Sunday, October 4, 2015

Executing x86 machine code from char array in C/C++

The idea is:
  1. create bunch of char array in a file containing the machine code(not as const)
  2. mark the memory as executable with mprotect or VirtualProtect.
  3. declare typedef function with their correct parameters and pointing to the address of our char array
  4. call it
It's just pretty straightforward for a simple function, like a function that do x+1 or a*b, but function calling another function(e.x.: function calling strlen, malloc, etc.) needs the pointer to be calculated at first or the function will call wrong code(undefined berhaviour).

Okay so let's start. I've been make simple function that XOR 7303014 by x. This is the code


#include <stdio.h>
 
int foo_bar(int baz) {
 int foo=7303014;
 return foo^baz;
}
 
int main() {
 int val=1904132;
 int out=foo_bar(val);
 printf("%s",&out);
 return 0;
}


It looks simple right. It only prints string "bar" in the console. Now what we need is the foo_bar function as machine code. Open Hex Editor and open the executable then find the function.

Note: We are using x86 byte code for this. Take care of this if you're try to compile it as 64-bit code.
The highlighted hex is the foo_bar function that we need to get their machine code representation and put them in our new code. This is our char array declaration

char foo_bar[]={0x55, 0x89, 0xE5, 0x83, 0xEC, 0x10, 0xC7, 0x45, 0xFC, 0x66, 0x6F, 0x6F, 0x00, 0x8B, 0x45, 0x08, 0x8B, 0x55, 0xFC, 0x31, 0xD0, 0xC9, 0xC3};

Now we have completed point 1 above, now let's make some code to finish point 2, 3, and 4

typedef int (*foo_bar_t)(int );
 
int main() {
 foo_bar_t foo=(foo_bar_t)(void*)foo_bar;
 DWORD old_protect;
 VirtualProtect(foo,sizeof(foo_bar),PAGE_EXECUTE_READWRITE,&old_protect);
 int out=foo(0);
 printf("%s",&out);
 return 0;
}

Q: Why cast to void* first then cast it to foo_bar_t?
A: Visual Studio doesn't like casting foo_bar_t directly from char* so, cast it to void* first. GCC works fine without cast to void* above.

Ok that's our complete main function. Now let's explain it.

The point 2 is in the VirtualProtect function. VirtualProtect function mark address pointed by foo variable to be executable, readable, and writeable(see PAGE_EXECUTE_READWRITE). Without this call, point 4 will very likely to fail(throws Segmentation Fault/Access Violation)(be sure to include Windows.h)
The point 3 is above the main. The typedef.
Then, point 4 is below point 2. Again, calling the function without setting the memory protection would likely causes your program stop working.

Alright, this is the complete code


#include <stdio.h>
#include <windows.h>
 
typedef int (*foo_bar_t)(int );
 
char foo_bar[]={0x55, 0x89, 0xE5, 0x83, 0xEC, 0x10, 0xC7, 0x45, 0xFC, 0x66, 0x6F, 0x6F, 0x00, 0x8B, 0x45, 0x08, 0x8B, 0x55, 0xFC, 0x31, 0xD0, 0xC9, 0xC3};
 
int main() {
 foo_bar_t foo=(foo_bar_t)(void*)foo_bar;
 DWORD old_protect;
 VirtualProtect(foo,sizeof(foo_bar),PAGE_EXECUTE_READWRITE,&old_protect);
 int out=foo(0);
 printf("%s",&out);
 return 0;
}


Now let's compile it and run it.







Our program run without error. That means we've been successfully run our machine code. That's for today.

Challenge: Compile both complete code above and find out why it prints "bar" and "foo".