On Wed, May 27, 2015 at 09:24:31AM +0300, Jani Nikula wrote:
> On Wed, 27 May 2015, Dave Airlie <airlied at gmail.com> wrote:
> > On 27 May 2015 at 01:17, Jani Nikula <jani.nikula at linux.intel.com> wrote:
> >> On Tue, 26 May 2015, Daniel Vetter <daniel at ffwll.ch> wrote:
> >>> On Tue, May 26, 2015 at 03:09:04PM +0200, Rainer Koenig wrote:
> >>>> Hi,
> >>>>
> >>>> a week ago I experienced problems on the skylake platform and got the
> >>>> adivce to try out the drm-intel-nightly branch. I tried and it was a
> >>>> success.
> >>>>
> >>>> So after the initial "git clone" of the tree I tried to keep updated by
> >>>> doing a "git pull" from time to time, but what's really strange is that
> >>>> I got merge conflicts, usually in the file integration-manifest, but
> >>>> sometimes also in source files.
> >>>>
> >>>> That's looks somewhat weird because I didn't touch any of the files in
> >>>> the tree and I thought that after cloning a frequent "git pull" will
> >>>> keep me up to date without the need to resolve merge conflicts.
> >>>>
> >>>> What is wrong with my thought? What did I do wrong?
> >>>
> >>> -nigthly is a rebasing tree, git pull does the wrong thing for that. The
> >>> proper way to track rebasing branches is (assuming you have no local
> >>> patches that you want to keep):
> >>>
> >>> $ git fetch origin
> >>> $ git reset --hard @{upstream}
> >>>
> >>>> Second, I pulled the "Linus"-tree today and found some log entries that
> >>>> said
> >>>> Merge branch 'drm-fixes-4.1' of
> >>>> git://people.freedesktop.org/~agd5f/linux into drm-fixes
> >>>>
> >>>> and
> >>>> Merge tag 'drm-intel-fixes-2015-05-21' of
> >>>> git://anongit.freedesktop.org/drm-intel into drm-fixes
> >>>>
> >>>> So I assumed, that the fix I proved to work one week ago now should also
> >>>> be available in the "vanilla" tree. So I compiled that on my test
> >>>> machine and got my bug back. :-(
> >>>>
> >>>> So my other question is, how do fixes from drm-intel-nightly find their
> >>>> weay into the "vanilla" linux tree? Is there some sort of process
> >>>> description.
> >>>
> >>> It takes a while. If the patch is in drm-intel-fixes, it will first got to
> >>> drm-fixes and then to vanilla upstream, then to stable kernels (if it's
> >>> cc: stable). You can check which branch a patch is in already with
> >>
> >> However we don't necessarily queue Skylake fixes to the current
> >> development kernels through drm-intel-fixes/drm-fixes, as the Skylake
> >> support there is anyway preliminary, the fix (I don't think we figured
> >> out which exact commit it was, did we?) may only end up upstream after
> >> the next merge window, i.e. at v4.2-rc1.
> >
> > So 4.1 won't cut it on skylake? I'm not sure everyone in Intel is aware.
>
> My approach has been that I don't queue fixes to current development
> kernels for platforms that still require i915.preliminary_hw_support=1
> or CONFIG_DRM_I915_PRELIMINARY_HW_SUPPORT=y. Obviously same for
> stable. I don't think this is unreasonable.
>
> Basically preliminary means the implementation does not have feature
> parity with the preceding platform, and is expected to be rough around
> th edges.
>
> Skylake still requires the preliminary flag even in
> drm-intel-nightly. Damien, should we drop preliminary for Skylake in
> v4.2? Or already in v4.1?
the planes abi in skl is still wrong, we expose both top plane both as the
new unified plane and through the cursor backwards interface the hw
provides. Until that's fixed (and that's not the case for 4.2) skl isn't
ready imo. It's a fairly small patch, we only need to stop registering the
legacy cursor plane and mark the top plane as the cursor one for backwards
abi purposes.
-Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
