Subject: Re: Wow! BIG jump, but now different problems...
To: Bruce Lane <kyrrin@bluefeathertech.com>
From: Chuck McManis <cmcmanis@mcmanis.com>
List: port-vax
Date: 03/27/1999 13:25:23
Wow, a question I know the answer too! One step up on the clue chain...

At 12:46 AM 3/27/99 -0800, Bruce Lane wrote:
>	1). At one point in the boot process, it says 'Boot device: <unknown>' and
>prompts me for 'Root device:' I have been responding 'qe0' sans quotes
>since the root is located on a Linux box.

Correct, the correct answer is qe0. This will continue until you have the
correct version of boot in your root directory with your kernel. Basically
the kernel didn't have the root device set so it had to ask.

>	2). The next thing it prompts me for is 'Dump device:' I've been telling
>it 'None.' Is this right?

Yup. Same deal. Just hit return.

>	3). Here's the kicker. It then asks me for 'file system' to which I've
>tried responding with both 'generic' (the default) and 'nfs.' The result is
>the same: It gets to this point...
>
>	exec /sbin/init: error 8
>	init not found
>	Panic: no init
>	(it then says 'Stopped at' and gives a hex address)
>	db>

This means that you have a post 1.3.2 kernel trying to boot 1.3.2 userland.
The error '8' is EXEC FORMAT ERROR because the new init uses a different
binary format than the snapshot kernels. You MUST use the snapshot userland
with kernels built on -current. Its a pain in the back side but it is
necessary.

>	As a final note: Once I get netbooting straightened out once and for all,
>is there a FAQ to do a local installation on a netbooted workstation? I'd
>rather not have to depend on the Linux box being a bootserver its entire life.

There are lots of things sprinkled around in various FAQs, not particularly
well organized but the information is there. Also you can "steal"
information from the other BSD projects since they are fundamentally quite
similar. 

Now depending on your disk devices you may run into the same problem I'm in
which is that the MSCP driver has a bug where it misses a handshake between
the controller and the CPU sits there waiting for a message and the
controller sits there waiting for a message queue to empty. I've made a few
changes to the mscp driver but I'm still not able to get to a point where I
can build a kernel yet, I did a 'make build' on the 1.3.3 sources and got a
build directory full of includes and the cleandir target ran as well but
the build target didn't (the make just stopped) so I'm guessing I need
either some environment vars set or something else. Eventually.

--Chuck