drm/i915/dsb: Pre allocate and late cleanup of cmd buffer

Pre-allocate command buffer in atomic_commit using intel_dsb_prepare
function which also includes pinning and map in cpu domain.

No functional change is dsb write/commit functions.

Now dsb get/put function is removed and ref-count mechanism is
not needed. Below dsb api added to do respective job mentioned
below.

intel_dsb_prepare - Allocate, pin and map the buffer.
intel_dsb_cleanup - Unpin and release the gem object.

RFC: Initial patch for design review.
v2: included _init() part in _prepare(). [Daniel, Ville]
v3: dsb_cleanup called after cleanup_planes. [Daniel]
v4: dsb structure is moved to intel_crtc_state from intel_crtc. [Maarten]
v5: dsb get/put/ref-count mechanism removed. [Maarten]
v6: Based on review feedback following changes are added,
- replaced intel_dsb structure by pointer in intel_crtc_state. [Maarten]
- passing intel_crtc_state to dsp-api to simplify the code. [Maarten]
- few dsb functions prototype modified to simplify code.
v7: added few cosmetic changes suggested by Jani and null check for
crtc_state in dsb_cleanup removed as suggested by Maarten.
v8: changed the function parameter to intel_crtc_state* of
ivb_load_lut_ext_max() from intel_crtc. [Maarten]
v9: error handling improved in _write() and prepare(). [Maarten]

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@intel.com>
Signed-off-by: Animesh Manna <animesh.manna@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200520130737.11240-1-animesh.manna@intel.com

drivers/gpu/drm/i915/display/intel_atomic.c

@@ -252,6 +252,7 @@ intel_crtc_duplicate_state(struct drm_crtc *crtc)
 	crtc_state->wm.need_postvbl_update = false;
 	crtc_state->fb_bits = 0;
 	crtc_state->update_planes = 0;
+	crtc_state->dsb = NULL;
 
 	return &crtc_state->uapi;
 }
@@ -292,6 +293,8 @@ intel_crtc_destroy_state(struct drm_crtc *crtc,
 {
 	struct intel_crtc_state *crtc_state = to_intel_crtc_state(state);
 
+	drm_WARN_ON(crtc->dev, crtc_state->dsb);
+
 	__drm_atomic_helper_crtc_destroy_state(&crtc_state->uapi);
 	intel_crtc_free_hw_state(crtc_state);
 	kfree(crtc_state);

drivers/gpu/drm/i915/display/intel_color.c

@@ -714,32 +714,30 @@ static void bdw_load_lut_10(struct intel_crtc *crtc,
 	intel_de_write(dev_priv, PREC_PAL_INDEX(pipe), 0);
 }
 
-static void ivb_load_lut_ext_max(struct intel_crtc *crtc)
+static void ivb_load_lut_ext_max(const struct intel_crtc_state *crtc_state)
 {
+	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
 	struct drm_i915_private *dev_priv = to_i915(crtc->base.dev);
-	struct intel_dsb *dsb = intel_dsb_get(crtc);
 	enum pipe pipe = crtc->pipe;
 
 	/* Program the max register to clamp values > 1.0. */
-	intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 0), 1 << 16);
-	intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 1), 1 << 16);
-	intel_dsb_reg_write(dsb, PREC_PAL_EXT_GC_MAX(pipe, 2), 1 << 16);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_EXT_GC_MAX(pipe, 0), 1 << 16);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_EXT_GC_MAX(pipe, 1), 1 << 16);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_EXT_GC_MAX(pipe, 2), 1 << 16);
 
 	/*
 	 * Program the gc max 2 register to clamp values > 1.0.
 	 * ToDo: Extend the ABI to be able to program values
 	 * from 3.0 to 7.0
 	 */
 	if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv)) {
-		intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 0),
+		intel_dsb_reg_write(crtc_state, PREC_PAL_EXT2_GC_MAX(pipe, 0),
 				    1 << 16);
-		intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 1),
+		intel_dsb_reg_write(crtc_state, PREC_PAL_EXT2_GC_MAX(pipe, 1),
 				    1 << 16);
-		intel_dsb_reg_write(dsb, PREC_PAL_EXT2_GC_MAX(pipe, 2),
+		intel_dsb_reg_write(crtc_state, PREC_PAL_EXT2_GC_MAX(pipe, 2),
 				    1 << 16);
 	}
-
-	intel_dsb_put(dsb);
 }
 
 static void ivb_load_luts(const struct intel_crtc_state *crtc_state)
@@ -753,15 +751,15 @@ static void ivb_load_luts(const struct intel_crtc_state *crtc_state)
 	} else if (crtc_state->gamma_mode == GAMMA_MODE_MODE_SPLIT) {
 		ivb_load_lut_10(crtc, degamma_lut, PAL_PREC_SPLIT_MODE |
 				PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 		ivb_load_lut_10(crtc, gamma_lut, PAL_PREC_SPLIT_MODE |
 				PAL_PREC_INDEX_VALUE(512));
 	} else {
 		const struct drm_property_blob *blob = gamma_lut ?: degamma_lut;
 
 		ivb_load_lut_10(crtc, blob,
 				PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 	}
 }
 
@@ -776,15 +774,15 @@ static void bdw_load_luts(const struct intel_crtc_state *crtc_state)
 	} else if (crtc_state->gamma_mode == GAMMA_MODE_MODE_SPLIT) {
 		bdw_load_lut_10(crtc, degamma_lut, PAL_PREC_SPLIT_MODE |
 				PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 		bdw_load_lut_10(crtc, gamma_lut, PAL_PREC_SPLIT_MODE |
 				PAL_PREC_INDEX_VALUE(512));
 	} else {
 		const struct drm_property_blob *blob = gamma_lut ?: degamma_lut;
 
 		bdw_load_lut_10(crtc, blob,
 				PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 	}
 }
 
@@ -877,7 +875,7 @@ static void glk_load_luts(const struct intel_crtc_state *crtc_state)
 		ilk_load_lut_8(crtc, gamma_lut);
 	} else {
 		bdw_load_lut_10(crtc, gamma_lut, PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 	}
 }
 
@@ -900,14 +898,12 @@ icl_load_gcmax(const struct intel_crtc_state *crtc_state,
 	       const struct drm_color_lut *color)
 {
 	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-	struct intel_dsb *dsb = intel_dsb_get(crtc);
 	enum pipe pipe = crtc->pipe;
 
 	/* FIXME LUT entries are 16 bit only, so we can prog 0xFFFF max */
-	intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 0), color->red);
-	intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 1), color->green);
-	intel_dsb_reg_write(dsb, PREC_PAL_GC_MAX(pipe, 2), color->blue);
-	intel_dsb_put(dsb);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_GC_MAX(pipe, 0), color->red);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_GC_MAX(pipe, 1), color->green);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_GC_MAX(pipe, 2), color->blue);
 }
 
 static void
@@ -916,7 +912,6 @@ icl_program_gamma_superfine_segment(const struct intel_crtc_state *crtc_state)
 	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
 	const struct drm_property_blob *blob = crtc_state->hw.gamma_lut;
 	const struct drm_color_lut *lut = blob->data;
-	struct intel_dsb *dsb = intel_dsb_get(crtc);
 	enum pipe pipe = crtc->pipe;
 	int i;
 
@@ -927,19 +922,17 @@ icl_program_gamma_superfine_segment(const struct intel_crtc_state *crtc_state)
 	 * 9 entries, corresponding to values 0, 1/(8 * 128 * 256),
 	 * 2/(8 * 128 * 256) ... 8/(8 * 128 * 256).
 	 */
-	intel_dsb_reg_write(dsb, PREC_PAL_MULTI_SEG_INDEX(pipe),
+	intel_dsb_reg_write(crtc_state, PREC_PAL_MULTI_SEG_INDEX(pipe),
 			    PAL_PREC_AUTO_INCREMENT);
 
 	for (i = 0; i < 9; i++) {
 		const struct drm_color_lut *entry = &lut[i];
 
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_MULTI_SEG_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_MULTI_SEG_DATA(pipe),
 					    ilk_lut_12p4_ldw(entry));
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_MULTI_SEG_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_MULTI_SEG_DATA(pipe),
 					    ilk_lut_12p4_udw(entry));
 	}
-
-	intel_dsb_put(dsb);
 }
 
 static void
@@ -949,7 +942,6 @@ icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state)
 	const struct drm_property_blob *blob = crtc_state->hw.gamma_lut;
 	const struct drm_color_lut *lut = blob->data;
 	const struct drm_color_lut *entry;
-	struct intel_dsb *dsb = intel_dsb_get(crtc);
 	enum pipe pipe = crtc->pipe;
 	int i;
 
@@ -963,12 +955,13 @@ icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state)
 	 * PAL_PREC_INDEX[0] and PAL_PREC_INDEX[1] map to seg2[1],
 	 * seg2[0] being unused by the hardware.
 	 */
-	intel_dsb_reg_write(dsb, PREC_PAL_INDEX(pipe), PAL_PREC_AUTO_INCREMENT);
+	intel_dsb_reg_write(crtc_state, PREC_PAL_INDEX(pipe),
+			    PAL_PREC_AUTO_INCREMENT);
 	for (i = 1; i < 257; i++) {
 		entry = &lut[i * 8];
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_DATA(pipe),
 					    ilk_lut_12p4_ldw(entry));
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_DATA(pipe),
 					    ilk_lut_12p4_udw(entry));
 	}
 
@@ -986,24 +979,22 @@ icl_program_gamma_multi_segment(const struct intel_crtc_state *crtc_state)
 	 */
 	for (i = 0; i < 256; i++) {
 		entry = &lut[i * 8 * 128];
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_DATA(pipe),
 					    ilk_lut_12p4_ldw(entry));
-		intel_dsb_indexed_reg_write(dsb, PREC_PAL_DATA(pipe),
+		intel_dsb_indexed_reg_write(crtc_state, PREC_PAL_DATA(pipe),
 					    ilk_lut_12p4_udw(entry));
 	}
 
 	/* The last entry in the LUT is to be programmed in GCMAX */
 	entry = &lut[256 * 8 * 128];
 	icl_load_gcmax(crtc_state, entry);
-	ivb_load_lut_ext_max(crtc);
-	intel_dsb_put(dsb);
+	ivb_load_lut_ext_max(crtc_state);
 }
 
 static void icl_load_luts(const struct intel_crtc_state *crtc_state)
 {
 	const struct drm_property_blob *gamma_lut = crtc_state->hw.gamma_lut;
 	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
-	struct intel_dsb *dsb = intel_dsb_get(crtc);
 
 	if (crtc_state->hw.degamma_lut)
 		glk_load_degamma_lut(crtc_state);
@@ -1018,11 +1009,10 @@ static void icl_load_luts(const struct intel_crtc_state *crtc_state)
 		break;
 	default:
 		bdw_load_lut_10(crtc, gamma_lut, PAL_PREC_INDEX_VALUE(0));
-		ivb_load_lut_ext_max(crtc);
+		ivb_load_lut_ext_max(crtc_state);
 	}
 
-	intel_dsb_commit(dsb);
-	intel_dsb_put(dsb);
+	intel_dsb_commit(crtc_state);
 }
 
 static u32 chv_cgm_degamma_ldw(const struct drm_color_lut *color)

drivers/gpu/drm/i915/display/intel_display.c

@@ -14996,8 +14996,24 @@ static int intel_atomic_check(struct drm_device *dev,
 
 static int intel_atomic_prepare_commit(struct intel_atomic_state *state)
 {
-	return drm_atomic_helper_prepare_planes(state->base.dev,
-						&state->base);
+	struct intel_crtc_state *crtc_state;
+	struct intel_crtc *crtc;
+	int i, ret;
+
+	ret = drm_atomic_helper_prepare_planes(state->base.dev, &state->base);
+	if (ret < 0)
+		return ret;
+
+	for_each_new_intel_crtc_in_state(state, crtc, crtc_state, i) {
+		bool mode_changed = needs_modeset(crtc_state);
+
+		if (mode_changed || crtc_state->update_pipe ||
+		    crtc_state->uapi.color_mgmt_changed) {
+			intel_dsb_prepare(crtc_state);
+		}
+	}
+
+	return 0;
 }
 
 u32 intel_crtc_get_vblank_counter(struct intel_crtc *crtc)
@@ -15427,15 +15443,27 @@ static void intel_atomic_commit_fence_wait(struct intel_atomic_state *intel_stat
 		    &wait_reset);
 }
 
+static void intel_cleanup_dsbs(struct intel_atomic_state *state)
+{
+	struct intel_crtc_state *old_crtc_state, *new_crtc_state;
+	struct intel_crtc *crtc;
+	int i;
+
+	for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state,
+					    new_crtc_state, i)
+		intel_dsb_cleanup(old_crtc_state);
+}
+
 static void intel_atomic_cleanup_work(struct work_struct *work)
 {
-	struct drm_atomic_state *state =
-		container_of(work, struct drm_atomic_state, commit_work);
-	struct drm_i915_private *i915 = to_i915(state->dev);
+	struct intel_atomic_state *state =
+		container_of(work, struct intel_atomic_state, base.commit_work);
+	struct drm_i915_private *i915 = to_i915(state->base.dev);
 
-	drm_atomic_helper_cleanup_planes(&i915->drm, state);
-	drm_atomic_helper_commit_cleanup_done(state);
-	drm_atomic_state_put(state);
+	intel_cleanup_dsbs(state);
+	drm_atomic_helper_cleanup_planes(&i915->drm, &state->base);
+	drm_atomic_helper_commit_cleanup_done(&state->base);
+	drm_atomic_state_put(&state->base);
 
 	intel_atomic_helper_free_state(i915);
 }
@@ -15565,6 +15593,13 @@ static void intel_atomic_commit_tail(struct intel_atomic_state *state)
 			modeset_put_power_domains(dev_priv, put_domains[i]);
 
 		intel_modeset_verify_crtc(crtc, state, old_crtc_state, new_crtc_state);
+
+		/*
+		 * DSB cleanup is done in cleanup_work aligning with framebuffer
+		 * cleanup. So copy and reset the dsb structure to sync with
+		 * commit_done and later do dsb cleanup in cleanup_work.
+		 */
+		old_crtc_state->dsb = fetch_and_zero(&new_crtc_state->dsb);
 	}
 
 	/* Underruns don't always raise interrupts, so check manually */
@@ -15714,8 +15749,15 @@ static int intel_atomic_commit(struct drm_device *dev,
 		intel_atomic_swap_global_state(state);
 
 	if (ret) {
+		struct intel_crtc_state *new_crtc_state;
+		struct intel_crtc *crtc;
+		int i;
+
 		i915_sw_fence_commit(&state->commit_ready);
 
+		for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i)
+			intel_dsb_cleanup(new_crtc_state);
+
 		drm_atomic_helper_cleanup_planes(dev, &state->base);
 		intel_runtime_pm_put(&dev_priv->runtime_pm, state->wakeref);
 		return ret;

drivers/gpu/drm/i915/display/intel_display_types.h

@@ -1067,6 +1067,9 @@ struct intel_crtc_state {
 
 	/* Only valid on TGL+ */
 	enum transcoder mst_master_transcoder;
+
+	/* For DSB related info */
+	struct intel_dsb *dsb;
 };
 
 enum intel_pipe_crc_source {
@@ -1136,9 +1139,6 @@ struct intel_crtc {
 	/* scalers available on this crtc */
 	int num_scalers;
 
-	/* per pipe DSB related info */
-	struct intel_dsb dsb;
-
 #ifdef CONFIG_DEBUG_FS
 	struct intel_pipe_crc pipe_crc;
 #endif