Re: 2.1.30 dies

Ingo Molnar (mingo@pc5829.hil.siemens.at)
Wed, 2 Apr 1997 13:06:25 +0200 (MET DST)


On Wed, 2 Apr 1997, Jochen Buehler wrote:

> after an upgrade from 2.1.29 (uptime ~20 days) to 2.1.30
> I noticed a slight overall slowdown of my system.
> After reading here that the SMP changes might affect
> performance of UP machines, I recompiled the 2.1.30
> kernel without SMP=1 defined in the toplevel
> Makefile.
> Unfortunately the resulting kernel does not seem to
> be very stable. My machine crashed twice after 5 minutes
> uptime (no load). It's just sitting there, not responding
> to pings or keyboard, screen frozen (X11). Only a hard reset
> can bring it back to life again. No messages logged anywhere.
> Needless to say, the kernel with SMP=1 runs fine...
>
> Anything I can do to see whats going on here ?

You could try installing the 'ktrace patch' i have posted a few days ago,
(check out http://www.ecsnet.com/ inofficial 2.1 patches section) and turn
'kernel debugging' and 'deadlock detection' on in 'Kernel Hacking'
options. That one is supposed to produce a nice oops instead of locking
up.

i'm running 2.1.30 kernels in all possible and impossible combinations,
and didnt have any lockups.

-- mingo