[Intel-gfx] [PATCH] drm/i915: Reset logical ring contexts' head and tail during GPU reset

[Daniel Vetter]

On Mon, Feb 16, 2015 at 04:12:53PM +0000, Thomas Daniel wrote:
> Work was getting left behind in LRC contexts during reset.  This causes a hang
> if the GPU is reset when HEAD==TAIL because the context's ringbuffer head and
> tail don't get reset and retiring a request doesn't alter them, so the ring
> still appears full.
> 
> Added a function intel_lr_context_reset() to reset head and tail on a LRC and
> its ringbuffer.
> 
> Call intel_lr_context_reset() for each context in i915_gem_context_reset() when
> in execlists mode.
> 
> Testcase: igt/pm_rps --run-subtest reset #bdw
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88096
> Signed-off-by: Thomas Daniel <thomas.daniel at intel.com>
> ---
>  drivers/gpu/drm/i915/i915_gem_context.c |   12 +++++++----
>  drivers/gpu/drm/i915/intel_lrc.c        |   34 +++++++++++++++++++++++++++++++
>  drivers/gpu/drm/i915/intel_lrc.h        |    2 ++
>  3 files changed, 44 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 8603bf4..70346b0 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -296,11 +296,15 @@ void i915_gem_context_reset(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  	int i;
>  
> -	/* In execlists mode we will unreference the context when the execlist
> -	 * queue is cleared and the requests destroyed.
> -	 */
> -	if (i915.enable_execlists)
> +	if (i915.enable_execlists) {
> +		struct intel_context *ctx;
> +
> +		list_for_each_entry(ctx, &dev_priv->context_list, link) {
> +			intel_lr_context_reset(dev, ctx);
> +		}
> +
>  		return;
> +	}

I'm somewhat voting for a vfunc for engine state reset since hiding the
lrc callback by going through legacy code doesn't look too pretty.
Follow-up patch if you're bored would be neat ;-)

>  
>  	for (i = 0; i < I915_NUM_RINGS; i++) {
>  		struct intel_engine_cs *ring = &dev_priv->ring[i];
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
> index aafcef3..1946bb9 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -1950,3 +1950,37 @@ error_unpin_ctx:
>  	drm_gem_object_unreference(&ctx_obj->base);
>  	return ret;
>  }
> +
> +void intel_lr_context_reset(struct drm_device *dev,
> +			struct intel_context *ctx)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct intel_engine_cs *ring;
> +	int i;
> +
> +	for_each_ring(ring, dev_priv, i) {
> +		struct drm_i915_gem_object *ctx_obj =
> +				ctx->engine[ring->id].state;
> +		if (ctx_obj) {

The common patter is
		
		if (!ctx_obj)
			continue;

which allows you to fold out one indent level. I couldn't resist and
done this while applying ...

Queued for -next, thanks for the patch.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch