Boot issues when RTG expansion enabled

[mark_k]

Hi,

I'm trying to set up WinUAE with RTG graphics. Emulated machine config:
- 68020 CPU, ECS, 1MB chip, 8MB (Zorro II) fast, A2000 chipset extra, Kickstart 1.3.
- Drives: HDF image file of my real Amiga hard disk, and shared directory on PC.

That seems to be working fine with the latest WinUAE beta. (The OFS boot partition startup-sequence loads Kick 3.1 using ReKick and reboots, then mounts FFS partitions and continues booting.)

However, when I enabled RTG in the Expansions page in WinUAE settings that seems to mess up the boot process. Specifically:

- With Kickstart 1.3 in ROM, no boot drive is found so it sits at the insert disk screen.

- With Kickstart 2.04, the HDF and shared directory appear in the boot menu which appears when you hold both mouse buttons. It runs the startup-sequence from the OFS partition. That uses ReKick to load Kickstart 3.1, which works and the computer reboots into Kick 3.1. But it sits at the insert disk screen, no booting from the HDF partition. If I get the 3.1 boot menu up, it just shows DF0: as the only boot device. And (maybe this is important?) the 3.1 Expansion Board Diagnostic shows this:

Board No. Manufacturer Product Status
   1         2011         1    Working
   2         2011        96    Defective

Booting with 2.04 in ROM, SysInfo autoconfig boards information showed:

Address		Size	Type		Product	Manufacturer	Serial No.
$00200000	4096K	Zorro II	1	2011		1
$10000000	2048K	Zorro III	96	2011		1
$00E90000	64K	Zorro II	82	2011		1

SysInfo board list booting with Kickstart 1.3 in ROM (note: HDF and shared directory are still enabled in WinUAE config):

$00200000	4096K	Zorro II	1	2011		1

Maybe the 1.3 expansion.library aborts the autoconfig process when it finds a Zorro III board? In which case, could making the Zorro III board get configured last help? And likewise for the 3.1 ReKick Kickstart (which probably doesn't have support for Zorro III boards).

An option to have the WinUAE RTG board be Zorro II (and so use the Z2 memory region) could possibly/probably get it to work with the soft-kicked 3.1. But just moving the board order so the Zorro III board is last in the chain could prevent the HDF and shared directories vanishing when booting with Kickstart 1.3 in ROM.

[Toni Wilen]

Yes, KS 1.x stops autoconfig process if any non-Z2 board is found and yes, some emulated Z2 boards come after Z3 boards causing missing harddrives etc under 1.3.

Ok, boards are reordered now, first all Z2 boards and then Z3 boards (if any).

Z2 compatible RTG board is not that easy, JIT direct requires that all memory boards' final base address is known before autoconfig starts.

btw, what are you exactly trying to do?
(that winuae.zip url updated too)

[mark_k]

Originally Posted by Toni Wilen

Yes, KS 1.x stops autoconfig process if any non-Z2 board is found and yes, some emulated Z2 boards come after Z3 boards causing missing harddrives etc under 1.3.

Ok, boards are reordered now, first all Z2 boards and then Z3 boards (if any).

Z2 compatible RTG board is not that easy, JIT direct requires that all memory boards' final base address is known before autoconfig starts.

btw, what are you exactly trying to do?
(that winuae.zip url updated too)

Thanks I'll give that a test.

Is is actually necessary to have the RTG board memory in Zorro 2/3 config space? You could configure it to have 64KB registers which would end up at $Ex0000, but non-autoconfig RTG RAM anywhere else. Assuming the uaegfx P96 driver were changed to know about that of course. (It could read one of the registers to get the RTG RAM base address.)

The unusual emulated hardware config is to reproduce as close as possible my old Amiga 2000 setup. I can't get close enough to fully boot the unmodified real Amiga hard disk image (because uaehf.device name isn't configurable, I mentioned that a while ago), but with a small change to the startup-sequence and an alternate mountlist I should be able to modify the image so it boots both my real Amiga and WinUAE.

My A2000 has Kickstart 1.3 in ROM and a GVP 68030 card (and a Piccolo graphics card). While I do have a 3.1 ROM chip (and 2.04, 2.05 in various other computers), I prefer to keep 1.3 since it works with more older games/demos. Plus using ReKick to load 3.1 into 32-bit fast RAM results in better performance than soft-kicking or remapping using the MMU. The system boots from an OFS partition then mounts the FFS partitions manually instead of having FFS in the RDB. (Because FFS versions later than 40.1 don't work under 1.3. So I'd be stuck with FFS 40.1 in the RDB unless I mount manually.)


Edit: I tested the latest winuae.zip and now with RTG enabled the emulated machine boots under 1.3, loads 3.1 and gets to Workbench correctly. I got a slightly strange result with the ReKicked Kickstart 3.1 though. In the Expansion Board Diagnostic, boards shown are

1   2011    1   Working
2   2011   82   Working
3   2011   96   Defective

That's what I expected. The low-level autoconfig code (which is actually in the ReKick program itself, not the ReKick-type Kickstart image) probably doesn't support Zorro III, since ReKick was designed for use with A500/A2000 machines. (Commodore had other solutions for other machines, e.g. maprom or A3000 SuperKickstart.)

Anyway, after letting it boot to Workbench I ran the Scout system monitor. The Expansions window gave this info:

Address    BoardAddr  Type  Manufacturer Product
$00004020  $00200000  $e0   UAE          emulated RAM expansion
$00004068  $00e90000  $d1   Prototype ID #82 (unknown)
$000040b0  $ff000000  $80   UAE          emulated graphics

So the RTG board is sort-of there, but judging from the board address is not configured properly. I double-clicked the uaegfx monitor driver icon. Running screenmode prefs showed the UAE RTG modes in the list. However, on selecting one of those the emulated machine crashed with the power LED flashing.

Next try, selecting an RTG mode gave a blank screen (colour 0 grey). The system still seemed to be up, I could press a hot-key to run Directory Opus but then WinUAE crashed.

Third try, I started Directory Opus and Imploder on custom screens. Changing Workbench to an RTG mode gave the blank screen, but I could move the mouse up and right and click on the (invisible) screen depth gadget which revealed the other (native chipset) screens.

It may not be possible to get WinUAE RTG working on Kickstarts which don't support Zorro III. But could the uaegfx driver init routine be changed to abort if it detects a bogus/misconfigured board address?

[Toni Wilen]

Originally Posted by mark_k

Is is actually necessary to have the RTG board memory in Zorro 2/3 config space? You could configure it to have 64KB registers which would end up at $Ex0000, but non-autoconfig RTG RAM anywhere else. Assuming the uaegfx P96 driver were changed to know about that of course. (It could read one of the registers to get the RTG RAM base address.)

Because it emulates proper Z3 RTG board. I am not going to introduce any hack boards that can cause compatibility problems.

Only correct way to solve this is to have Z2 RTG option. Maybe someday.

But could the uaegfx driver init routine be changed to abort if it detects a bogus/misconfigured board address?

Done

[Toni Wilen]

I was bored -> RTG ZorroII/III option is now done

Note that GUI does not yet restrict RTG RAM size.
(Same URL)

[mark_k]

Cool, I got that working. Now you just need to add a 6MB Zorro II RAM option! (Useful for 6MB RAM + 2MB RTG memory in the Zorro II region.)

[Toni Wilen]

Originally Posted by mark_k

Cool, I got that working. Now you just need to add a 6MB Zorro II RAM option! (Useful for 6MB RAM + 2MB RTG memory in the Zorro II region.)

Unfortunately single 6MB autoconfig board is impossible. Two boards (4M + 2M) needed. (2 logical boards in single physical board or two physical boards)