so_inject.c

#include <stdio.h>
#include <stdlib.h>
#include <sys/user.h>
#include <asm/ptrace.h>
#include <sys/ptrace.h>
#include <sys/wait.h>
#include <sys/mman.h>
#include <dlfcn.h>
#include <dirent.h>
#include <unistd.h>
#include <string.h>
#include <elf.h>
#include <android/log.h>
#include <sys/uio.h>

#if defined(__i386__)
#define pt_regs         user_regs_struct
#elif defined(__aarch64__)
#define pt_regs         user_pt_regs
#define uregs    regs
#define ARM_pc    pc
#define ARM_sp    sp
#define ARM_cpsr    pstate
#define ARM_lr        regs[30]
#define ARM_r0        regs[0]
#define PTRACE_GETREGS PTRACE_GETREGSET
#define PTRACE_SETREGS PTRACE_SETREGSET
#endif

#define ENABLE_DEBUG 1

#if ENABLE_DEBUG
#define  LOG_TAG "INJECT"
#define  LOGD(fmt, args...)  __android_log_print(ANDROID_LOG_DEBUG,LOG_TAG, fmt, ##args)
#define DEBUG_PRINT(format,args...) \
    LOGD(format, ##args)
#else
#define DEBUG_PRINT(format,args...)
#endif

#define CPSR_T_MASK     ( 1u << 5 )

#if defined(__aarch64__)
const char *libc_path = "/system/lib64/libc.so";
const char *linker_path = "/system/bin/linker64";
#else
const char *libc_path = "/system/lib/libc.so";
const char *linker_path = "/system/bin/linker";
#endif

int ptrace_readdata(pid_t pid,  uint8_t *src, uint8_t *buf, size_t size)
{
    long i, j, remain;
    uint8_t *laddr;
    size_t bytes_width = sizeof(long);

    union u {
        long val;
        char chars[bytes_width];
    } d;

    j = size / bytes_width;
    remain = size % bytes_width;

    laddr = buf;

    for (i = 0; i < j; i ++) {
        d.val = ptrace(PTRACE_PEEKTEXT, pid, src, 0);
        memcpy(laddr, d.chars, bytes_width);
        src += bytes_width;
        laddr += bytes_width;
    }

    if (remain > 0) {
        d.val = ptrace(PTRACE_PEEKTEXT, pid, src, 0);
        memcpy(laddr, d.chars, remain);
    }

    return 0;
}

/*
Func : 将size字节的data数据写入到pid进程的dest地址处
@param dest: 目的进程的栈地址
@param data: 需要写入的数据的起始地址
@param size: 需要写入的数据的大小，以字节为单位
*/
int ptrace_writedata(pid_t pid, uint8_t *dest, uint8_t *data, size_t size)
{
    long i, j, remain;
    uint8_t *laddr;
    size_t bytes_width = sizeof(long);

    //很巧妙的联合体，这样就可以方便的以字节为单位写入4字节数据，再以long为单位ptrace_poketext到栈中 
    union u {
        long val;
        char chars[bytes_width];
    } d;

    j = size / bytes_width;
    remain = size % bytes_width;

    laddr = data;

    //先以4字节为单位进行数据写入

    for (i = 0; i < j; i ++) {
        memcpy(d.chars, laddr, bytes_width);
        ptrace(PTRACE_POKETEXT, pid, dest, d.val);

        dest  += bytes_width;
        laddr += bytes_width;
    }

    if (remain > 0) {
        //为了最大程度的保持原栈的数据，先读取dest的long数据，然后只更改其中的前remain字节，再写回
        d.val = ptrace(PTRACE_PEEKTEXT, pid, dest, 0);
        for (i = 0; i < remain; i ++) {
            d.chars[i] = *laddr ++;
        }

        ptrace(PTRACE_POKETEXT, pid, dest, d.val);
    }

    return 0;
}

/*
功能总结：
1，将要执行的指令写入寄存器中，指令长度大于4个long的话，需要将剩余的指令通过ptrace_writedata函数写入栈中；
2，使用ptrace_continue函数运行目的进程，直到目的进程返回状态值0xb7f（对该值的分析见后面红字）；
3，函数执行完之后，目标进程挂起，使用ptrace_getregs函数获取当前的所有寄存器值，方便后面使用ptrace_retval函数获取函数的返回值。
*/
#if defined(__arm__) || defined(__aarch64__)
int ptrace_call(pid_t pid, uintptr_t addr, long *params, int num_params, struct pt_regs* regs)
{
    int i;
#if defined(__arm__)
    int num_param_registers = 4;
#elif defined(__aarch64__)
    int num_param_registers = 8;
#endif

    for (i = 0; i < num_params && i < num_param_registers; i ++) {
        regs->uregs[i] = params[i];
    }

    //
    // push remained params onto stack
    //
    if (i < num_params) {
        regs->ARM_sp -= (num_params - i) * sizeof(long) ;
        ptrace_writedata(pid, (void *)regs->ARM_sp,(uint8_t *)& params[i], (num_params - i) * sizeof(long));
    }
    //将PC寄存器值设为目标函数的地址
    regs->ARM_pc = addr;
    //进行指令集判断 
    if (regs->ARM_pc & 1) {
        /* thumb */
        regs->ARM_pc &= (~1u);
        // #define CPSR_T_MASK  ( 1u << 5 )  CPSR为程序状态寄存器
        regs->ARM_cpsr |= CPSR_T_MASK;
    } else {
        /* arm */
        regs->ARM_cpsr &= ~CPSR_T_MASK;
    }

    //设置子程序的返回地址为空，以便函数执行完后，返回到null地址，产生SIGSEGV错误，详细作用见后面的红字分析
    regs->ARM_lr = 0;

    /*
    *Ptrace_setregs就是将修改后的regs写入寄存器中，然后调用ptrace_continue来执行我们指定的代码
    */
    if (ptrace_setregs(pid, regs) == -1
            || ptrace_continue(pid) == -1) {
        printf("error\n");
        return -1;
    }

    int stat = 0;
    waitpid(pid, &stat, WUNTRACED);
    /* WUNTRACED告诉waitpid，如果子进程进入暂停状态，那么就立即返回。如果是被ptrace的子进程，那么即使不提供WUNTRACED参数，也会在子进程进入暂停状态的时候立即返回。
    对于使用ptrace_cont运行的子进程，它会在3种情况下进入暂停状态：①下一次系统调用；②子进程退出；③子进程的执行发生错误。这里的0xb7f就表示子进程进入了暂停状态，且发送的错误信号为11(SIGSEGV)，它表示试图访问未分配给自己的内存, 或试图往没有写权限的内存地址写数据。那么什么时候会发生这种错误呢？显然，当子进程执行完注入的函数后，由于我们在前面设置了regs->ARM_lr = 0，它就会返回到0地址处继续执行，这样就会产生SIGSEGV了！*/

    //这个循环是否必须我还不确定。因为目前每次ptrace_call调用必定会返回0xb7f，不过在这也算是增加容错性吧~

    //通过看ndk的源码sys/wait.h以及man waitpid可以知道这个0xb7f的具体作用。首先说一下stat的值：高2字节用于表示导致子进程的退出或暂停状态信号值，低2字节表示子进程是退出(0x0)还是暂停(0x7f)状态。0xb7f就表示子进程为暂停状态，导致它暂停的信号量为11即sigsegv错误。
    while (stat != 0xb7f) {
        if (ptrace_continue(pid) == -1) {
            printf("error\n");
            return -1;
        }
        waitpid(pid, &stat, WUNTRACED);
    }

    return 0;
}

#elif defined(__i386__)
long ptrace_call(pid_t pid, uintptr_t addr, long *params, int num_params, struct user_regs_struct * regs)
{
    regs->esp -= (num_params) * sizeof(long) ;
    ptrace_writedata(pid, (void *)regs->esp, (uint8_t *)params, (num_params) * sizeof(long));

    long tmp_addr = 0x00;
    regs->esp -= sizeof(long);
    ptrace_writedata(pid, regs->esp, (char *)&tmp_addr, sizeof(tmp_addr));

    regs->eip = addr;

    if (ptrace_setregs(pid, regs) == -1
            || ptrace_continue( pid) == -1) {
        printf("error\n");
        return -1;
    }

    int stat = 0;
    waitpid(pid, &stat, WUNTRACED);
    while (stat != 0xb7f) {
        if (ptrace_continue(pid) == -1) {
            printf("error\n");
            return -1;
        }
        waitpid(pid, &stat, WUNTRACED);
    }

    return 0;
}
#else
#error "Not supported"
#endif

int ptrace_getregs(pid_t pid, struct pt_regs * regs)
{
#if defined (__aarch64__)
        int regset = NT_PRSTATUS;
        struct iovec ioVec;

        ioVec.iov_base = regs;
        ioVec.iov_len = sizeof(*regs);
    if (ptrace(PTRACE_GETREGSET, pid, (void*)regset, &ioVec) < 0) {
        perror("ptrace_getregs: Can not get register values");
        printf(" io %llx, %d", ioVec.iov_base, ioVec.iov_len);
        return -1;
    }

    return 0;
#else
    if (ptrace(PTRACE_GETREGS, pid, NULL, regs) < 0) {
        perror("ptrace_getregs: Can not get register values");
        return -1;
    }

    return 0;
#endif
}

int ptrace_setregs(pid_t pid, struct pt_regs * regs)
{
#if defined (__aarch64__)
        int regset = NT_PRSTATUS;
        struct iovec ioVec;

        ioVec.iov_base = regs;
        ioVec.iov_len = sizeof(*regs);
    if (ptrace(PTRACE_SETREGSET, pid, (void*)regset, &ioVec) < 0) {
        perror("ptrace_setregs: Can not get register values");
        return -1;
    }

    return 0;
#else
    if (ptrace(PTRACE_SETREGS, pid, NULL, regs) < 0) {
        perror("ptrace_setregs: Can not set register values");
        return -1;
    }

    return 0;
#endif
}

int ptrace_continue(pid_t pid)
{
    if (ptrace(PTRACE_CONT, pid, NULL, 0) < 0) {
        perror("ptrace_cont");
        return -1;
    }

    return 0;
}

int ptrace_attach(pid_t pid)
{
    if (ptrace(PTRACE_ATTACH, pid, NULL, 0) < 0) {
        perror("ptrace_attach");
        return -1;
    }

    int status = 0;
    waitpid(pid, &status , WUNTRACED);

    return 0;
}

int ptrace_detach(pid_t pid)
{
    if (ptrace(PTRACE_DETACH, pid, NULL, 0) < 0) {
        perror("ptrace_detach");
        return -1;
    }

    return 0;
}


//显然，这里面核心的就是get_module_base函数：
/*
此函数的功能就是通过遍历/proc/pid/maps文件，来找到目的module_name的内存映射起始地址。
由于内存地址的表达方式是startAddrxxxxxxx-endAddrxxxxxxx的，所以会在后面使用strtok(line,"-")来分割字符串
如果pid = -1,表示获取本地进程的某个模块的地址，
否则就是pid进程的某个模块的地址。
*/

void* get_module_base(pid_t pid, const char* module_name)
{
    FILE *fp;
    long addr = 0;
    char *pch;
    char filename[32];
    char line[1024];

    if (pid < 0) {
        /* self process */
        snprintf(filename, sizeof(filename), "/proc/self/maps", pid);
    } else {
        snprintf(filename, sizeof(filename), "/proc/%d/maps", pid);
    }

    fp = fopen(filename, "r");

    if (fp != NULL) {
        while (fgets(line, sizeof(line), fp)) {
            if (strstr(line, module_name)) {
                //分解字符串为一组字符串。line为要分解的字符串，"-"为分隔符字符串。
                pch = strtok( line, "-" );
                //将参数pch字符串根据参数base(表示进制)来转换成无符号的长整型数 
                addr = strtoull( pch, NULL, 16 );

                if (addr == 0x8000)
                    addr = 0;

                break;
            }
        }

        fclose(fp) ;
    }

    return (void *)addr;
}

/*
该函数为一个封装函数，通过调用get_module_base函数来获取目的进程的某个模块的起始地址，然后通过公式计算出指定函数在目的进程的起始地址。
*/
void* get_remote_addr(pid_t target_pid, const char* module_name, void* local_addr)
{
    void* local_handle, *remote_handle;

    //获取本地某个模块的起始地址
    local_handle = get_module_base(-1, module_name);
    //获取远程pid的某个模块的起始地址
    remote_handle = get_module_base(target_pid, module_name);

    DEBUG_PRINT("[+] get_remote_addr: local[%llx], remote[%llx]\n", local_handle, remote_handle);
    /*这需要我们好好理解：local_addr - local_handle的值为指定函数(如mmap)在该模块中的偏移量，然后再加上rempte_handle，结果就为指定函数在目的进程的虚拟地址*/
    void * ret_addr = (void *)((uintptr_t)local_addr + (uintptr_t)remote_handle - (uintptr_t)local_handle);

#if defined(__i386__)
    if (!strcmp(module_name, libc_path)) {
        ret_addr += 2;
    }
#endif
    return ret_addr;
}

//根据name找到pid
int find_pid_of(const char *process_name)
{
    int id;
    pid_t pid = -1;
    DIR* dir;
    FILE *fp;
    char filename[32];
    char cmdline[256];

    struct dirent * entry;

    if (process_name == NULL)
        return -1;

    dir = opendir("/proc");
    if (dir == NULL)
        return -1;

    while((entry = readdir(dir)) != NULL) {
        id = atoi(entry->d_name);
        if (id != 0) {
            sprintf(filename, "/proc/%d/cmdline", id);
            fp = fopen(filename, "r");
            if (fp) {
                fgets(cmdline, sizeof(cmdline), fp);
                fclose(fp);

                if (strcmp(process_name, cmdline) == 0) {
                    /* process found */
                    pid = id;
                    break;
                }
            }
        }
    }

    closedir(dir);
    return pid;
}

uint64_t ptrace_retval(struct pt_regs * regs)
{
#if defined(__arm__) || defined(__aarch64__)
    return regs->ARM_r0;
#elif defined(__i386__)
    return regs->eax;
#else
#error "Not supported"
#endif
}

uint64_t ptrace_ip(struct pt_regs * regs)
{
#if defined(__arm__) || defined(__aarch64__)
    return regs->ARM_pc;
#elif defined(__i386__)
    return regs->eip;
#else
#error "Not supported"
#endif
}

//总结一下ptrace_call_wrapper，它的完成两个功能：
//一是调用ptrace_call函数来执行指定函数，执行完后将子进程挂起；
//二是调用ptrace_getregs函数获取所有寄存器的值，主要是为了获取r0即函数的返回值。
int ptrace_call_wrapper(pid_t target_pid, const char * func_name, void * func_addr, long * parameters, int param_num, struct pt_regs * regs)
{
    DEBUG_PRINT("[+] Calling %s in target process.\n", func_name);
    if (ptrace_call(target_pid, (uintptr_t)func_addr, parameters, param_num, regs) == -1)
        return -1;

    if (ptrace_getregs(target_pid, regs) == -1)
        return -1;
    DEBUG_PRINT("[+] Target process returned from %s, return value=%llx, pc=%llx \n",
            func_name, ptrace_retval(regs), ptrace_ip(regs));
    return 0;
}

//远程注入
int inject_remote_process(pid_t target_pid, const char *library_path, const char *function_name, const char *param, size_t param_size)
{
    int ret = -1;
    void *mmap_addr, *dlopen_addr, *dlsym_addr, *dlclose_addr, *dlerror_addr;
    void *local_handle, *remote_handle, *dlhandle;
    uint8_t *map_base = 0;
    uint8_t *dlopen_param1_ptr, *dlsym_param2_ptr, *saved_r0_pc_ptr, *inject_param_ptr, *remote_code_ptr, *local_code_ptr;

    struct pt_regs regs, original_regs;
    long parameters[10];

    DEBUG_PRINT("[+] Injecting process: %d\n", target_pid);

    //①ATTATCH，指定目标进程，开始调试
    if (ptrace_attach(target_pid) == -1)
        goto exit;

    //②GETREGS，获取目标进程的寄存器，保存现场
    if (ptrace_getregs(target_pid, &regs) == -1)
        goto exit;

    /* save original registers */
    memcpy(&original_regs, &regs, sizeof(regs));

    //③通过get_remote_addr函数获取目的进程的mmap函数的地址，以便为libxxx.so分配内存

    /*
        需要对(void*)mmap进行说明：这是取得inject本身进程的mmap函数的地址，由于mmap函数在libc.so 
        库中，为了将libxxx.so加载到目的进程中，就需要使用目的进程的mmap函数，所以需要查找到libc.so库在目的进程的起始地址。
    */
    mmap_addr = get_remote_addr(target_pid, libc_path, (void *)mmap);
    DEBUG_PRINT("[+] Remote mmap address: %llx\n", mmap_addr);

    /* call mmap (null, 0x4000, PROT_READ | PROT_WRITE | PROT_EXEC,
                             MAP_ANONYMOUS | MAP_PRIVATE, 0, 0);
    匿名申请一块0x4000大小的内存
    */
    parameters[0] = 0;  // addr
    parameters[1] = 0x4000; // size
    parameters[2] = PROT_READ | PROT_WRITE | PROT_EXEC;  // prot
    parameters[3] = MAP_ANONYMOUS | MAP_PRIVATE; // flags
    parameters[4] = 0; //fd
    parameters[5] = 0; //offset

    if (ptrace_call_wrapper(target_pid, "mmap", mmap_addr, parameters, 6, &regs) == -1)
        goto exit;

    //⑤从寄存器中获取mmap函数的返回值，即申请的内存首地址：
    map_base = ptrace_retval(&regs);

    //⑥依次获取linker中dlopen、dlsym、dlclose、dlerror函数的地址:
    dlopen_addr = get_remote_addr( target_pid, linker_path, (void *)dlopen );
    dlsym_addr = get_remote_addr( target_pid, linker_path, (void *)dlsym );
    dlclose_addr = get_remote_addr( target_pid, linker_path, (void *)dlclose );
    dlerror_addr = get_remote_addr( target_pid, linker_path, (void *)dlerror );

    DEBUG_PRINT("[+] Get imports: dlopen: %llx, dlsym: %llx, dlclose: %llx, dlerror: %llx\n",
            dlopen_addr, dlsym_addr, dlclose_addr, dlerror_addr);

    printf("library path = %s\n", library_path);
    //⑦调用dlopen函数：
    /*
    ①将要注入的so名写入前面mmap出来的内存
    ②写入dlopen代码
    ③执行dlopen("libxxx.so", RTLD_NOW ! RTLD_GLOBAL)
    RTLD_NOW之类的参数作用可参考：
    http://baike.baidu.com/view/2907309.htm?fr=aladdin
    ④取得dlopen的返回值，存放在sohandle变量中
    */
    ptrace_writedata(target_pid, map_base, library_path, strlen(library_path) + 1);

    parameters[0] = map_base;
    parameters[1] = RTLD_NOW| RTLD_GLOBAL;

    if (ptrace_call_wrapper(target_pid, "dlopen", dlopen_addr, parameters, 2, &regs) == -1)
        goto exit_with_error;

    void * sohandle = ptrace_retval(&regs);
    if(!sohandle) {
        if (ptrace_call_wrapper(target_pid, "dlerror", dlerror_addr, 0, 0, &regs) != -1)
        {
            uint8_t *errret = ptrace_retval(&regs);
            uint8_t errbuf[250];
            ptrace_readdata(target_pid, errret, errbuf, 250);
            printf("%s\n", errbuf);
        }
        goto exit_with_error;
      }

    //⑧调用dlsym函数
    /*
    等同于hook_entry_addr = (void *)dlsym(sohandle, "hook_entry");
    */
#define FUNCTION_NAME_ADDR_OFFSET       0x100
    ptrace_writedata(target_pid, map_base + FUNCTION_NAME_ADDR_OFFSET, function_name, strlen(function_name) + 1);
    parameters[0] = sohandle;
    parameters[1] = map_base + FUNCTION_NAME_ADDR_OFFSET;

    if (ptrace_call_wrapper(target_pid, "dlsym", dlsym_addr, parameters, 2, &regs) == -1)
        goto exit_with_error;

    void * hook_entry_addr = ptrace_retval(&regs);
    DEBUG_PRINT("hook_entry_addr = %p\n", hook_entry_addr);

    //⑨调用hook_entry函数：
#define FUNCTION_PARAM_ADDR_OFFSET      0x200
    ptrace_writedata(target_pid, map_base + FUNCTION_PARAM_ADDR_OFFSET, param, strlen(param) + 1);
    parameters[0] = map_base + FUNCTION_PARAM_ADDR_OFFSET;

    if (ptrace_call_wrapper(target_pid, "hook_entry", hook_entry_addr, parameters, 1, &regs) == -1)
        goto exit_with_error;

/*
    printf("Press enter to dlclose and detach\n");
    getchar();
*/

    parameters[0] = sohandle;

    //⑩调用dlclose关闭lib:
    if (ptrace_call_wrapper(target_pid, "dlclose", dlclose, parameters, 1, &regs) == -1)
        goto exit_with_error;

exit_with_error:
    /* restore */
    //⑪恢复现场并退出ptrace:
    ptrace_setregs(target_pid, &original_regs);
    ptrace_detach(target_pid);
    ret = 0;

exit:
    return ret;
}

void print_help(char * self)
{
    printf("SO INJECT for Android API 23\n");
    printf("Usage:\n\t%s <PID/Process> <so_path> [hook_entry] [parameter]", self);
    printf("\nOptimized by H.K.T\n");
}

int main(int argc, char** argv) {
    pid_t target_pid;
    char szBuf[256];
    char pathBuf[256];
    char paraBuf[256]; // = "I'm parameter!";
    char entryBuf[256]; // = "hook_entry";
    char* lpFlag = NULL;

    if (argc < 3)
    {
        print_help(argv[0]);
        return 0;
    }
    else if(argc == 3)
    {
        strcpy(entryBuf, "hook_entry");
        strcpy(paraBuf, "I'm parameter!");
    }
    else if (argc == 4)
    {
        strcpy(entryBuf, argv[3]);
    }
    else
    {
        strcpy(entryBuf, argv[3]);
        strcpy(paraBuf, argv[4]);
    }

    strcpy(szBuf, argv[1]);
    strcpy(pathBuf, argv[2]);
    lpFlag = strstr(szBuf, "#");

    if (lpFlag != NULL)
    {
        *lpFlag = '\0';
    }

    target_pid = find_pid_of(szBuf);

    if (-1 == target_pid) {
        printf("Can't find the process\n");
        return -1;
    }

    printf("pid : %d\npath : %s\nentry: %s\npara : %s\n", target_pid, pathBuf, entryBuf, paraBuf);
    inject_remote_process(target_pid, pathBuf, entryBuf, paraBuf, strlen(paraBuf));
    return 0;
}