Re: 2.6.15-mm1 - locks solid when starting KDE (EDAC errors)

[Dave Jones <davej@redhat.com>]

On Fri, Jan 06, 2006 at 04:40:12PM -0800, Andrew Morton wrote:
 > Dave Jones <davej@redhat.com> wrote:
 > >
 > > On Sat, Jan 07, 2006 at 01:25:22AM +0100, Jesper Juhl wrote:
 > >  > On 1/6/06, Andrew Morton <akpm@osdl.org> wrote:
 > >  > > Jesper Juhl <jesper.juhl@gmail.com> wrote:
 > >  > >
 > >  > > >  Reverted that one patch, then rebuild/reinstalled the kernel
 > >  > > >  (with the same .config) and booted it - no change. It still locks up
 > >  > > >  in the exact same spot.
 > >  > > >  X starts & runs fine (sort of) since I can play around at the kdm
 > >  > > >  login screen all I want, it's only once I actually login and KDE
 > >  > > >  proper starts that it locks up.
 > >  > >
 > >  > > Oh bugger.  No serial console/netconsole or such?
 > >  > >
 > >  > > Or are you able log in and then quickly do the alt-ctrl-F1 thing, see if we
 > >  > > get an oops?
 > >  > >
 > >  > I switched to tty1 right after logging in, and after a few seconds
 > >  > (corresponding pretty well with the time it takes to hit the same spot
 > >  > where it crashed all previous times) I got a lot of nice crash info
 > >  > scrolling by.
 > >  > Actually a *lot* scrolled by, a rough guestimate says some 4-6 (maybe
 > >  > more) screens scrolled by, and since the box locks up solid I couldn't
 > >  > scroll up to get at the initial parts :(  So all I have for you is the
 > >  > final block - hand copied from the screen using pen and paper
 > >  > ...
 > >  > It never makes it to the logs, and as mentioned previously I don't
 > >  > have another machine to capture on via netconsole or serial, so if you
 > >  > have any good ideas as to how to capture it all, then I'm all ears.
 > > 
 > > If only someone did a patch to pause the text output after the first oops..
 > > 
 > > Oh wait! Someone did!
 > > 
 > 
 > umm, it'd be more helpful if you'd actually sent the patch so Jesper could
 > apply it so we can find this bug.
 > 
 > I think I did one of those too.  It required a new kernel boot option
 > `halt-after-oops' or some such.  Sounds like a good idea?

I thought Jesper had made a comment in that thread, so was aware of it.
Looking at the archive, I see was mistaken.

Jesper: http://lkml.org/lkml/2006/1/4/534
(unmunged diff is at http://lkml.org/lkml/diff/2006/1/4/534/1)

		Dave