Arbitrary Read/Write

Arbitrary Read

fwrite

  • Set the _fileno to the file descriptor of stdout

  • Set _flag & ~_IO_NO_WRITES

  • Set _flag |= _IO_CURRENTLY_PUTTING

  • Set the write_base and write_ptr to memory address which you want to read

  • Set _IO_read_end equal to _IO_write_base

or though not always reliable, in pwntools:

fp = FileStructure(0x0)
payload = fp.read(addr, 0x120)
io.send(payload)

fclose

upon fclose(), if buffer is not empty, it will be flushed (written to the _fileno)

  • Set the _fileno to the file descriptor of choice

  • Set write_base to the start of write address

  • Set write_ptr to the end of write address

be careful of closing if _fileno is set to stdout or stdin

explained in-depth in this writeup:

we can trace the function calls as the following,

int
_IO_new_fclose (FILE *fp)
{
  int status;

  CHECK_FILE(fp, EOF);

  // ... 

  /* First unlink the stream.  */
  if (fp->_flags & _IO_IS_FILEBUF)
    _IO_un_link ((struct _IO_FILE_plus *) fp);

  _IO_acquire_lock (fp);
  
  // ...
  
  _IO_FINISH (fp); // <-- OUR INTEREST
  
  // ...
  
  _IO_deallocate_file (fp);
  return status;
}

versioned_symbol (libc, _IO_new_fclose, _IO_fclose, GLIBC_2_1);
strong_alias (_IO_new_fclose, __new_fclose)
versioned_symbol (libc, __new_fclose, fclose, GLIBC_2_1);

which then calls _IO_FINISH(),

typedef void (*_IO_finish_t) (FILE *, int); /* finalize */
#define _IO_FINISH(FP) JUMP1 (__finish, FP, 0)
#define _IO_WFINISH(FP) WJUMP1 (__finish, FP, 0)

referring to the jump table, we can see that the implementation for __finish is _IO_new_file_finish()

void
_IO_new_file_finish (FILE *fp, int dummy)
{
  if (_IO_file_is_open (fp))
    {
      _IO_do_flush (fp); // <-- OUR INTEREST
      if (!(fp->_flags & _IO_DELETE_DONT_CLOSE))
	_IO_SYSCLOSE (fp);
    }
  _IO_default_finish (fp, 0);
}
libc_hidden_ver (_IO_new_file_finish, _IO_file_finish)

which will call the flush macro _IO_do_flush() and is defined as follows:

#define _IO_do_flush(_f) \
  ((_f)->_mode <= 0							      \
   ? _IO_do_write(_f, (_f)->_IO_write_base,				      \
		  (_f)->_IO_write_ptr-(_f)->_IO_write_base)		      \
   : _IO_wdo_write(_f, (_f)->_wide_data->_IO_write_base,		      \
		   ((_f)->_wide_data->_IO_write_ptr			      \
		    - (_f)->_wide_data->_IO_write_base)))

which writes write_base up until write_ptr


puts

refer to fwrite.


Arbitrary Write

fread

  • Set the _fileno to file descriptor of stdin

  • Set _flag &~ _IO_NO_READS

  • Set _flag |= _IO_CURRENTLY_PUTTING

  • Set read_base equals to read_ptr to NULL

  • Set the buf_base and buf_end to memory address which you want to write

  • buf_end - buf_base > size of fread (BUFFER SIZE > READ SIZE) (or other function, scanf, etc)

or though not always reliable, in pwntools:

fp = FileStructure(0x0)
payload = fp.write(addr, 0x120)
io.send(payload)

fwrite

fwrite(const void *buf, size_t size, size_t count, FILE *fp);

  • Set fp->_flags to _IO_MAGIC | ~_IO_LINE_BUF | ~_IO_CURRENTLY_PUTTING

  • Set buf to the data wish to be written

  • Set fp->write_ptr to the start of write address

  • Set fp->write_end to the end of write address

we can trace the function calls as the following,

size_t
_IO_fwrite (const void *buf, size_t size, size_t count, FILE *fp)
{
  // ...
  _IO_acquire_lock (fp);
  if (_IO_vtable_offset (fp) != 0 || _IO_fwide (fp, -1) == -1)
    written = _IO_sputn (fp, (const char *) buf, request); // <-- OUR INTEREST
  // ...
}
libc_hidden_def (_IO_fwrite)

which calls _IO_sputn() macro:

// https://elixir.bootlin.com/glibc/glibc-2.31/source/libio/libioP.h#L379
#define _IO_sputn(__fp, __s, __n) _IO_XSPUTN (__fp, __s, __n)

// https://elixir.bootlin.com/glibc/glibc-2.31/source/libio/libioP.h#L176
#define _IO_XSPUTN(FP, DATA, N) JUMP2 (__xsputn, FP, DATA, N)
#define _IO_WXSPUTN(FP, DATA, N) WJUMP2 (__xsputn, FP, DATA, N)

referring to the jump table, we can see that the implementation for __xsputn is _IO_new_file_xsputn()

size_t
_IO_new_file_xsputn (FILE *f, const void *data, size_t n)
{
  // ....
  size_t to_do = n;
  // ....
  size_t count = 0;
  // ....
  
  if ((f->_flags & _IO_LINE_BUF) && (f->_flags & _IO_CURRENTLY_PUTTING)) // [0]
    {
      // .. 
    }
  else if (f->_IO_write_end > f->_IO_write_ptr) // <-- NEEDS TO BE SATISFIED [1]
    count = f->_IO_write_end - f->_IO_write_ptr; /* Space available. */

  /* Then fill the buffer. */
  if (count > 0) // <-- NEEDS TO BE SATISFIED [2]
    {
      if (count > to_do) // <-- NEEDS TO BE SATISFIED [3]
	count = to_do;
      f->_IO_write_ptr = __mempcpy (f->_IO_write_ptr, s, count); // <-- OUR INTEREST
      s += count;
      to_do -= count;
    }
  if (to_do + must_flush > 0)
    {
     // ...  
    }
  return n - to_do;
}
libc_hidden_ver (_IO_new_file_xsputn, _IO_file_xsputn)

our interest lies in __mempcpy (f->_IO_write_ptr, s, count); where it will do memcpy of destination write_ptr and source s (buf).

first, note that initially count and to_do is set to 0 as default, then modified over the course of the function execution.

at [3] we need count to be bigger than to_do which can be easily done because it is 0 as default, so we just need to somehow able to change count to be bigger than 0 which also a condition in [2]

though the block at [0] also modifies count, I think its the route at [1] is more trivial and intuitive. so we need to satisfy [1] while dissatisfy [0].

when all of those requirements are fulfilled, it will then call memcpy which corresponds to write profit.


Last updated