Subject: Re: SCSI support
To: None <port-next68k@netbsd.org>
From: Goran Koruga <goran.koruga@hermes.si>
List: port-next68k
Date: 05/31/2002 09:52:37
On Thu, 30 May 2002 20:31:42 +0200, Christian Limpach wrote:

Hi.

Thanks for your reply.

> Hi!
> 
> > I have tried the snapshot from netbsd.org in NetBSD-daily dir and
> > your SCSI kernel with console support. It worked for a while, then I
> > get the following error :
> >
> > xe0: out of receive DMA buffers
> 
> hmm, I hope this is unfortunate :-(  By design this panic indicates
> that the machine was not handling incoming packets fast enough and
> that the queue had filled up.  I've only had this happen once or
> twice.  In this case, I think there is some condition which somehow
> shuts down network receiving.  One reason might be that too much code
> runs at too high interrupt levels and this locks up or that some
> structures are not properly protected from being updated from
> interrupt handlers.
> 
> Does this happen all the time?  Is your network busy?  Was the machine
> doing anything at the time?

It happens each time I try to use it. It's up for several minutes and
then the above happens. Machine itself wasn't doing anything, it was
waiting for my input since I've run disklabel -e -I sd0. As for the
network, yes it's rather busy. Every now and then I also see kernel
messages about some ARP issues (I think it's about mismatched addresses,
but I didn't pay attention to it). Perhaps it would help if I try to
hook it up to a switch where it would see less traffic ? Note that I
didn't see such issues when I was using kernel w/o SCSI support (I think
it was 1.5.1) and doing FTP transfers and such. But I guess this doesn't
mean much since only network card was active at the time.

> > Is this specifically for booting from disk ?
> 
> no.  When I tried current a couple of weeks ago I couldn't boot any
> new kernel at all since my bootloader was very old and didn't support
> ELF.  I thought that other people might have the same problem since
> there were some things broken in current for a couple of months which
> prevented new kernels from booting at all (like the trampoline jump in
> locore.s which got optimized to a pc-relative branch with a recent
> assembler upgrade...)

Oh OK, I guess I don't need this since it boots for me.

Thanks,
    Goran

-- 

Writing about music is like dancing about architecture.
    -- Frank Zappa