Subject: Re: Newer kernel?
To: Andrew Gillham <gillhaa@ghost.whirlpool.com>
From: Darrin B. Jewell <dbj@netbsd.org>
List: port-next68k
Date: 03/12/1999 14:56:35
>xe0 at intio0
>xe0 at MAC address 00:00:0f:01:10:94
>isrdispatch_autovec: stray level 3 interrupt
>isrdispatch_autovec: stray level 3 interrupt
>[.. tons of these ..]
>isrdispatch_autovec: stray level 3 interrupt
>isrdispatch_autovec: stray level 3 interrupt
>panic: isr_dispatch_autovec: too many stray interrupts
>Stopped in  at  _Debugger+0x6:  unlk    a6
>
>What is the current status of SCSI, etc?

These stray level 3 interrupts are scsi interrupts that are ocurring
with no scsi device driver to handle them.  This happens sometimes if
you interrupted a disk boot to get back to the rom monitor.  The
current sources shouldn't have this problem, since they have a scsi
driver that works well enough to probe and configure the scsi bus.
Unfortunately, the scsi driver currently drops out data so it isn't
useful to mount disks yet.  I am working on it daily. Unfortunately,
writing drivers by trial and error leads to lots of mysterious race
conditions.

>
>I've tried to cross compile from my i386 with 'doit.m68k', but
>I get an assembler error like this:
>/usr/local/bin/m68k-unknown-netbsd1.3E-gcc -x assembler-with-cpp -traditional-cp
>p -c -m68040 -I/usr/src/sys/arch/m68k/fpsp -o res_func.o res_func.s
>res_func.s: Assembler messages:
>res_func.s:1524: Error: parse error -- statement `fcmpd #:0x41dfffffffc00000,fp0
>' ignored
>res_func.s:1527: Error: parse error -- statement `fcmpd #:0xc1e0000000000000,fp0
>' ignored
>[.. more lines ..]

Unfortunately, the cross compiler doesn't like the floating point constants
in the fpsp library very well.  To get around this problem, you can use
the precompiled object files in
ftp://ftp.netbsd.org/pub/NetBSD/arch/next68k/tools/fpsp.tgz
and just "touch" them to bring them up to date if needed.

>
>Any help would be appreciated.
>
>-Andrew
>--

I will put up a new kernel from the current sources this evening.
Alternately, the kernel from the december snapshot should work if you
don't let the rom try to boot from the scsi disk before interrupting
it.

Darrin