Re: [PATCH] drm/i915/gen9: Only copy WM results for changed pipes to skl_hw

[Lyude Paul <cpaul@redhat.com>]

Just posting this so it's on the ML.

NAK on this patch, this causes leftover cursors on SKL. Additionally, there's a
noticeable lag every time I move the cursor from one monitor to another, and the
flickering from the broken watermarks still happens.

On Thu, 2016-07-07 at 11:37 -0700, Matt Roper wrote:
> When we write watermark values to the hardware, those values are stored
> in dev_priv->wm.skl_hw.  However with recent watermark changes, the
> results structure we're copying from only contains valid watermark and
> DDB values for the pipes that are actually changing; the values for
> other pipes remain 0.  Thus a blind copy of the entire skl_wm_values
> structure will clobber the values for unchanged pipes...we need to be
> more selective and only copy over the values for the changing pipes.
> 
> This mistake was hidden until recently due to another bug that caused us
> to erroneously re-calculate watermarks for all active pipes rather than
> changing pipes.  Only when that bug was fixed was the impact of this bug
> discovered (e.g., modesets failing with "Requested display configuration
> exceeds system watermark limitations" messages and leaving watermarks
> non-functional, even ones initiated by intel_fbdev_restore_mode).
> 
> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Cc: Lyude Paul <cpaul@redhat.com>
> Fixes: 734fa01f3a17 ("drm/i915/gen9: Calculate watermarks during atomic
> 'check' (v2)")
> Fixes: 9b6130227495 ("drm/i915/gen9: Re-allocate DDB only for changed pipes")
> Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
> ---
>  drivers/gpu/drm/i915/intel_pm.c | 22 ++++++++++++++++++++--
>  1 file changed, 20 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index 5a8ee0c..2913535 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -4018,8 +4018,10 @@ static void skl_update_wm(struct drm_crtc *crtc)
>  	struct drm_device *dev = crtc->dev;
>  	struct drm_i915_private *dev_priv = to_i915(dev);
>  	struct skl_wm_values *results = &dev_priv->wm.skl_results;
> +	struct skl_wm_values *hw_vals = &dev_priv->wm.skl_hw;
>  	struct intel_crtc_state *cstate = to_intel_crtc_state(crtc->state);
>  	struct skl_pipe_wm *pipe_wm = &cstate->wm.skl.optimal;
> +	int pipe;
>  
>  	if ((results->dirty_pipes & drm_crtc_mask(crtc)) == 0)
>  		return;
> @@ -4031,8 +4033,24 @@ static void skl_update_wm(struct drm_crtc *crtc)
>  	skl_write_wm_values(dev_priv, results);
>  	skl_flush_wm_values(dev_priv, results);
>  
> -	/* store the new configuration */
> -	dev_priv->wm.skl_hw = *results;
> +	/*
> +	 * Store the new configuration (but only for the pipes that have
> +	 * changed; the other values weren't recomputed).
> +	 */
> +	for_each_pipe_masked(dev_priv, pipe, results->dirty_pipes) {
> +		hw_vals->wm_linetime[pipe] = results->wm_linetime[pipe];
> +		hw_vals->ddb.pipe[pipe] = results->ddb.pipe[pipe];
> +		memcpy(&hw_vals->ddb.plane[pipe], &results->ddb.plane[pipe],
> +		       sizeof(hw_vals->ddb.plane[pipe]));
> +		memcpy(&hw_vals->ddb.y_plane[pipe], &results-
> >ddb.y_plane[pipe],
> +		       sizeof(hw_vals->ddb.y_plane[pipe]));
> +		memcpy(&hw_vals->plane[pipe],
> +		       &results->plane[pipe],
> +		       sizeof(results->plane[pipe]));
> +		memcpy(&hw_vals->plane_trans[pipe],
> +		       &results->plane_trans[pipe],
> +		       sizeof(results->plane_trans[pipe]));
> +	}
>  
>  	mutex_unlock(&dev_priv->wm.wm_mutex);
>  }
-- 

Cheers,
	Lyude
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx