diff mbox

[RFC,8/8] drm/i915/vlv: Add intermediate field in intel_crtc_wm_state and handlers for two-level watermark

Message ID 1466672279-16236-9-git-send-email-chix.ding@intel.com (mailing list archive)
State New, archived
Headers show

Commit Message

chix.ding@intel.com June 23, 2016, 8:57 a.m. UTC
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

Rename vlv_compute_wm to vlv_compute_pipe_wm to compute optimal watermark
Add vlv_compute_intermediate_wm to computer intermediate watermark
Add vlv_initial_watermarks to write intermediate watermark into hardware
Add vlv_optimize_watermarks to write optimal watermark into hardware
Change valleyview_crtc_enable to call .initial_watermarks handler

This patch adds the handlers for two-level atomic watermark for VLV/CHV.
It makes use of the optimal and intermediate watermark fields added in
the previous commits to calculate the optimal and intermediate state.
It sets the intermediate watermark which is the safer value of the
currently active and the optimal watermark pre-vblank. Then it sets the
optimal watermark after-vblank.

v2:
- use macro drm_atomic_crtc_state_for_each_plane_state
- remove redundant debug statements in vlv_pipe_set_fifo_size

v3:
- use macro drm_atomic_crtc_state_for_each_plane_state to simplify the code
- check !new_state->active || modeset in computing intermediate watermark

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Chi Ding <chix.ding@intel.com>
---
 drivers/gpu/drm/i915/intel_display.c |   7 +-
 drivers/gpu/drm/i915/intel_drv.h     |   3 +-
 drivers/gpu/drm/i915/intel_pm.c      | 150 +++++++++++++++++++++++++++--------
 3 files changed, 121 insertions(+), 39 deletions(-)
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 3f8987b..431d082 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -4598,8 +4598,6 @@  static void intel_post_plane_update(struct intel_crtc_state *old_crtc_state)
 
 	intel_frontbuffer_flip(dev, pipe_config->fb_bits);
 
-	crtc->wm.cxsr_allowed = true;
-
 	if (pipe_config->update_wm_post && pipe_config->base.active)
 		intel_update_watermarks(&crtc->base);
 
@@ -4645,7 +4643,6 @@  static void intel_pre_plane_update(struct intel_crtc_state *old_crtc_state)
 	}
 
 	if (pipe_config->disable_cxsr) {
-		crtc->wm.cxsr_allowed = false;
 
 		/*
 		 * Vblank time updates from the shadow to live plane control register
@@ -6092,7 +6089,7 @@  static void valleyview_crtc_enable(struct drm_crtc *crtc)
 
 	intel_color_load_luts(&pipe_config->base);
 
-	intel_update_watermarks(crtc);
+	dev_priv->display.initial_watermarks(pipe_config);
 	intel_enable_pipe(intel_crtc);
 
 	assert_vblank_disabled(crtc);
@@ -14310,8 +14307,6 @@  static void intel_crtc_init(struct drm_device *dev, int pipe)
 	intel_crtc->cursor_cntl = ~0;
 	intel_crtc->cursor_size = ~0;
 
-	intel_crtc->wm.cxsr_allowed = true;
-
 	BUG_ON(pipe >= ARRAY_SIZE(dev_priv->plane_to_crtc_mapping) ||
 	       dev_priv->plane_to_crtc_mapping[intel_crtc->plane] != NULL);
 	dev_priv->plane_to_crtc_mapping[intel_crtc->plane] = &intel_crtc->base;
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 2d3dfec..c207148 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -447,6 +447,7 @@  struct intel_crtc_wm_state {
 		} ilk;
 
 		struct {
+			struct vlv_wm_state intermediate;
 			struct vlv_wm_state optimal;
 		} vlv;
 
@@ -681,8 +682,6 @@  struct intel_crtc {
 			struct vlv_wm_state vlv;
 		} active;
 
-		/* allow CxSR on this pipe */
-		bool cxsr_allowed;
 	} wm;
 
 	int scanline_offset;
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index ceded5b..d9d1ecb 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -30,6 +30,7 @@ 
 #include "intel_drv.h"
 #include "../../../platform/x86/intel_ips.h"
 #include <linux/module.h>
+#include <drm/drm_atomic_helper.h>
 
 /**
  * DOC: RC6
@@ -964,6 +965,7 @@  static uint16_t vlv_compute_wm_level(struct intel_plane *plane,
 	clock = cstate->base.adjusted_mode.crtc_clock;
 	htotal = cstate->base.adjusted_mode.crtc_htotal;
 	width = cstate->pipe_src_w;
+
 	if (WARN_ON(htotal == 0))
 		htotal = 1;
 
@@ -986,27 +988,28 @@  static uint16_t vlv_compute_wm_level(struct intel_plane *plane,
 static void vlv_compute_fifo(struct intel_crtc_state *cstate,
 				struct vlv_wm_state *wm_state)
 {
-	struct intel_crtc *crtc = to_intel_crtc(cstate->base.crtc);
-	struct drm_device *dev = crtc->base.dev;
-	struct intel_plane *plane;
 	unsigned int total_rate = 0;
 	const int fifo_size = 512 - 1;
 	int fifo_extra, fifo_left = fifo_size;
 	int rate[I915_MAX_PLANES] = {};
 	int i;
+	const struct drm_plane_state *pstate;
+	struct drm_plane *plane;
 
-	for_each_intel_plane_on_crtc(dev, crtc, plane) {
+	drm_atomic_crtc_state_for_each_plane_state(plane, pstate, &cstate->base) {
 		struct intel_plane_state *state =
-			to_intel_plane_state(plane->base.state);
+			to_intel_plane_state(pstate);
+		struct intel_plane *intel_plane =
+			to_intel_plane(plane);
 
-		if (plane->base.type == DRM_PLANE_TYPE_CURSOR)
+		if (intel_plane->base.type == DRM_PLANE_TYPE_CURSOR)
 			continue;
 
 		if (state->visible) {
 			wm_state->num_active_planes++;
-			rate[wm_plane_id(plane)] =
+			rate[wm_plane_id(intel_plane)] =
 			drm_format_plane_cpp(state->base.fb->pixel_format, 0);
-			total_rate += rate[wm_plane_id(plane)];
+			total_rate += rate[wm_plane_id(intel_plane)];
 		}
 	}
 
@@ -1089,18 +1092,19 @@  static void vlv_invert_wms(struct intel_crtc *crtc,
 	}
 }
 
-static int vlv_compute_wm(struct intel_crtc_state *cstate)
+static int vlv_compute_pipe_wm(struct intel_crtc_state *cstate)
 {
 	struct intel_crtc *crtc = to_intel_crtc(cstate->base.crtc);
 	struct drm_device *dev = crtc->base.dev;
 	struct vlv_wm_state *wm_state = &cstate->wm.vlv.optimal;
-	struct intel_plane *plane;
+	struct drm_plane *plane;
 	int sr_fifo_size = INTEL_INFO(dev)->num_pipes * 512 - 1;
 	int level;
+	const struct drm_plane_state *pstate;
 
 	memset(wm_state, 0, sizeof(*wm_state));
 
-	wm_state->cxsr = crtc->pipe != PIPE_C && crtc->wm.cxsr_allowed;
+	wm_state->cxsr = crtc->pipe != PIPE_C;
 	wm_state->num_levels = to_i915(dev)->wm.max_level + 1;
 
 	wm_state->num_active_planes = 0;
@@ -1117,31 +1121,33 @@  static int vlv_compute_wm(struct intel_crtc_state *cstate)
 		}
 	}
 
-	for_each_intel_plane_on_crtc(dev, crtc, plane) {
+	drm_atomic_crtc_state_for_each_plane_state(plane, pstate, &cstate->base) {
 		struct intel_plane_state *state =
-			to_intel_plane_state(plane->base.state);
+			to_intel_plane_state(pstate);
+		struct intel_plane *intel_plane =
+			to_intel_plane(plane);
 
 		if (!state->visible)
 			continue;
 
 		/* normal watermarks */
 		for (level = 0; level < wm_state->num_levels; level++) {
-			int wm = vlv_compute_wm_level(plane, cstate, state, level);
-			int max_wm = plane->base.type == DRM_PLANE_TYPE_CURSOR ? 63 : 511;
+			int wm = vlv_compute_wm_level(intel_plane, cstate, state, level);
+			int max_wm = intel_plane->base.type == DRM_PLANE_TYPE_CURSOR ? 63 : 511;
 
 			if (level == 0 && wm > max_wm) {
 				DRM_DEBUG_KMS("Requested display configuration "
 				"exceeds system watermark limitations\n");
 				DRM_DEBUG_KMS("Plane %d.%d: blocks required = %u/%u\n",
-					crtc->pipe,
-					drm_plane_index(&plane->base), wm, max_wm);
+				      crtc->pipe,
+				      drm_plane_index(&intel_plane->base), wm, max_wm);
 				return -EINVAL;
 			}
 
-			if (wm > wm_state->fifo_size[wm_plane_id(plane)])
+			if (wm > wm_state->fifo_size[wm_plane_id(intel_plane)])
 				break;
 
-			switch (plane->base.type) {
+			switch (intel_plane->base.type) {
 				int sprite;
 			case DRM_PLANE_TYPE_CURSOR:
 				wm_state->wm[level].cursor = wm;
@@ -1150,7 +1156,7 @@  static int vlv_compute_wm(struct intel_crtc_state *cstate)
 				wm_state->wm[level].primary = wm;
 				break;
 			case DRM_PLANE_TYPE_OVERLAY:
-				sprite = plane->plane;
+				sprite = intel_plane->plane;
 				wm_state->wm[level].sprite[sprite] = wm;
 				break;
 			}
@@ -1162,7 +1168,7 @@  static int vlv_compute_wm(struct intel_crtc_state *cstate)
 			continue;
 
 		/* maxfifo watermarks */
-		switch (plane->base.type) {
+		switch (intel_plane->base.type) {
 			int sprite, level;
 		case DRM_PLANE_TYPE_CURSOR:
 			for (level = 0; level < wm_state->num_levels; level++)
@@ -1176,7 +1182,7 @@  static int vlv_compute_wm(struct intel_crtc_state *cstate)
 					    wm_state->wm[level].primary);
 			break;
 		case DRM_PLANE_TYPE_OVERLAY:
-			sprite = plane->plane;
+			sprite = intel_plane->plane;
 			for (level = 0; level < wm_state->num_levels; level++)
 				wm_state->sr[level].plane =
 					min(wm_state->sr[level].plane,
@@ -1196,6 +1202,65 @@  static int vlv_compute_wm(struct intel_crtc_state *cstate)
 	return 0;
 }
 
+
+static int vlv_compute_intermediate_wm(struct drm_device *dev,
+				       struct intel_crtc *intel_crtc,
+				       struct intel_crtc_state *newstate)
+{
+	struct vlv_wm_state *a = &newstate->wm.vlv.intermediate;
+	struct vlv_wm_state *b = &intel_crtc->wm.active.vlv;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	int level, max_level = dev_priv->wm.max_level;
+
+	/*
+	 * Start with the final, target watermarks, then combine with the
+	 * currently active watermarks to get values that are safe both before
+	 * and after the vblank.
+	 */
+	*a = newstate->wm.vlv.optimal;
+
+	if (!newstate->base.active || drm_atomic_crtc_needs_modeset(&newstate->base))
+		return 0;
+
+	a->num_levels = min(a->num_levels, b->num_levels);
+	a->cxsr &= b->cxsr;
+
+	for (level = 0; level < a->num_levels; level++) {
+		struct vlv_pipe_wm *a_wm = &a->wm[level];
+		const struct vlv_pipe_wm *b_wm = &b->wm[level];
+		struct vlv_sr_wm *a_sr = &a->sr[level];
+		const struct vlv_sr_wm *b_sr = &b->sr[level];
+
+		a_wm->primary = min(a_wm->primary, b_wm->primary);
+		a_wm->sprite[0] = min(a_wm->sprite[0], b_wm->sprite[0]);
+		a_wm->sprite[1] = min(a_wm->sprite[1], b_wm->sprite[1]);
+		a_wm->cursor = min(a_wm->cursor, b_wm->cursor);
+
+		if (a->cxsr) {
+			a_sr->plane = min(a_sr->plane, b_sr->plane);
+			a_sr->cursor = min(a_sr->cursor, b_sr->cursor);
+		} else {
+			a_sr->plane = b_sr->plane;
+			a_sr->cursor = b_sr->cursor;
+		}
+	}
+
+	/* reset the invalid levels */
+	for (level = a->num_levels; level <= max_level; level++) {
+		memset(&a->wm[level], 0, sizeof(struct vlv_pipe_wm));
+		memset(&a->sr[level], 0, sizeof(struct vlv_sr_wm));
+	}
+
+	/*
+	 * If our intermediate WM are identical to the final WM, then we can
+	 * omit the post-vblank programming; only update if it's different.
+	 */
+	if (memcmp(a, &newstate->wm.ilk.optimal, sizeof(*a)) == 0)
+		newstate->wm.need_postvbl_update = false;
+
+	return 0;
+}
+
 #define VLV_FIFO(plane, value) \
 	(((value) << DSPARB_ ## plane ## _SHIFT_VLV) & DSPARB_ ## plane ## _MASK_VLV)
 
@@ -1330,11 +1395,7 @@  static void vlv_update_wm(struct drm_crtc *crtc)
 	enum pipe pipe = intel_crtc->pipe;
 	struct vlv_wm_values wm = {};
 
-	vlv_compute_wm(intel_crtc->config);
-	mutex_lock(&dev_priv->wm.wm_mutex);
-	intel_crtc->wm.active.vlv = intel_crtc->config->wm.vlv.optimal;
 	vlv_merge_wm(dev, &wm);
-	mutex_unlock(&dev_priv->wm.wm_mutex);
 
 	if (memcmp(&dev_priv->wm.vlv, &wm, sizeof(wm)) == 0) {
 		/* FIXME should be part of crtc atomic commit */
@@ -1378,6 +1439,29 @@  static void vlv_update_wm(struct drm_crtc *crtc)
 	dev_priv->wm.vlv = wm;
 }
 
+
+static void vlv_initial_watermarks(struct intel_crtc_state *cstate)
+{
+	struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev);
+	struct intel_crtc *intel_crtc = to_intel_crtc(cstate->base.crtc);
+
+	mutex_lock(&dev_priv->wm.wm_mutex);
+	intel_crtc->wm.active.vlv = cstate->wm.vlv.intermediate;
+	vlv_update_wm(cstate->base.crtc);
+	mutex_unlock(&dev_priv->wm.wm_mutex);
+}
+
+static void vlv_optimize_watermarks(struct intel_crtc_state *cstate)
+{
+	struct drm_i915_private *dev_priv = to_i915(cstate->base.crtc->dev);
+	struct intel_crtc *intel_crtc = to_intel_crtc(cstate->base.crtc);
+
+	mutex_lock(&dev_priv->wm.wm_mutex);
+	intel_crtc->wm.active.vlv = cstate->wm.vlv.optimal;
+	vlv_update_wm(cstate->base.crtc);
+	mutex_unlock(&dev_priv->wm.wm_mutex);
+}
+
 #define single_plane_enabled(mask) is_power_of_2(mask)
 
 static void g4x_update_wm(struct drm_crtc *crtc)
@@ -7406,12 +7490,16 @@  void intel_init_pm(struct drm_device *dev)
 			DRM_DEBUG_KMS("Failed to read display plane latency. "
 				      "Disable CxSR\n");
 		}
-	} else if (IS_CHERRYVIEW(dev)) {
-		vlv_setup_wm_latency(dev);
-		dev_priv->display.update_wm = vlv_update_wm;
-	} else if (IS_VALLEYVIEW(dev)) {
+	} else if (IS_CHERRYVIEW(dev) || IS_VALLEYVIEW(dev)) {
 		vlv_setup_wm_latency(dev);
-		dev_priv->display.update_wm = vlv_update_wm;
+		dev_priv->display.compute_pipe_wm =
+			vlv_compute_pipe_wm;
+		dev_priv->display.compute_intermediate_wm =
+			vlv_compute_intermediate_wm;
+		dev_priv->display.initial_watermarks =
+			vlv_initial_watermarks;
+		dev_priv->display.optimize_watermarks =
+			vlv_optimize_watermarks;
 	} else if (IS_PINEVIEW(dev)) {
 		if (!intel_get_cxsr_latency(IS_PINEVIEW_G(dev),
 					    dev_priv->is_ddr3,