On Tue, Aug 02, 2016 at 10:00:12AM +0200, Daniel Vetter wrote:
On Tue, Aug 2, 2016 at 4:26 AM, Linus Torvalds torvalds@linux-foundation.org wrote:
On Mon, Aug 1, 2016 at 9:32 PM, Dave Airlie airlied@linux.ie wrote:
This is the main drm pull request for 4.8, I'm down with a cold at the moment so hopefully this isn't in too bad a state, I finished pulling stuff last week mostly (nouveau fixes just went in today), so only this message should be influenced by illness. Apologies to anyone who's major feature I missed :-)
i915: BXT support enabled by default GVT-g infrastructure GuC command submission and fixes BXT workarounds SKL/BKL workarounds Demidlayering device registration Thundering herd fixes Missing pci ids Atomic updates
Hmm. I did the merge and pushed it out, but testing it on my laptop shows some very annoying flickering problem.
The screen goes dark for a very short while (one frame? Who knows? Seems longer occasionally). I have no idea what triggers it, but it happens quite a lot when it happens. Like once every second or two. And it seems to happen most of the time, although right now it happens to be behaving nicely, so sometimes it goes for a while without the flickering.
Things *work*, but the flickering is nasty enough to make the end result painful to use.
The only thing I see in dmesg that looks bad is
[drm:intel_set_cpu_fifo_underrun_reporting [i915]] *ERROR* uncleared fifo underrun on pipe A [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
but I've seen that before, and it happens a couple of times during boot. Not once per second.
This is my old Vaio 11 Pro, now running Fedora 24 (up-to-date as of today).
So it's bog-standard intel graphics (i5-4200U - Haswell ULT).
Suggestions to try?
psr or fbc are the likely culprits. More likely fbc if the underruns correlate with the flicker (but note that by default we only report them once per modest, needs to be reset with a dpms or suspend/resume). Driver should even pick up the new module option settings at runtime (but again you need to force a modeset, just suspend/resume quickly), so fast to test.
I think FBC is still disabled by default on HSW.
So PSR seems more likely. The underruns might point at some watermark fail though :(
I have a couple of pending PSR patches you may want to try as well, if i915.enable_psr=0 helps.
First set is here: git://github.com/vsyrjala/linux.git psr_setup_time_2 This should be perfectly safe to go in actually, as it will only result in disabling PSR with certain panels.
The second set is here: git://github.com/vsyrjala/linux.git psr_fixes_2 This one I think is causing some kind of slight regression on one machine in our CI system. Still not sure what's going on there.