Merge tag 'drm-vc4-next-2016-02-17' of github.com:anholt/linux into drm-next

This pull request brings in overlay plane support for vc4.

* tag 'drm-vc4-next-2016-02-17' of github.com:anholt/linux:
  drm/vc4: Add support for YUV planes.
  drm/vc4: Add support a few more RGB display plane formats.
  drm/vc4: Add support for scaling of display planes.
  drm/vc4: Fix which value is being used for source image size.
  drm/vc4: Add more display planes to each CRTC.
  drm/vc4: Make the CRTCs cooperate on allocating display lists.
  drm/vc4: Add a proper short-circut path for legacy cursor updates.
  drm/vc4: Move the plane clipping/scaling setup to a separate function.
  drm/vc4: Add missing __iomem annotation to hw_dlist.
  drm/vc4: Improve comments on vc4_plane_state members.

drivers/gpu/drm/vc4/vc4_crtc.c

@@ -49,22 +49,27 @@ struct vc4_crtc {
 	/* Which HVS channel we're using for our CRTC. */
 	int channel;
 
-	/* Pointer to the actual hardware display list memory for the
-	 * crtc.
-	 */
-	u32 __iomem *dlist;
-
-	u32 dlist_size; /* in dwords */
-
 	struct drm_pending_vblank_event *event;
 };
 
+struct vc4_crtc_state {
+	struct drm_crtc_state base;
+	/* Dlist area for this CRTC configuration. */
+	struct drm_mm_node mm;
+};
+
 static inline struct vc4_crtc *
 to_vc4_crtc(struct drm_crtc *crtc)
 {
 	return (struct vc4_crtc *)crtc;
 }
 
+static inline struct vc4_crtc_state *
+to_vc4_crtc_state(struct drm_crtc_state *crtc_state)
+{
+	return (struct vc4_crtc_state *)crtc_state;
+}
+
 struct vc4_crtc_data {
 	/* Which channel of the HVS this pixelvalve sources from. */
 	int hvs_channel;
@@ -319,11 +324,13 @@ static void vc4_crtc_enable(struct drm_crtc *crtc)
 static int vc4_crtc_atomic_check(struct drm_crtc *crtc,
 				 struct drm_crtc_state *state)
 {
+	struct vc4_crtc_state *vc4_state = to_vc4_crtc_state(state);
 	struct drm_device *dev = crtc->dev;
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	struct drm_plane *plane;
-	struct vc4_crtc *vc4_crtc = to_vc4_crtc(crtc);
+	unsigned long flags;
 	u32 dlist_count = 0;
+	int ret;
 
 	/* The pixelvalve can only feed one encoder (and encoders are
 	 * 1:1 with connectors.)
@@ -346,18 +353,12 @@ static int vc4_crtc_atomic_check(struct drm_crtc *crtc,
 
 	dlist_count++; /* Account for SCALER_CTL0_END. */
 
-	if (!vc4_crtc->dlist || dlist_count > vc4_crtc->dlist_size) {
-		vc4_crtc->dlist = ((u32 __iomem *)vc4->hvs->dlist +
-				   HVS_BOOTLOADER_DLIST_END);
-		vc4_crtc->dlist_size = ((SCALER_DLIST_SIZE >> 2) -
-					HVS_BOOTLOADER_DLIST_END);
-
-		if (dlist_count > vc4_crtc->dlist_size) {
-			DRM_DEBUG_KMS("dlist too large for CRTC (%d > %d).\n",
-				      dlist_count, vc4_crtc->dlist_size);
-			return -EINVAL;
-		}
-	}
+	spin_lock_irqsave(&vc4->hvs->mm_lock, flags);
+	ret = drm_mm_insert_node(&vc4->hvs->dlist_mm, &vc4_state->mm,
+				 dlist_count, 1, 0);
+	spin_unlock_irqrestore(&vc4->hvs->mm_lock, flags);
+	if (ret)
+		return ret;
 
 	return 0;
 }
@@ -368,47 +369,29 @@ static void vc4_crtc_atomic_flush(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	struct vc4_crtc *vc4_crtc = to_vc4_crtc(crtc);
+	struct vc4_crtc_state *vc4_state = to_vc4_crtc_state(crtc->state);
 	struct drm_plane *plane;
 	bool debug_dump_regs = false;
-	u32 __iomem *dlist_next = vc4_crtc->dlist;
+	u32 __iomem *dlist_start = vc4->hvs->dlist + vc4_state->mm.start;
+	u32 __iomem *dlist_next = dlist_start;
 
 	if (debug_dump_regs) {
 		DRM_INFO("CRTC %d HVS before:\n", drm_crtc_index(crtc));
 		vc4_hvs_dump_state(dev);
 	}
 
-	/* Copy all the active planes' dlist contents to the hardware dlist.
-	 *
-	 * XXX: If the new display list was large enough that it
-	 * overlapped a currently-read display list, we need to do
-	 * something like disable scanout before putting in the new
-	 * list.  For now, we're safe because we only have the two
-	 * planes.
-	 */
+	/* Copy all the active planes' dlist contents to the hardware dlist. */
 	drm_atomic_crtc_for_each_plane(plane, crtc) {
 		dlist_next += vc4_plane_write_dlist(plane, dlist_next);
 	}
 
-	if (dlist_next == vc4_crtc->dlist) {
-		/* If no planes were enabled, use the SCALER_CTL0_END
-		 * at the start of the display list memory (in the
-		 * bootloader section).  We'll rewrite that
-		 * SCALER_CTL0_END, just in case, though.
-		 */
-		writel(SCALER_CTL0_END, vc4->hvs->dlist);
-		HVS_WRITE(SCALER_DISPLISTX(vc4_crtc->channel), 0);
-	} else {
 	writel(SCALER_CTL0_END, dlist_next);
 	dlist_next++;
 
-		HVS_WRITE(SCALER_DISPLISTX(vc4_crtc->channel),
-			  (u32 __iomem *)vc4_crtc->dlist -
-			  (u32 __iomem *)vc4->hvs->dlist);
+	WARN_ON_ONCE(dlist_next - dlist_start != vc4_state->mm.size);
 
-		/* Make the next display list start after ours. */
-		vc4_crtc->dlist_size -= (dlist_next - vc4_crtc->dlist);
-		vc4_crtc->dlist = dlist_next;
-	}
+	HVS_WRITE(SCALER_DISPLISTX(vc4_crtc->channel),
+		  vc4_state->mm.start);
 
 	if (debug_dump_regs) {
 		DRM_INFO("CRTC %d HVS after:\n", drm_crtc_index(crtc));
@@ -573,6 +556,36 @@ static int vc4_page_flip(struct drm_crtc *crtc,
 		return drm_atomic_helper_page_flip(crtc, fb, event, flags);
 }
 
+static struct drm_crtc_state *vc4_crtc_duplicate_state(struct drm_crtc *crtc)
+{
+	struct vc4_crtc_state *vc4_state;
+
+	vc4_state = kzalloc(sizeof(*vc4_state), GFP_KERNEL);
+	if (!vc4_state)
+		return NULL;
+
+	__drm_atomic_helper_crtc_duplicate_state(crtc, &vc4_state->base);
+	return &vc4_state->base;
+}
+
+static void vc4_crtc_destroy_state(struct drm_crtc *crtc,
+				   struct drm_crtc_state *state)
+{
+	struct vc4_dev *vc4 = to_vc4_dev(crtc->dev);
+	struct vc4_crtc_state *vc4_state = to_vc4_crtc_state(state);
+
+	if (vc4_state->mm.allocated) {
+		unsigned long flags;
+
+		spin_lock_irqsave(&vc4->hvs->mm_lock, flags);
+		drm_mm_remove_node(&vc4_state->mm);
+		spin_unlock_irqrestore(&vc4->hvs->mm_lock, flags);
+
+	}
+
+	__drm_atomic_helper_crtc_destroy_state(crtc, state);
+}
+
 static const struct drm_crtc_funcs vc4_crtc_funcs = {
 	.set_config = drm_atomic_helper_set_config,
 	.destroy = vc4_crtc_destroy,
@@ -581,8 +594,8 @@ static const struct drm_crtc_funcs vc4_crtc_funcs = {
 	.cursor_set = NULL, /* handled by drm_mode_cursor_universal */
 	.cursor_move = NULL, /* handled by drm_mode_cursor_universal */
 	.reset = drm_atomic_helper_crtc_reset,
-	.atomic_duplicate_state = drm_atomic_helper_crtc_duplicate_state,
-	.atomic_destroy_state = drm_atomic_helper_crtc_destroy_state,
+	.atomic_duplicate_state = vc4_crtc_duplicate_state,
+	.atomic_destroy_state = vc4_crtc_destroy_state,
 };
 
 static const struct drm_crtc_helper_funcs vc4_crtc_helper_funcs = {
@@ -644,9 +657,9 @@ static int vc4_crtc_bind(struct device *dev, struct device *master, void *data)
 	struct vc4_dev *vc4 = to_vc4_dev(drm);
 	struct vc4_crtc *vc4_crtc;
 	struct drm_crtc *crtc;
-	struct drm_plane *primary_plane, *cursor_plane;
+	struct drm_plane *primary_plane, *cursor_plane, *destroy_plane, *temp;
 	const struct of_device_id *match;
-	int ret;
+	int ret, i;
 
 	vc4_crtc = devm_kzalloc(dev, sizeof(*vc4_crtc), GFP_KERNEL);
 	if (!vc4_crtc)
@@ -675,27 +688,49 @@ static int vc4_crtc_bind(struct device *dev, struct device *master, void *data)
 		goto err;
 	}
 
-	cursor_plane = vc4_plane_init(drm, DRM_PLANE_TYPE_CURSOR);
-	if (IS_ERR(cursor_plane)) {
-		dev_err(dev, "failed to construct cursor plane\n");
-		ret = PTR_ERR(cursor_plane);
-		goto err_primary;
-	}
-
-	drm_crtc_init_with_planes(drm, crtc, primary_plane, cursor_plane,
+	drm_crtc_init_with_planes(drm, crtc, primary_plane, NULL,
 				  &vc4_crtc_funcs, NULL);
 	drm_crtc_helper_add(crtc, &vc4_crtc_helper_funcs);
 	primary_plane->crtc = crtc;
-	cursor_plane->crtc = crtc;
 	vc4->crtc[drm_crtc_index(crtc)] = vc4_crtc;
 	vc4_crtc->channel = vc4_crtc->data->hvs_channel;
 
+	/* Set up some arbitrary number of planes.  We're not limited
+	 * by a set number of physical registers, just the space in
+	 * the HVS (16k) and how small an plane can be (28 bytes).
+	 * However, each plane we set up takes up some memory, and
+	 * increases the cost of looping over planes, which atomic
+	 * modesetting does quite a bit.  As a result, we pick a
+	 * modest number of planes to expose, that should hopefully
+	 * still cover any sane usecase.
+	 */
+	for (i = 0; i < 8; i++) {
+		struct drm_plane *plane =
+			vc4_plane_init(drm, DRM_PLANE_TYPE_OVERLAY);
+
+		if (IS_ERR(plane))
+			continue;
+
+		plane->possible_crtcs = 1 << drm_crtc_index(crtc);
+	}
+
+	/* Set up the legacy cursor after overlay initialization,
+	 * since we overlay planes on the CRTC in the order they were
+	 * initialized.
+	 */
+	cursor_plane = vc4_plane_init(drm, DRM_PLANE_TYPE_CURSOR);
+	if (!IS_ERR(cursor_plane)) {
+		cursor_plane->possible_crtcs = 1 << drm_crtc_index(crtc);
+		cursor_plane->crtc = crtc;
+		crtc->cursor = cursor_plane;
+	}
+
 	CRTC_WRITE(PV_INTEN, 0);
 	CRTC_WRITE(PV_INTSTAT, PV_INT_VFP_START);
 	ret = devm_request_irq(dev, platform_get_irq(pdev, 0),
 			       vc4_crtc_irq_handler, 0, "vc4 crtc", vc4_crtc);
 	if (ret)
-		goto err_cursor;
+		goto err_destroy_planes;
 
 	vc4_set_crtc_possible_masks(drm, crtc);
 
@@ -703,10 +738,12 @@ static int vc4_crtc_bind(struct device *dev, struct device *master, void *data)
 
 	return 0;
 
-err_cursor:
-	cursor_plane->funcs->destroy(cursor_plane);
-err_primary:
-	primary_plane->funcs->destroy(primary_plane);
+err_destroy_planes:
+	list_for_each_entry_safe(destroy_plane, temp,
+				 &drm->mode_config.plane_list, head) {
+		if (destroy_plane->possible_crtcs == 1 << drm_crtc_index(crtc))
+		    destroy_plane->funcs->destroy(destroy_plane);
+	}
 err:
 	return ret;
 }

drivers/gpu/drm/vc4/vc4_drv.h

@@ -149,7 +149,17 @@ struct vc4_v3d {
 struct vc4_hvs {
 	struct platform_device *pdev;
 	void __iomem *regs;
-	void __iomem *dlist;
+	u32 __iomem *dlist;
+
+	/* Memory manager for CRTCs to allocate space in the display
+	 * list.  Units are dwords.
+	 */
+	struct drm_mm dlist_mm;
+	/* Memory manager for the LBM memory used by HVS scaling. */
+	struct drm_mm lbm_mm;
+	spinlock_t mm_lock;
+
+	struct drm_mm_node mitchell_netravali_filter;
 };
 
 struct vc4_plane {

drivers/gpu/drm/vc4/vc4_hvs.c

@@ -100,12 +100,76 @@ int vc4_hvs_debugfs_regs(struct seq_file *m, void *unused)
 }
 #endif
 
+/* The filter kernel is composed of dwords each containing 3 9-bit
+ * signed integers packed next to each other.
+ */
+#define VC4_INT_TO_COEFF(coeff) (coeff & 0x1ff)
+#define VC4_PPF_FILTER_WORD(c0, c1, c2)				\
+	((((c0) & 0x1ff) << 0) |				\
+	 (((c1) & 0x1ff) << 9) |				\
+	 (((c2) & 0x1ff) << 18))
+
+/* The whole filter kernel is arranged as the coefficients 0-16 going
+ * up, then a pad, then 17-31 going down and reversed within the
+ * dwords.  This means that a linear phase kernel (where it's
+ * symmetrical at the boundary between 15 and 16) has the last 5
+ * dwords matching the first 5, but reversed.
+ */
+#define VC4_LINEAR_PHASE_KERNEL(c0, c1, c2, c3, c4, c5, c6, c7, c8,	\
+				c9, c10, c11, c12, c13, c14, c15)	\
+	{VC4_PPF_FILTER_WORD(c0, c1, c2),				\
+	 VC4_PPF_FILTER_WORD(c3, c4, c5),				\
+	 VC4_PPF_FILTER_WORD(c6, c7, c8),				\
+	 VC4_PPF_FILTER_WORD(c9, c10, c11),				\
+	 VC4_PPF_FILTER_WORD(c12, c13, c14),				\
+	 VC4_PPF_FILTER_WORD(c15, c15, 0)}
+
+#define VC4_LINEAR_PHASE_KERNEL_DWORDS 6
+#define VC4_KERNEL_DWORDS (VC4_LINEAR_PHASE_KERNEL_DWORDS * 2 - 1)
+
+/* Recommended B=1/3, C=1/3 filter choice from Mitchell/Netravali.
+ * http://www.cs.utexas.edu/~fussell/courses/cs384g/lectures/mitchell/Mitchell.pdf
+ */
+static const u32 mitchell_netravali_1_3_1_3_kernel[] =
+	VC4_LINEAR_PHASE_KERNEL(0, -2, -6, -8, -10, -8, -3, 2, 18,
+				50, 82, 119, 155, 187, 213, 227);
+
+static int vc4_hvs_upload_linear_kernel(struct vc4_hvs *hvs,
+					struct drm_mm_node *space,
+					const u32 *kernel)
+{
+	int ret, i;
+	u32 __iomem *dst_kernel;
+
+	ret = drm_mm_insert_node(&hvs->dlist_mm, space, VC4_KERNEL_DWORDS, 1,
+				 0);
+	if (ret) {
+		DRM_ERROR("Failed to allocate space for filter kernel: %d\n",
+			  ret);
+		return ret;
+	}
+
+	dst_kernel = hvs->dlist + space->start;
+
+	for (i = 0; i < VC4_KERNEL_DWORDS; i++) {
+		if (i < VC4_LINEAR_PHASE_KERNEL_DWORDS)
+			writel(kernel[i], &dst_kernel[i]);
+		else {
+			writel(kernel[VC4_KERNEL_DWORDS - i - 1],
+			       &dst_kernel[i]);
+		}
+	}
+
+	return 0;
+}
+
 static int vc4_hvs_bind(struct device *dev, struct device *master, void *data)
 {
 	struct platform_device *pdev = to_platform_device(dev);
 	struct drm_device *drm = dev_get_drvdata(master);
 	struct vc4_dev *vc4 = drm->dev_private;
 	struct vc4_hvs *hvs = NULL;
+	int ret;
 
 	hvs = devm_kzalloc(&pdev->dev, sizeof(*hvs), GFP_KERNEL);
 	if (!hvs)
@@ -119,6 +183,33 @@ static int vc4_hvs_bind(struct device *dev, struct device *master, void *data)
 
 	hvs->dlist = hvs->regs + SCALER_DLIST_START;
 
+	spin_lock_init(&hvs->mm_lock);
+
+	/* Set up the HVS display list memory manager.  We never
+	 * overwrite the setup from the bootloader (just 128b out of
+	 * our 16K), since we don't want to scramble the screen when
+	 * transitioning from the firmware's boot setup to runtime.
+	 */
+	drm_mm_init(&hvs->dlist_mm,
+		    HVS_BOOTLOADER_DLIST_END,
+		    (SCALER_DLIST_SIZE >> 2) - HVS_BOOTLOADER_DLIST_END);
+
+	/* Set up the HVS LBM memory manager.  We could have some more
+	 * complicated data structure that allowed reuse of LBM areas
+	 * between planes when they don't overlap on the screen, but
+	 * for now we just allocate globally.
+	 */
+	drm_mm_init(&hvs->lbm_mm, 0, 96 * 1024);
+
+	/* Upload filter kernels.  We only have the one for now, so we
+	 * keep it around for the lifetime of the driver.
+	 */
+	ret = vc4_hvs_upload_linear_kernel(hvs,
+					   &hvs->mitchell_netravali_filter,
+					   mitchell_netravali_1_3_1_3_kernel);
+	if (ret)
+		return ret;
+
 	vc4->hvs = hvs;
 	return 0;
 }
@@ -129,6 +220,12 @@ static void vc4_hvs_unbind(struct device *dev, struct device *master,
 	struct drm_device *drm = dev_get_drvdata(master);
 	struct vc4_dev *vc4 = drm->dev_private;
 
+	if (vc4->hvs->mitchell_netravali_filter.allocated)
+		drm_mm_remove_node(&vc4->hvs->mitchell_netravali_filter);
+
+	drm_mm_takedown(&vc4->hvs->dlist_mm);
+	drm_mm_takedown(&vc4->hvs->lbm_mm);
+
 	vc4->hvs = NULL;
 }
 

drivers/gpu/drm/vc4/vc4_kms.c

@@ -49,6 +49,15 @@ vc4_atomic_complete_commit(struct vc4_commit *c)
 
 	drm_atomic_helper_commit_modeset_enables(dev, state);
 
+	/* Make sure that drm_atomic_helper_wait_for_vblanks()
+	 * actually waits for vblank.  If we're doing a full atomic
+	 * modeset (as opposed to a vc4_update_plane() short circuit),
+	 * then we need to wait for scanout to be done with our
+	 * display lists before we free it and potentially reallocate
+	 * and overwrite the dlist memory with a new modeset.
+	 */
+	state->legacy_cursor_update = false;
+
 	drm_atomic_helper_wait_for_vblanks(dev, state);
 
 	drm_atomic_helper_cleanup_planes(dev, state);

drivers/gpu/drm/vc4/vc4_regs.h

@@ -503,7 +503,12 @@ enum hvs_pixel_format {
 	HVS_PIXEL_FORMAT_RGB888 = 5,
 	HVS_PIXEL_FORMAT_RGBA6666 = 6,
 	/* 32bpp */
-	HVS_PIXEL_FORMAT_RGBA8888 = 7
+	HVS_PIXEL_FORMAT_RGBA8888 = 7,
+
+	HVS_PIXEL_FORMAT_YCBCR_YUV420_3PLANE = 8,
+	HVS_PIXEL_FORMAT_YCBCR_YUV420_2PLANE = 9,
+	HVS_PIXEL_FORMAT_YCBCR_YUV422_3PLANE = 10,
+	HVS_PIXEL_FORMAT_YCBCR_YUV422_2PLANE = 11,
 };
 
 /* Note: the LSB is the rightmost character shown.  Only valid for
@@ -536,6 +541,21 @@ enum hvs_pixel_format {
 #define SCALER_CTL0_ORDER_MASK			VC4_MASK(14, 13)
 #define SCALER_CTL0_ORDER_SHIFT			13
 
+#define SCALER_CTL0_SCL1_MASK			VC4_MASK(10, 8)
+#define SCALER_CTL0_SCL1_SHIFT			8
+
+#define SCALER_CTL0_SCL0_MASK			VC4_MASK(7, 5)
+#define SCALER_CTL0_SCL0_SHIFT			5
+
+#define SCALER_CTL0_SCL_H_PPF_V_PPF		0
+#define SCALER_CTL0_SCL_H_TPZ_V_PPF		1
+#define SCALER_CTL0_SCL_H_PPF_V_TPZ		2
+#define SCALER_CTL0_SCL_H_TPZ_V_TPZ		3
+#define SCALER_CTL0_SCL_H_PPF_V_NONE		4
+#define SCALER_CTL0_SCL_H_NONE_V_PPF		5
+#define SCALER_CTL0_SCL_H_NONE_V_TPZ		6
+#define SCALER_CTL0_SCL_H_TPZ_V_NONE		7
+
 /* Set to indicate no scaling. */
 #define SCALER_CTL0_UNITY			BIT(4)
 
@@ -551,6 +571,12 @@ enum hvs_pixel_format {
 #define SCALER_POS0_START_X_MASK		VC4_MASK(11, 0)
 #define SCALER_POS0_START_X_SHIFT		0
 
+#define SCALER_POS1_SCL_HEIGHT_MASK		VC4_MASK(27, 16)
+#define SCALER_POS1_SCL_HEIGHT_SHIFT		16
+
+#define SCALER_POS1_SCL_WIDTH_MASK		VC4_MASK(11, 0)
+#define SCALER_POS1_SCL_WIDTH_SHIFT		0
+
 #define SCALER_POS2_ALPHA_MODE_MASK		VC4_MASK(31, 30)
 #define SCALER_POS2_ALPHA_MODE_SHIFT		30
 #define SCALER_POS2_ALPHA_MODE_PIPELINE		0
@@ -564,6 +590,80 @@ enum hvs_pixel_format {
 #define SCALER_POS2_WIDTH_MASK			VC4_MASK(11, 0)
 #define SCALER_POS2_WIDTH_SHIFT			0
 
+/* Color Space Conversion words.  Some values are S2.8 signed
+ * integers, except that the 2 integer bits map as {0x0: 0, 0x1: 1,
+ * 0x2: 2, 0x3: -1}
+ */
+/* bottom 8 bits of S2.8 contribution of Cr to Blue */
+#define SCALER_CSC0_COEF_CR_BLU_MASK		VC4_MASK(31, 24)
+#define SCALER_CSC0_COEF_CR_BLU_SHIFT		24
+/* Signed offset to apply to Y before CSC. (Y' = Y + YY_OFS) */
+#define SCALER_CSC0_COEF_YY_OFS_MASK		VC4_MASK(23, 16)
+#define SCALER_CSC0_COEF_YY_OFS_SHIFT		16
+/* Signed offset to apply to CB before CSC (Cb' = Cb - 128 + CB_OFS). */
+#define SCALER_CSC0_COEF_CB_OFS_MASK		VC4_MASK(15, 8)
+#define SCALER_CSC0_COEF_CB_OFS_SHIFT		8
+/* Signed offset to apply to CB before CSC (Cr' = Cr - 128 + CR_OFS). */
+#define SCALER_CSC0_COEF_CR_OFS_MASK		VC4_MASK(7, 0)
+#define SCALER_CSC0_COEF_CR_OFS_SHIFT		0
+#define SCALER_CSC0_ITR_R_601_5			0x00f00000
+#define SCALER_CSC0_ITR_R_709_3			0x00f00000
+#define SCALER_CSC0_JPEG_JFIF			0x00000000
+
+/* S2.8 contribution of Cb to Green */
+#define SCALER_CSC1_COEF_CB_GRN_MASK		VC4_MASK(31, 22)
+#define SCALER_CSC1_COEF_CB_GRN_SHIFT		22
+/* S2.8 contribution of Cr to Green */
+#define SCALER_CSC1_COEF_CR_GRN_MASK		VC4_MASK(21, 12)
+#define SCALER_CSC1_COEF_CR_GRN_SHIFT		12
+/* S2.8 contribution of Y to all of RGB */
+#define SCALER_CSC1_COEF_YY_ALL_MASK		VC4_MASK(11, 2)
+#define SCALER_CSC1_COEF_YY_ALL_SHIFT		2
+/* top 2 bits of S2.8 contribution of Cr to Blue */
+#define SCALER_CSC1_COEF_CR_BLU_MASK		VC4_MASK(1, 0)
+#define SCALER_CSC1_COEF_CR_BLU_SHIFT		0
+#define SCALER_CSC1_ITR_R_601_5			0xe73304a8
+#define SCALER_CSC1_ITR_R_709_3			0xf2b784a8
+#define SCALER_CSC1_JPEG_JFIF			0xea34a400
+
+/* S2.8 contribution of Cb to Red */
+#define SCALER_CSC2_COEF_CB_RED_MASK		VC4_MASK(29, 20)
+#define SCALER_CSC2_COEF_CB_RED_SHIFT		20
+/* S2.8 contribution of Cr to Red */
+#define SCALER_CSC2_COEF_CR_RED_MASK		VC4_MASK(19, 10)
+#define SCALER_CSC2_COEF_CR_RED_SHIFT		10
+/* S2.8 contribution of Cb to Blue */
+#define SCALER_CSC2_COEF_CB_BLU_MASK		VC4_MASK(19, 10)
+#define SCALER_CSC2_COEF_CB_BLU_SHIFT		10
+#define SCALER_CSC2_ITR_R_601_5			0x00066204
+#define SCALER_CSC2_ITR_R_709_3			0x00072a1c
+#define SCALER_CSC2_JPEG_JFIF			0x000599c5
+
+#define SCALER_TPZ0_VERT_RECALC			BIT(31)
+#define SCALER_TPZ0_SCALE_MASK			VC4_MASK(28, 8)
+#define SCALER_TPZ0_SCALE_SHIFT			8
+#define SCALER_TPZ0_IPHASE_MASK			VC4_MASK(7, 0)
+#define SCALER_TPZ0_IPHASE_SHIFT		0
+#define SCALER_TPZ1_RECIP_MASK			VC4_MASK(15, 0)
+#define SCALER_TPZ1_RECIP_SHIFT			0
+
+/* Skips interpolating coefficients to 64 phases, so just 8 are used.
+ * Required for nearest neighbor.
+ */
+#define SCALER_PPF_NOINTERP			BIT(31)
+/* Replaes the highest valued coefficient with one that makes all 4
+ * sum to unity.
+ */
+#define SCALER_PPF_AGC				BIT(30)
+#define SCALER_PPF_SCALE_MASK			VC4_MASK(24, 8)
+#define SCALER_PPF_SCALE_SHIFT			8
+#define SCALER_PPF_IPHASE_MASK			VC4_MASK(6, 0)
+#define SCALER_PPF_IPHASE_SHIFT			0
+
+#define SCALER_PPF_KERNEL_OFFSET_MASK		VC4_MASK(13, 0)
+#define SCALER_PPF_KERNEL_OFFSET_SHIFT		0
+#define SCALER_PPF_KERNEL_UNCACHED		BIT(31)
+
 #define SCALER_SRC_PITCH_MASK			VC4_MASK(15, 0)
 #define SCALER_SRC_PITCH_SHIFT			0