Re: Re: For intel 865 motherboards

[Observer <obs...@mo...>]

> On Wednesday 01 December 2004 20:27, observer wrote:
> > Hi
> > After a lot of trouble i was able to make ruby work perfectly with
an
> > intel 865gbf and 865perl motherboards, this was after apliying this
> > patch to the kernel:
> > 
> > diff -Naur linux-2.6.8/arch/i386/pci/fixup.c
> > linux-2.6.8.orig/arch/i386/pci/fixup.c
> > --- linux-2.6.8/arch/i386/pci/fixup.c   2004-08-14
00:37:26.000000000
> > -0500
> > +++ linux-2.6.8.orig/arch/i386/pci/fixup.c      2004-11-19
> > 00:30:59.000000000 -0600
> > @@ -191,9 +191,9 @@
> >   */
> >  static void __devinit pci_fixup_transparent_bridge(struct pci_dev
*dev)
> >  {
> > -       if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI &&
> > +       /*if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI &&
> >             (dev->device & 0xff00) == 0x2400)
> > -               dev->transparent = 1;
> > +               dev->transparent = 1;*/
> >  }
>
>                                                                                 
> >  /*
> > 
> > What i"ve found is that this fix was intended for the first
motherboards
> > with this pci bridges but (i845 i think) but it seems in newer
> > motherboards the problem has changed, this also solved a problem
with a
> > Davicom 21x4x ethernet card and some non ruby dual x configurations,
of
> > course with such a complicated patch there is a chance that some
system
> > burn in flames, but i hope someone more qualified can give a better
> > solution.
> > Hope this helps someone.
> 
> Can You show more details. Why trouble pop up under Linux-ruby , 
> but not under Linux at all ?
> If this is wide trouble, then You should notify kernel mailing list.
> 
> Aivils
 
There is trouble in Linux but it seems that they quite rare and maybe
they are taken as hardware problems or incompatibilities, I've seen a
post in the Kernel list pointing the same but no answer was given.
The problems i mention about the davicom card were found months before i
ever installed ruby but i simply changed the card, but testing the
patched kernel i found that it worked with the fix removed.
Also until this patch i haven't been able to run a multi head config
with more than 2 (sometimes not even 2) nvidia cards , no matter if
using Ruby-Linux or simply Linux but after the patch there is no
trouble, but who was going to report this as a pci bug when every
xinerama howto says that it may or may not work and that some cards mix
better than others?. 
With some other brands of cards it seems that a xserver normally has no
trouble to work but when running with the isolate patch (which is more
related to pci than to ruby) more than 2 equal cards the system freezes.
I also found something that may be unrelated but in the pci id list in
sourceforge and in the kernel 2.6.9 the definition of component
8086:244e, which is a pci bridge and is present in both motherboards,
has changed but those definitions are there only to show the names
instead of the numbers of the components.
Looking in the kernel list it seems the fix creates similar troubles (in
this motherboards) to those that it was mean to fix, anyway just
removing the code might be dangerous since all components in the board
have this: dev->device & 0xff00) == 0x2400 and i don't know how will
this affect to other components, my system have been running without
trouble for about a month with this.