逆向过elf程序都知道,GCC的canary,x86_64下从fs:0x28偏移处获取,32位下从gs:0x14偏移处获取。但知道canary如何产生,为什么在这里取的人比较少。
下面以x86_64平台为例,通过glibc源码分析一下。
看第一个问题:为什么从%fs:0x28处取。%fs寄存器被glibc定义为存放tls信息,查看tls结构:
typedef struct
{
void *tcb; /* Pointer to the TCB. Not necessarily the
thread descriptor used by libpthread. */
dtv_t *dtv;
void *self; /* Pointer to the thread descriptor. */
int multiple_threads;
int gscope_flag;
uintptr_t sysinfo;
uintptr_t stack_guard; /* canary,0x28偏移 */
uintptr_t pointer_guard;
……
} tcbhead_t;
可以看到%fs:0x28实际取的是当前线程控制块的stack_guard变量,这个变量在线程创建时已经固定。下面看第二个问题,stack_guard如何赋值的。
Linux加载器完成elf加载后,会将入口设置为_start,并在栈上为_start提供入参。_start的代码在sysdeps/x86_64/start.S文件中。
_start从栈上取参数,然后调用__libc_start_main()函数,这个函数也是在main()函数之前执行:
58 ENTRY (_start)
59 /* Clearing frame pointer is insufficient, use CFI. */
60 cfi_undefined (rip)
61 /* Clear the frame pointer. The ABI suggests this be done, to mark
62 the outermost frame obviously. */
63 xorl %ebp, %ebp
64
65 /* Extract the arguments as encoded on the stack and set up
66 the arguments for __libc_start_main (int (*main) (int, char **, char **),
67 int argc, char *argv,
68 void (*init) (void), void (*fini) (void),
69 void (*rtld_fini) (void), void *stack_end).
70 The arguments are passed via registers and on the stack:
71 main: %rdi
72 argc: %rsi
73 argv: %rdx
74 init: %rcx
75 fini: %r8
76 rtld_fini: %r9
77 stack_end: stack. */
__libc_start_main()首先以_dl_random这个全局变量为入参,生成canary,然后通过THREAD_SET_STACK_GUARD宏将canary赋值给tls的stack_guard变量。
198 /* Set up the stack checker's canary. */
199 uintptr_t stack_chk_guard = _dl_setup_stack_chk_guard (_dl_random);
200 # ifdef THREAD_SET_STACK_GUARD
201 THREAD_SET_STACK_GUARD (stack_chk_guard);
202 # else
203 __stack_chk_guard = stack_chk_guard;
204 # endif
看下_dl_random哪里来的,在glibc源码中,有2处但实现大致相同:
126 /* Random data provided by the kernel. */
127 void *_dl_random;
288 case AT_RANDOM:
289 _dl_random = (void *) av->a_un.a_val;
注意av这个变量,逆向跟踪发现其最终来自__libc_start_main()的argv参数。也就是_dl_random是由加载器提供的。而AT_RANDOM表示内核提供了接口,支持canary的随机数生成。可以使用下面命令查看:
kiiim@ubuntu :~/glibc-2.22$ LD_SHOW_AUXV=1 /bin/true grep AT_RANDOM
AT_RANDOM: 0x7fffdaf776e9
看下实际代码中,这个内核接口指的是什么,canary值又如何取。
rand_size = CONFIG_SECURITY_AUXV_RANDOM_SIZE * sizeof(unsigned long);
u_rand_bytes = NULL;
if (rand_size) {
unsigned char k_rand_bytes[CONFIG_SECURITY_AUXV_RANDOM_SIZE * sizeof(unsigned long)];
get_random_bytes(k_rand_bytes, rand_size);
u_rand_bytes = (elf_addr_t __user *)STACK_ALLOC(p, rand_size);
if (__copy_to_user(u_rand_bytes, k_rand_bytes, rand_size))
return -EFAULT;
}
发现在内核中通过get_random_bytes()接口产生,并copy_to_user()到用户空间。而内核中的安全随机数,也推荐使用get_random_bytes()生成。下面看下实现:
http://lxr.free-electrons.com/source/drivers/char/random.c
void get_random_bytes(void *buf, int nbytes)
{
#if DEBUG_RANDOM_BOOT > 0
if (unlikely(nonblocking_pool.initialized == 0))
printk(KERN_NOTICE "random: %pF get_random_bytes called "
"with %d bits of entropy available
",
(void *) _RET_IP_,
nonblocking_pool.entropy_total);
#endif
trace_get_random_bytes(nbytes, _RET_IP_);
extract_entropy(&nonblocking_pool, buf, nbytes, 0, 0);
}
EXPORT_SYMBOL(get_random_bytes);
而看一下read /dev/urandom的内核实现:
static ssize_t
urandom_read(struct file *file, char __user *buf, size_t nbytes, loff_t *ppos)
{
int ret;
if (unlikely(nonblocking_pool.initialized == 0))
printk_once(KERN_NOTICE "random: %s urandom read "
"with %d bits of entropy available
",
current->comm, nonblocking_pool.entropy_total);
nbytes = min_t(size_t, nbytes, INT_MAX >> (ENTROPY_SHIFT + 3));
ret = extract_entropy_user(&nonblocking_pool, buf, nbytes);
trace_urandom_read(8 * nbytes, ENTROPY_BITS(&nonblocking_pool),
ENTROPY_BITS(&input_pool));
return ret;
}
可以看到get_random_bytes()与read /dev/urandom实现是相同的,都是通过extract_entropy*从"entropy pool"中取的随机数。只不过一个在内核空间用,将结果返回到一块内核buffer,一个在用户空间使用,将结果返回到一块用户buffer。
下面再来看下,程序中如何使用这个canary。分析a()函数:
void a() {
int a = 3;
char str[16];
}
x86_64平台汇编如下:
(gdb) disass a
Dump of assembler code for function a:
0x000000000040055d <+0>: push %rbp
0x000000000040055e <+1>: mov %rsp,%rbp
0x0000000000400561 <+4>: sub $0x30,%rsp
0x0000000000400565 <+8>: mov %fs:0x28,%rax
0x000000000040056e <+17>: mov %rax,-0x8(%rbp)
0x0000000000400572 <+21>: xor %eax,%eax
0x0000000000400574 <+23>: movl $0x3,-0x24(%rbp) ;变量重排,a的地址低于str地址
0x000000000040057b <+30>: mov -0x8(%rbp),%rax
0x000000000040057f <+34>: xor %fs:0x28,%rax
0x0000000000400588 <+43>: je 0x40058f <a+50>
0x000000000040058a <+45>: callq 0x400440 <__stack_chk_fail@plt>
0x000000000040058f <+50>: leaveq
0x0000000000400590 <+51>: retq
End of assembler dump.
可以看到,GCC的栈保护还实现了变量重排。但与微软实现不同,GCC取出canary后并没有与ebp异或,直接放到栈上。也就是说,同一线程中,所有的canary值都是相同的,通过调试验证也中如此:
Breakpoint 1, 0x000000000040056e in a () at 1.c:4
4 void a() {
(gdb) p/x $rax
$1 = 0xc609d364696f6000
(gdb) c
Continuing.
Breakpoint 2, 0x00000000004005a2 in b () at 1.c:9
9 void b() {
(gdb) p/x $rax
$2 = 0xc609d364696f6000
(gdb) c
Continuing.
Breakpoint 1, 0x000000000040056e in a () at 1.c:4
4 void a() {
(gdb) p/x $rax
$3 = 0xc609d364696f6000
(gdb)