omsa5 segfault

[Greg Dickie]

Unfortunately have not had a chance to test it out yet. I will let you
know the results.

Greg

On Fri, 2006-09-01 at 09:49 -0500, Matt Domsch wrote:
> On Mon, Aug 28, 2006 at 09:29:09AM -0500, Matt Domsch wrote:
> > On Mon, Aug 28, 2006 at 09:40:10AM -0400, Greg Dickie wrote:
> > 
> > > > If that's actually causing a noticable slowdown in applications,
> > > > that's not great.  In mainline kernels we made one modification which
> > > > hasn't been rolled back into the OMSA releases which replaces a call
> > > > in the driver from udelay(1) to schedule(), such that when it is busy
> > > > waiting (at a very low priority), other tasks get a chance to run if
> > > > they want.  This slows down the IPMI driver, such that it takes longer
> > > > to poll sensors, perform firmware updates, and the like, by a bit, but
> > > > makes other apps more responsive while the IPMI controller is busy.
> > >
> > > We are also seeing some behavior that could easily be explained by the
> > > scenario you outline below. Would it be possible to at least make a
> > > patch available to fix that?
> > 
> > Sure.  cd into your /usr/src/openipmi-* directory appropriate for your
> > distro/version, and do:
> > 
> > # sed -e 's:udelay(1):schedule():' *
> 
> Did this help?
> 
-- 
Greg Dickie
just a guy
Maximum Throughput