Re: [PATCH v2 10/16] drm/i915: add struct ctx_reset_state

[Ian Romanick <idr@freedesktop.org>]

On 03/14/2013 08:52 AM, Mika Kuoppala wrote:
> To count context losses, add struct ctx_reset_state for
> both i915_hw_context and drm_i915_file_private.
> drm_i915_file_private is used when there is no context.
>
> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
> ---
>   drivers/gpu/drm/i915/i915_dma.c         |    4 +++-
>   drivers/gpu/drm/i915/i915_drv.h         |   19 +++++++++++++++++++
>   drivers/gpu/drm/i915/i915_gem_context.c |   11 +++++++++++
>   3 files changed, 33 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
> index e16099b..7902d97 100644
> --- a/drivers/gpu/drm/i915/i915_dma.c
> +++ b/drivers/gpu/drm/i915/i915_dma.c
> @@ -1792,7 +1792,7 @@ int i915_driver_open(struct drm_device *dev, struct drm_file *file)
>   	struct drm_i915_file_private *file_priv;
>
>   	DRM_DEBUG_DRIVER("\n");
> -	file_priv = kmalloc(sizeof(*file_priv), GFP_KERNEL);
> +	file_priv = kzalloc(sizeof(*file_priv), GFP_KERNEL);
>   	if (!file_priv)
>   		return -ENOMEM;
>
> @@ -1801,6 +1801,8 @@ int i915_driver_open(struct drm_device *dev, struct drm_file *file)
>   	spin_lock_init(&file_priv->mm.lock);
>   	INIT_LIST_HEAD(&file_priv->mm.request_list);
>
> +	i915_gem_context_init_reset_state(dev, &file_priv->reset_state);
> +
>   	idr_init(&file_priv->context_idr);
>
>   	return 0;
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index a54c507..d004548 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -433,6 +433,19 @@ struct i915_hw_ppgtt {
>   	void (*cleanup)(struct i915_hw_ppgtt *ppgtt);
>   };
>
> +struct ctx_reset_state {
> +	/* guilty and reset counts when context initialized */
> +	unsigned long guilty_cnt;
> +	unsigned long reset_cnt;

I think we can afford to spell out "count."  The first time I saw cnt, 
it looked like a dirty word. :)

I think this structure could you some better description of the overall 
architecture.  It's not completely obvious from the individual pieces... 
and that makes it really hard to evaluate.

reset_cnt is the number of resets since start-up.  What is guilty_cnt? 
What are innocent and guilty (below)?

All of this makes it difficult for me to tell whether or not the logic 
in patch 16 is correct... and I don't think it is.

> +
> +	unsigned innocent;
> +	unsigned guilty;
> +

         /* Time when this context was last blamed for a GPU reset.
          */
> +	unsigned long last_guilty_reset;
> +
> +	/* banned to submit more work */
> +	bool banned;
> +};
>
>   /* This must match up with the value previously used for execbuf2.rsvd1. */
>   #define DEFAULT_CONTEXT_ID 0
> @@ -443,6 +456,7 @@ struct i915_hw_context {
>   	struct drm_i915_file_private *file_priv;
>   	struct intel_ring_buffer *ring;
>   	struct drm_i915_gem_object *obj;
> +	struct ctx_reset_state reset_state;
>   };
>
>   enum no_fbc_reason {
> @@ -805,6 +819,7 @@ struct i915_gpu_error {
>
>   	unsigned long last_reset;
>
> +	unsigned long guilty_cnt;
>   	/**
>   	 * State variable and reset counter controlling the reset flow
>   	 *
> @@ -1257,6 +1272,8 @@ struct drm_i915_file_private {
>   		struct list_head request_list;
>   	} mm;
>   	struct idr context_idr;
> +
> +	struct ctx_reset_state reset_state;
>   };
>
>   #define INTEL_INFO(dev)	(((struct drm_i915_private *) (dev)->dev_private)->info)
> @@ -1677,6 +1694,8 @@ struct i915_hw_context * __must_check
>   i915_switch_context(struct intel_ring_buffer *ring,
>   		    struct drm_file *file, int to_id);
>   void i915_gem_context_free(struct kref *ctx_ref);
> +void i915_gem_context_init_reset_state(struct drm_device *dev,
> +				       struct ctx_reset_state *rs);
>   int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
>   				  struct drm_file *file);
>   int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
> diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
> index 8fb4d3c..dbd14b8 100644
> --- a/drivers/gpu/drm/i915/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/i915_gem_context.c
> @@ -145,6 +145,15 @@ static void do_destroy(struct i915_hw_context *ctx)
>   	kfree(ctx);
>   }
>
> +void i915_gem_context_init_reset_state(struct drm_device *dev,
> +				       struct ctx_reset_state *rs)
> +{
> +	struct drm_i915_private *dev_priv = dev->dev_private;
> +
> +	rs->reset_cnt = atomic_read(&dev_priv->gpu_error.reset_counter);
> +	rs->guilty_cnt = dev_priv->gpu_error.guilty_cnt;
> +}
> +
>   static struct i915_hw_context *
>   create_hw_context(struct drm_device *dev,
>   		  struct drm_i915_file_private *file_priv)
> @@ -177,6 +186,8 @@ create_hw_context(struct drm_device *dev,
>
>   	ctx->file_priv = file_priv;
>
> +	i915_gem_context_init_reset_state(dev, &ctx->reset_state);
> +
>   again:
>   	if (idr_pre_get(&file_priv->context_idr, GFP_KERNEL) == 0) {
>   		ret = -ENOMEM;
>