Re: [9fans] Plan 9 buffer overflow exploit explained in Phrack Volume 0x0b, Issue 0x3e, Phile #0x09 of 0x0f

[Dan Cross <cross@math.psu.edu>]

> it's not just mpm.  i can run almost every binary that has
> ever been compiled on plan 9 (for my current architecture).
> that's useful occasionally, and it's 100 lines of code to implement.
> if you don't want to carry it around, you could chop it out
> of your copy, but there are much bigger things you could drop
> instead.

Oh, I don't know; it depends on what the compatibility routines do.  If
they just marshall arguments around (for instance, dealing with the
widening of types and the like), that's one thing.  On the other hand,
if they represent entirely different implementations of common system
calls, such as in cases where semantics have changed and some amount of
compatibility with the old semantics was desired, that's another.  If
we find the system call table filling up with the latter, I'd say it's
time to start thinking about recompiling things that are usefully
brought forward and chopping out the compatibility stuff.  Yes, there
are bigger fish to fry, but if you can kill this one while it's still a
tadpole, it seems like a good idea.

That said, I'd note that some of the more interesting programs in,
e.g., 2nd edition are affected not by the lack or change in system
calls, but by fundamental changes in subsystems (ie, the introduction
of /dev/draw as a replacement for /dev/bitblt and friends, which
affects a number of things).

	- Dan C.

(Ps- I really do think it would be nice to get mpm out and give people
the opportunity to either rewrite it or recompile it with g++.)