[Intel-gfx] Wrong fb_bar under 64bit address

[Eric Anholt]

On Fri, 2008-12-05 at 08:50 +0800, Edward Shu wrote:
> Eric Anholt: 
> > On Thu, 2008-12-04 at 16:58 +0800, Edward Shu wrote:
> >   
> > > Hi,
> > >   I am investigating a PCI specific issue for Intel DDX drivers.
> > > 
> > >   In i830_driver.c:i830_detect_chipset()
> > > 
> > > 1323	 /* Now that we know the chipset, figure
> > >          out the resource base addrs */
> > > 1324	if (IS_I9XX(pI830)) {
> > > 1325	fb_bar = 2;
> > > 1326	mmio_bar = 0;
> > > 1327	} else {
> > > 1328	fb_bar = 0;
> > > 1329	mmio_bar = 1;
> > > 1330	}
> > > 
> > > fb_bar is an BAR index in PCI configure space. In G35,
> > > the GMADR spans from 18~1F which can have 64 bit address
> > > in G35.  And fb_bar should be 1 instead of 2 following
> > > PCI convention.  Why does this wrong fb_bar work by now?
> > > I guess BIOS never give GMADDR a 64bit address so far
> > > tough it has 64bit address ability. However,  64bit address
> > > for GMADR will definitely come.  What do you think of this?
> > >     
> > 
> > PCI BAR numbers are based on the offset, not the index in the list of
> > BARs the devie has.
> >   
> If it is that way, how can it deal with 64bit PCI address? 
> combining two consecutive BARS together?

That's how it's done (by your OS).

-- 
Eric Anholt
eric at anholt.net                         eric.anholt at intel.com


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20081204/198771c6/attachment.sig>