pci rework not quite ready yet?

[Keith Packard]

After converting the intel driver to the pciaccess API, I've discovered
a couple of shortcomings that may need to be addressed before this code
is really ready for use.

The first problem was that we continue to have a legacy code in the
server which passes bus addresses around, like xf86MapVidMem.

The second problem is that only xf86MapVidMem has the mtrr manipulation
code required for X servers to run on x86 machines. libpciaccess must
perform mtrr (or PAT, at some point) manipulation as a part of the
memory mapping/unmapping process.

Finally, I've been unable to figure out why DRI-based Mesa locks up my
machines; switching back to the legacy code without changing the kernel
module, DRM library or Mesa driver makes things work again, so it's
clearly something in the X server or my driver causing the problem.
Examining log files shows no obvious cause.

I've pushed a 'no-pci-rework' branch to the X server repository so that
people can continue to use master for their work in areas not related to
fixing the pciaccess code. I'm wondering if it makes sense to make that
'master' until pciaccess resolves these issues.

-- 
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg/attachments/20070828/4d46e3b9/attachment.pgp>