Quoting Mika Kuoppala (2018-05-22 13:49:24)
> From: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
>
> When checking if engine is idling on a kernel context,
> the last request emitted to it could have been the exact
> request to switch into kernel context.
>
> Do not bail out early even if engine has requests,
> if the last request was for kernel context.
>
> Fixes: a89d1f921c15 ("drm/i915: Split i915_gem_timeline into individual timelines")
> Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@xxxxxxxxx>
> ---
> drivers/gpu/drm/i915/i915_gem_context.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index b69b18ef8120..3fe1212b0f7e 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -595,7 +595,10 @@ static bool engine_has_idle_kernel_context(struct intel_engine_cs *engine)
> lockdep_assert_held(&engine->i915->drm.struct_mutex);
>
> list_for_each_entry(ring, active_rings, active_link) {
> - if (last_request_on_engine(ring->timeline, engine))
> + struct i915_request *rq =
> + last_request_on_engine(ring->timeline, engine);
> +
> + if (rq && rq->gem_context != engine->i915->kernel_context)
> return false;
Ah, this isn't enough yet. The challenge is that we don't want to report
false when there is an active batch followed by the kernel context.
I think we can rely on the kernel request priority being the lowest
here...
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
