-current and sparc SMP

[Hauke Fath <hauke%Espresso.Rhein-Neckar.DE@localhost>]

Hi,

I just gave a -current snapshot a try on my ss20 (2x SM71) - the boot logs
are here
<http://la.causeuse.org/hauke/NetBSD/sparc-smp/2010-06-18_5_99_30_SMP.txt.gz>.
The second kernel is a supposedly-UP config:

<snip>
# $Id: PIZZA_UP_PF,v 1.2 2010/06/11 12:15:52 hf Exp $

# pizza.causeuse.org single-processor (UP) kernel
# with pf(4) support in place of ipfilter.

include "arch/sparc/conf/PIZZA"

## Disable multi-processor support

no options      MULTIPROCESSOR          # include multiprocessor support

## disable inlining __cpu_simple_lock()
## the locore version attempts to detect stuck locks
## not inlining saves about 60K on the kernel's text
no options      __CPU_SIMPLE_LOCK_NOINLINE

no cpu*         at mainbus0             # declare additional CPUs

## Choose pf over ipfilter as packet filter

no pseudo-device        ipfilter

pseudo-device   pf                      # PF packet filter
pseudo-device   pflog                   # PF log if
</snip>

which nevertheless died in about the same multi-user boot phase as the MP
kernel. So, unless I actually have to physically remove the second cpu (do
I?), there may be more issues than just the sparc MP problem.

I haven't even tried netbsd-5, figuring that any sparc relevant MP code
improvements would appear on HEAD.

How about your SMP experiences, plans? Anything I could try with the
-current kernels while they're hot?

        hauke



--
"It's never straight up and down"     (DEVO)