[Intel-gfx] regression bisected: display sync drops after 7221fc33

[dann frazier]

On Tue, Feb 02, 2016 at 06:04:32PM -0800, Matt Roper wrote:
> On Tue, Feb 02, 2016 at 05:35:49PM -0700, dann frazier wrote:
> > On Tue, Feb 02, 2016 at 02:51:20PM -0800, Matt Roper wrote:
> > > On Tue, Feb 02, 2016 at 03:21:30PM -0700, dann frazier wrote:
> > > > On Tue, Jan 26, 2016 at 09:40:27AM -0700, dann frazier wrote:
> > > > > On Tue, Jan 26, 2016 at 08:10:04AM -0700, dann frazier wrote:
> > > > > > On Tue, Jan 26, 2016 at 01:17:01PM +0200, Ville Syrjälä wrote:
> > > > > > > On Mon, Jan 25, 2016 at 09:35:26AM -0700, dann frazier wrote:
> > > > > > > > I've got a Lenovo T410 w/ an external display (DisplayPort<->HDMI). After
> > > > > > > > upgrading from a 4.2 kernel to a 4.3, I started hitting an issue where
> > > > > > > > the display thinks it has lost input for a few moments. Oddly, it
> > > > > > > > usually happens when I start typing in a terminal after a short period
> > > > > > > > of inactivity.
> > > > > > > > 
> > > > > > > > I see no OS-side symptoms - there are no new messages in dmesg/syslog/Xorg
> > > > > > > > log, etc.
> > > > > > > 
> > > > > > > Any fifo underrun reports in dmesg in general?
> > > > > > 
> > > > > > Yes, several:
> > > > > >   [ 1886.547409] [drm:ironlake_irq_handler [i915]] *ERROR* PCH transcoder A FIFO underrun
> > > > > > 
> > > > > > > > I reproduced with latest Linus master (@ 3e1e21c7b), and then
> > > > > > > > bisected down to the following commit:
> > > > > > > > 
> > > > > > > > commit 7221fc333dbe1743a3dff155b03527fda90d4ec1
> > > > > > > > Author: Matt Roper <matthew.d.roper at intel.com>
> > > > > > > > Date:   Thu Sep 24 15:53:08 2015 -0700
> > > > > > > > 
> > > > > > > >     drm/i915: Eliminate usage of pipe_wm_parameters from ILK-style WM (v2)
> > > > > > > 
> > > > > > > That's not in 4.3.
> > > > > > 
> > > > > > Ah, you're right. I was going by the version in the Makefile at that commit,
> > > > > > which says 4.3.0-rc3. git naiveté.
> > > > > > 
> > > > > > > As far as currentl Linus tree goes, you may want to try these:
> > > > > > > http://patchwork.freedesktop.org/patch/msgid/1452776015-22076-2-git-send-email-ville.syrjala@linux.intel.com
> > > > > > > http://patchwork.freedesktop.org/patch/msgid/1452776015-22076-1-git-send-email-ville.syrjala@linux.intel.com
> > > > > > 
> > > > > > Building - I'll test and report back.
> > > > > 
> > > > > fyi, the problem persists with 4.5-rc1 + these 2 patches.
> > > > 
> > > > Hi - Is there any further information I can provide to assist with this?
> > > > 
> > > >   -dann
> > > 
> > > Hi Dann.  Could you test with 
> > > 
> > > http://patchwork.freedesktop.org/patch/72194/
> > > 
> > > and see if it has any effect on the behavior you see?
> > 
> > Sure - what tree should I apply it to? (Doesn't seem to apply to
> > 4.5-rc2).
> > 
> >   -dann
> 
> The patch is written against branch drm-intel-nightly from
> git://anongit.freedesktop.org/drm-intel
> 
> Once the patch (or an updated version of it based on review feedback)
> gets merged to -nightly, it will get cherry-picked into our -fixes
> tree as well so that it shows up in an -rc release.

I'm not seeing the issue with this fix applied; and reverting just
this fix reintroduces the problem.

Tested-by: dann frazier <dannf at dannf.org>

Thanks!