[PlanetCCRMA] kernel 2.6.14 and Alsa 1.0.10 for FC3

Fernando Lopez-Lezcano nando@ccrma.Stanford.EDU
Fri Nov 18 18:51:01 2005

On Sat, 2005-11-19 at 00:57 +0000, Michele Spinolo wrote:
> I noticed these two nice updates are out, but are not available for FC3 
> PlanetCCRMA@Home, which is still based on 2.6.10 kernel and Alsa 1.0.9 
> modules.
> Fernando, will we have an update?:-)

I have been working for quite a while on a 2.6.14 based kernel with Ingo
Molnar's realtime preemption patches (this time configured with the more
risky PREEMPT_RT option) - and before that with 2.6.13+. 

At this point I have reached a dead end, there is a problem with Jack
and the current dual core machines I have here at CCRMA. The [Mark's
answer just got here, I'll keep writing anyway :-] problem manifests
itself through warnings in Jack that do not translate into xruns, but
should not happen anyway[*]. 

There is no solution at this point, I'll have to post to jack-devel to
see what the gurus there think. 

I'm running 2.6.14-0.7.rrt.rhfc4.ccrma and alsa 1.0.10 right now, but
I'm not sure if I want to mess everybody else's machines as well :-)

-- Fernando

[*] long answer: each cpu has its own timer and apparently they can
drift from each other as time goes by... Jack hops from processor to
processor depending on how things are scheduled, if it is in the same
processor from which it originally got the initial "time reference" all
is fine, if it is on the other processor eventually the time measurement
goes wrong and you get warnings that have no basis - lots of them. One
suggested "solution" is to pin jackd to one of the processors but that
is just a hack IMHO. Another would be to use a different timing source,
there is one but that one is more expensive in processor cycles.
Probably the only acceptable solution I've heard so far...