[ubuntu-x] Status of kernel X drivers

[Bryce Harrington]

On Thu, Feb 18, 2010 at 12:04:00PM +0100, Geir Ove Myhr wrote:
> On Thu, Feb 18, 2010 at 8:26 AM, Bryce Harrington <bryce at canonical.com> wrote:
> > On Thu, Feb 18, 2010 at 08:16:53AM +0100, Geir Ove Myhr wrote:
> >> > * ? - For Intel/KMS, we probably need to blacklist all the 8xx cards.
> >> > ? ? ?Some of them might work, but users could then selectively use LBM
> >> > ? ? ?(2.6.33) in those cases.
> >> Could you elaborate on this? What does blacklisting mean in this case?
> >> What kind of 8xx problems would blacklisting solve?
> > Blacklisting KMS. ?Although on reflection we need to blacklist and force
> > to vesa.
> 
> This doesn't sound like something our 8xx users will be very happy
> about. Is it the freeze problems [1] or other problems that is the
> reason for blacklisting? Is it supposed to be only temporary? Wouldn't
> it be better to stick with -intel 2.9.1 and re-enable UMS? Are there
> any bugs that can only be solved by using 2.10.0?

Well, you probably have a better eye on KMS problems with 8xx than I;
I know the 8xx cards don't get tested as thoroughly by upstream so am
assuming users have been running into problems with KMS, and so we'll
want a way to turn it on or off for particular cards.

If we ship -intel 2.9.1, then with KMS turned off it'll just fall back
to UMS.  That should be no problem.

If we ship 2.10, then since it doesn't support UMS any longer, we'd need
to have xserver load vesa in these cases.

I guess the real question is, "how well does KMS work on the various 8xx
cards"?  Geir, could you take a look through the current 8xx bug reports
and evaluate how good or bad it is?  If KMS is working fine across the
board, then we don't need to make any special provisions, which would be
nice.

This would also help drive an answer to whether to move to 2.10 as
planned or stick with 2.9.1.  Merging 2.10 is near the top of my todo
list but I'm open to sticking with 2.9.x if we have solid reasons to do
so.  But if we do that, we'd need to assemble a blocker list of bug
reports that we can give to Intel.

2.10 does bring some fixes, but hardware enablement is probably the
bigger driver for us.  I also suspect that upstream is going to require
any upstreamed bug reports be re-tested against 2.10, but I suppose as
long as it's available in a ppa somewhere, that shouldn't be a huge
issue for us.

Bryce