Re: kill -9 <pid of X>

Jon M. Taylor (taylorj@ecs.csus.edu)
Wed, 12 Aug 1998 14:15:48 -0700 (PDT)


On Wed, 12 Aug 1998, MOLNAR Ingo wrote:

>
> On Wed, 12 Aug 1998, Jon M. Taylor wrote:
>
> > > The X server can avoid even kill -9 from a normal user. As it stands,
> > > it doesn't do that, and I personally think it's a bug in X.
> >
> > I agree.
>
> [...]
>
> > I stand by my claim. If userspace cannot, by its nature, properly
> > guarantee the atomicity of those critical sections, video card programming
> > cannot be done properly in userspace. Thus, it *is* the kernel's job.
>
> now do you aggree or do you stand by your claim? You cannot have it both
> ways ...

Why not? The fact that a non-root process cannot kill -9 a root
process is in no way related to the ability of userspace code to guarantee
atomicity of critical sections.

> as explained, the KILL signal _is_ being handled and can be masked if
> another user sends it. The problem with the current X server is that it
> runs under the uid it was started.

I agree.

> there _are_ things that are kludgy with the current X server. But it's not
> at all that bad as you try to make it up,

It isn't "bad" at all. It is fast and even the crashes that do
occur don't happen all that often. It _is_ suboptimal.

> and the current X server solves
> alot of much tougher problems, in a nightmarish environment (PC video
> hardware that is).

It doesn't have to be nightmarish if all the horrible brokenness
of current video hardware is hidden behind a nice clean driver API.

> Mode switching just takes up some 2-10% complexity of a
> video driver, and is used in 0.0001% of the time. Please come up with a
> working and superior alternative instead of denouncing the present one ...

We (GGI) have. The fbcon people also have. The only missing
ingredient is acceleration, and that is on the way.

> Just remember the GGI flamewar.

How can I ever forget it (them)?

> Many many words and assumptions and bad
> feelings, then came the FBCON code silently, it was clean and people
> wanted it, so it went into the kernel.

Let's not gild the lily here. fbcon was merged when the m68k port
was merged. It *had* to be merged, because the m68k people had been using
it heavily for years. It was also compatible with the existing Linux
console and KGI's console wasn't. fbcon is a nice, clean design but it
is not KGI. It isn't supposed to be.

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html