[LAU] updates to real-time kernel

[Fernando Lopez-Lezcano]

On Wed, 2008-07-23 at 17:34 +0200, Robin Gareus wrote:
> Esben Stien wrote:
> > Susan Cragin <susancragin at earthlink.net> writes:
> > 
> >> It's stuck at 2.6.24-19, and the generic kernel is up to 2.6.26-4
> > 
> > http://kerneltrap.org/Linux/Removing_the_Big_Kernel_Lock
> > 
> > It seems like we're stuck with 2.6.25. (I'm stuck with 2.6.23 cause of
> > some weird audio problems with xmms2 on 2.6.24 and 2.6.25). 
> > 
> > Anyone know what the foreseeable future holds?
> 
> 2.6.24 was very stable for audio, at least over here more so than
> 2.6.25!

Same here in my experience. I have not tried 2.6.26 yet. 

> I've got a 2.6.26-rt1 working from
> git://git.kernel.org/pub/scm/linux/kernel/git/clrkwllms/rt-linux.git
> 
> It supports Rui's rtirq setup and jackd/USB-UA25 ran >5 mins without
> xruns at 128frames/period, wifi working, it suspends to RAM and
> everything. BUT I did not yet try any longer since it spits out the
> following message once per second:
> 
>  =======================
> BUG: sleeping function called from invalid context swapper(0) at
> kernel/rtmutex.c:743
> in_atomic():1 [00000001], irqs_disabled():0
> Pid: 0, comm: swapper Not tainted 2.6.26-rt1 #2
>  [<c01250cb>] __might_sleep+0xeb/0xf0
>  [<c034199f>] __rt_spin_lock+0x5f/0x70
>  [<c03419b8>] rt_spin_lock+0x8/0x10
>  [<c01498d0>] pm_qos_requirement+0x10/0x30
>  [<f886afc9>] acpi_processor_idle+0x407/0x530 [processor]
>  [<f886abc2>] ? acpi_processor_idle+0x0/0x530 [processor]
>  [<c0102cf8>] cpu_idle+0x68/0xf0
>  [<c033253c>] rest_init+0x6c/0x70
>  =======================
> 
> It does not seem to be a real Bug.. but it's an annoying /mosquito/. 

I asked Ingo about this and never heard back from him. 
This patch stops the mosquito noise for me :-)

  http://lkml.org/lkml/2008/6/8/59

But as there are no comments from the gurus it is possibly a bad idea
(although it has not been bad in my case, AFAICT)...

-- Fernando

> I guess it's related to NO_HZ or some ACPI_XXX config. I could not get the
> -tip branch to boot at all, but I did not try too hard just yet.
> 
> stay tuned,