Re: X server support?

[KIYOHARA Takashi <kiyohara%kk.iij4u.or.jp@localhost>]

Hi! Michael,


From: Michael Lorenz <macallan%netbsd.org@localhost>
Date: Wed, 17 Sep 2008 10:25:51 -0400

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hello,
> 
> On Sep 16, 2008, at 2:06 PM, KIYOHARA Takashi wrote:
> 
> > From: Michael Lorenz <macallan%netbsd.org@localhost>
> > Date: Thu, 11 Sep 2008 00:42:11 -0400
> >
> >> Then X tries to mmap the card's framebuffer aperture, apparently
> >> believing it's at 0xfc000000 but I see no such range reported in the
> >> PCI scan. Could you please post the card's pcictl dump output?
> >> I suspect that the card either doesn't have a BAR for its aperture or
> >> the driver doesn't look at it and tries to get the aperture address
> >> from the graphics chip's IO registers which returns bogus since we
> >> failed to map IO space.
> >> If the card doesn't use a BAR for its aperture we have a problem
> >> since drivers like genfb will only let you mmap ranges that are
> >> either known to belong to the graphics chip or PCI IO.
> >
> > I builded xserver for bebox on last weekend.  It was not able to load
> > pcidata correctly though it made *_ drv.o for X.
> >
> >
> > Can this problem be solved?  Should I overwrite the module of ofppc or
> > macppc again?
> 
> That's a known problem, Magnus Henoch posted a workaround on port- 
> macppc a while ago ( it 's a hack in ld.so_elf / rtld.c and avoids  
> using __builtin_return_addr(0) which occasionally returns bogus on  
> powerpc, so far xorg's module loader is the only known program that  
> triggers it. I'm not aware of a similar problem on any other CPU  
> architecture though.


I correct following and rebuild (./build.sh -u release)
And, there was no change though ld.elf.so was replaced.  Moreover, there
was no change though base.tgz, xbase.tgz, and xserver.tgz were replaced.

  http://mail-index.netbsd.org/tech-kern/2008/05/15/msg001401.html


Do you understand where I should confirm?

Thanks,
--
kiyohara