[Intel-gfx] [PATCH 3/8] drm/i915: save/restore DSPARB only for chip before gen4

[Jesse Barnes]

On Mon, 21 Mar 2011 17:27:14 +0800
Zhenyu Wang <zhenyuw at linux.intel.com> wrote:

> Signed-off-by: Zhenyu Wang <zhenyuw at linux.intel.com>
> ---
>  drivers/gpu/drm/i915/i915_suspend.c |   12 ++++--------
>  1 files changed, 4 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_suspend.c b/drivers/gpu/drm/i915/i915_suspend.c
> index 422981b..7f00f8b 100644
> --- a/drivers/gpu/drm/i915/i915_suspend.c
> +++ b/drivers/gpu/drm/i915/i915_suspend.c
> @@ -603,7 +603,8 @@ void i915_save_display(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  
>  	/* Display arbitration control */
> -	dev_priv->saveDSPARB = I915_READ(DSPARB);
> +	if (dev_priv->info->gen < 4)
> +		dev_priv->saveDSPARB = I915_READ(DSPARB);
>  
>  	/* This is only meaningful in non-KMS mode */
>  	/* Don't save them in KMS mode */
> @@ -695,7 +696,8 @@ void i915_restore_display(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  
>  	/* Display arbitration */
> -	I915_WRITE(DSPARB, dev_priv->saveDSPARB);
> +	if (dev_priv->info->gen < 4)
> +		I915_WRITE(DSPARB, dev_priv->saveDSPARB);

Was it Cantiga that started doing automatic FIFO sizing for us?
Checking the docs...  Yeah, only Bearlake-C (G33?) and Cantiga do this
automatically.  So a blanket gen4 check probably isn't right.

>  
>  	/* Display port ratios (must be done before clock is set) */
>  	if (SUPPORTS_INTEGRATED_DP(dev)) {
> @@ -795,9 +797,6 @@ int i915_save_state(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  	int i;
>  
> -	/* Hardware status page */
> -	dev_priv->saveHWS = I915_READ(HWS_PGA);
> -
>  	i915_save_display(dev);
>  
>  	/* Interrupt state */
> @@ -842,9 +841,6 @@ int i915_restore_state(struct drm_device *dev)
>  	struct drm_i915_private *dev_priv = dev->dev_private;
>  	int i;
>  
> -	/* Hardware status page */
> -	I915_WRITE(HWS_PGA, dev_priv->saveHWS);
> -
>  	i915_restore_display(dev);
>  
>  	/* Interrupt state */


These hunks are unrelated to DSPARB; assuming we actually want to do
this it should be a separate patch.  Based on earlier reports, it
sounds like we're not properly idling across suspend/resume, in that
we're trying to execute commands at resume before setting up the HWS
again.


-- 
Jesse Barnes, Intel Open Source Technology Center