[Intel-gfx] Video freezes continue to plague me

[Chris Wilson]

On Tue, Mar 15, 2016 at 07:29:22PM -0500, Chris wrote:
> On Tue, 2016-03-15 at 19:17 -0400, Felix Miata wrote:
> > Chris composed on 2016-03-15 17:43 (UTC-0500):
> > 
> > > On Tue, 2016-03-15 at 16:37 -0400, Felix Miata wrote:
> > 
> > >> Chris composed on 2016-03-15 14:50 (UTC-0500):
> > 
> > >> > My system is a Dell Optiplex 780.
> > 
> > >> Which? Tower? Desktop? SFF? USFF?
> > 
> > > Sorry, it's a desktop
> > 
> > >> > I've filed several bug reports both
> > >> > with the Intel Graphics Bugzilla and the Ubuntu Launchpad pages.
> > >> > Initially back when this started in 2014 every 'freeze' would result in
> > >> > an error of [drm:i915_hangcheck_elapsed [i915]] *ERROR* Hangcheck timer
> > >> > elapsed... render ring idle' in syslog. It would happen every few days
> > >> > or so and nothing that was done would correct the problem....
> > 
> > >> Was ensuring heat is a non-issue also done? SFF & USFF models tend to run
> > >> rather hot, so accumulated dust, and especially tobacco tar, within,
> > >> especially within the power supply, can become an intermittent problem for
> > >> software.
> > 
> > > Currently it's 87F outside. I have a window open directly behind the
> > > system and a fan blowing on the front of it. We're just not ready to
> > > turn the AC on yet. CPU0 is running around 116F and CPU1 109F. This
> > > whole issue of the video freezing whether it's with or without the
> > > Hangcheck error has been ongoing here on my system since 13 Sept 2014.
> > > The last time the video froze with the Hangcheck error was Jan 17 2016.
> > > Since then the video has frozen without the error on 1/21/16; 1/27/16;
> > > 2/25/16; 3/08/16; 3/11/16; and today 3/15/16 at 09:50am. I came back to
> > > the desktop at around 11am to find it frozen at 9:50. I moved the mouse,
> > > the video flickered and the clock caught up and came unfrozen. The only
> > > thing in my syslog for that time is:
> > 
> > > 09:50:07 localhost kernel: [347769.284331] [drm:drm_mode_addfb2] [FB:60]
> > > 09:50:07 localhost kernel: [347769.303256] [drm:drm_mode_addfb2] [FB:62]
> > > 09:50:07 localhost kernel: [347769.324369] [drm:drm_mode_addfb2] [FB:63]
> > 
> > Ambient temps tolerated by humans really shouldn't matter. Excess heat inside 
> > the case is often a problem elsewhere than where temp probes lie when dust 
> > has materially accumulated. What I was really asking is, is its inside 
> > reasonably clean, including inside the PSU? In 2014 it would have been around 
> > 5 or 6 years old. Also, is the PSU fan always moving some air?
> > 
> > Maybe it's time to give it an overnight memtest workout. Could be there's 
> > become an iffy bit in the video RAM area.
> > 
> > I have a 780 too, and a 760, and a 745, and several 620s, but they are all 
> > SFF, and don't often get left on more than a couple of hours at a time. Video 
> > lockups I don't remember ever seeing on any of them. Gnome isn't installed on 
> > any. Kubuntu on one 620 might be 14.10, but only openSUSE 13.1, 13.2, 42.1 
> > and Tumbleweed are on my 780. 13.2 uses kernel 3.16.7, 42.1 4.1.15, and TW 
> > (currently) 4.4.3.
> 
> Thanks Felix, I'll see if giving it a good will make any difference.

I finally tracked down a bug in the Xserver that would prevent Clients
from being woken up if the DRI2 Drawable was destroyed by another
Client. That is fixed in xorg-xserver-1.18.2.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre