Path: utzoo!utgpu!water!watmath!clyde!rutgers!umd5!brl-adm!husc6!sri-unix!quintus!ok
From: ok@quintus.UUCP (Richard A. O'Keefe)
Newsgroups: comp.arch
Subject: Re: self-modifying code
Message-ID: <674@cresswell.quintus.UUCP>
Date: 22 Feb 88 03:10:00 GMT
References: <486@taux01.UUCP> <666@cresswell.quintus.UUCP> <815@spar.SPAR.SLB.COM> <2169@bsu-cs.UUCP>
Organization: Quintus Computer Systems, Mountain View, CA
Lines: 29

In article <2169@bsu-cs.UUCP>, dhesi@bsu-cs.UUCP (Rahul Dhesi) writes:
> [re: how to overcome the code versus data distinction]
> For those implementors who wish to allow data to be safely treated as
> code, I suggest a system call that will accept a range of memory
> addresses and a request to change it from code to data or from data to
> code.  E.g.:
> 
>      #include <modech.h>
>      if (modech(startadd, endadd, TO_CODE) == -1)
>         error ("could not change data into code for execution");
> 
It won't work in a separate I&D machine.  To start with, that range of
data addresses may *already* be in use in the code space, or it may not
be valid.  (Imagine a machine where code space is negative and data
space is positive.)  Another thing that can go wrong is that, as on the
B6700, code and data may be different *kinds* of address.  If you can
use the same addresses for code and data as Rahul Dhesi suggests, your
problem is just an unco-operative operating systems.  That's why I
suggested "load this .obj file where you like, and tell me where you put
it", because that *does* make sense in genuinely separate I&D machines.
Of course it doesn't have to be a .obj file; the thing you hand to the
kernel might be a chunk of data memory to copy, or for that matter a
Postscript program to set individual bits (:-).  The important thing is
that it has to be a copy, because "separate I&D" means *separate*.

An interesting point of dynamic *extension* of programs is that (if the
code space allocator grain size is a multiple of the instruction cache
entry size) you don't need to flush the instruction cache, because valid
entries are still valid.