[linux-audio-user] Realtime kernel does not boot

[Florian Schmidt]

On Friday 22 September 2006 15:43, Mark Knecht wrote:
> On 9/22/06, Florian Schmidt <mista.tapas at gmx.net> wrote:
> <SNIP>
>
> > Well, if you use a closed source driver, there's no way for Ingo to fix
> > it. Period. That's why noone will look at it. The place to report kernel
> > troubles to, when using closed source drivers, is the provider of these
> > kernel drivers. In this case ATI.
> >
> > Regards,
> > Flo
>
>    This conversation is just silly and a waste of time. 

No, it's not. There's an important point to be made.

>    Just because 
> an ATI driver is on the machine doesn't mean you don't report a
> problem with the with the kernel to Ingo. 

It exactly does mean that, because he will ignore the bug report for good 
reasons.

> What about when I boot the 
> machine, the kernel has a problem at boot time and I'm not even in X?
> What about when the ATI driver isn't being loaded and the machine is
> in text mode? What about when I *do* have the ATI driver loaded and
> the problem is clearly kernel related? This conversation is just silly
> and a waste of time.

Well, if the ATI driver isn't loaded and the kernel shows a bug, then report 
it.

>    As I stated earlier, I do not report problems to Ingo that are ATI
> driver related. I did not report this specific problem (machine
> crashed when exiting X) to him since it was possibly ATI related. I
> reported other problems. Please read the thread.

The problem with closed source drivers is:

There's no process seperation between different parts of the kernel. Any part 
of the kernel can write memory anywhere else in the kernel. If i wanted to, i 
could write a kernel module that simply overwrites the whole kernel in RAM 
with 0's. 

If you have a closed source kernel module it could be a bad one.  It could do 
something really bad and there could be bugs showing up that _look_ like they 
are caused by some other part of the kernel, but really they are produced by 
the closed source driver. As there's no way to find out, as the driver is 
closed source, it is a simple manner of efficiency to simply rule this out, 
by simply not caring about bug reports with closed source drivers loaded.

So, if you think the kernel has a problem, run your setup without the closed 
source driver. If the problem can be reproduced -> Great, you found a bug 
that can be fixed. If it cannot be reproduced -> too bad, it seems your 
closed source driver might be faulty -> report to the closed source driver 
author.

I hope this clears the issue up a bit..

Regards,
Flo

P.S.: using closed source drivers with -rt kernels is asking for trouble 
anyways, as the -rt model relies on rewriting parts of the kernel (by 
altering macros, inline functions, etc.). The closed source binary driver 
cannot be altered in the same way. Thus it is highly probable, that bad 
things might happen if running a closed source driver.

-- 
Palimm Palimm!
http://tapas.affenbande.org