In 2.6, sparc64 has pci domains. This breaks X.org, without the following patch,
backwards-compatible with 2.4.
Created attachment 442 [details] [review]
Credit to Daniel Seyffer <firstname.lastname@example.org> and Ciaran McCreesh
It appears that a similar fix has already been integrated. See
What if the graphics card is not on domain 0000?
I have an Radeon at 00003 and it is not being found.
XFree86 has modified linuxPci.c significantly to
traverse PCI domains -- looks like this is exactly
what is needed to detect graphics cards on higher
numbered domains. Could the maintainer kindly consider
folding in similar changes?
The license is always questionable .. but it was Marc Aurele La France who
Definately would like to know if anyone is working on this. Can we use the code
from Xfree86 legally (if I use it and get it working, can it be accepted), or do
I have to find my own solution? Is anyone else working on this?
The "fixed" code is absolutely no use on my box at all. I'd like to see what
everyone else's /proc/bus/pci looks like to make sure I don't rule anyone out
when I go hacking away ... assuming we can't just plug in the Xfree version.
As far as I know, all code distributed by XFree86 can be reused by anyone legally,
as long as you follow the license terms. However, X.Org has made the decision
that the terms of the XFree86 1.1 license do not meet X.Org's goals, and will not
accept code under that license into the X.Org CVS. Redistributors may choose to
add code under that license into their own packages, but are responsible for
ensuring they meet the terms imposed by the license. Code under other licenses
in the XFree86 tree may be acceptable to import to X.Org (and much of it has been
in the original XFree86 4.4RC2 import for X.Org X11R6.7 release).
There is a PCI domain patch listed in bug #2368 that applies cleanly to my 6.8.2
tree. It does find my video card in domain 5, but then crashes leaving the
video card in an unknown state. I don't know which bug is a duplicate of which
since 2368 isn't sparc specific, this bug is, but I don't think the problem is
sparc specific, but thats what I'm testing on.
It only takes me about 3-4 hours to recompile X, so I'm gonna see if I figure
I've not tested the patch listed in the Xfree86 tree. Has anyone else?
The patch in bug #2368 will solve the remaining issues with PCI domains on
Sparc. I have a Sun Ultra E450 running kernel 2.6 and have tested the patch
listed in bug #2368 in multiple PCI domains. Works well and stable!
dup of 2368
*** This bug has been marked as a duplicate of 2368 ***