[Intel-gfx] [PATCH 05/20] drm/i915: Separate GPU hang waitqueue from advance

[Tvrtko Ursulin]

On 01/07/16 12:22, Chris Wilson wrote:
> Currently __i915_wait_request uses a per-engine wait_queue_t for the dual
> purpose of waking after the GPU advances or for waking after an error.
> In the future, we may add even more wake sources and require greater
> separation, but for now we can conceptually simplify wakeups by separating
> the two sources. In particular, this allows us to use different wait-queues
> (e.g. one on the engine advancement, a global one for errors and one on
> each requests) without any hassle.
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>   drivers/gpu/drm/i915/i915_drv.h |  6 ++++++
>   drivers/gpu/drm/i915/i915_gem.c |  5 +++++
>   drivers/gpu/drm/i915/i915_irq.c | 19 ++++---------------
>   3 files changed, 15 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 4948c90c9bd4..0d0e4ac4dadb 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1410,6 +1410,12 @@ struct i915_gpu_error {
>   #define I915_WEDGED			(1 << 31)
>
>   	/**
> +	 * Waitqueue to signal when a hang is detected. Used to for waiters
> +	 * to release the struct_mutex for the reset to procede.
> +	 */
> +	wait_queue_head_t wait_queue;
> +
> +	/**
>   	 * Waitqueue to signal when the reset has completed. Used by clients
>   	 * that wait for dev_priv->mm.wedged to settle.
>   	 */
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 34f724cc40b8..b607493a8d3a 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1455,6 +1455,7 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
>   	const bool irq_test_in_progress =
>   		ACCESS_ONCE(dev_priv->gpu_error.test_irq_rings) & intel_engine_flag(engine);
>   	int state = interruptible ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE;
> +	DEFINE_WAIT(reset);
>   	DEFINE_WAIT(wait);
>   	unsigned long timeout_expire;
>   	s64 before = 0; /* Only to silence a compiler warning. */
> @@ -1499,6 +1500,7 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
>   		goto out;
>   	}
>
> +	add_wait_queue(&dev_priv->gpu_error.wait_queue, &reset);
>   	for (;;) {
>   		struct timer_list timer;
>
> @@ -1551,6 +1553,8 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
>   			destroy_timer_on_stack(&timer);
>   		}
>   	}
> +	remove_wait_queue(&dev_priv->gpu_error.wait_queue, &reset);
> +
>   	if (!irq_test_in_progress)
>   		engine->irq_put(engine);
>
> @@ -5281,6 +5285,7 @@ i915_gem_load_init(struct drm_device *dev)
>   			  i915_gem_retire_work_handler);
>   	INIT_DELAYED_WORK(&dev_priv->mm.idle_work,
>   			  i915_gem_idle_work_handler);
> +	init_waitqueue_head(&dev_priv->gpu_error.wait_queue);
>   	init_waitqueue_head(&dev_priv->gpu_error.reset_queue);
>
>   	dev_priv->relative_constants_mode = I915_EXEC_CONSTANTS_REL_GENERAL;
> diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> index 83f40baeb1f3..6c17596d75dd 100644
> --- a/drivers/gpu/drm/i915/i915_irq.c
> +++ b/drivers/gpu/drm/i915/i915_irq.c
> @@ -2488,11 +2488,8 @@ static irqreturn_t gen8_irq_handler(int irq, void *arg)
>   	return ret;
>   }
>
> -static void i915_error_wake_up(struct drm_i915_private *dev_priv,
> -			       bool reset_completed)
> +static void i915_error_wake_up(struct drm_i915_private *dev_priv)
>   {
> -	struct intel_engine_cs *engine;
> -
>   	/*
>   	 * Notify all waiters for GPU completion events that reset state has
>   	 * been changed, and that they need to restart their wait after
> @@ -2501,18 +2498,10 @@ static void i915_error_wake_up(struct drm_i915_private *dev_priv,
>   	 */
>
>   	/* Wake up __wait_seqno, potentially holding dev->struct_mutex. */
> -	for_each_engine(engine, dev_priv)
> -		wake_up_all(&engine->irq_queue);
> +	wake_up_all(&dev_priv->gpu_error.wait_queue);
>
>   	/* Wake up intel_crtc_wait_for_pending_flips, holding crtc->mutex. */
>   	wake_up_all(&dev_priv->pending_flip_queue);
> -
> -	/*
> -	 * Signal tasks blocked in i915_gem_wait_for_error that the pending
> -	 * reset state is cleared.
> -	 */
> -	if (reset_completed)
> -		wake_up_all(&dev_priv->gpu_error.reset_queue);
>   }
>
>   /**
> @@ -2577,7 +2566,7 @@ static void i915_reset_and_wakeup(struct drm_i915_private *dev_priv)
>   		 * Note: The wake_up also serves as a memory barrier so that
>   		 * waiters see the update value of the reset counter atomic_t.
>   		 */
> -		i915_error_wake_up(dev_priv, true);
> +		wake_up_all(&dev_priv->gpu_error.reset_queue);
>   	}
>   }
>
> @@ -2714,7 +2703,7 @@ void i915_handle_error(struct drm_i915_private *dev_priv,
>   		 * ensure that the waiters see the updated value of the reset
>   		 * counter atomic_t.
>   		 */
> -		i915_error_wake_up(dev_priv, false);
> +		i915_error_wake_up(dev_priv);
>   	}
>
>   	i915_reset_and_wakeup(dev_priv);
>

Looks ok to me.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>

Regards,

Tvrtko