Subject: Re: Complete(ish) binary snapshot, 20000714
To: Leo Smiers <L.Smiers@cable.A2000.nl>
From: Ben Harris <bjh21@netbsd.org>
List: port-arm26
Date: 07/18/2000 08:38:18
On Mon, 17 Jul 2000, Leo Smiers wrote:

> I have downloaded all the sets and tried this on my A310. Did you now that
> that the arm26 base set is around 32 MB's larger than the arm32 base set? 

That's not terribly surprising.  NetBSD/arm26 doesn't currently do shared
libraries, while NetBSD/arm32 does.

> It seems that each time I boot into netbsd is different. At least the
> register values are different (including R15) and sometimes I get a the
> repeating fault: can't handle fault. This lines keeps repeating.
> I also get panic: prefetch abort in kernel mode.

Hmm.  I need to remember quite what all these messages mean.  In the
meantime, since I accidentally compiled the kernel with UVM history
logging turned on, I'd be interested to see the last screenful or so of
output from typing "call uvm_host(3)" at the debugger prompt after an
error.

For what it's worth, I've been seeing similar behaviour on my test
machine, but usually only after the system has booted a fair way and
started needing to swap.  Of course, your machine has rather less RAM than
mine, so it'll start trying to swap sooner.  It may be that I need a
slimmer kernel for smaller machines.

> If the kdb starts I get the following output of the db> ps command:
> PID	PPID	PGRP	UID	S	FLAGS	COMMAND		WAIT
>  8	0	0	0	3	0X20204	ioflush		syncer
>  7	0	0	0	3	0X20204	reaper		reaper
>  6	0	0	0	3	0X20204	pagedeamon	deamon_
>  5	0	0	0	3	0X20284	nfsio		nfsidl
>  4	0	0	0	3	0X20284	nfsio		nfsidl
>  3	0	0	0	3	0X20284	nfsio		nfsidl
>  2	0	0	0	3	0X20284	nfsio		nfsidl
> >1	0	0	0	7	0X4	init
>  0	-1	0	0	3	0X20204	swapper		schedul
> This seems consistent for all errors (address exception and prefetch)

Right, so it's always dying at the same point.

> Last time you asked me if I was booting in single user mode. How can I
> force a single user boot (something like bbbb -s ?).

That's correct.

> I do think however that
> this won't help because the init process hasn't procedeed that far. 

I'm inclined to agree.
 
> If you want more information please let me know.

I'm busy giving a Linux (spit!) sysadmin course at the moment.  I'll try
to look at the problem more seriously at the weekend.

-- 
Ben Harris                                                   <bjh21@netbsd.org>
Portmaster, NetBSD/arm26               <URL:http://www.netbsd.org/Ports/arm26/>