Re: Console on 9000/425t with A1659A graphics

[Chris Hanson <cmhanson%eschatologist.net@localhost>]

On Nov 30, 2018, at 3:33 PM, Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost> wrote:

> (sorry for a late response)

I wouldn’t call it late. :)

>> I have a 9000/425t that included A1659A graphics in the SGC slot, and I only seem to be able to use serial console with NetBSD 8.0.
> 
> Wow.  Currently SGC sti(4) is tested only on 425e, and
> sti(4) on 425t is one of todo list items.
> (I have 425t but it has only DIO topcat)

Ah, interesting. It does seem to show up at least!

>> Booting the netbsd-GENERIC kernel via NFS, and using serial console, I see the sti(4) driver come up and recognize the 1280x1024@8bpp framebuffer attached via SGC, and it looks like wscons wants to use it (or is using it), but whenever I've tried to boot using local console I get nothing.
> 
> Could you post full dmesg on the serial console?
> How did you setup to get screen console?
> In that case which console is used on bootloader?

On startup, the ROM will use local graphical console if I have it set to, including SYS_UBOOT or SYS_INST. As soon as the kernel starts, however, output switches to serial regardless of how whether I have ROM set to use local or serial console. (At first I thought boot was hanging, I attached a serial cable on a whim and discovered that’s what was happening.)

I don’t have any dmesg output at the moment because I need to reinstall the machine; I wound up making / too small. Once I have NetBSD reinstalled I’ll get this. For the install itself I put all the main packages (not debug.tgz or tests.tgz) on the system.

One thing I did notice is that I got a message at startup that there was no /dev/console found, before the message that the overlay file system was mounted at /dev. Perhaps when I previously installed—which involved manually doing a “tar -zpxf base.tgz” etc.—I didn’t wind up with a complete /dev.

I also didn’t set anything up to get a console because I have no idea what to do in order to do that (whether editing /etc/wscons.conf or /dev/ttys or anything else). I’m relatively new to NetBSD, wscons, and so on—I followed 386BSD at the beginning, but put that aside around 1994. I do know what I’m doing with NEXTSTEP and Mac OS X (from the old days through the latest) and know enough to be dangerous when it comes to SunOS 4.

> There is a console detection code in sys/arch/hp300/hp300/audoconf.c:
> https://nxr.netbsd.org/diff/src/sys/arch/hp300/hp300/autoconf.c?r2=%2Fsrc%2Fsys%2Farch%2Fhp300%2Fhp300%2Fautoconf.c%401.105&r1=%2Fsrc%2Fsys%2Farch%2Fhp300%2Fhp300%2Fautoconf.c%401.104
> https://nxr.netbsd.org/diff/src/sys/arch/hp300/dev/sti_sgc.c?r2=%2Fsrc%2Fsys%2Farch%2Fhp300%2Fdev%2Fsti_sgc.c%401.2&r1=%2Fsrc%2Fsys%2Farch%2Fhp300%2Fdev%2Fsti_sgc.c%401.1
> but maybe there is some quirk on sgc sti(4) on 425t.

I strongly suspect that SGC is going to be pretty close to the same on 425t and 425e (and early 700 series as well); it appears that SGC was only even an option on late-production 4XXt systems.

>> Any ideas? I'd love to actually get X11 up on this system. It does work fine with HP-UX 9.10 after all. ;)
> 
> One problem to get Xorg server is there is no information to
> control color palette of sti(4) framebuffer.  I tried to write
> RAMDAC registers directly but it looked some more sti(4) specific
> control was required.
> 
> Note drawing framebuffer against sti(4) on 425e using mlterm-fb worked:
> https://twitter.com/tsutsuii/status/503007544420233217
> https://twitter.com/tsutsuii/status/503008051658366979
> https://twitter.com/tsutsuii/status/503009014855127041

Interesting! I have no idea what to do to get to that point, I’ve never configured wscons, mbterm-fb, or anything else.

  — Chris