[Intel-gfx] ✗ Fi.CI.IGT: failure for series starting with [01/46] drm/i915: Return immediately if trylock fails for direct-reclaim

Chris Wilson chris at chris-wilson.co.uk
Mon Jan 7 17:19:26 UTC 2019


Quoting Patchwork (2019-01-07 17:10:25)
> #### Possible regressions ####
> 
>   * igt at i915_missed_irq:
>     - shard-snb:          PASS -> FAIL
>     - shard-kbl:          PASS -> FAIL
>     - shard-hsw:          PASS -> FAIL
>     - shard-skl:          PASS -> FAIL
>     - shard-glk:          PASS -> FAIL
>     - shard-apl:          PASS -> FAIL

It's not really a bug, just a failure in how we test. (Not that I could
think of a better test.) The issue is that since we have coupled
together the waiting + dma_fence, every time we retire/signal the
fences, we remove the waiter. So in this test, the retire worker runs in
the background signaling all the fences before the
intel_breadcrumbs_hangcheck runs -- so we never see the missed interrupt
as we do asynchronous polling.

My conclusion is a patch to just remove the fake_irq as we do not need
it anymore.
-Chris


More information about the Intel-gfx mailing list