On Mon, 2010-09-13 at 18:48 +0100, Matthew Garrett wrote:
> I've looked into this issue more closely and think I've worked out the
> underlying problem. The system in question appears to have two GPUs and
> exposes two ACPI backlight devices. Both of these are associated with
> existing PCI devices, so we don't ignore either of them because of that.
> Further, one of them (the AMD one) implements the spec properly and
> should work. We don't seem to perform a more fine-grained check to
> identify whether every ACPI backlight has all the required methods, and
> so as a result we provide both the working one and the non-working one.
>
> Having thought about this some more, I don't think this is the right
> approach. We should be ensuring that every backlight ahs all the
> required methods and then dropping the one that doesn't. This should be
> replaced with a native i915 backlight, and I sent patches to do that
> last week.
I agree. Your proposed design is good, and I have successfully tested
your proposed patches[1] (after minor porting changes to Ubuntu
Maverick's 2.6.35). Thanks very much Matthew!
FYI, I have published an experimental Ubuntu Maverick PPA kernel[2]
which includes your patches, plus my dell_laptop tweaks to inhibit the
broken dell_backlight by a module param or dmi blacklist table (in lieu
of a yet to be implemented more fine-grained check).
-Kamal Mostafa <kamal@xxxxxxxxxxxxx>
[1] 2010-09-08 [Intel-gfx] [PATCH] i915: Add native backlight control
2010-09-08 [Intel-gfx] [PATCH] Backlight: Add backlight type
[2] https://launchpad.net/~kamalmostafa/+archive/linux-kamal-mjgbacklight
Attachment:
signature.asc
Description: This is a digitally signed message part
