Re: [PATCH stable-5.1 0/3] drm/i915: Prevent screen from flickering when the CDCLK changes

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Fri, Jun 14, 2019 at 01:59:51PM +0800, Jian-Hong Pan wrote:
> Ville Syrjälä <ville.syrjala@linux.intel.com> 於 2019年6月13日 週四 下午6:05寫道：
> >
> > On Thu, Jun 13, 2019 at 04:37:48PM +0800, Jian-Hong Pan wrote:
> > > Greg KH <gregkh@linuxfoundation.org> 於 2019年6月13日 週四 下午3:52寫道：
> > > >
> > > > On Mon, Jun 10, 2019 at 02:01:39PM +0800, Jian-Hong Pan wrote:
> > > > > Hi,
> > > > >
> > > > > After apply the commit "drm/i915: Force 2*96 MHz cdclk on glk/cnl when audio
> > > > > power is enabled", it induces the screen to flicker when the CDCLK changes on
> > > > > the laptop like ASUS E406MA. [1]
> > > > >
> > > > > So, we need these commits to prevent that:
> > > > > commit 48d9f87ddd21 drm/i915: Save the old CDCLK atomic state
> > > > > commit 2b21dfbeee72 drm/i915: Remove redundant store of logical CDCLK state
> > > > > commit 59f9e9cab3a1 drm/i915: Skip modeset for cdclk changes if possible
> > > > >
> > > > > [1]: https://bugzilla.kernel.org/show_bug.cgi?id=203623#c12
> > > > >
> > > > > Jian-Hong Pan
> > > > >
> > > > > Imre Deak (2):
> > > > >   drm/i915: Save the old CDCLK atomic state
> > > > >   drm/i915: Remove redundant store of logical CDCLK state
> > > > >
> > > > > Ville Syrjälä (1):
> > > > >   drm/i915: Skip modeset for cdclk changes if possible
> > > > >
> > > > >  drivers/gpu/drm/i915/i915_drv.h      |   3 +-
> > > > >  drivers/gpu/drm/i915/intel_cdclk.c   | 155 ++++++++++++++++++++++-----
> > > > >  drivers/gpu/drm/i915/intel_display.c |  48 +++++++--
> > > > >  drivers/gpu/drm/i915/intel_drv.h     |  18 +++-
> > > > >  4 files changed, 186 insertions(+), 38 deletions(-)
> > > >
> > > > These are all big patches, I would like to get an ack from the i915
> > > > developer(s) that these are acceptable for the stable tree before
> > > > applying them...
> > > >
> > > > thanks,
> > > >
> > > > greg k-h
> > >
> > > Hi Intel friends,
> > >
> > > We have laptops like ASUS E406MA, which hits the issue: The audio
> > > playback does not work anymore after suspend & resume [1]
> > > Thanks for your contribution.  We found the commit "drm/i915: Force
> > > 2*96 MHz cdclk on glk/cnl when audio power is enabled" can fix it.
> > > But, it induces the screen to flicker when the CDCLK changes on.  So,
> > > we need these commits to be back ported to Linux stable 5.1.x tree to
> > > prevent flickering:
> >
> > Pardon the delay.
> >
> > Now that I refreshed my memory a bit I can't really see how
> > this could fix anything, except by luck. Before these patches
> > we always forced cdclk to be >=2*96 MHz so audio should never
> > have hit this particular problem.
> >
> > The reason for adding this extra complexity was to claw back
> > a few milliwatts of power by allowing cdclk to drop below that
> > magic limit when audio isn't needed.
> >
> > I *think* these patches should probably work in 5.1, but for
> > now I don't see this as anything more than bandaid for an
> > unknown issue somewhere else. So I'd rather like a new bug
> > filed at https://bugs.freedesktop.org/enter_bug.cgi?product=DRI&component=DRM/Intel
> > with a full dmesg with drm.debug=0xe (+ some audio debug
> > knob(s) which show when it's trying to poke the hw during
> > suspend/resume) attached.
> 
> Thanks for your attention!
> I have filed the bug on freedesktop's bugzilla.
> https://bugs.freedesktop.org/show_bug.cgi?id=110916
> 
> I can upload more log information with guidance.

OK, I think I see what's going on. We would just need to change a few 
lines of code to fix that (basically just s/0/2*96000/ in two places)
but we don't want that upstream anymore due to these patches.

So I guess we need to go ahead with the backport. I read through them
and checked the 5.1 baseline and couldn't spot any problems. So for
the backport to 5.1:
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

-- 
Ville Syrjälä
Intel