Arbitrary Read/Write
Last updated
Last updated
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:
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,
which then calls _IO_FINISH(),
which will call the flush macro _IO_do_flush() and is defined as follows:
which writes write_base up until write_ptr
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:
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,
which calls _IO_sputn() macro:
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.
referring to the , we can see that the implementation for __finish is _IO_new_file_finish()
refer to .
referring to the , we can see that the implementation for __xsputn is _IO_new_file_xsputn()
