[PATCH 2/6] drm/i915/guc: Fix issues with live_preempt_cancel

[John Harrison]

On 7/27/2022 19:42, John.C.Harrison at Intel.com wrote:
> From: Matthew Brost <matthew.brost at intel.com>
>
> Having semaphores results in different behavior when a dependent request
> is cancelled. In the case of semaphores the request could be on the HW
> and complete successfully while without the request is held in the
> driver and the error from the dependent request is propagated. Fix
> live_preempt_cancel to take this behavior into account.
>
> Also update live_preempt_cancel to use new function intel_context_ban
> rather than intel_context_set_banned.
>
> Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> Signed-off-by: John Harrison <John.C.Harrison at Intel.com>
Reviewed-by: John Harrison <John.C.Harrison at Intel.com>

> ---
>   drivers/gpu/drm/i915/gt/selftest_execlists.c | 16 +++++++++++-----
>   1 file changed, 11 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/selftest_execlists.c b/drivers/gpu/drm/i915/gt/selftest_execlists.c
> index 02fc97a0ab502..015f8cd3463e2 100644
> --- a/drivers/gpu/drm/i915/gt/selftest_execlists.c
> +++ b/drivers/gpu/drm/i915/gt/selftest_execlists.c
> @@ -2087,7 +2087,7 @@ static int __cancel_active0(struct live_preempt_cancel *arg)
>   		goto out;
>   	}
>   
> -	intel_context_set_banned(rq->context);
> +	intel_context_ban(rq->context, rq);
>   	err = intel_engine_pulse(arg->engine);
>   	if (err)
>   		goto out;
> @@ -2146,7 +2146,7 @@ static int __cancel_active1(struct live_preempt_cancel *arg)
>   	if (err)
>   		goto out;
>   
> -	intel_context_set_banned(rq[1]->context);
> +	intel_context_ban(rq[1]->context, rq[1]);
>   	err = intel_engine_pulse(arg->engine);
>   	if (err)
>   		goto out;
> @@ -2229,7 +2229,7 @@ static int __cancel_queued(struct live_preempt_cancel *arg)
>   	if (err)
>   		goto out;
>   
> -	intel_context_set_banned(rq[2]->context);
> +	intel_context_ban(rq[2]->context, rq[2]);
>   	err = intel_engine_pulse(arg->engine);
>   	if (err)
>   		goto out;
> @@ -2244,7 +2244,13 @@ static int __cancel_queued(struct live_preempt_cancel *arg)
>   		goto out;
>   	}
>   
> -	if (rq[1]->fence.error != 0) {
> +	/*
> +	 * The behavior between having semaphores and not is different. With
> +	 * semaphores the subsequent request is on the hardware and not cancelled
> +	 * while without the request is held in the driver and cancelled.
> +	 */
> +	if (intel_engine_has_semaphores(rq[1]->engine) &&
> +	    rq[1]->fence.error != 0) {
>   		pr_err("Normal inflight1 request did not complete\n");
>   		err = -EINVAL;
>   		goto out;
> @@ -2292,7 +2298,7 @@ static int __cancel_hostile(struct live_preempt_cancel *arg)
>   		goto out;
>   	}
>   
> -	intel_context_set_banned(rq->context);
> +	intel_context_ban(rq->context, rq);
>   	err = intel_engine_pulse(arg->engine); /* force reset */
>   	if (err)
>   		goto out;