Re: [git pull] drm fixes for 4.10-rc6 (just missed rc5 tagging :-)

[Daniel Vetter]

On Thu, Jan 26, 2017 at 05:33:57AM +0100, Lukas Wunner wrote:
> On Wed, Jan 25, 2017 at 01:54:32PM +0100, Daniel Vetter wrote:
> > On Wed, Jan 25, 2017 at 06:10:57PM +0900, Michel Dänzer wrote:
> > > On 25/01/17 05:33 PM, Markus Trippelsdorf wrote:
> > > > On 2017.01.23 at 09:38 +1000, Dave Airlie wrote:
> > > >> ----------------------------------------------------------------
> > > >> Alex Deucher (8):
> > > >>       drm/radeon/si: load special ucode for certain MC configs
> > > >>       drm/amdgpu/si: load special ucode for certain MC configs
> > > >>       drm/amdgpu: drop oland quirks
> > > >>       drm/amdgpu: drop the mclk quirk for hainan
> > > >>       drm/radeon: drop oland quirks
> > > >>       drm/radeon: drop the mclk quirk for hainan
> > > >>       drm/radeon: add support for new hainan variants
> > > >>       drm/amdgpu: add support for new hainan variants
> > > > 
> > > > Since the merge I get the following warning during boot:
> > > 
> > > [...]
> > > 
> > > > [    2.627043] WARNING: CPU: 0 PID: 1 at ./include/drm/drm_crtc.h:857 drm_kms_helper_poll_init+0x127/0x140
> > > 
> > > This is likely due to
> > > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=3846fd9b86001bea171943cc3bb9222cb6da6b42
> > > 
> > > Daniel, please take a look.
> > 
> > Yup, I butchered this badly. The patch is correct, it also references the
> > right commits and if those commits would all be in 4.10 it would even
> > work. But the connector_list locking rework is only in drm-next, which
> > means it's totally not working :(
> > 
> > Dave, can you pls revert this in drm-fixes and re-apply to drm-next, and
> > then when you backmerge -fixes into -next make sure git doesn't drop it on
> > the floor? Or maybe for safety do it the other way round:
> > - first revert on drm-fixes
> > - then backmerge
> > - then reapply to drm-next.
> 
> It's not sufficient to only fix this in drm-next.  Without 3846fd9b8600,
> certain Optimus laptops won't (runtime) resume:
> https://bugzilla.kernel.org/show_bug.cgi?id=190861
> 
> Bjorn Helgaas therefore had queued up a revert for PCIe port runtime PM
> and only dequeued it once 3846fd9b8600 landed in Linus' tree.
> 
> Please re-submit 3846fd9b8600 and its prerequisites for drm-fixes at
> your earliest convenience.

That's way, way too much depencies for 4.10, the connector_locking rework
is by far not simple. We'll do this properly in 4.11, and for 4.10 the
most we'll be a quick hack to make nouveau happy (which we'll revert for
4.11 again). This probably means we need to delay the pci change until
4.11 too :(
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch