Subject: Re: Isn't "db>" on boot bad? :P
To: leam <>
From: Andrey Petrov <>
List: port-sparc64
Date: 02/11/2003 10:59:53
On Mon, Feb 10, 2003 at 09:10:14PM -0500, leam wrote:
> I'm working to dump and restore from one disk to another, and the system 
> won't boot from the second disk. Particulars:
>    NetBSD 1.6 (Sparc64), generic re-compiled kernel
>    root, /usr, and /var were dump'd and restore'd with:
> 	dump 0uf - /dev/rsd0a | ( cd /mnt; restore rvf - )
>         { where a is "/", and /dev/rsd1a is mounted on /mnt}
> After I did those three (/export is the only other filesystem) I:
>   /usr/sbin/installboot -m sparc64 -t ffs /dev/rsd1a /usr/mdec/bootblk 
> primary
> When it tried to boot it was locked in a loop saying:
>   trap: textfault at 109ffc!! sending SIGILL due to trap at 16: illegal 
> instruction
> After either "stop-a" or "^c" (it is late for me and near my nappy time) 
> I got:
>   kdb breakpoint at 12c52ac
>   stopped in pid 1 (init) at cpu_Debugger+0x4: nop
>   db>
> The machine is currently at the "db>" prompt.
> My priority on this is "educational"; that is, the machines sole-purpose 
> is to let me learn NetBSD, so if it stays there for a bit that's fine. 
> If you have a solution, I'd appreciate it. If you just point me to a 
> particular section of the documentation i've either missed or 
> misunderstood, that is just as fine.
Looks like init got corrupted.

What machine is that? if you have ultrasparc2 (not 2e or 2i) processor
then it's quite possible to be iommu bug which was fixed in -current and
the bug actually manifested on dump|restore operation. You can try
pax instead of dump. And you can try to boot into single-user and
compare init with original.