[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-smp
Subject:    Re: Stuck on TLB IPI
From:       Jorge Nerin <comandante () zaralinux ! com>
Date:       2000-10-16 18:14:55
[Download RAW message or body]

Dominique wrote:
> 
> Jorge Nerin wrote:
> 
>         Buenos dias!
> 
> > > > Send more info: kernel version, mobo, HDs (ide, scsi?), cpus (stepping),
> > > > irq list (are there shared irqs?),...
> > >
> > >         If it a non-smp kernel, everything is fine, but as soon as it is
> > > compiled in smp...  It appens with 2.2.14, 2.2.16 and 17.  The mb is an
> > > intergraph with an aic7850 built-in.  2 HD scsi-2 and a scsi cd.  I'll
> > > check the stepping tomorrow.  In Linux, is there a way I can get it, as
> > > well as how the kernel affects the irq's (I know, basic questions, but
> > > I'm a little new in linux.)
> > > >
> > > > Is there any Oops, or something you can pass througth ksymoops?
> > >
> > >         Again, where/how can I find that?
> >
> > Well, you can check the stepping by doing cat /proc/cpuinfo, and if
> > didn't know if you have suffered Oops, then probably you didn't have
> > suffered it, but also you may watch the logs (/var/log/messages) for
> > strings like Oops, BUG, Unable handle kernel NULL pointer,...
> 
>         Nothing, everything is fine on that side.
> 
>         family 6, model 1, stepping 7 (the other one is 6), CPU ID level 2.
> The mb allows me, if I want, to set the cpu's at 2 different speeds.

First make sure the two run at the same speed, I think there is partial
support for asymetric cpu speed by dissabling TSC, but I remmeber it has
its problems (I think).

Then you could also check in Intel's web site if the two steppings can
work together without problemes, somewhere in their website they have a
table of stepping that tells which can be mixed, which cannot and which
shouldn't.

I think that you have no problem and that some driver is causing this.
This message means that somewhere the interrupts have been disabled too
long, the loop used to measure this does a fixed count of loops, so if
the cpu speed is very fast the loop waits less. Also on this side the
PIII 500 seems to execute the loop much faster than other cpus, don't
know why.

Perhaps this could be caused by a very long write flushed to the disk?
Tell us what your /proc/interrupts looks like, just to see if there are
shared irq with the scsi interface.

> 
>                 Gracias,
> 
>                         Dominique

De nada

> >
> > --
> > Jorge Nerin
> > <comandante@zaralinux.com>
> 
>         ?Commandante de donde?

Es un apodo, de los tiempos de msdos del comandante norton, lo usaba
mucho.
It's a nick from the msdos times, from norton commander, I used it
heavily.

-- 
Jorge Nerin
<comandante@zaralinux.com>
-
To unsubscribe from this list: send the line "unsubscribe linux-smp" in
the body of a message to majordomo@vger.kernel.org

[prev in list] [next in list] [prev in thread] [next in thread]