Re: NetBSD installation woes on Sun Blade 100

[Martin Husemann <martin%duskware.de@localhost>]

On Mon, Jul 20, 2020 at 04:23:03PM +0200, Connor McLaughlan wrote:
> I still need support in building my own sparc64 kernel for a Sun Blade
> 100 and similar machines.
> Setting SSIZE to 4 seems to be not enough for the mach64 driver to
> initialize correctly.

The kernel stack size should be totally unrelated here - there was an
issue with reading EDID data for radeon cards (XVR100) that got
recently fixed (and now is fixed in netbsd-9 as well) and which may
have been worked around with more stack.

However, there is something very wrong with mach64. I fixed my U5
recentish and see this at boot (with serial console, using -current):

[   1.0000000] machfb0 at pci1 dev 2 function 0: vendor 1002 product 4754 (rev. 0x9a)
[   1.0000000] machfb0: 16 MB aperture at 0xe1000000, 4 KB registers at 0x00000000
[   1.0000000] machfb0: 128 KB ROM at 0xe1020000
[   1.0000000] machfb0: 2048 KB SGRAM 62.999 MHz, maximum RAMDAC clock 170 MHz
[   1.0000000] machfb0: unable to use preferred mode
[   1.0000000] machfb0: initializing the DSP
[   1.0000000] machfb0: initial resolution 1152x720 at 8 bpp
[   1.0000000] wsdisplay1 at machfb0
[   1.0000000] direct rendering for machfb0 unsupported


OF uses 1280x1024 x 60Hz with my display and when I boot with machfb as
console the display goes out of range as soon as machfb attaches.

This is with output-device set to "screen".

Unfortunately this also happens with "screen:r1280x1024x60", and dmesg
says:

machfb0 at pci1 dev 2 function 0: vendor 1002 product 4754 (rev. 0x9a)
machfb0: 16 MB aperture at 0xe1000000, 4 KB registers at 0x00000000
machfb0: 128 KB ROM at 0xe1020000
machfb0: 2048 KB SGRAM 62.999 MHz, maximum RAMDAC clock 170 MHz
machfb0: autoconfiguration error: unable to use preferred mode
machfb0: initializing the DSP
machfb0: initial resolution 1152x720 at 8 bpp
wsdisplay0 at machfb0: console (default, dumb emulation)

and again my display goes out of range. But the machine comes up and I can
log in via the network, so testing is easy and non destructive.

Michael, Julian - any ideas? Ping me if you need some debug info.

Martin