Re: [PATCH] drm/i915: Always recompute watermarks when distrust_bios_wm is set

[Mahesh Kumar <mahesh1.kumar@intel.com>]

Hi,


On Wednesday 31 May 2017 01:26 PM, Maarten Lankhorst wrote:
> On some systems there can be a race condition in which no crtc state is
> added to the first atomic commit. This results in all crtc's having a
> null DDB allocation, causing a FIFO underrun on any update until the
> first modeset.
>
> Reported-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Inspired-by: Mahesh Kumar <mahesh1.kumar@intel.com>
> Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Fixes: 98d39494d375 ("drm/i915/gen9: Compute DDB allocation at atomic
> check time (v4)")
> Cc: <stable@vger.kernel.org> # v4.8+
> Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
> Cc: Matt Roper <matthew.d.roper@intel.com>
> ---
>   drivers/gpu/drm/i915/intel_pm.c | 18 ++++++++++++++++++
>   1 file changed, 18 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index a488e068c3d6..8daa382b47f3 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -5037,10 +5037,27 @@ skl_compute_wm(struct drm_atomic_state *state)
>   	struct drm_crtc_state *cstate;
>   	struct intel_atomic_state *intel_state = to_intel_atomic_state(state);
>   	struct skl_wm_values *results = &intel_state->wm_results;
> +	struct drm_device *dev = state->dev;
>   	struct skl_pipe_wm *pipe_wm;
>   	bool changed = false;
>   	int ret, i;
>   
> +	if (to_i915(dev)->wm.distrust_bios_wm) {
> +		changed = true;
> +
> +		/*
> +		 * The first commit after driver load must update the ddb
> +		 * allocations, so make sure we at least hold a lock to
> +		 * inspect the state required for watermark updates.
> +		 *
> +		 * Below we'll grab the required crtc state.
> +		 */
> +		ret = drm_modeset_lock(&dev->mode_config.connection_mutex,
> +				       state->acquire_ctx);
Do we need to get lock here?
We will anyway go ahead & get the connection_mutex lock during 
skl_compute_ddb based on distruct_bios_wm flag.

-Mahesh
> +		if (ret)
> +			return ret;
> +	}
> +
>   	/*
>   	 * If this transaction isn't actually touching any CRTC's, don't
>   	 * bother with watermark calculation.  Note that if we pass this
> @@ -5051,6 +5068,7 @@ skl_compute_wm(struct drm_atomic_state *state)
>   	 */
>   	for_each_new_crtc_in_state(state, crtc, cstate, i)
>   		changed = true;
> +
>   	if (!changed)
>   		return 0;
>