Re: [TEST] PREEMPT on PPC

[Sven Luther <sven.luther@wanadoo.fr>]

On Thu, Mar 25, 2004 at 05:47:20PM +1100, Benjamin Herrenschmidt wrote:
> Hi !
> 
> I've start looking into the various issues of running
> CONFIG_PREEMPT on ppc32. I found some problems, there
> may be more, but here's a first patch that should apply on a
> recent 2.6.x.

BTW, on non-pmac powerpc hardware, using PREEMPT with a cpu-driven sound
driver (builtin via sound for example), result in sound problems, which
looses the continuity of the sound. The same happened on my x86 box too
last time i tested the preempt stuff, but it was some time ago.

So, my question is, is this a bug, or something inherent in the way the
preempt patch works, and there is no solution ? 

I understand that the preempt patch allows the kernel drivers to be
preempted, thus resulting in the problematic behavior. Would it be
possible to make it skip some drivers, or something such ?

Friendly,

Sven Luther