drm/i915: Always recompute watermarks when distrust_bios_wm is set, v2.
authorMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Wed, 31 May 2017 15:42:36 +0000 (17:42 +0200)
committerMaarten Lankhorst <maarten.lankhorst@linux.intel.com>
Thu, 1 Jun 2017 07:46:37 +0000 (09:46 +0200)
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.

Changes since v1:
- Do not take the connection_mutex, this is already done below.

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>
Link: http://patchwork.freedesktop.org/patch/msgid/20170531154236.27180-1-maarten.lankhorst@linux.intel.com
Reviewed-by: Mahesh Kumar <mahesh1.kumar@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
drivers/gpu/drm/i915/intel_pm.c

index 936eef1..fce4bc5 100644 (file)
@@ -4819,11 +4819,19 @@ 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;
 
        /*
+        * When we distrust bios wm we always need to recompute to set the
+        * expected DDB allocations for each CRTC.
+        */
+       if (to_i915(dev)->wm.distrust_bios_wm)
+               changed = true;
+
+       /*
         * If this transaction isn't actually touching any CRTC's, don't
         * bother with watermark calculation.  Note that if we pass this
         * test, we're guaranteed to hold at least one CRTC state mutex,
@@ -4833,6 +4841,7 @@ skl_compute_wm(struct drm_atomic_state *state)
         */
        for_each_new_crtc_in_state(state, crtc, cstate, i)
                changed = true;
+
        if (!changed)
                return 0;