Re: iSeries buffer overflow immunity?

On Mon, 2004-11-08 at 10:10, Bruce Vining wrote:
> 
> 
> Correct, you cannot assume that admin_flag0 and admin_flag1 are physically
> located around buffer.  Defining them as subfields within a data structure
> would.
> 
> I used your definitions and defined some integer fields (diff1 and diff2).
> diff1 = %addr(Buffer) - %addr(admin_flag0) resulted in 8; diff2 =
> %addr(admin_flag1) - %addr(Buffer) resulted in -7.  This suggests that with
> my current release, PTF level, etc. that storage has been laid out as
> admin_flag0, admin_flag1, filler, buffer.   You cannot however rely on this
> -- a PTF and recompile of your program, or recompiling on a different
> release, could alter all of this...

That's good news.  So, in my opinion, the likelihood that an attacker
could hit an exact byte in RPG static storage is getting extremely low -
The buffer has to be accessible via a pointer, unprotected by program
code, and the storage layout from the compiler is arbitrary, so the
important 'admin_flag' is not necessarily accessible past the end of the
buffer.

Thanks for your insight.
--
Regards,
Rich

Current Conditions in Des Moines, IA
Broken Clouds
Temp 42.8F
Winds out of the Southwest at 9mph