00000000 08:01 828061 /bin/cat
08c9b000-08cbc000 rw-p 00000000 00:00 0 [heap]
前两个数字是区域的边界(第一个字节的地址和最后一个字节,以hexa表示).
下一列包含权限,如果这是文件映射,则有关于文件(偏移量,设备,inode和名称)的一些信息.
有关更多信息,请参见proc(5)手册页或Understanding Linux /proc/id/maps.
result->int64_value = *((int64_t *)&peekbuf.cache[reqaddr - peekbuf.base]); /*lint !e826 */
原因：
在夜神模拟器中要使用x86架构的程序，如果使用armeabi-v7a架构的程序会崩溃
//----------------- searchregions
void *address = NULL;
unsigned offset, nread = 0;
for (offset = 0; offset < nread; offset++)
{
address = r->start + offset;
peekdata(vars->target, address, &data_value) //第一次地址：0x8000 第二次地址：0x8001
}
//----------------- peekdata
for (i = 0; i < shift_size1; i += sizeof(long))// 循环一次到两次
{
long ptraced_long = ptrace(PTRACE_PEEKDATA, pid, ptrace_address, NULL); //读4字节
... ....
*((long *)&peekbuf.cache[peekbuf.size]) = ptraced_long;
peekbuf.size += sizeof(long);
}
result->int64_value = *((int64_t *)&peekbuf.cache[reqaddr - peekbuf.base]); /*lint !e826 */
//结果存的是8字节
//第一次 result->int64_value = *((int64_t *)&peekbuf.cache[0]);
//第二次 result->int64_value = *((int64_t *)&peekbuf.cache[1]);
#include <stdarg.h>
void appendToLogFile(const char *format, ...)
{
char content[1024];
memset(content, 0, 1024);
va_list args;
va_start(args, format);
vsprintf(content, format, args);
#if 1 //添加换行
strcat(content, "\n");
#endif
va_end(args);
FILE* log_file = fopen("/data/local/tmp/scan.log", "a+");
if (log_file != NULL) {
fwrite(content, strlen(content), 1, log_file);
fflush(log_file);
fclose(log_file);
}
}
使用：
appendToLogFile("## tag2.1.3 peekbuf.base : %0x", peekbuf.base);
在内存中存放的数据是不会区分类型的，在某个地址取出来的数据，强制转换之后数据是可以直观理解的数据
float double类型,在内存中的存储方式
unsigned char uip_buf1[] = { 0x5E, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00 };//内存中的十六进制数据
char* resultChar = (char*)(&(uip_buf1)); //取出1字节
int* resultInt = (int*)(&(uip_buf1)); //取出4字节
float* resultFlo = (float*)(&(uip_buf1)); //取出4字节
double* resultDou = (double*)(&(uip_buf1)); //取出8字节
printf("%c\n", *resultChar); //^(对应ascii码0x5E)
printf("%d\n", *resultInt); //94
printf("%f %lf %e %.8e\n", *resultFlo, *resultFlo, *resultFlo, *resultFlo); //0.000000 0.000000 1.317221e-43 1.31722056e-43
printf("%f %lf %e %.8e\n", *resultDou, *resultDou, *resultDou, *resultDou); //0.000000 0.000000 4.243992e-313 4.24399159e-313
char strMsg[512];
memset(strMsg, 0, 512);
sprintf(strMsg, "%c, %d, %.8e, %.8e\n", *resultChar, *resultInt, *resultFlo, *resultDou);
printf("%s", strMsg); //^, 94, 1.31722056e-43, 4.24399159e-313
long类型表示long int，一般简写为long，注意long不表示long double。
虽然有时候会有sizeof(long)=sizeof(int)，long int与int是不同的：
16位系统：long是4字节，int是2字节
32位系统：long是4字节，int是4字节
64位系统：long也是4字节，int是4字节
标准只规定long不小于int的长度，int不小于short的长度。
即sizeof(long)>=sizeof(int)
#include <stdarg.h>
void dumpMem(long data)
{
FILE* log_file = fopen("/data/local/tmp/memRegion", "a+");
if (log_file != NULL)
{
fwrite(&data, 1, 4, log_file);
fflush(log_file);
fclose(log_file);
}
}
long ptraced_long = ptrace(PTRACE_PEEKDATA, pid, ptrace_address, NULL); //ptrace获取到的数据是long类型的
dumpMem(ptraced_long);
element_t *n = vars->regions->head;
region_t *r;
while (n)
{
unsigned offset, nread = 0;
r = n->data;
nread = r->size;
for (offset = 0; offset < nread; offset++)
{
value_t data_value;
address = r->start + offset;
peekdata(vars->target, address, &data_value);
}
}
bool peekdata(pid_t pid, void *addr, value_t * result)
{
if ()//有缓存时，直接读取 返回
{
case 1;
result->int64_value = *((int64_t *)&peekbuf.cache[reqaddr - peekbuf.base]);
return true;
}
else if ()//没有缓存，计算shift_size1 后面ptrace(PTRACE_PEEKDATA
{
case 2;
//做数据的迁移
}
else if ()//开始时调用一次
{
case 3;
}
for (i = 0; i < shift_size1; i += sizeof(long))
{
long ptraced_long = ptrace(PTRACE_PEEKDATA, pid, ptrace_address, NULL);
}
result->int64_value = *((int64_t *)&peekbuf.cache[reqaddr - peekbuf.base]);
return true;
}
读内存
case 3
case 3
case 2
case 1
case 1
case 1
case 2
case 1
case 1
case 1
case 2
case 1
case 1
case 1
... ...
在调用ptrace(PTRACE_ATTACH )之前，需要先执行su，有管理权限调用才会成功。
if (system("su") == -1)
LOGE( "system(\"su\") failed, command was not executed.\n");
else
LOGE("system(\"su\") sucess, command was executed.\n");
Android Studio下面调试（ndk）上面的代码，调用system("su")是会signal---SIGCHLD
ptrace只调用PTRACE_ATTACH游戏后，回到游戏，游戏会卡死黑屏。
#define APP_PKG_NAME "com/gameassist/plugin/nativeutils"
# define NELEM(x) ((int) (sizeof(x) / sizeof((x)[0])))
int register_GodBridge(JNIEnv *env)
{
static const JNINativeMethod methods[] = {
{ "init", "(I)V",(void *)init_bridge},
{ "initNativePtr", "(II)V", (void *)initNativePtr_bridge },
{ "nativeProcessCheat", "(IIII)I", (int *)nativeProcessCheat_bridge }
};
jclass clazz;
clazz = env->FindClass(APP_PKG_NAME"/NativeUtils");
return env->RegisterNatives(clazz, methods, NELEM(methods));
}
注意"(IIII)I"，括号里面的IIII直接不要加分号，不然会register失败
Ljava/lang/String;后面是要加分号的
加了extern "C" 才能通过dlsym找到so里面的函数
extern "C"
{
__attribute__((visibility("default"))) int jniInitNativePtr(JNIEnv *env, jobject thiz, int arg0, int arg1)
{
LOGE("######### jniInitNativePtr begin.\n");
LOGE("######### jniInitNativePtr end.\n");
return 0;
}
__attribute__((visibility("default"))) int jniDoProcessCheat(JNIEnv *env, jobject thiz, int arg0, int arg1, int arg2, int arg3)
{
LOGE("######### jniDoProcessCheat begin.\n");
LOGE("######### jniDoProcessCheat end.\n");
return 0;
}
}
int (*f_JNIOnLoad)(JavaVM* vm, void* reserved);
f_JNIOnLoad = (jint(*)(JavaVM*, void*))dlsym(dll, "JNI_OnLoad");
JNI_OnLoad的实现都是这样写：
__attribute__((visibility("default")))
jint JNI_OnLoad(JavaVM* vm, void* reserved)
{
}
so在IDA中export可以看到的函数都是可以dlysm的，在Functions window中看到的函数名和export的可能是不一样的
f_initNativePtr = (int(*)(JNIEnv *, jobject, unsigned int, unsigned int))dlsym(dll, "_Z16jniInitNativePtrP7_JNIEnvP7_jclassii");
f_nativeProcessCheat = (int(*)(JNIEnv *, jobject, int, int, int, int))dlsym(dll, "_Z17jniDoProcessCheatP7_JNIEnvP7_jclassiiii");
//info.dli_fbase 加载so时的地址，info.dli_saddr 符号的实际地址
//info.dli_fbase + 查看IDA .text段 符号的偏移 = 符号的实际地址
Dl_info info;
dladdr((void*)f_JNIOnLoad, &info);
appendToLogFile("f_JNIOnLoad Dl_info: %x %x %x", info.dli_fbase, info.dli_saddr, (void*)(uint32_t(info.dli_saddr) - uint32_t(info.dli_fbase)));
dladdr((void*)f_initNativePtr, &info);
appendToLogFile("f_initNativePtr Dl_info: %x %x %x", info.dli_fbase, info.dli_saddr, (void*)(uint32_t(info.dli_saddr) - uint32_t(info.dli_fbase)));
dladdr((void*)f_nativeProcessCheat, &info);
appendToLogFile("f_nativeProcessCheat Dl_info: %x %x %x", info.dli_fbase, info.dli_saddr, (void*)(uint32_t(info.dli_saddr) - uint32_t(info.dli_fbase)));
(void*)(uint32_t(info.dli_saddr) - uint32_t(info.dli_fbase))在armeabi-v7a x86下可以编译通过， 在x86_64下编译不报错
error: 'Dl_info' was not declared in this scope Dl_info info; 在Application.mk中添加APP_PLATFORM := android-9
打印的日志：
######### init_bridge begin.
dlsym dlopen sucess
dlsym dll != NULL
find f_JNIOnLoad.
call f_JNIOnLoad.
find jniInitNativePtr.
find jniDoProcessCheat.
f_JNIOnLoad Dl_info: c4da0000 c4da539d 539d
f_initNativePtr Dl_info: c4da0000 c4da5301 5301
f_nativeProcessCheat Dl_info: c4da0000 c4da52b9 52b9
######### init_bridge end.
Dl_info info; 需要在Application.mk中使用APP_PLATFORM := android-9指定android平台版本。
默认情况下，如果使用android-3并失败，出现找不到Dl_info