Re: [LAU] Any word on an rt kernel for Fedora 9?

On Mon, 2008-10-27 at 09:56 +0000, Rui Nuno Capela wrote:
> Cassiel wrote:
> > Hi Fernando,
> >
> > I confirm midi is working with 2.6.26-8 on debian lenny but jackd
> > freezes trying to stop when with snd_seq. (DLs state in the output of
> > `ps -ax`).
> >
> > I didn't test the other alternatives (not a midi fan... simple user) but
> > I am going to and feedback to this thread.
> >
>
> (am i the only one to be aware of this?)
>
> midi is seriously broken on 2.6.26-rt, specially when driven outboard;
> events delivered out of time if at all, outrageous jitter and
> unpredictable drop outs;

I don't get any messages at all. Well, I do get _one_ and that's it. And
the software that uses MIDI (kmidimonitor) is left is a bad hanging
unkillable state. I tried several kernel options to try to get my
realtime kernel to be as close as the Fedora kernel (which works) but no
results so far.

The timing problems might be related to the new RTC interface in newer
kernels. Apparently ALSA cannot/does not use it, and that might lead to
bad timing. As MIDI does not work I could not test for timing but maybe
somebody else can try to use CONFIG_RTC instead of CONFIG_RTC_CLASS, if
you use the former then ALSA can use CONFIG_SND_RTCTIMER... But this is
just a guess on my part.

-- Fernando

> internally sequenced material (eg. queueing
> events to a softsynth) might work barely but that is just little comfort.
>
> note that this badness applies to PREEMPT_RT kernels >= 2.6.26, vanilla
> is fine afaict.
>
> all that trouble seems to be related to a regression in 2.6.26-rc2 that
> just made the bkl non preemptible anymore, leaving the -rt people with a
> herculean task to comply--look after Ingo Molnar's "Kill the BKL" tree
> for news...

_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
Received on Wed Oct 29 00:15:01 2008