[LinuxBIOS] RE : Re: VGA bios and i/o access for powerpc

[jean-francois simon]

Hi,
  Maybe there is a way to do without the emulation layer altogether:
  Since in our case we use PCI graphics, I can record all the PCI cycles made by the VGA BIOS to the graphic card on a BIOS based system (using a PCI analyser), put them into a listing, run a perl script on it and include them in the firmware of the non BIOS system....an redo the process whenever we need to use a different graphic which doesn't happen very often 
  -jf simon

I think a sure way is to record all the PCI cycles generated by the VGA BIOS 
jean-francois simon  a écrit :
      Those legacy IO ports are NOT necessary mapped to the PCI IO address.
Some VGA card does and some others come with some kind of remapping
or offset. Your system chipset should have a way to map legacy VGA IO
ports to memory address. You have to read the doc to the PPC chipset 
carefully to find it out.

I have read the chipset doc and there is no such VGA support.
  But could you plse explain more on the different ways for VGA cards to map the VGA registers? Since the registers are located on a PCI agent (the graphic card), there must be a way to access them by generating PCI cycles?
  Thanks
  -jf simon

---------------------------------
 Découvrez un nouveau moyen de poser toutes vos questions quelque soit le sujet ! Yahoo! Questions/Réponses pour partager vos connaissances, vos opinions et vos expériences. Cliquez ici.

[Ronald G Minnich]

jean-francois simon wrote:
> Hi,
> Maybe there is a way to do without the emulation layer altogether:
> Since in our case we use PCI graphics, I can record all the PCI cycles 
> made by the VGA BIOS to the graphic card on a BIOS based system (using a 
> PCI analyser), put them into a listing, run a perl script on it and 
> include them in the firmware of the non BIOS system....an redo the 
> process whenever we need to use a different graphic which doesn't happen 
> very often

we've had poor luck with that approach. Plus, even in graphics card of 
same model, memory programming can change quite a bit between cards. I 
am not sure this will work.

ron

[jf simon]

Ronald G Minnich wrote:
> jean-francois simon wrote:
> > Hi,
> > Maybe there is a way to do without the emulation layer altogether:
> > Since in our case we use PCI graphics, I can record all the PCI 
> > cycles made by the VGA BIOS to the graphic card on a BIOS based 
> > system (using a PCI analyser), put them into a listing, run a perl 
> > script on it and include them in the firmware of the non BIOS 
> > system....an redo the process whenever we need to use a different 
> > graphic which doesn't happen very often
>
> we've had poor luck with that approach. Plus, even in graphics card of 
> same model, memory programming can change quite a bit between cards. I 
> am not sure this will work.
>
> ron

Thanks. I'll give it a quick try. If it works I'll let you know.
-jfs

___________________________________________________________________________ 
D�couvrez un nouveau moyen de poser toutes vos questions quelque soit le sujet ! 
Yahoo! Questions/R�ponses pour partager vos connaissances, vos opinions et vos exp�riences. 
http://fr.answers.yahoo.com

[Ronald G Minnich]

jf simon wrote:
> Ronald G Minnich wrote:
> > jean-francois simon wrote:
> > > Hi,
> > > Maybe there is a way to do without the emulation layer altogether:
> > > Since in our case we use PCI graphics, I can record all the PCI 
> > > cycles made by the VGA BIOS to the graphic card on a BIOS based 
> > > system (using a PCI analyser), put them into a listing, run a perl 
> > > script on it and include them in the firmware of the non BIOS 
> > > system....an redo the process whenever we need to use a different 
> > > graphic which doesn't happen very often
> >
> > we've had poor luck with that approach. Plus, even in graphics card of 
> > same model, memory programming can change quite a bit between cards. I 
> > am not sure this will work.
> >
> > ron
>
> Thanks. I'll give it a quick try. If it works I'll let you know.
> -jfs

keep in mind that, if it works on one *instance* of a card from vendor 
X, it may fail on the same model card from vendor X, purchased on the 
same day, from the same store, on the same shelf.

PC hardware is wonderful.

ron

[jf simon]

Ronald G Minnich wrote:
> keep in mind that, if it works on one *instance* of a card from vendor 
> X, it may fail on the same model card from vendor X, purchased on the 
> same day, from the same store, on the same shelf.
>
> PC hardware is wonderful.

Thanks. I will run  a trace and see first how large it is (100's or 
1000's of cycles). Maybe there are pci cycles that are checking  from 
flags to be set on the vga chipset, and these could explain the behavior 
differences you are mentionning, in case the flags hasn't changed to the 
expected value before the init process goes on. Or timings between 
cycles that need to be respected. I think I can write some perl script 
that will extract these  informations from the analyser trace and 
produce the right C code. I hope the pci graphic card doesn't rise 
interrupts during that process though..
-jfs

___________________________________________________________________________ 
Découvrez un nouveau moyen de poser toutes vos questions quelque soit le sujet ! 
Yahoo! Questions/Réponses pour partager vos connaissances, vos opinions et vos expériences. 
http://fr.answers.yahoo.com

[ollie]

On Wed, 2006-09-06 at 07:47 +0200, jean-francois simon wrote:
> Hi,
> Maybe there is a way to do without the emulation layer altogether:
> Since in our case we use PCI graphics, I can record all the PCI cycles
> made by the VGA BIOS to the graphic card on a BIOS based system (using
> a PCI analyser), put them into a listing, run a perl script on it and
> include them in the firmware of the non BIOS system....an redo the
> process whenever we need to use a different graphic which doesn't
> happen very often 
> -jf simon

Unfortunately, this is no going to work. The VGA BIOS still use the 
legacy IO ports. All you will get in the listing is in/out to port
0x3xx. You just have to figure out how to forward IO to legacy ports.

Ollie
>

[jf simon]

ollie wrote:
> Unfortunately, this is no going to work. The VGA BIOS still use the 
> legacy IO ports. All you will get in the listing is in/out to port
> 0x3xx. You just have to figure out how to forward IO to legacy ports.
>
> Ollie

If you snoop the PCI bus right on the PCI slot for the graphic card, you 
see "real"  pci i/o cycles bound to the PCI i/o BAR of the graphic card. 
If the address is outside of the BAR the access will not be decoded byte 
graphic PCI agent, I would think.
-jf simon

___________________________________________________________________________ 
Découvrez un nouveau moyen de poser toutes vos questions quelque soit le sujet ! 
Yahoo! Questions/Réponses pour partager vos connaissances, vos opinions et vos expériences. 
http://fr.answers.yahoo.com