Subject: Re: Is _any_body using NetBSD 1.6+ on a series 400 machine with
To: None <bsieker@freenet.de>
From: Havard Eidnes <he@netbsd.org>
List: port-hp300
Date: 07/27/2003 23:40:33
Hi,

> I tried to install -current today on a HP9000 425s, and it still
> has the same problem as it did with 1.6: When it tries to initialise
> the framebuffer console, immediately after it prints a message
> about "Entrypoint" it fails.
>
> But, whereas in 1.6 it stuck solid, and I had to do a hardware
> reset, -current resets the machine and tries to reboot.

I can't offer much of direct assitance at present, but I've done some
experimentation to try to narrow down the source of the problem.

I've been using uboot.lif from 1.6.1 as SYS_UBOOT (i.e. I net-boot my
40MB 425t with Catseye graphics, which normally runs, uhm...  1.4.3).

It appears that if you try to use a serial console, the boot code has
no problems loading the newer kernels.  I've not experimented
sufficiently to find out what I'd ned to do to get a login on the
graphics terminal if I use a serial console.

What I have done is that I've cross-compiled a number of hp300 kernels
of various vintage from between the branch points for 1.5 and 1.6.
Somewhere in the middle there, we made the transition to ELF, and the
newish boot code is required to boot the ELF kernels.

Mostly what happens when I boot the ELF kernels with the graphics
display as the console, is that instead of a reset or hang, I get the
following messages from the boot loader (instead of it starting the
kernel):

Start @ fd803400 [1=3D0xfd991508-0x1df1a0]...
Entry point: 0xfd803400
Got unexpected trap: format=3Df vector=3Dd94 ps=3D2704 pc=3D37737008
dregs: 1f4000 fd9e2000 c fd802000 85 1e0000 a0000006 0
aregs: fd945b04 fd005788 fffff000 fb00581c fd9e2000 fd802000 ffffef70 f=
d945a9c

This is from the July 1st 2001 kernel.  I got basically the same
message when trying to boot the other ELF kernels on the graphics
console, ranging from 1 Jul 2000, 1.5B (had to tweak to compile) to 15
Dec 2001, 1.5Z.

However, the 1 Jul 2001 1.5W kernel compiled natively on a hp300
running 1.5.3, i.e. so that the kernel ends up being a.out .... boots
up Just Fine on the graphics console!

I suspect something is Wrong in the boot loader wrt. the handling of
the graphics console, i.e. this is probably not a kernel problem, but
that's as far as I've managed to narrow it down so far.

Regards,

- H=E5vard