[ubuntu-x] Lucid support for Intel(R) HD Graphics

[Peter Clifton]

On Fri, 2010-02-12 at 12:35 -0800, Bryce Harrington wrote:

> The PPA approach is best.  We're aware that there is stuff in 2.6.33
> that may require backporting onto 2.6.32 and we're currently trying to
> figure out how to handle it.  So for now, if people just verify that
> booting the mainline kernel solves it will be enough.  I think Intel can
> give us guidance on what to backport so no need to have bug reporters
> use up their time bisecting on this issue.

With regards back-porting, I've got the entire intel-drm/for-linus
branch running on top of 2.6.32 at the moment, and the back-porting
effort was quite manageable.

I have cheated somewhat, and am running the entire drm/ sub-tree from
that repository, along with the include/drm sub-tree swapped out from
the 2.6.32 headers, and replaced with those for "for-linus".

Once built, it just requires swapping out
the /lib/modules/*/kernel/drivers/drm/* modules - so could conceivably
be done from a Debian package providing some diversions?

There were two commits which rely on API changes outside the drm tree:

commit 2c761270d5520dd84ab0b4e47c24d99ff8503c38
Author: Dave Chinner <david at fromorbit.com>
Date:   Tue Jan 12 17:39:16 2010 +1100

    lib: Introduce generic list_sort function

And:

commit 4bdadb9785696439c6e2b3efe34aa76df1149c83
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date:   Wed Jan 27 13:36:32 2010 +0000

    drm/i915: Selectively enable self-reclaim

Which uses the new function new read_cache_page_gfp(). I was able to
pull in a bunch of code into drivers/gpu/drm/i915/i915_gem.c which
implements this function.


With the latest libdrm fix in Xorg edgers, and this kernel - things seem
generally ok. I've still got a few lid close -> GPU hang issues though,
and the occasional time the computer will come back from suspend with
its back-light stuck off.