Re: Kernel config for Sun 3/50

[Trent McNair <tmcnair%gmail.com@localhost>]

Thanks!

1.1m is as small as I've been able to go. This is smaller than my
SunOS kernel which made me curious.

Sun-3 Architecture manual
(http://www.bitsavers.org/pdf/sun/sun3/Sun-3_Architecture_Manual_Ver_1.0_Jan85.pdf)
states:

        Main memory shadowes video, memory in the range of physical addresses
        starting at 1 megabytes and extending for 128 kilobytes. This
area of main memory
        is called the copy region. If the copy enable bit in the
system enable register is
        set then data written into this copy region is also written
into the video memory at
        the same location within the 128K region. A read from the copy
region returns the
        data in main memory and does not affect the video memory

This would seem to indicate that if I'm not using the framebuffer
(which I am not) I could comment out the Sun 3/50 size check. I'll
give it a try over the weekend.

On Tue, Jan 10, 2023 at 12:24 PM David Brownlee <abs%absd.org@localhost> wrote:
>
> On Mon, 9 Jan 2023 at 06:13, Trent McNair <tmcnair%gmail.com@localhost> wrote:
> >
> > I've set up a cross-compile toolchain to build 9.3 for my Sun 3/50. I am able to build a kernel and userland, but I'm running into the "kernel too large" error in pmap.c:
> >
> >         if (cpu_machine_id == ID_SUN3_50) {
> >                 hole_start = m68k_trunc_page(OBMEM_BW50_ADDR);
> >                 hole_size  = m68k_round_page(OBMEM_BW2_SIZE);
> >                 if (avail_start > hole_start) {
> >                         mon_printf("kernel too large for Sun3/50 %d %d %d\n", hole_start, hole_size, avail_start);
> >                         sunmon_abort();
> >                 }
> >         }
> >
> > I've looked through old mailing list threads and I guess this limit has to do with the location of the 3/50's frame buffer. I can't seem to remove enough
> > functionality to get a small enough kernel to fit below hole_start. Can a working kernel be built for this machine?
>
> It's likely to be a challenge to get something small enough to fit.
> Sun3 GENERIC already has COPTS="-Os -fno-unwind-tables
> -fno-omit-frame-pointer", which could be tweaked up to COPTS="-Os
> -fno-unwind-tables" which might shave a fraction.
>
> I'd be tempted to build a test kernel without any scsi and/or network
> support just to confirm it will boot to get a baseline
>
> David