[Intel-gfx] [PATCH v3 26/33] drm/i915: Treat kernel context as initialised

[Joonas Lahtinen]

On pe, 2016-06-03 at 15:37 +0100, Chris Wilson wrote:
> The kernel context exists simply as a placeholder and should never be
> executed with a render context. It does not need the golden render
> state, as that will always be applied to a user context. By skipping the
> initialisation we can avoid issues in attempting to program the golden
> render context when trying to make the hardware idle.
> 
> v2: Rebase
> 
> Testcase: igt/drm_module_reload_basic #byt
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95634
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> ---
>  drivers/gpu/drm/i915/i915_gem_context.c | 14 +++++++++-----
>  drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++++++++
>  2 files changed, 19 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index c620fe6c9d96..d0076838e788 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -434,13 +434,17 @@ void i915_gem_context_lost(struct drm_i915_private *dev_priv)
>  			i915_gem_context_unpin(engine->last_context, engine);
>  			engine->last_context = NULL;
>  		}
> -
> -		/* Force the GPU state to be reinitialised on enabling */
> -		dev_priv->kernel_context->engine[engine->id].initialised =
> -			engine->init_context == NULL;
>  	}
>  
> -	/* Force the GPU state to be reinitialised on enabling */
> +	/* Force the GPU state to be restored on enabling */
> +	if (!i915.enable_execlists) {
> +		for_each_engine(engine, dev_priv) {
> +			struct intel_context *ce =

Here I would name it as "kce", to make explicit it's about kernel
context engines.

Reviewed-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>

> +				&dev_priv->kernel_context->engine[engine->id];
> +
> +			ce->initialised = true;
> +		}
> +	}
>  	dev_priv->kernel_context->remap_slice = ALL_L3_SLICES(dev_priv);
>  }
>  
> diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> index 92bb376e5039..1a389d0dcdd2 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> @@ -2257,6 +2257,16 @@ static int intel_ring_context_pin(struct i915_gem_context *ctx,
>  			goto error;
>  	}
>  
> +	/* The kernel context is only used as a placeholder for flushing the
> +	 * active context. It is never used for submitting user rendering and
> +	 * as such never requires the golden render context, and so we can skip
> +	 * emitting it when we switch to the kernel context. This is required
> +	 * as during eviction, we cannot allocate and pin the renderstate in
> +	 * order to initialise the context.
> +	 */
> +	if (ctx == ctx->i915->kernel_context)
> +		ce->initialised = true;
> +
>  	i915_gem_context_reference(ctx);
>  	return 0;
>  
-- 
Joonas Lahtinen
Open Source Technology Center
Intel Corporation