Re: EPIA-M "Magic Potion"?

[Dave Ashley]

Ron says:
> dave, can you send me a diff -u and I'll apply that patch.

Between what and what? Your source and my source have forked hopelessly
apart. A human has to step in and integrate the changes. I certainly don't
have time to do it. In fact I had thought when I passed my patch back to
this group July 2003 my version would have been integrated. If I send you
a patch now, maybe that will just gather dust again and we'll have to go
through this all again later.

Meanwhile I explained in the email what had to be patched and how, start
with that. This should get VGA working at least. Maybe that's all that's
left. Note I'm not actually building the latest cvs version of linuxbios,
I just looked at the files and saw the vga init code wasn't there.

-Dave

EPIA-M "Magic Potion"?

[John Laur]

Dave, Ron, Dmitry,

The patches Dave sent were great! The earlysetup.inc successfully gets
VGA going (the pci hard reset is not necessary in my configuration), and
the flash application helped me realize what a POS this winbond flash
chip in my RD1-PL is... The automatic verification in the tool is very
useful. Thank you, Dave.

I am happy to put together patches to current CVS providing I can get
something working well. Dmitry has just sent a patch for splash that
hopefully will help me set some things up so that I can hand over to
vesafb or viafb. Dmitry, did I see somewhere that you had a patch for
making viafb work correctly when compiled into the kernel? 2.6
vesafb+vm86d might also be a possibility to get framebuffer init that I
will investigate after exhausting alternatives. Anyway, if I can
consolidate all these patches, I'll make a unified one that can
hopefully fix CVS if that is OK with everybody.

Anyway, on to the last current problems:

Even though the sytem is more functional, I still can't do much;
however, because my PCI devices still do not seem to be getting assigned
IRQ's. Is there something I'm just missing here like a kernel patch, or
maybe there is still something else that is not in the current freebios
CVS?

Next, is there a way to hardcode the amount of ram in v1? I know
autodetection does not work, but even if I could just manually have the
machine set up the full 512MB that would help me a lot until the v2 port
progresses.

I will try working with Dmitry's changes that he has sent this evening
and post results soon. I'm hoping I'll find the IRQ problems..

Thanks again, everyone for all the help!

John
> -----Original Message-----
> From: [e-mail] [mailto:linuxbios-
> [e-mail]] On Behalf Of Dave Ashley
> Sent: Sunday, March 28, 2004 6:55 PM
> To: [e-mail]
> Subject: Re: EPIA-M "Magic Potion"?
>
> Ron says:
> > dave, can you send me a diff -u and I'll apply that patch.
>
> Between what and what? Your source and my source have forked

hopelessly
> apart. A human has to step in and integrate the changes. I certainly

don't
> have time to do it. In fact I had thought when I passed my patch back

to
> this group July 2003 my version would have been integrated. If I send

you
> a patch now, maybe that will just gather dust again and we'll have to

go
> through this all again later.
>
> Meanwhile I explained in the email what had to be patched and how,

start
> with that. This should get VGA working at least. Maybe that's all

that's
> left. Note I'm not actually building the latest cvs version of

linuxbios,
> I just looked at the files and saw the vga init code wasn't there.
>
> -Dave
> _______________________________________________
> Linuxbios mailing list
> [e-mail]
> http://www.clustermatic.org/mailman/listinfo/linuxbios

EPIA-M "Magic Potion"?

[Dmitry Borisov]

> vesafb or viafb. Dmitry, did I see somewhere that you had a patch for
> making viafb work correctly when compiled into the kernel? 2.6

Yes I have a patch for 2.4.23-epia1 that integrates viafb into kernel.
Should work with all kernels though.
> hopefully fix CVS if that is OK with everybody.

That would be cool.
> Even though the sytem is more functional, I still can't do much;
> however, because my PCI devices still do not seem to be getting assigned
> IRQ's. Is there something I'm just missing here like a kernel patch, or

I don't have external PCI devices, All built ins working properly( except
CLE266 which points to improper initialization ).
> Next, is there a way to hardcode the amount of ram in v1? I know
> autodetection does not work, but even if I could just manually have the
> machine set up the full 512MB that would help me a lot until the v2 port
> progresses.

I tried that but in fact ran into serious problems with DFB. Something went
wrong... So I stuck with 96M at this time...

Dmitry/

EPIA-M "Magic Potion"?

[John Laur]

Thank you, Dmitry. Your patches seem to fix a lot of my remaining
problems.

The kernel patch to make viafb work as compiled-in gets a framebuffer
going very soon after the kernel loads. I extended it a little bit to
allow for full selection of all module parameters from the kernel
commandline. I guess since it didn't work, nobody ever used the kernel
commandline params :) The vga stuff is now working perfectly except for
hardware-assisted.

The other thing that fixed a lot of issues is that I switched to an
earlier vga bios (extracted from bios v1.10F). I was trying to use the
latest one released with BIOS version 1.13 and it seems to not work
correctly. I'm going to try again now that I have some other stuff
working to verify that this is the case. Something bad running in the
1.13 vga bios I believe could have been the cause of my IRQ issues,
which are solved for the moment as well.

Whenever I can make the dust clear, I'll create a diff for CVS including
either amending the existing HOWTO with the information I have learned
or creating a new one especially for EPIA-M. I'll also include the viafb
patch from Dmitry + my extra changes in kernel-patches.

The final three non-memory related problems I'm looking at are:

1) trying to load the agpgart module hangs the system
2) applications hang when accessing the hardware clock
3) via_v4l_drv (hardware MPEG acceleration) wants to be loaded before
viafb in order to work unfortunately

Again, thanks everyone!

John
> -----Original Message-----
> From: [e-mail] [mailto:linuxbios-
> [e-mail]] On Behalf Of Dmitry Borisov
> Sent: Sunday, March 28, 2004 10:54 PM
> To: [e-mail]
> Subject: Re: EPIA-M "Magic Potion"?
> > vesafb or viafb. Dmitry, did I see somewhere that you had a patch

for
> > making viafb work correctly when compiled into the kernel? 2.6
>
> Yes I have a patch for 2.4.23-epia1 that integrates viafb into kernel.
> Should work with all kernels though.
> > hopefully fix CVS if that is OK with everybody.
>
> That would be cool.
> > Even though the sytem is more functional, I still can't do much;
> > however, because my PCI devices still do not seem to be getting

assigned
> > IRQ's. Is there something I'm just missing here like a kernel patch,

or
> I don't have external PCI devices, All built ins working properly(

except
> CLE266 which points to improper initialization ).
> > Next, is there a way to hardcode the amount of ram in v1? I know
> > autodetection does not work, but even if I could just manually have

the
> > machine set up the full 512MB that would help me a lot until the v2

port
> > progresses.
>
> I tried that but in fact ran into serious problems with DFB. Something
> went
> wrong... So I stuck with 96M at this time...
>
> Dmitry/
>
> Linuxbios mailing list
> [e-mail]
> http://www.clustermatic.org/mailman/listinfo/linuxbios

EPIA-M "Magic Potion"?

[Dmitry Borisov]

Subject: RE: EPIA-M "Magic Potion"?
> Thank you, Dmitry. Your patches seem to fix a lot of my remaining
> problems.

Good.
> The kernel patch to make viafb work as compiled-in gets a framebuffer
> going very soon after the kernel loads. I extended it a little bit to
> allow for full selection of all module parameters from the kernel
> commandline. I guess since it didn't work, nobody ever used the kernel
> commandline params :) The vga stuff is now working perfectly except for
> hardware-assisted.

All params works in my case... But still I would be interested in your
changes...
> via_v4l_drv (hardware MPEG acceleration) wants to be loaded before
>
> viafb in order to work unfortunately

I'm going to release MPEG-2 support builtin for ffmpeg/mplayer under DFB. No
need for X/v4l/etc.
It is working already, but still I want to scrub something. Will let you
know at earliest.

Dmitry/

EPIA-M "Magic Potion"?

[ron minnich]

On Sun, 28 Mar 2004, John Laur wrote:
> I am happy to put together patches to current CVS providing I can get
> something working well.

Go for it. I thought I had managed to bring all Dave's stuff in but 
clearly that did not happen.
> Even though the sytem is more functional, I still can't do much;
> however, because my PCI devices still do not seem to be getting assigned
> IRQ's. Is there something I'm just missing here like a kernel patch, or
> maybe there is still something else that is not in the current freebios
> CVS?

Let's get your first patch in, then look at IRQs.

ron

EPIA-M "Magic Potion"?

[ron minnich]

On Sun, 28 Mar 2004, Dave Ashley wrote:
> Between what and what? Your source and my source have forked hopelessly
> apart. A human has to step in and integrate the changes. I certainly
> don't have time to do it. In fact I had thought when I passed my patch
> back to this group July 2003 my version would have been integrated. If I
> send you a patch now, maybe that will just gather dust again and we'll
> have to go through this all again later.

no, I committed that last patch, and I thought I had closed the loop, so 
I'm not sure why that part got missed.

OK, I'll just go with your last post.

ron