Re: [MiNT] Cookie Jar problem

[Frank Naumann <fnaumann@mail.CS.Uni-Magdeburg.De>]

Hello!

> > > global area established by gluestik, and make it unsafe by definition this
> > > way, or integrate STIK emulation into kernel (I'd strongly dislike that).
> >
> > I don't know if you noticed but that's why gluestik is placed in global
> > memory.
>
> I did. The point is, as I understand, how to make the gluestik work with
> the non-global cookie jar, right? You proposed the Ssystem() expansion,
> but this would make the entire modification pointless (the point of making
> jar private is avoid a situation when a buggy or maliciously written
> program destroys the jar thus influencing all processes which would look
> it up; with Ssystem() modifying all the copies this would still be
> possible).

I don't aggree. Ssystem is a system call. If the system call destroy
something the kernel have for sure a bad bug. Every process have still
it's private cookie jar.

Btw. how do you managed private cookie jar? Do you exchange on every
context switch 0x5a0?

> If gluestik would be an XDD, it would be able to place its cookie into
> kernels cookie jar. This way the STiK cookie would become available for
> all processes, and no kludges in Ssystem() are needed ("wolf/ox"
> solution).

Oh, I don't think you know how much work this would be. At least it's not
so simple as you describe. As an xdd run in kernel space you must first
write lot of new interfaces or helper routines to push a wrapper table
into user space that can then cleanly wrap over the xdd.


Tschuess
   ...Frank

--
ATARI FALCON 040 // MILAN 060
-----------------------------------------
http://www.cs.uni-magdeburg.de/~fnaumann/
e-Mail: fnaumann@freemint.de