On Thu, Dec 29, 2005 at 09:45:27PM -0600, Peter Seebach wrote:
> In message <20051229200715.GA1084@antioche.eu.org>, Manuel Bouyer writes:
> >I'm seeing these on occasions too, on various systems but all with the
> >same CPU/motherboard so I suspect hardware issue. My workaround is
> >a patch which retries the IPI if it failed (with a counter to not look up if
> >it repeteadly fails)
> 
> >Sorry I can't get this patch before next week ... you should be able to
> >find it in one of my posts to tech-smp or port-i386 though.
> 
> Oh, fascinating; just retrying seems to work?  Encouraging, only not.

Yes, and as I did't see other reports of this issue I though about a
hardware issue (some interrupts being missed, or cache coherency not always
right). Now, if this can show up on other hardware too, it's more likely
a race condition in the kernel.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--