Re: NetBSD 8.10 does not boot on 7043-140

[Ulrich Teichert <lists%ulrich-teichert.de@localhost>]

Hi T,

> my guess would be to turn on the DDB or KGDB option in the _COM version 
> of the kernel (compile) and try to debug it that way.
> http://www.netbsd.org/docs/kernel/kgdb.html

OK, but I like to figure it out how it should run before firing up a debugger.

> I have recently tried to compile with the KGDB kernel option enabled, 
> but it fails without a few workarounds (I will be making a separate post 
> on this shortly).

Sounds good, at least we are attacking the issue from different angles.

> For this _COM issue, it might be best (once KGDB is 
> fully working) to setup the 7043 to boot from network BOOTP/TFTP and use 
> a separate (more powerful) machine for KGDB REMOTE, cross-compiling, and 
> TFTP/BOOTP server. This should allow you to cross-compile the kernel on 
> the REMOTE machine and push it to the 7043 by simply copying the 
> cross-compiled kernel to the /tftp directory on the same machine. The 
> kernel source would also be on the same machine for KGDB to use, so it 
> seems like this is probably an ideal setup for a physical machine.

I am already cross-compiling and netbooting ;-)

> There's probably a way to do it through QEMU, but I don't really have 
> any experience with it.

I've tried with:

qemu-system-ppc -m size=128 -M prep -bios qprepofw-serial-svn-3738.rom -cpu 603 -cdrom soft/netbsd/prep/usr/src/distrib/prep/cdroms/installcd/obj/NetBSD-8.0-prep.iso -fda soft/netbsd/prep/usr/src/distrib/prep/cdroms/installcd/obj/cdrom/prep/installation/floppy/sysinst_com0.fs -boot order=adc -nographic

But the process just sits there after it spits out:

WARNING: Image format was not specified for 'soft/netbsd/prep/usr/src/distrib/prep/cdroms/installcd/obj/cdrom/prep/installation/floppy/sysinst_com0.fs' and probing guessed raw.
         Automatically detecting the format is dangerous for raw images, write operations on block 0 will be restricted.
         Specify the 'raw' format explicitly to remove the restrictions.
qemu-system-ppc: Machine type 'prep' is deprecated: use 40p machine type instead

Type 'i' to interrupt stand-init sequence
First stand-init:
Calibrate
CIF buffers
DHCP init
memory node
Instruction cache on
Decrementer
Enable machine check exceptions
Set Memory Map
MMU

... and seems to hang there. It eats a CPU until I kill it.

> I've only seen this problem with the _COM versions of the kernel, so 
> when you setup KGDB, you might want to have it use COM1 and have TERM on 
> COM0 to see what's going on.

I only have a serial console attached anyway. Attaching a second one
should be no problem.

So far, my feeling is that the interrupts are getting enabled before
some workarounds from cpu_startup() -> prep_init() are in place.
Still trying to figure out where they get enabled other than in
init_intr(). All this code is in sys/arch/prep/prep/machdep.c and
should get called from main, but it looks like we're never getting there
because of the interrupt. But it's still very murky to me...

CU,
Uli
-- 
Dipl. Inf. Ulrich Teichert|e-mail: Ulrich.Teichert%gmx.de@localhost | Listening to:
Stormweg 24               | Syndrome 81: Le Condor De La Rade, Rixe:
24539 Neumuenster, Germany| Infatigables, The Bellrays: Bad Reaction