drm/i915/gen9: unconditionally apply the memory bandwidth WA

Mahesh Kumar is already working on a proper implementation for the
workaround, but while we still don't have it, let's just
unconditionally apply the workaround for everybody and we hope we can
close all those numerous bugzilla tickets. Also, I'm not sure how easy
it will be to backport the final implementation to the stable Kernels,
and this patch here is probably easier to backport.

At the present moment I still don't have confirmation that this patch
fixes any of the bugs listed below, but we should definitely try
testing all of them again.

v2: s/intel_needs_memory_bw_wa/skl_needs_memory_bw_wa/ (Lyude).
v3: Rebase (dev -> dev_priv change on ilk_wm_max_level).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94337
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94605
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94884
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95010
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96226
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96828
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97450
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97830
Cc: stable@vger.kernel.org
Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Lyude <cpaul@redhat.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Lyude <cpaul@redhat.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1476210338-9797-1-git-send-email-paulo.r.zanoni@intel.com

diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 9ae23384976c4c7efa8b5258bff1e615c4427091..8b528b279aaa18e828c83ad860a52b85334a9fe0 100644 (file)
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2883,6 +2883,21 @@ skl_wm_plane_id(const struct intel_plane *plane)
        }
 }
 
+/*
+ * FIXME: We still don't have the proper code detect if we need to apply the WA,
+ * so assume we'll always need it in order to avoid underruns.
+ */
+static bool skl_needs_memory_bw_wa(struct intel_atomic_state *state)
+{
+       struct drm_i915_private *dev_priv = to_i915(state->base.dev);
+
+       if (IS_SKYLAKE(dev_priv) || IS_BROXTON(dev_priv) ||
+           IS_KABYLAKE(dev_priv))
+               return true;
+
+       return false;
+}
+
 static bool
 intel_has_sagv(struct drm_i915_private *dev_priv)
 {
@@ -3003,9 +3018,10 @@ bool intel_can_enable_sagv(struct drm_atomic_state *state)
        struct drm_device *dev = state->dev;
        struct drm_i915_private *dev_priv = to_i915(dev);
        struct intel_atomic_state *intel_state = to_intel_atomic_state(state);
-       struct drm_crtc *crtc;
+       struct intel_crtc *crtc;
+       struct intel_plane *plane;
        enum pipe pipe;
-       int level, plane;
+       int level, id, latency;
 
        if (!intel_has_sagv(dev_priv))
                return false;
@@ -3023,27 +3039,36 @@ bool intel_can_enable_sagv(struct drm_atomic_state *state)
 
        /* Since we're now guaranteed to only have one active CRTC... */
        pipe = ffs(intel_state->active_crtcs) - 1;
-       crtc = dev_priv->pipe_to_crtc_mapping[pipe];
+       crtc = to_intel_crtc(dev_priv->pipe_to_crtc_mapping[pipe]);
 
-       if (crtc->state->mode.flags & DRM_MODE_FLAG_INTERLACE)
+       if (crtc->base.state->mode.flags & DRM_MODE_FLAG_INTERLACE)
                return false;
 
-       for_each_plane(dev_priv, pipe, plane) {
+       for_each_intel_plane_on_crtc(dev, crtc, plane) {
+               id = skl_wm_plane_id(plane);
+
                /* Skip this plane if it's not enabled */
-               if (intel_state->wm_results.plane[pipe][plane][0] == 0)
+               if (intel_state->wm_results.plane[pipe][id][0] == 0)
                        continue;
 
                /* Find the highest enabled wm level for this plane */
                for (level = ilk_wm_max_level(dev_priv);
-                    intel_state->wm_results.plane[pipe][plane][level] == 0; --level)
+                    intel_state->wm_results.plane[pipe][id][level] == 0; --level)
                     { }
 
+               latency = dev_priv->wm.skl_latency[level];
+
+               if (skl_needs_memory_bw_wa(intel_state) &&
+                   plane->base.state->fb->modifier[0] ==
+                   I915_FORMAT_MOD_X_TILED)
+                       latency += 15;
+
                /*
                 * If any of the planes on this pipe don't enable wm levels
                 * that incur memory latencies higher then 30µs we can't enable
                 * the SAGV
                 */
-               if (dev_priv->wm.skl_latency[level] < SKL_SAGV_BLOCK_TIME)
+               if (latency < SKL_SAGV_BLOCK_TIME)
                        return false;
        }
 
@@ -3558,12 +3583,18 @@ static int skl_compute_plane_wm(const struct drm_i915_private *dev_priv,
        uint32_t width = 0, height = 0;
        uint32_t plane_pixel_rate;
        uint32_t y_tile_minimum, y_min_scanlines;
+       struct intel_atomic_state *state =
+               to_intel_atomic_state(cstate->base.state);
+       bool apply_memory_bw_wa = skl_needs_memory_bw_wa(state);
 
        if (latency == 0 || !cstate->base.active || !intel_pstate->base.visible) {
                *enabled = false;
                return 0;
        }
 
+       if (apply_memory_bw_wa && fb->modifier[0] == I915_FORMAT_MOD_X_TILED)
+               latency += 15;
+
        width = drm_rect_width(&intel_pstate->base.src) >> 16;
        height = drm_rect_height(&intel_pstate->base.src) >> 16;
 
@@ -3616,6 +3647,8 @@ static int skl_compute_plane_wm(const struct drm_i915_private *dev_priv,
                                 plane_blocks_per_line);
 
        y_tile_minimum = plane_blocks_per_line * y_min_scanlines;
+       if (apply_memory_bw_wa)
+               y_tile_minimum *= 2;
 
        if (fb->modifier[0] == I915_FORMAT_MOD_Y_TILED ||
            fb->modifier[0] == I915_FORMAT_MOD_Yf_TILED) {