Re: Catching NForce2 lockup with NMI watchdog

From: Ian Kumlien
Date: Tue Dec 09 2003 - 18:14:15 EST


On Tue, 2003-12-09 at 23:04, Craig Bradney wrote:
> On Tue, 2003-12-09 at 19:12, Ian Kumlien wrote:
> > Bob wrote:
> > > Using a patch that fixes a number of people's nforce2
> > > lockups while enabling io-apic edge timer, I can now
> > > use nmi_watchdog=2 but not =1
> >
> > Why regurgitate patches that are outdated, Personally i find int
> > outdated after Ross made his patches available and they DO enable
> > nmi_watchdog=1. (I have seen the old patches mentioned more than once,
> > if something better comes along, please move to that instead.)
> >
> > http://marc.theaimsgroup.com/?l=linux-kernel&m=107080280512734&w=2
> >
> > Anyways, Is there anyway to detect if the cpu is "disconnected" or, is
> > there anyway to see when the kernel sends it's halts that triggers the
> > disconnect? (or is it automagic?)
> >
> > If there was a way to check, then thats all thats needed, all delays can
> > be removed and the code can be more generalized.
> >
> > (Since doubt that this is apic torment. It's more apic trying to talk to
> > a disconnected cpu... (which both approaches hints at imho))
>
> Have these patches been submitted for review for inclusion into the main
> kernel?

No, there is no final patch in anyway, there are just dodgy workarounds.
I just deem this better with working nmi_watchdog=1

> I'm still running the old IO-APIC patch (Uptime 3d 20h) and having no
> issues whatsoever.

They fix the same problem..

> Are all of the patches at that address you provide necessary?

nope, but they are all nforce2 related.

> What do the IDE ones claim to fix? I have had no real issue with IDE at
> all.. being able to burn CDs, DVDs, use my ATA133 drive for hdparm,
> greps, compilation, and general use.....

it's just a cleanup afair.

Anyways, I think that if we find someway to detect cpu disconnect, then
we just need that "detection" prior to the apic ack...
(just a guess though)

--
Ian Kumlien <pomac () vapor ! com> -- http://pomac.netswarm.net

Attachment: signature.asc
Description: This is a digitally signed message part