kernel log buffer size (also default HZ)...

Tim Gardner tim.gardner at canonical.com
Mon Sep 28 15:39:24 BST 2009


Daniel J Blueman wrote:
> Hi guys,
> 
> CONFIG_LOG_BUF_SHIFT=17 is too small when debugging certain types of
> kernel issue. I've raised a report at:
> https://bugs.launchpad.net/ubuntu/+source/linux-meta/+bug/424810
> 
> Since it's a zero-risk and lean change, how much chance is there to
> get it in Karmic, in order to make bug-catching easier, as it's easy
> to stumble on this?
> 

See
https://bugs.edge.launchpad.net/ubuntu/karmic/+source/linux/+bug/424810/comments/2

> Separately, upstream moved to using HZ=100 on x86(-64) (see benchmark
> improvement at http://lwn.net/Articles/331607/), due to all
> performance-affecting timers being asynchronous (via hrtimers and
> tickless). I raised the discussion at:
> 
> http://groups.google.com/group/linux.kernel/browse_thread/thread/e6b370e71ddfe153/f164d6068360b034?#f164d6068360b034
> 
> Is thus seems reasonable to bring the benefit to Ubuntu, keeping it
> more competitive with other distros/OSs. Is this possible?
> 

I've actually been considering this for amd64 because you're pretty much
guaranteed to have an HPET. I think I would just as soon leave i386 and
lpia as they are since many of those platforms either have buggy HRES
timer sources, or have none at all. I'll have to rely on other folks to
tell me about the ports architectures (sparc, ia64, powerpc).

Please start a tracking bug for this issue since we are in beta freeze.

rtg
-- 
Tim Gardner tim.gardner at canonical.com



More information about the kernel-team mailing list