On Mon, 26 May 2025, Bernhard Beschow wrote:
Am 25. Mai 2025 00:03:15 UTC schrieb BALATON Zoltan <bala...@eik.bme.hu>:
On Sun, 28 Jul 2024, BALATON Zoltan wrote:
On Sun, 28 Jul 2024, quan wrote:
I am playing around with the parallel port on the pegasos 2 and I noticed
the following issue:
The parallel port on startup is set to iobase 0x378, then later it gets
remapped to iobase 0x3bc.
When I tried writing to the 0x3bc location with
trace:memory_region_ops_write on, I get logs about writing to the location
'parallel', so far so good, but the writes don't make it through (it never
calls parallel_ioport_write* functions)
I traced it through where it calls portio_write, and inside the
portio_write code:
This line always fails for me:
const MemoryRegionPortio *mrp = find_portio(mrpio, addr, size, true);
This is due to the mrpio has the offset of 0x44 (0x3bc-0x378) in the list,
and the 'addr' in the code ranges from (0-15) (address relative to 0x3bc),
so it can never match the parallel io function to execute.
This seems like a bug to me. I think there is some issue where the
portio_list is not getting updated properly after the remapping to 0x3bc.
The other devices (Serial, RTC, PM, VGA) responds properly for me, so it is
only the parallel port that has this issue.
Is this something that broke recently or can the same be reproduced with QEMU
v8.2.0. The relocation of these devices was implemented around commit
35a6380b4ed27f (and the ones before that). Adding Bernhard to cc as well.
This commit also seems to break display with some VGA ROMs as can be reproduced
with:
qemu-system-ppc -machine pegasos2 -bios pegasos2.rom -serial stdio -vga none
-device ati-vga,romfile=ati-rage128progl16mb.VBI
where the romfile is something from
http://vgamuseum.info/index.php/companies/item/111-ati-rage-128-pro for
example). Until commit 35a6380b4ed27f^ you get picture on emulated VGA but not
after (although I have no idea what a VGA BIOS has to do with the parallel
port). Any idea how to fix this?
Here an observation: Executing `info mtree` reveals that the parallel
port gets mapped to 0x3bc - 0x3c3 which overlaps with some vga region
(0x3c0 - 0x3cf). So the commit seems to expose a firmware bug in
pegasos2.rom or in its default configuration. On real hardware both
devices might respond such that the bug might not materialize there like
in QEMU, but that's just a guess. Maybe one could work around that by
decreasing the priority of the parallel io ports.
Hmm, I haven't noticed that (only looked at the diff between working and
non-working and saw it moved) but according to e.g.
https://www.stanislavs.org/helppc/ports.html The original parallel
interface had less ports so there was no overlap, maybe it's because we
emulate EPP+ECP or other advanced features that the message from the Linux
list mentions as not working at the default address on real machine
either. So we may need to disable additional ports which probably do not
exist on real machine unless enabled (which the firmware does not seem to
do) or maybe changing priority as you say could help. Anybody wants to
come up with a patch? I don't have time now and may forget later.
Regards,
BALATON Zoltan
