[ubuntu-x] Status of kernel X drivers
chase.douglas at ubuntu.com
Thu Feb 18 15:34:12 UTC 2010
Timo Aaltonen wrote:
> On Thu, 18 Feb 2010, Christopher James Halse Rogers wrote:
>> As far as nouveau goes, this is pretty obviously the best solution; we
>> won't be supporting nouveau except via the lbm packages, so there's no
>> real packaging overhead. Also, if we wanted to ship a kernel component
>> newer than that found in 2.6.33 we'd need to deal with the API bump that
>> has been threatened for ages an has now been made. This would be
>> difficult, but not impossible, to deal with. At this, post
>> feature-freeze stage, I don't think it's worth the effort.
If nouveau is provided through lbm, and the (what I presume)
linux-backports-modules-nouveau package is installed by default, then should we
be considering how this will affect the SRU process for lbm as well? The current
SRU process for lbm is much looser than for the linux-image package. I assume
this is because lbm isn't installed by default and isn't fully supported in the
same manner. If we are going to be supporting nouveau to the same degree as we
do drivers in linux-image, perhaps we should think about splitting the nouveau
drivers out as a separate package. This would ensure that the support levels of
the lbm drivers are uniform instead of the mix that would entail if we put
nouveau in lbm.
A second proposal I'd like to make would be integrating some of these drivers
into jockey. Right now jockey proposes proprietary, marginally supported drivers
for ubuntu. This is similar to lbm in that the drivers provided are not
supported to the same level as the default modules provided by linux-image.
There are many cases, however, where the lbm drivers may solve issues for ubuntu
users. Unfortunately, lbm is not really advertised anywhere. I believe it would
be useful for users to have the option through jockey to install backport
modules, with all the disclaimers of level of support. This helps us not only by
providing a possibly better experience for end users when lbm fixes their
issues, but also through testing. We would be more likely to see bug reports
like, "My hardware doesn't work in lucid until I install lbm". We can find and
vet individual fixes more easily and integrate them into the linux-image package.
This would be very useful for cases where the intel drivers in linux-image can't
be fixed due to non-trivial backporting of patches, but the updated drivers
could still be provided through lbm. Jockey could look for graphics chipsets
that we know function better through lbm and then suggest them to the user just
like it's done today for proprietary drivers.
More information about the kernel-team