[Intel-gfx] ✗ Fi.CI.BAT: failure for drm/i915: Rest of my DSI and DPLL stuff

[Ville Syrjälä]

On Thu, Apr 14, 2016 at 10:24:23AM +0300, Tomi Sarvela wrote:
> On Wednesday 13 April 2016 19:29:25 Ville Syrjälä wrote:
> [ILK-hp8440p problems]
> > ...
> > 
> > [  265.316531] drm/i915: Resetting chip after gpu hang
> > [  265.316585] [drm:i915_reset [i915]] *ERROR* Failed to reset chip: -110
> > 
> > Looks like the GPU died for some reason.
> > 
> > > Test kms_force_connector_basic:
> > >         Subgroup prune-stale-modes:
> > >                 pass       -> SKIP       (ivb-t430s)
> > 
> > No clue what happened here. The change isn't visible in the results
> > visualization page, nor can I see this subtest even being run in
> > the dmesg for Patchwork_1881. The dmesg ends at gtt-blt, but based on
> > the json there were a ton of tests run after that. In fact the json
> > and dmesg seem totally out of sync.
> > 
> > Tomi, any idea what's going on here?
> 
> The link from Patchwork page is hosed, but the full dmesgs are at
> http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/Patchwork_1881/ilk-hp8440p/[1] 
> http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/Patchwork_1881/ivb-t430s/[2] 

Not sure what you mean with hosed. It took me to the same place where I
was before (ie. [2]). But if you look at the dmesg vs. results.json the
order of the tests is totally different. I thought the json would show
them in the order they were run, but maybe I'm wrong?

I did find the skipped kms_force_connector_basic/prune-stale-modes in
the dmesg eventually, but it was a bit more difficult since it skipped
already from a fixture, and so the subtest name was never recorded in
the dmesg, which is a bit unfortunate. As far as the skipping goes, it
look like the VGA connector flips to 'connected' after every suspend
for some reason, and then goes back to 'disconnected' at some point. So
it feels like we have a bug in the connector resume handling still.

> 
> The dmesg is synced, but if the reset happens with delay, then it might be in wrong 
> place. I'm suggesting to search for  test name and looking upwards from there.
> 
> > > Results at /archive/results/CI_IGT_test/Patchwork_1881/
> 
> Tomi
> 
> --------
> [1] http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/Patchwork_1881/ilk-hp8440p/
> [2] http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/Patchwork_1881/ivb-t430s/

-- 
Ville Syrjälä
Intel OTC