[PATCH 12/18] drm/i915/guc: Ensure request ordering via completion fences

[Daniele Ceraolo Spurio]

On 7/20/2021 3:39 PM, Matthew Brost wrote:
> If two requests are on the same ring, they are explicitly ordered by the
> HW. So, a submission fence is sufficient to ensure ordering when using
> the new GuC submission interface. Conversely, if two requests share a
> timeline and are on the same physical engine but different context this
> doesn't ensure ordering on the new GuC submission interface. So, a
> completion fence needs to be used to ensure ordering.
>
> v2:
>   (Daniele)
>    - Don't delete spin lock
>
> Signed-off-by: John Harrison <John.C.Harrison at Intel.com>
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> ---
>   drivers/gpu/drm/i915/i915_request.c | 12 ++++++++++--
>   1 file changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index ef26724fe980..3ecdc9180d8f 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -432,6 +432,7 @@ void i915_request_retire_upto(struct i915_request *rq)
>   
>   	do {
>   		tmp = list_first_entry(&tl->requests, typeof(*tmp), link);
> +		GEM_BUG_ON(!i915_request_completed(tmp));
>   	} while (i915_request_retire(tmp) && tmp != rq);
>   }
>   
> @@ -1380,6 +1381,9 @@ i915_request_await_external(struct i915_request *rq, struct dma_fence *fence)
>   	return err;
>   }
>   
> +static int
> +i915_request_await_request(struct i915_request *to, struct i915_request *from);
> +

I missed it in the previous rev, but this forward decl seems unneeded.
With it dropped:

Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>

Daniele

>   int
>   i915_request_await_execution(struct i915_request *rq,
>   			     struct dma_fence *fence)
> @@ -1463,7 +1467,8 @@ i915_request_await_request(struct i915_request *to, struct i915_request *from)
>   			return ret;
>   	}
>   
> -	if (is_power_of_2(to->execution_mask | READ_ONCE(from->execution_mask)))
> +	if (!intel_engine_uses_guc(to->engine) &&
> +	    is_power_of_2(to->execution_mask | READ_ONCE(from->execution_mask)))
>   		ret = await_request_submit(to, from);
>   	else
>   		ret = emit_semaphore_wait(to, from, I915_FENCE_GFP);
> @@ -1622,6 +1627,8 @@ __i915_request_add_to_timeline(struct i915_request *rq)
>   	prev = to_request(__i915_active_fence_set(&timeline->last_request,
>   						  &rq->fence));
>   	if (prev && !__i915_request_is_complete(prev)) {
> +		bool uses_guc = intel_engine_uses_guc(rq->engine);
> +
>   		/*
>   		 * The requests are supposed to be kept in order. However,
>   		 * we need to be wary in case the timeline->last_request
> @@ -1632,7 +1639,8 @@ __i915_request_add_to_timeline(struct i915_request *rq)
>   			   i915_seqno_passed(prev->fence.seqno,
>   					     rq->fence.seqno));
>   
> -		if (is_power_of_2(READ_ONCE(prev->engine)->mask | rq->engine->mask))
> +		if ((!uses_guc && is_power_of_2(READ_ONCE(prev->engine)->mask | rq->engine->mask)) ||
> +		    (uses_guc && prev->context == rq->context))
>   			i915_sw_fence_await_sw_fence(&rq->submit,
>   						     &prev->submit,
>   						     &rq->submitq);