Hi,

Thanks for the response.

I've been trying without the l2 cache for the moment just to remove one
variable.  The L3 cache is out of the motherboard.

I'm not 100% sure how to disable speculative processing.  What I did do
though is change the code in powerpc/oei/cpu_subr.c that looks like:

 case MPC7400:
        case MPC7410:
                hid0 &= ~HID0_SPD;


so that HIDO_SPD is enabled rather than disabled.  My reading of the 7400 
cpu manual is that this disables speculative processing.  Of course, my
reading could be completely wrong.

What this does is get beyond something that boots to something that stays
up a while.

What it does not do is run very well.  If I start either a lot of network
traffic and/or disk traffic (say untar a file across nfs) we will 
hang and/or crash.

Right now I'm waiting for a current userland and kernel to finish building
with -mcpu=7400 -O3 and we'll see if that helps.

If no I'll probably just give up.  200mhz 604e that stays up and part
of the cluster beats a 350mhz g4 that has to keep rebooting :-(

Thanks again!

cheers

bruce

On Sat, Jun 25, 2005 at 04:57:09PM -0400, Chris Tribo wrote:
> 
> Is this one of those situations where speculative processing needs to be
> disabled? Any chance your L2 cache size/type/frequency multiplier are
> wrong? Did you remove the L3 (if any) from the motherboard?
> 
> On Sat, 25 Jun 2005, Bruce O'Neel wrote:
> 
> > Hi,
> >
> > I've recently gotten a Sonnet G4 /350 PCI Upgrade card for my 7300 and I am
> > not having the greatest luck getting it to work.
> >
> > It boots, and runs for a while, but, once I load it down (say start a
> >  build.sh and untar a file from a nfs disk) it crashes.  It never seems to
> > crash with the same result and things like the compiler start segfaulting
> > and gzip complains about bad CRCs.
> >
> > It could be bad memory, I guess, but I've swapped the memory in and out
> > and it all seems ok.  I don't have MacOS so I didn't ever load the
> > Sonnet install diskettes, that might be the problem.  I also thought
> > maybe it might be network traffic but that turns out not to be the case.
> >
> > I'm running current built with -mtune=7400 (both kernel and userland)
> > and I also have a kernel built -mcpu=7400 with the L2 cache turned
> > on.
> >
> > Am I missing something obvious, or, is this just a case where sometimes
> > a PCI upgrade just doesn't work.
> >
> > Thanks very any ideas in advance!
> >
> > cheers
> >
> > bruce
> >
> >
> 
> -- 
> The nice thing about Windows is - It does not just crash, it displays a
> dialog box and lets you press 'OK' first.
> 						-- Arno Schaefer

-- 
edoneel@sdf.lonestar.org
SDF Public Access UNIX System - http://sdf.lonestar.org