[Intel-gfx] [PATCH] Revert "drm/i915: use a separate context for gpu relocs"

[Chris Wilson]

Quoting Daniele Ceraolo Spurio (2019-12-03 22:19:07)
> 
> 
> On 11/29/19 4:48 AM, Chris Wilson wrote:
> > Since commit c45e788d95b4 ("drm/i915/tgl: Suspend pre-parser across GTT
> > invalidations"), we now disable the advanced preparser on Tigerlake for the
> > invalidation phase at the start of the batch, we no longer need to emit
> > the GPU relocations from a second context as they are now flushed inlined.
> > 
> 
> c45e788d95b4 only applies to the RCS though and IIRC I've seen issues 
> with the relocations on other engines as well, although they were much 
> rarer. Also, the comment left in intel_lrc.c still references reloc_gpu().

The tests we have don't discriminate between the engines, and the
so far we've only observed the issue on RCS. Look at igt_cs_tlb and see
what it might be missing wrt to triggering the issue on the other
engines.
-Chris