Merge branch 'topic/vsp1' into patchwork

* topic/vsp1: (36 commits)
  [media] v4l: vsp1: wpf: Add flipping support
  [media] v4l: vsp1: rwpf: Support runtime modification of controls
  [media] v4l: vsp1: Simplify alpha propagation
  [media] v4l: vsp1: clu: Support runtime modification of controls
  [media] v4l: vsp1: lut: Support runtime modification of controls
  [media] v4l: vsp1: Support runtime modification of controls
  [media] v4l: vsp1: Add Cubic Look Up Table (CLU) support
  [media] v4l: vsp1: lut: Expose configuration through a control
  [media] v4l: vsp1: lut: Initialize the mutex
  [media] v4l: vsp1: dl: Don't free fragments with interrupts disabled
  [media] v4l: vsp1: Set entities functions
  [media] v4l: vsp1: Don't create LIF entity when the userspace API is enabled
  [media] v4l: vsp1: Don't register media device when userspace API is disabled
  [media] v4l: vsp1: Base link creation on availability of entities
  [media] media: Add video statistics computation functions
  [media] media: Add video processing entity functions
  [media] v4l: vsp1: sru: Fix intensity control ID
  [media] v4l: vsp1: Stop the pipeline upon the first STREAMOFF
  [media] v4l: vsp1: Constify operation structures
  [media] v4l: vsp1: pipe: Fix typo in comment
  ...

31 files changed, 1217 insertions, 334 deletions

diff --git a/drivers/gpu/drm/rcar-du/rcar_du_vsp.c b/drivers/gpu/drm/rcar-du/rcar_du_vsp.c
index e671a7cd3463..6ac717f2056f 100644
--- a/drivers/gpu/drm/rcar-du/rcar_du_vsp.c
+++ b/drivers/gpu/drm/rcar-du/rcar_du_vsp.c
@@ -148,40 +148,39 @@ static void rcar_du_vsp_plane_setup(struct rcar_du_vsp_plane *plane)
 	struct rcar_du_vsp_plane_state *state =
 		to_rcar_vsp_plane_state(plane->plane.state);
 	struct drm_framebuffer *fb = plane->plane.state->fb;
-	struct v4l2_rect src;
-	struct v4l2_rect dst;
-	dma_addr_t paddr[2] = { 0, };
-	u32 pixelformat = 0;
+	struct vsp1_du_atomic_config cfg = {
+		.pixelformat = 0,
+		.pitch = fb->pitches[0],
+		.alpha = state->alpha,
+		.zpos = state->zpos,
+	};
 	unsigned int i;
 
-	src.left = state->state.src_x >> 16;
-	src.top = state->state.src_y >> 16;
-	src.width = state->state.src_w >> 16;
-	src.height = state->state.src_h >> 16;
+	cfg.src.left = state->state.src_x >> 16;
+	cfg.src.top = state->state.src_y >> 16;
+	cfg.src.width = state->state.src_w >> 16;
+	cfg.src.height = state->state.src_h >> 16;
 
-	dst.left = state->state.crtc_x;
-	dst.top = state->state.crtc_y;
-	dst.width = state->state.crtc_w;
-	dst.height = state->state.crtc_h;
+	cfg.dst.left = state->state.crtc_x;
+	cfg.dst.top = state->state.crtc_y;
+	cfg.dst.width = state->state.crtc_w;
+	cfg.dst.height = state->state.crtc_h;
 
 	for (i = 0; i < state->format->planes; ++i) {
 		struct drm_gem_cma_object *gem;
 
 		gem = drm_fb_cma_get_gem_obj(fb, i);
-		paddr[i] = gem->paddr + fb->offsets[i];
+		cfg.mem[i] = gem->paddr + fb->offsets[i];
 	}
 
 	for (i = 0; i < ARRAY_SIZE(formats_kms); ++i) {
 		if (formats_kms[i] == state->format->fourcc) {
-			pixelformat = formats_v4l2[i];
+			cfg.pixelformat = formats_v4l2[i];
 			break;
 		}
 	}
 
-	WARN_ON(!pixelformat);
-
-	vsp1_du_atomic_update(plane->vsp->vsp, plane->index, pixelformat,
-			      fb->pitches[0], paddr, &src, &dst);
+	vsp1_du_atomic_update(plane->vsp->vsp, plane->index, &cfg);
 }
 
 static int rcar_du_vsp_plane_atomic_check(struct drm_plane *plane,
@@ -220,8 +219,7 @@ static void rcar_du_vsp_plane_atomic_update(struct drm_plane *plane,
 	if (plane->state->crtc)
 		rcar_du_vsp_plane_setup(rplane);
 	else
-		vsp1_du_atomic_update(rplane->vsp->vsp, rplane->index, 0, 0, 0,
-				      NULL, NULL);
+		vsp1_du_atomic_update(rplane->vsp->vsp, rplane->index, NULL);
 }
 
 static const struct drm_plane_helper_funcs rcar_du_vsp_plane_helper_funcs = {
@@ -269,6 +267,7 @@ static void rcar_du_vsp_plane_reset(struct drm_plane *plane)
 		return;
 
 	state->alpha = 255;
+	state->zpos = plane->type == DRM_PLANE_TYPE_PRIMARY ? 0 : 1;
 
 	plane->state = &state->state;
 	plane->state->plane = plane;
@@ -283,6 +282,8 @@ static int rcar_du_vsp_plane_atomic_set_property(struct drm_plane *plane,
 
 	if (property == rcdu->props.alpha)
 		rstate->alpha = val;
+	else if (property == rcdu->props.zpos)
+		rstate->zpos = val;
 	else
 		return -EINVAL;
 
@@ -299,6 +300,8 @@ static int rcar_du_vsp_plane_atomic_get_property(struct drm_plane *plane,
 
 	if (property == rcdu->props.alpha)
 		*val = rstate->alpha;
+	else if (property == rcdu->props.zpos)
+		*val = rstate->zpos;
 	else
 		return -EINVAL;
 
@@ -378,6 +381,8 @@ int rcar_du_vsp_init(struct rcar_du_vsp *vsp)
 
 		drm_object_attach_property(&plane->plane.base,
 					   rcdu->props.alpha, 255);
+		drm_object_attach_property(&plane->plane.base,
+					   rcdu->props.zpos, 1);
 	}
 
 	return 0;
diff --git a/drivers/gpu/drm/rcar-du/rcar_du_vsp.h b/drivers/gpu/drm/rcar-du/rcar_du_vsp.h
index df3bf3805c69..510dcc9c6816 100644
--- a/drivers/gpu/drm/rcar-du/rcar_du_vsp.h
+++ b/drivers/gpu/drm/rcar-du/rcar_du_vsp.h
@@ -44,6 +44,7 @@ static inline struct rcar_du_vsp_plane *to_rcar_vsp_plane(struct drm_plane *p)
  * @state: base DRM plane state
  * @format: information about the pixel format used by the plane
  * @alpha: value of the plane alpha property
+ * @zpos: value of the plane zpos property
  */
 struct rcar_du_vsp_plane_state {
 	struct drm_plane_state state;
@@ -51,6 +52,7 @@ struct rcar_du_vsp_plane_state {
 	const struct rcar_du_format_info *format;
 
 	unsigned int alpha;
+	unsigned int zpos;
 };
 
 static inline struct rcar_du_vsp_plane_state *
diff --git a/drivers/media/platform/Kconfig b/drivers/media/platform/Kconfig
index 2c2670c44333..129dfde1da1a 100644
--- a/drivers/media/platform/Kconfig
+++ b/drivers/media/platform/Kconfig
@@ -278,10 +278,24 @@ config VIDEO_RENESAS_JPU
 	  To compile this driver as a module, choose M here: the module
 	  will be called rcar_jpu.
 
+config VIDEO_RENESAS_FCP
+	tristate "Renesas Frame Compression Processor"
+	depends on ARCH_RENESAS || COMPILE_TEST
+	depends on OF
+	---help---
+	  This is a driver for the Renesas Frame Compression Processor (FCP).
+	  The FCP is a companion module of video processing modules in the
+	  Renesas R-Car Gen3 SoCs. It handles memory access for the codec,
+	  VSP and FDP modules.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called rcar-fcp.
+
 config VIDEO_RENESAS_VSP1
 	tristate "Renesas VSP1 Video Processing Engine"
 	depends on VIDEO_V4L2 && VIDEO_V4L2_SUBDEV_API && HAS_DMA
 	depends on (ARCH_RENESAS && OF) || COMPILE_TEST
+	depends on !ARM64 || VIDEO_RENESAS_FCP
 	select VIDEOBUF2_DMA_CONTIG
 	---help---
 	  This is a V4L2 driver for the Renesas VSP1 video processing engine.
diff --git a/drivers/media/platform/Makefile b/drivers/media/platform/Makefile
index c6b3b9259a03..21771c1a13fb 100644
--- a/drivers/media/platform/Makefile
+++ b/drivers/media/platform/Makefile
@@ -46,6 +46,7 @@ obj-$(CONFIG_VIDEO_SH_VOU)		+= sh_vou.o
 
 obj-$(CONFIG_SOC_CAMERA)		+= soc_camera/
 
+obj-$(CONFIG_VIDEO_RENESAS_FCP) 	+= rcar-fcp.o
 obj-$(CONFIG_VIDEO_RENESAS_JPU) 	+= rcar_jpu.o
 obj-$(CONFIG_VIDEO_RENESAS_VSP1)	+= vsp1/
 
diff --git a/drivers/media/platform/rcar-fcp.c b/drivers/media/platform/rcar-fcp.c
new file mode 100644
index 000000000000..6a7bcc3028b1
--- /dev/null
+++ b/drivers/media/platform/rcar-fcp.c
@@ -0,0 +1,181 @@
+/*
+ * rcar-fcp.c  --  R-Car Frame Compression Processor Driver
+ *
+ * Copyright (C) 2016 Renesas Electronics Corporation
+ *
+ * Contact: Laurent Pinchart (laurent.pinchart@ideasonboard.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/device.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/slab.h>
+
+#include <media/rcar-fcp.h>
+
+struct rcar_fcp_device {
+	struct list_head list;
+	struct device *dev;
+};
+
+static LIST_HEAD(fcp_devices);
+static DEFINE_MUTEX(fcp_lock);
+
+/* -----------------------------------------------------------------------------
+ * Public API
+ */
+
+/**
+ * rcar_fcp_get - Find and acquire a reference to an FCP instance
+ * @np: Device node of the FCP instance
+ *
+ * Search the list of registered FCP instances for the instance corresponding to
+ * the given device node.
+ *
+ * Return a pointer to the FCP instance, or an ERR_PTR if the instance can't be
+ * found.
+ */
+struct rcar_fcp_device *rcar_fcp_get(const struct device_node *np)
+{
+	struct rcar_fcp_device *fcp;
+
+	mutex_lock(&fcp_lock);
+
+	list_for_each_entry(fcp, &fcp_devices, list) {
+		if (fcp->dev->of_node != np)
+			continue;
+
+		/*
+		 * Make sure the module won't be unloaded behind our back. This
+		 * is a poor man's safety net, the module should really not be
+		 * unloaded while FCP users can be active.
+		 */
+		if (!try_module_get(fcp->dev->driver->owner))
+			fcp = NULL;
+
+		goto done;
+	}
+
+	fcp = ERR_PTR(-EPROBE_DEFER);
+
+done:
+	mutex_unlock(&fcp_lock);
+	return fcp;
+}
+EXPORT_SYMBOL_GPL(rcar_fcp_get);
+
+/**
+ * rcar_fcp_put - Release a reference to an FCP instance
+ * @fcp: The FCP instance
+ *
+ * Release the FCP instance acquired by a call to rcar_fcp_get().
+ */
+void rcar_fcp_put(struct rcar_fcp_device *fcp)
+{
+	if (fcp)
+		module_put(fcp->dev->driver->owner);
+}
+EXPORT_SYMBOL_GPL(rcar_fcp_put);
+
+/**
+ * rcar_fcp_enable - Enable an FCP
+ * @fcp: The FCP instance
+ *
+ * Before any memory access through an FCP is performed by a module, the FCP
+ * must be enabled by a call to this function. The enable calls are reference
+ * counted, each successful call must be followed by one rcar_fcp_disable()
+ * call when no more memory transfer can occur through the FCP.
+ *
+ * Return 0 on success or a negative error code if an error occurs. The enable
+ * reference count isn't increased when this function returns an error.
+ */
+int rcar_fcp_enable(struct rcar_fcp_device *fcp)
+{
+	if (!fcp)
+		return 0;
+
+	return pm_runtime_get_sync(fcp->dev);
+}
+EXPORT_SYMBOL_GPL(rcar_fcp_enable);
+
+/**
+ * rcar_fcp_disable - Disable an FCP
+ * @fcp: The FCP instance
+ *
+ * This function is the counterpart of rcar_fcp_enable(). As enable calls are
+ * reference counted a disable call may not disable the FCP synchronously.
+ */
+void rcar_fcp_disable(struct rcar_fcp_device *fcp)
+{
+	if (fcp)
+		pm_runtime_put(fcp->dev);
+}
+EXPORT_SYMBOL_GPL(rcar_fcp_disable);
+
+/* -----------------------------------------------------------------------------
+ * Platform Driver
+ */
+
+static int rcar_fcp_probe(struct platform_device *pdev)
+{
+	struct rcar_fcp_device *fcp;
+
+	fcp = devm_kzalloc(&pdev->dev, sizeof(*fcp), GFP_KERNEL);
+	if (fcp == NULL)
+		return -ENOMEM;
+
+	fcp->dev = &pdev->dev;
+
+	pm_runtime_enable(&pdev->dev);
+
+	mutex_lock(&fcp_lock);
+	list_add_tail(&fcp->list, &fcp_devices);
+	mutex_unlock(&fcp_lock);
+
+	platform_set_drvdata(pdev, fcp);
+
+	return 0;
+}
+
+static int rcar_fcp_remove(struct platform_device *pdev)
+{
+	struct rcar_fcp_device *fcp = platform_get_drvdata(pdev);
+
+	mutex_lock(&fcp_lock);
+	list_del(&fcp->list);
+	mutex_unlock(&fcp_lock);
+
+	pm_runtime_disable(&pdev->dev);
+
+	return 0;
+}
+
+static const struct of_device_id rcar_fcp_of_match[] = {
+	{ .compatible = "renesas,fcpv" },
+	{ },
+};
+
+static struct platform_driver rcar_fcp_platform_driver = {
+	.probe		= rcar_fcp_probe,
+	.remove		= rcar_fcp_remove,
+	.driver		= {
+		.name	= "rcar-fcp",
+		.of_match_table = rcar_fcp_of_match,
+		.suppress_bind_attrs = true,
+	},
+};
+
+module_platform_driver(rcar_fcp_platform_driver);
+
+MODULE_ALIAS("rcar-fcp");
+MODULE_AUTHOR("Laurent Pinchart <laurent.pinchart@ideasonboard.com>");
+MODULE_DESCRIPTION("Renesas FCP Driver");
+MODULE_LICENSE("GPL");
diff --git a/drivers/media/platform/vsp1/Makefile b/drivers/media/platform/vsp1/Makefile
index 95b3ac2ea7ef..1328e1bd2143 100644
--- a/drivers/media/platform/vsp1/Makefile
+++ b/drivers/media/platform/vsp1/Makefile
@@ -1,7 +1,8 @@
 vsp1-y					:= vsp1_drv.o vsp1_entity.o vsp1_pipe.o
 vsp1-y					+= vsp1_dl.o vsp1_drm.o vsp1_video.o
 vsp1-y					+= vsp1_rpf.o vsp1_rwpf.o vsp1_wpf.o
-vsp1-y					+= vsp1_hsit.o vsp1_lif.o vsp1_lut.o
+vsp1-y					+= vsp1_clu.o vsp1_hsit.o vsp1_lut.o
 vsp1-y					+= vsp1_bru.o vsp1_sru.o vsp1_uds.o
+vsp1-y					+= vsp1_lif.o
 
 obj-$(CONFIG_VIDEO_RENESAS_VSP1)	+= vsp1.o
diff --git a/drivers/media/platform/vsp1/vsp1.h b/drivers/media/platform/vsp1/vsp1.h
index 46738b6c5f72..06a2ec7e5ad4 100644
--- a/drivers/media/platform/vsp1/vsp1.h
+++ b/drivers/media/platform/vsp1/vsp1.h
@@ -25,11 +25,13 @@
 
 struct clk;
 struct device;
+struct rcar_fcp_device;
 
 struct vsp1_drm;
 struct vsp1_entity;
 struct vsp1_platform_data;
 struct vsp1_bru;
+struct vsp1_clu;
 struct vsp1_hsit;
 struct vsp1_lif;
 struct vsp1_lut;
@@ -45,6 +47,9 @@ struct vsp1_uds;
 #define VSP1_HAS_LUT		(1 << 1)
 #define VSP1_HAS_SRU		(1 << 2)
 #define VSP1_HAS_BRU		(1 << 3)
+#define VSP1_HAS_CLU		(1 << 4)
+#define VSP1_HAS_WPF_VFLIP	(1 << 5)
+#define VSP1_HAS_WPF_HFLIP	(1 << 6)
 
 struct vsp1_device_info {
 	u32 version;
@@ -62,12 +67,10 @@ struct vsp1_device {
 	const struct vsp1_device_info *info;
 
 	void __iomem *mmio;
-	struct clk *clock;
-
-	struct mutex lock;
-	int ref_count;
+	struct rcar_fcp_device *fcp;
 
 	struct vsp1_bru *bru;
+	struct vsp1_clu *clu;
 	struct vsp1_hsit *hsi;
 	struct vsp1_hsit *hst;
 	struct vsp1_lif *lif;
diff --git a/drivers/media/platform/vsp1/vsp1_bru.c b/drivers/media/platform/vsp1/vsp1_bru.c
index b1068c018011..8268b87727a7 100644
--- a/drivers/media/platform/vsp1/vsp1_bru.c
+++ b/drivers/media/platform/vsp1/vsp1_bru.c
@@ -249,7 +249,7 @@ static int bru_set_selection(struct v4l2_subdev *subdev,
 	return 0;
 }
 
-static struct v4l2_subdev_pad_ops bru_pad_ops = {
+static const struct v4l2_subdev_pad_ops bru_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = bru_enum_mbus_code,
 	.enum_frame_size = bru_enum_frame_size,
@@ -259,7 +259,7 @@ static struct v4l2_subdev_pad_ops bru_pad_ops = {
 	.set_selection = bru_set_selection,
 };
 
-static struct v4l2_subdev_ops bru_ops = {
+static const struct v4l2_subdev_ops bru_ops = {
 	.pad    = &bru_pad_ops,
 };
 
@@ -269,13 +269,16 @@ static struct v4l2_subdev_ops bru_ops = {
 
 static void bru_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_bru *bru = to_bru(&entity->subdev);
 	struct v4l2_mbus_framefmt *format;
 	unsigned int flags;
 	unsigned int i;
 
+	if (!full)
+		return;
+
 	format = vsp1_entity_get_pad_format(&bru->entity, bru->entity.config,
 					    bru->entity.source_pad);
 
@@ -390,7 +393,8 @@ struct vsp1_bru *vsp1_bru_create(struct vsp1_device *vsp1)
 	bru->entity.type = VSP1_ENTITY_BRU;
 
 	ret = vsp1_entity_init(vsp1, &bru->entity, "bru",
-			       vsp1->info->num_bru_inputs + 1, &bru_ops);
+			       vsp1->info->num_bru_inputs + 1, &bru_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_COMPOSER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/drivers/media/platform/vsp1/vsp1_clu.c b/drivers/media/platform/vsp1/vsp1_clu.c
new file mode 100644
index 000000000000..b63d2dbe5ea3
--- /dev/null
+++ b/drivers/media/platform/vsp1/vsp1_clu.c
@@ -0,0 +1,292 @@
+/*
+ * vsp1_clu.c  --  R-Car VSP1 Cubic Look-Up Table
+ *
+ * Copyright (C) 2015-2016 Renesas Electronics Corporation
+ *
+ * Contact: Laurent Pinchart (laurent.pinchart@ideasonboard.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/device.h>
+#include <linux/slab.h>
+
+#include <media/v4l2-subdev.h>
+
+#include "vsp1.h"
+#include "vsp1_clu.h"
+#include "vsp1_dl.h"
+
+#define CLU_MIN_SIZE				4U
+#define CLU_MAX_SIZE				8190U
+
+/* -----------------------------------------------------------------------------
+ * Device Access
+ */
+
+static inline void vsp1_clu_write(struct vsp1_clu *clu, struct vsp1_dl_list *dl,
+				  u32 reg, u32 data)
+{
+	vsp1_dl_list_write(dl, reg, data);
+}
+
+/* -----------------------------------------------------------------------------
+ * Controls
+ */
+
+#define V4L2_CID_VSP1_CLU_TABLE			(V4L2_CID_USER_BASE | 0x1001)
+#define V4L2_CID_VSP1_CLU_MODE			(V4L2_CID_USER_BASE | 0x1002)
+#define V4L2_CID_VSP1_CLU_MODE_2D		0
+#define V4L2_CID_VSP1_CLU_MODE_3D		1
+
+static int clu_set_table(struct vsp1_clu *clu, struct v4l2_ctrl *ctrl)
+{
+	struct vsp1_dl_body *dlb;
+	unsigned int i;
+
+	dlb = vsp1_dl_fragment_alloc(clu->entity.vsp1, 1 + 17 * 17 * 17);
+	if (!dlb)
+		return -ENOMEM;
+
+	vsp1_dl_fragment_write(dlb, VI6_CLU_ADDR, 0);
+	for (i = 0; i < 17 * 17 * 17; ++i)
+		vsp1_dl_fragment_write(dlb, VI6_CLU_DATA, ctrl->p_new.p_u32[i]);
+
+	spin_lock_irq(&clu->lock);
+	swap(clu->clu, dlb);
+	spin_unlock_irq(&clu->lock);
+
+	vsp1_dl_fragment_free(dlb);
+	return 0;
+}
+
+static int clu_s_ctrl(struct v4l2_ctrl *ctrl)
+{
+	struct vsp1_clu *clu =
+		container_of(ctrl->handler, struct vsp1_clu, ctrls);
+
+	switch (ctrl->id) {
+	case V4L2_CID_VSP1_CLU_TABLE:
+		clu_set_table(clu, ctrl);
+		break;
+
+	case V4L2_CID_VSP1_CLU_MODE:
+		clu->mode = ctrl->val;
+		break;
+	}
+
+	return 0;
+}
+
+static const struct v4l2_ctrl_ops clu_ctrl_ops = {
+	.s_ctrl = clu_s_ctrl,
+};
+
+static const struct v4l2_ctrl_config clu_table_control = {
+	.ops = &clu_ctrl_ops,
+	.id = V4L2_CID_VSP1_CLU_TABLE,
+	.name = "Look-Up Table",
+	.type = V4L2_CTRL_TYPE_U32,
+	.min = 0x00000000,
+	.max = 0x00ffffff,
+	.step = 1,
+	.def = 0,
+	.dims = { 17, 17, 17 },
+};
+
+static const char * const clu_mode_menu[] = {
+	"2D",
+	"3D",
+	NULL,
+};
+
+static const struct v4l2_ctrl_config clu_mode_control = {
+	.ops = &clu_ctrl_ops,
+	.id = V4L2_CID_VSP1_CLU_MODE,
+	.name = "Mode",
+	.type = V4L2_CTRL_TYPE_MENU,
+	.min = 0,
+	.max = 1,
+	.def = 1,
+	.qmenu = clu_mode_menu,
+};
+
+/* -----------------------------------------------------------------------------
+ * V4L2 Subdevice Pad Operations
+ */
+
+static int clu_enum_mbus_code(struct v4l2_subdev *subdev,
+			      struct v4l2_subdev_pad_config *cfg,
+			      struct v4l2_subdev_mbus_code_enum *code)
+{
+	static const unsigned int codes[] = {
+		MEDIA_BUS_FMT_ARGB8888_1X32,
+		MEDIA_BUS_FMT_AHSV8888_1X32,
+		MEDIA_BUS_FMT_AYUV8_1X32,
+	};
+
+	return vsp1_subdev_enum_mbus_code(subdev, cfg, code, codes,
+					  ARRAY_SIZE(codes));
+}
+
+static int clu_enum_frame_size(struct v4l2_subdev *subdev,
+			       struct v4l2_subdev_pad_config *cfg,
+			       struct v4l2_subdev_frame_size_enum *fse)
+{
+	return vsp1_subdev_enum_frame_size(subdev, cfg, fse, CLU_MIN_SIZE,
+					   CLU_MIN_SIZE, CLU_MAX_SIZE,
+					   CLU_MAX_SIZE);
+}
+
+static int clu_set_format(struct v4l2_subdev *subdev,
+			  struct v4l2_subdev_pad_config *cfg,
+			  struct v4l2_subdev_format *fmt)
+{
+	struct vsp1_clu *clu = to_clu(subdev);
+	struct v4l2_subdev_pad_config *config;
+	struct v4l2_mbus_framefmt *format;
+
+	config = vsp1_entity_get_pad_config(&clu->entity, cfg, fmt->which);
+	if (!config)
+		return -EINVAL;
+
+	/* Default to YUV if the requested format is not supported. */
+	if (fmt->format.code != MEDIA_BUS_FMT_ARGB8888_1X32 &&
+	    fmt->format.code != MEDIA_BUS_FMT_AHSV8888_1X32 &&
+	    fmt->format.code != MEDIA_BUS_FMT_AYUV8_1X32)
+		fmt->format.code = MEDIA_BUS_FMT_AYUV8_1X32;
+
+	format = vsp1_entity_get_pad_format(&clu->entity, config, fmt->pad);
+
+	if (fmt->pad == CLU_PAD_SOURCE) {
+		/* The CLU output format can't be modified. */
+		fmt->format = *format;
+		return 0;
+	}
+
+	format->code = fmt->format.code;
+	format->width = clamp_t(unsigned int, fmt->format.width,
+				CLU_MIN_SIZE, CLU_MAX_SIZE);
+	format->height = clamp_t(unsigned int, fmt->format.height,
+				 CLU_MIN_SIZE, CLU_MAX_SIZE);
+	format->field = V4L2_FIELD_NONE;
+	format->colorspace = V4L2_COLORSPACE_SRGB;
+
+	fmt->format = *format;
+
+	/* Propagate the format to the source pad. */
+	format = vsp1_entity_get_pad_format(&clu->entity, config,
+					    CLU_PAD_SOURCE);
+	*format = fmt->format;
+
+	return 0;
+}
+
+/* -----------------------------------------------------------------------------
+ * V4L2 Subdevice Operations
+ */
+
+static const struct v4l2_subdev_pad_ops clu_pad_ops = {
+	.init_cfg = vsp1_entity_init_cfg,
+	.enum_mbus_code = clu_enum_mbus_code,
+	.enum_frame_size = clu_enum_frame_size,
+	.get_fmt = vsp1_subdev_get_pad_format,
+	.set_fmt = clu_set_format,
+};
+
+static const struct v4l2_subdev_ops clu_ops = {
+	.pad    = &clu_pad_ops,
+};
+
+/* -----------------------------------------------------------------------------
+ * VSP1 Entity Operations
+ */
+
+static void clu_configure(struct vsp1_entity *entity,
+			  struct vsp1_pipeline *pipe,
+			  struct vsp1_dl_list *dl, bool full)
+{
+	struct vsp1_clu *clu = to_clu(&entity->subdev);
+	struct vsp1_dl_body *dlb;
+	unsigned long flags;
+	u32 ctrl = VI6_CLU_CTRL_AAI | VI6_CLU_CTRL_MVS | VI6_CLU_CTRL_EN;
+
+	/* The format can't be changed during streaming, only verify it at
+	 * stream start and store the information internally for future partial
+	 * reconfiguration calls.
+	 */
+	if (full) {
+		struct v4l2_mbus_framefmt *format;
+
+		format = vsp1_entity_get_pad_format(&clu->entity,
+						    clu->entity.config,
+						    CLU_PAD_SINK);
+		clu->yuv_mode = format->code == MEDIA_BUS_FMT_AYUV8_1X32;
+		return;
+	}
+
+	/* 2D mode can only be used with the YCbCr pixel encoding. */
+	if (clu->mode == V4L2_CID_VSP1_CLU_MODE_2D && clu->yuv_mode)
+		ctrl |= VI6_CLU_CTRL_AX1I_2D | VI6_CLU_CTRL_AX2I_2D
+		     |  VI6_CLU_CTRL_OS0_2D | VI6_CLU_CTRL_OS1_2D
+		     |  VI6_CLU_CTRL_OS2_2D | VI6_CLU_CTRL_M2D;
+
+	vsp1_clu_write(clu, dl, VI6_CLU_CTRL, ctrl);
+
+	spin_lock_irqsave(&clu->lock, flags);
+	dlb = clu->clu;
+	clu->clu = NULL;
+	spin_unlock_irqrestore(&clu->lock, flags);
+
+	if (dlb)
+		vsp1_dl_list_add_fragment(dl, dlb);
+}
+
+static const struct vsp1_entity_operations clu_entity_ops = {
+	.configure = clu_configure,
+};
+
+/* -----------------------------------------------------------------------------
+ * Initialization and Cleanup
+ */
+
+struct vsp1_clu *vsp1_clu_create(struct vsp1_device *vsp1)
+{
+	struct vsp1_clu *clu;
+	int ret;
+
+	clu = devm_kzalloc(vsp1->dev, sizeof(*clu), GFP_KERNEL);
+	if (clu == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	spin_lock_init(&clu->lock);
+
+	clu->entity.ops = &clu_entity_ops;
+	clu->entity.type = VSP1_ENTITY_CLU;
+
+	ret = vsp1_entity_init(vsp1, &clu->entity, "clu", 2, &clu_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_LUT);
+	if (ret < 0)
+		return ERR_PTR(ret);
+
+	/* Initialize the control handler. */
+	v4l2_ctrl_handler_init(&clu->ctrls, 2);
+	v4l2_ctrl_new_custom(&clu->ctrls, &clu_table_control, NULL);
+	v4l2_ctrl_new_custom(&clu->ctrls, &clu_mode_control, NULL);
+
+	clu->entity.subdev.ctrl_handler = &clu->ctrls;
+
+	if (clu->ctrls.error) {
+		dev_err(vsp1->dev, "clu: failed to initialize controls\n");
+		ret = clu->ctrls.error;
+		vsp1_entity_destroy(&clu->entity);
+		return ERR_PTR(ret);
+	}
+
+	v4l2_ctrl_handler_setup(&clu->ctrls);
+
+	return clu;
+}
diff --git a/drivers/media/platform/vsp1/vsp1_clu.h b/drivers/media/platform/vsp1/vsp1_clu.h
new file mode 100644
index 000000000000..036e0a2f1a42
--- /dev/null
+++ b/drivers/media/platform/vsp1/vsp1_clu.h
@@ -0,0 +1,48 @@
+/*
+ * vsp1_clu.h  --  R-Car VSP1 Cubic Look-Up Table
+ *
+ * Copyright (C) 2015 Renesas Corporation
+ *
+ * Contact: Laurent Pinchart (laurent.pinchart@ideasonboard.com)
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+#ifndef __VSP1_CLU_H__
+#define __VSP1_CLU_H__
+
+#include <linux/spinlock.h>
+
+#include <media/media-entity.h>
+#include <media/v4l2-ctrls.h>
+#include <media/v4l2-subdev.h>
+
+#include "vsp1_entity.h"
+
+struct vsp1_device;
+struct vsp1_dl_body;
+
+#define CLU_PAD_SINK				0
+#define CLU_PAD_SOURCE				1
+
+struct vsp1_clu {
+	struct vsp1_entity entity;
+
+	struct v4l2_ctrl_handler ctrls;
+
+	bool yuv_mode;
+	spinlock_t lock;
+	unsigned int mode;
+	struct vsp1_dl_body *clu;
+};
+
+static inline struct vsp1_clu *to_clu(struct v4l2_subdev *subdev)
+{
+	return container_of(subdev, struct vsp1_clu, entity.subdev);
+}
+
+struct vsp1_clu *vsp1_clu_create(struct vsp1_device *vsp1);
+
+#endif /* __VSP1_CLU_H__ */
diff --git a/drivers/media/platform/vsp1/vsp1_dl.c b/drivers/media/platform/vsp1/vsp1_dl.c
index e238d9b9376b..37c3518aa2a8 100644
--- a/drivers/media/platform/vsp1/vsp1_dl.c
+++ b/drivers/media/platform/vsp1/vsp1_dl.c
@@ -15,6 +15,7 @@
 #include <linux/dma-mapping.h>
 #include <linux/gfp.h>
 #include <linux/slab.h>
+#include <linux/workqueue.h>
 
 #include "vsp1.h"
 #include "vsp1_dl.h"
@@ -92,11 +93,13 @@ enum vsp1_dl_mode {
  * @index: index of the related WPF
  * @mode: display list operation mode (header or headerless)
  * @vsp1: the VSP1 device
- * @lock: protects the active, queued and pending lists
+ * @lock: protects the free, active, queued, pending and gc_fragments lists
  * @free: array of all free display lists
  * @active: list currently being processed (loaded) by hardware
  * @queued: list queued to the hardware (written to the DL registers)
  * @pending: list waiting to be queued to the hardware
+ * @gc_work: fragments garbage collector work struct
+ * @gc_fragments: array of display list fragments waiting to be freed
  */
 struct vsp1_dl_manager {
 	unsigned int index;
@@ -108,6 +111,9 @@ struct vsp1_dl_manager {
 	struct vsp1_dl_list *active;
 	struct vsp1_dl_list *queued;
 	struct vsp1_dl_list *pending;
+
+	struct work_struct gc_work;
+	struct list_head gc_fragments;
 };
 
 /* -----------------------------------------------------------------------------
@@ -262,21 +268,10 @@ static struct vsp1_dl_list *vsp1_dl_list_alloc(struct vsp1_dl_manager *dlm)
 	return dl;
 }
 
-static void vsp1_dl_list_free_fragments(struct vsp1_dl_list *dl)
-{
-	struct vsp1_dl_body *dlb, *next;
-
-	list_for_each_entry_safe(dlb, next, &dl->fragments, list) {
-		list_del(&dlb->list);
-		vsp1_dl_body_cleanup(dlb);
-		kfree(dlb);
-	}
-}
-
 static void vsp1_dl_list_free(struct vsp1_dl_list *dl)
 {
 	vsp1_dl_body_cleanup(&dl->body0);
-	vsp1_dl_list_free_fragments(dl);
+	list_splice_init(&dl->fragments, &dl->dlm->gc_fragments);
 	kfree(dl);
 }
 
@@ -311,7 +306,16 @@ static void __vsp1_dl_list_put(struct vsp1_dl_list *dl)
 	if (!dl)
 		return;
 
-	vsp1_dl_list_free_fragments(dl);
+	/* We can't free fragments here as DMA memory can only be freed in
+	 * interruptible context. Move all fragments to the display list
+	 * manager's list of fragments to be freed, they will be
+	 * garbage-collected by the work queue.
+	 */
+	if (!list_empty(&dl->fragments)) {
+		list_splice_init(&dl->fragments, &dl->dlm->gc_fragments);
+		schedule_work(&dl->dlm->gc_work);
+	}
+
 	dl->body0.num_entries = 0;
 
 	list_add_tail(&dl->list, &dl->dlm->free);
@@ -550,6 +554,40 @@ void vsp1_dlm_reset(struct vsp1_dl_manager *dlm)
 	dlm->pending = NULL;
 }
 
+/*
+ * Free all fragments awaiting to be garbage-collected.
+ *
+ * This function must be called without the display list manager lock held.
+ */
+static void vsp1_dlm_fragments_free(struct vsp1_dl_manager *dlm)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&dlm->lock, flags);
+
+	while (!list_empty(&dlm->gc_fragments)) {
+		struct vsp1_dl_body *dlb;
+
+		dlb = list_first_entry(&dlm->gc_fragments, struct vsp1_dl_body,
+				       list);
+		list_del(&dlb->list);
+
+		spin_unlock_irqrestore(&dlm->lock, flags);
+		vsp1_dl_fragment_free(dlb);
+		spin_lock_irqsave(&dlm->lock, flags);
+	}
+
+	spin_unlock_irqrestore(&dlm->lock, flags);
+}
+
+static void vsp1_dlm_garbage_collect(struct work_struct *work)
+{
+	struct vsp1_dl_manager *dlm =
+		container_of(work, struct vsp1_dl_manager, gc_work);
+
+	vsp1_dlm_fragments_free(dlm);
+}
+
 struct vsp1_dl_manager *vsp1_dlm_create(struct vsp1_device *vsp1,
 					unsigned int index,
 					unsigned int prealloc)
@@ -568,6 +606,8 @@ struct vsp1_dl_manager *vsp1_dlm_create(struct vsp1_device *vsp1,
 
 	spin_lock_init(&dlm->lock);
 	INIT_LIST_HEAD(&dlm->free);
+	INIT_LIST_HEAD(&dlm->gc_fragments);
+	INIT_WORK(&dlm->gc_work, vsp1_dlm_garbage_collect);
 
 	for (i = 0; i < prealloc; ++i) {
 		struct vsp1_dl_list *dl;
@@ -589,8 +629,12 @@ void vsp1_dlm_destroy(struct vsp1_dl_manager *dlm)
 	if (!dlm)
 		return;
 
+	cancel_work_sync(&dlm->gc_work);
+
 	list_for_each_entry_safe(dl, next, &dlm->free, list) {
 		list_del(&dl->list);
 		vsp1_dl_list_free(dl);
 	}
+
+	vsp1_dlm_fragments_free(dlm);
 }
diff --git a/drivers/media/platform/vsp1/vsp1_drm.c b/drivers/media/platform/vsp1/vsp1_drm.c
index fc4bbc401e67..fe9665e57b3b 100644
--- a/drivers/media/platform/vsp1/vsp1_drm.c
+++ b/drivers/media/platform/vsp1/vsp1_drm.c
@@ -230,42 +230,33 @@ EXPORT_SYMBOL_GPL(vsp1_du_atomic_begin);
  * vsp1_du_atomic_update - Setup one RPF input of the VSP pipeline
  * @dev: the VSP device
  * @rpf_index: index of the RPF to setup (0-based)
- * @pixelformat: V4L2 pixel format for the RPF memory input
- * @pitch: number of bytes per line in the image stored in memory
- * @mem: DMA addresses of the memory buffers (one per plane)
- * @src: the source crop rectangle for the RPF
- * @dst: the destination compose rectangle for the BRU input
- * @alpha: global alpha value for the input
- * @zpos: the Z-order position of the input
+ * @cfg: the RPF configuration
  *
- * Configure the VSP to perform composition of the image referenced by @mem
- * through RPF @rpf_index, using the @src crop rectangle and the @dst
+ * Configure the VSP to perform image composition through RPF @rpf_index as
+ * described by the @cfg configuration. The image to compose is referenced by
+ * @cfg.mem and composed using the @cfg.src crop rectangle and the @cfg.dst
  * composition rectangle. The Z-order is configurable with higher @zpos values
  * displayed on top.
  *
- * Image format as stored in memory is expressed as a V4L2 @pixelformat value.
- * As a special case, setting the pixel format to 0 will disable the RPF. The
- * @pitch, @mem, @src and @dst parameters are ignored in that case. Calling the
+ * If the @cfg configuration is NULL, the RPF will be disabled. Calling the
  * function on a disabled RPF is allowed.
  *
- * The memory pitch is configurable to allow for padding at end of lines, or
- * simple for images that extend beyond the crop rectangle boundaries. The
- * @pitch value is expressed in bytes and applies to all planes for multiplanar
- * formats.
+ * Image format as stored in memory is expressed as a V4L2 @cfg.pixelformat
+ * value. The memory pitch is configurable to allow for padding at end of lines,
+ * or simply for images that extend beyond the crop rectangle boundaries. The
+ * @cfg.pitch value is expressed in bytes and applies to all planes for
+ * multiplanar formats.
  *
  * The source memory buffer is referenced by the DMA address of its planes in
- * the @mem array. Up to two planes are supported. The second plane DMA address
- * is ignored for formats using a single plane.
+ * the @cfg.mem array. Up to two planes are supported. The second plane DMA
+ * address is ignored for formats using a single plane.
  *
  * This function isn't reentrant, the caller needs to serialize calls.
  *
  * Return 0 on success or a negative error code on failure.
  */
-int vsp1_du_atomic_update_ext(struct device *dev, unsigned int rpf_index,
-			      u32 pixelformat, unsigned int pitch,
-			      dma_addr_t mem[2], const struct v4l2_rect *src,
-			      const struct v4l2_rect *dst, unsigned int alpha,
-			      unsigned int zpos)
+int vsp1_du_atomic_update(struct device *dev, unsigned int rpf_index,
+			  const struct vsp1_du_atomic_config *cfg)
 {
 	struct vsp1_device *vsp1 = dev_get_drvdata(dev);
 	const struct vsp1_format_info *fmtinfo;
@@ -276,7 +267,7 @@ int vsp1_du_atomic_update_ext(struct device *dev, unsigned int rpf_index,
 
 	rpf = vsp1->rpf[rpf_index];
 
-	if (pixelformat == 0) {
+	if (!cfg) {
 		dev_dbg(vsp1->dev, "%s: RPF%u: disable requested\n", __func__,
 			rpf_index);
 
@@ -287,38 +278,39 @@ int vsp1_du_atomic_update_ext(struct device *dev, unsigned int rpf_index,
 	dev_dbg(vsp1->dev,
 		"%s: RPF%u: (%u,%u)/%ux%u -> (%u,%u)/%ux%u (%08x), pitch %u dma { %pad, %pad } zpos %u\n",
 		__func__, rpf_index,
-		src->left, src->top, src->width, src->height,
-		dst->left, dst->top, dst->width, dst->height,
-		pixelformat, pitch, &mem[0], &mem[1], zpos);
+		cfg->src.left, cfg->src.top, cfg->src.width, cfg->src.height,
+		cfg->dst.left, cfg->dst.top, cfg->dst.width, cfg->dst.height,
+		cfg->pixelformat, cfg->pitch, &cfg->mem[0], &cfg->mem[1],
+		cfg->zpos);
 
 	/* Store the format, stride, memory buffer address, crop and compose
 	 * rectangles and Z-order position and for the input.
 	 */
-	fmtinfo = vsp1_get_format_info(pixelformat);
+	fmtinfo = vsp1_get_format_info(cfg->pixelformat);
 	if (!fmtinfo) {
 		dev_dbg(vsp1->dev, "Unsupport pixel format %08x for RPF\n",
-			pixelformat);
+			cfg->pixelformat);
 		return -EINVAL;
 	}
 
 	rpf->fmtinfo = fmtinfo;
 	rpf->format.num_planes = fmtinfo->planes;
-	rpf->format.plane_fmt[0].bytesperline = pitch;
-	rpf->format.plane_fmt[1].bytesperline = pitch;
-	rpf->alpha = alpha;
+	rpf->format.plane_fmt[0].bytesperline = cfg->pitch;
+	rpf->format.plane_fmt[1].bytesperline = cfg->pitch;
+	rpf->alpha = cfg->alpha;
 
-	rpf->mem.addr[0] = mem[0];
-	rpf->mem.addr[1] = mem[1];
+	rpf->mem.addr[0] = cfg->mem[0];
+	rpf->mem.addr[1] = cfg->mem[1];
 	rpf->mem.addr[2] = 0;
 
-	vsp1->drm->inputs[rpf_index].crop = *src;
-	vsp1->drm->inputs[rpf_index].compose = *dst;
-	vsp1->drm->inputs[rpf_index].zpos = zpos;
+	vsp1->drm->inputs[rpf_index].crop = cfg->src;
+	vsp1->drm->inputs[rpf_index].compose = cfg->dst;
+	vsp1->drm->inputs[rpf_index].zpos = cfg->zpos;
 	vsp1->drm->inputs[rpf_index].enabled = true;
 
 	return 0;
 }
-EXPORT_SYMBOL_GPL(vsp1_du_atomic_update_ext);
+EXPORT_SYMBOL_GPL(vsp1_du_atomic_update);
 
 static int vsp1_du_setup_rpf_pipe(struct vsp1_device *vsp1,
 				  struct vsp1_rwpf *rpf, unsigned int bru_input)
@@ -499,8 +491,10 @@ void vsp1_du_atomic_flush(struct device *dev)
 
 		vsp1_entity_route_setup(entity, pipe->dl);
 
-		if (entity->ops->configure)
-			entity->ops->configure(entity, pipe, pipe->dl);
+		if (entity->ops->configure) {
+			entity->ops->configure(entity, pipe, pipe->dl, true);
+			entity->ops->configure(entity, pipe, pipe->dl, false);
+		}
 
 		/* The memory buffer address must be applied after configuring
 		 * the RPF to make sure the crop offset are computed.
diff --git a/drivers/media/platform/vsp1/vsp1_drv.c b/drivers/media/platform/vsp1/vsp1_drv.c
index e2d779fac0eb..e1377ffe3d68 100644
--- a/drivers/media/platform/vsp1/vsp1_drv.c
+++ b/drivers/media/platform/vsp1/vsp1_drv.c
@@ -19,12 +19,15 @@
 #include <linux/of.h>
 #include <linux/of_device.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 #include <linux/videodev2.h>
 
+#include <media/rcar-fcp.h>
 #include <media/v4l2-subdev.h>
 
 #include "vsp1.h"
 #include "vsp1_bru.h"
+#include "vsp1_clu.h"
 #include "vsp1_dl.h"
 #include "vsp1_drm.h"
 #include "vsp1_hsit.h"
@@ -145,7 +148,7 @@ static int vsp1_uapi_create_links(struct vsp1_device *vsp1)
 			return ret;
 	}
 
-	if (vsp1->info->features & VSP1_HAS_LIF) {
+	if (vsp1->lif) {
 		ret = media_create_pad_link(&vsp1->wpf[0]->entity.subdev.entity,
 					    RWPF_PAD_SOURCE,
 					    &vsp1->lif->entity.subdev.entity,
@@ -168,19 +171,15 @@ static int vsp1_uapi_create_links(struct vsp1_device *vsp1)
 
 	for (i = 0; i < vsp1->info->wpf_count; ++i) {
 		/* Connect the video device to the WPF. All connections are
-		 * immutable except for the WPF0 source link if a LIF is
-		 * present.
+		 * immutable.
 		 */
 		struct vsp1_rwpf *wpf = vsp1->wpf[i];
-		unsigned int flags = MEDIA_LNK_FL_ENABLED;
-
-		if (!(vsp1->info->features & VSP1_HAS_LIF) || i != 0)
-			flags |= MEDIA_LNK_FL_IMMUTABLE;
 
 		ret = media_create_pad_link(&wpf->entity.subdev.entity,
 					    RWPF_PAD_SOURCE,
 					    &wpf->video->video.entity, 0,
-					    flags);
+					    MEDIA_LNK_FL_IMMUTABLE |
+					    MEDIA_LNK_FL_ENABLED);
 		if (ret < 0)
 			return ret;
 	}
@@ -204,7 +203,8 @@ static void vsp1_destroy_entities(struct vsp1_device *vsp1)
 	}
 
 	v4l2_device_unregister(&vsp1->v4l2_dev);
-	media_device_unregister(&vsp1->media_dev);
+	if (vsp1->info->uapi)
+		media_device_unregister(&vsp1->media_dev);
 	media_device_cleanup(&vsp1->media_dev);
 
 	if (!vsp1->info->uapi)
@@ -252,6 +252,16 @@ static int vsp1_create_entities(struct vsp1_device *vsp1)
 		list_add_tail(&vsp1->bru->entity.list_dev, &vsp1->entities);
 	}
 
+	if (vsp1->info->features & VSP1_HAS_CLU) {
+		vsp1->clu = vsp1_clu_create(vsp1);
+		if (IS_ERR(vsp1->clu)) {
+			ret = PTR_ERR(vsp1->clu);
+			goto done;
+		}
+
+		list_add_tail(&vsp1->clu->entity.list_dev, &vsp1->entities);
+	}
+
 	vsp1->hsi = vsp1_hsit_create(vsp1, true);
 	if (IS_ERR(vsp1->hsi)) {
 		ret = PTR_ERR(vsp1->hsi);
@@ -268,7 +278,11 @@ static int vsp1_create_entities(struct vsp1_device *vsp1)
 
 	list_add_tail(&vsp1->hst->entity.list_dev, &vsp1->entities);
 
-	if (vsp1->info->features & VSP1_HAS_LIF) {
+	/* The LIF is only supported when used in conjunction with the DU, in
+	 * which case the userspace API is disabled. If the userspace API is
+	 * enabled skip the LIF, even when present.
+	 */
+	if (vsp1->info->features & VSP1_HAS_LIF && !vsp1->info->uapi) {
 		vsp1->lif = vsp1_lif_create(vsp1);
 		if (IS_ERR(vsp1->lif)) {
 			ret = PTR_ERR(vsp1->lif);
@@ -379,14 +393,15 @@ static int vsp1_create_entities(struct vsp1_device *vsp1)
 	/* Register subdev nodes if the userspace API is enabled or initialize
 	 * the DRM pipeline otherwise.
 	 */
-	if (vsp1->info->uapi)
+	if (vsp1->info->uapi) {
 		ret = v4l2_device_register_subdev_nodes(&vsp1->v4l2_dev);
-	else
-		ret = vsp1_drm_init(vsp1);
-	if (ret < 0)
-		goto done;
+		if (ret < 0)
+			goto done;
 
-	ret = media_device_register(mdev);
+		ret = media_device_register(mdev);
+	} else {
+		ret = vsp1_drm_init(vsp1);
+	}
 
 done:
 	if (ret < 0)
@@ -462,35 +477,16 @@ static int vsp1_device_init(struct vsp1_device *vsp1)
 /*
  * vsp1_device_get - Acquire the VSP1 device
  *
- * Increment the VSP1 reference count and initialize the device if the first
- * reference is taken.
+ * Make sure the device is not suspended and initialize it if needed.
  *
  * Return 0 on success or a negative error code otherwise.
  */
 int vsp1_device_get(struct vsp1_device *vsp1)
 {
-	int ret = 0;
-
-	mutex_lock(&vsp1->lock);
-	if (vsp1->ref_count > 0)
-		goto done;
-
-	ret = clk_prepare_enable(vsp1->clock);
-	if (ret < 0)
-		goto done;
-
-	ret = vsp1_device_init(vsp1);
-	if (ret < 0) {
-		clk_disable_unprepare(vsp1->clock);
-		goto done;
-	}
-
-done:
-	if (!ret)
-		vsp1->ref_count++;
+	int ret;
 
-	mutex_unlock(&vsp1->lock);
-	return ret;
+	ret = pm_runtime_get_sync(vsp1->dev);
+	return ret < 0 ? ret : 0;
 }
 
 /*
@@ -501,12 +497,7 @@ done:
  */
 void vsp1_device_put(struct vsp1_device *vsp1)
 {
-	mutex_lock(&vsp1->lock);
-
-	if (--vsp1->ref_count == 0)
-		clk_disable_unprepare(vsp1->clock);
-
-	mutex_unlock(&vsp1->lock);
+	pm_runtime_put_sync(vsp1->dev);
 }
 
 /* -----------------------------------------------------------------------------
@@ -518,14 +509,8 @@ static int vsp1_pm_suspend(struct device *dev)
 {
 	struct vsp1_device *vsp1 = dev_get_drvdata(dev);
 
-	WARN_ON(mutex_is_locked(&vsp1->lock));
-
-	if (vsp1->ref_count == 0)
-		return 0;
-
 	vsp1_pipelines_suspend(vsp1);
-
-	clk_disable_unprepare(vsp1->clock);
+	pm_runtime_force_suspend(vsp1->dev);
 
 	return 0;
 }
@@ -534,21 +519,39 @@ static int vsp1_pm_resume(struct device *dev)
 {
 	struct vsp1_device *vsp1 = dev_get_drvdata(dev);
 
-	WARN_ON(mutex_is_locked(&vsp1->lock));
+	pm_runtime_force_resume(vsp1->dev);
+	vsp1_pipelines_resume(vsp1);
 
-	if (vsp1->ref_count == 0)
-		return 0;
+	return 0;
+}
+#endif
 
-	clk_prepare_enable(vsp1->clock);
+static int vsp1_pm_runtime_suspend(struct device *dev)
+{
+	struct vsp1_device *vsp1 = dev_get_drvdata(dev);
 
-	vsp1_pipelines_resume(vsp1);
+	rcar_fcp_disable(vsp1->fcp);
 
 	return 0;
 }
-#endif
+
+static int vsp1_pm_runtime_resume(struct device *dev)
+{
+	struct vsp1_device *vsp1 = dev_get_drvdata(dev);
+	int ret;
+
+	if (vsp1->info) {
+		ret = vsp1_device_init(vsp1);
+		if (ret < 0)
+			return ret;
+	}
+
+	return rcar_fcp_enable(vsp1->fcp);
+}
 
 static const struct dev_pm_ops vsp1_pm_ops = {
 	SET_SYSTEM_SLEEP_PM_OPS(vsp1_pm_suspend, vsp1_pm_resume)
+	SET_RUNTIME_PM_OPS(vsp1_pm_runtime_suspend, vsp1_pm_runtime_resume, NULL)
 };
 
 /* -----------------------------------------------------------------------------
@@ -559,7 +562,8 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 	{
 		.version = VI6_IP_VERSION_MODEL_VSPS_H2,
 		.gen = 2,
-		.features = VSP1_HAS_BRU | VSP1_HAS_LUT | VSP1_HAS_SRU,
+		.features = VSP1_HAS_BRU | VSP1_HAS_CLU | VSP1_HAS_LUT
+			  | VSP1_HAS_SRU | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
 		.uds_count = 3,
 		.wpf_count = 4,
@@ -568,9 +572,9 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPR_H2,
 		.gen = 2,
-		.features = VSP1_HAS_BRU | VSP1_HAS_SRU,
+		.features = VSP1_HAS_BRU | VSP1_HAS_SRU | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
-		.uds_count = 1,
+		.uds_count = 3,
 		.wpf_count = 4,
 		.num_bru_inputs = 4,
 		.uapi = true,
@@ -580,22 +584,24 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 		.features = VSP1_HAS_BRU | VSP1_HAS_LIF | VSP1_HAS_LUT,
 		.rpf_count = 4,
 		.uds_count = 1,
-		.wpf_count = 4,
+		.wpf_count = 1,
 		.num_bru_inputs = 4,
 		.uapi = true,
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPS_M2,
 		.gen = 2,
-		.features = VSP1_HAS_BRU | VSP1_HAS_LUT | VSP1_HAS_SRU,
+		.features = VSP1_HAS_BRU | VSP1_HAS_CLU | VSP1_HAS_LUT
+			  | VSP1_HAS_SRU | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
-		.uds_count = 3,
+		.uds_count = 1,
 		.wpf_count = 4,
 		.num_bru_inputs = 4,
 		.uapi = true,
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPI_GEN3,
 		.gen = 3,
-		.features = VSP1_HAS_LUT | VSP1_HAS_SRU,
+		.features = VSP1_HAS_CLU | VSP1_HAS_LUT | VSP1_HAS_SRU
+			  | VSP1_HAS_WPF_HFLIP | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 1,
 		.uds_count = 1,
 		.wpf_count = 1,
@@ -603,7 +609,7 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPBD_GEN3,
 		.gen = 3,
-		.features = VSP1_HAS_BRU,
+		.features = VSP1_HAS_BRU | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
 		.wpf_count = 1,
 		.num_bru_inputs = 5,
@@ -611,7 +617,8 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPBC_GEN3,
 		.gen = 3,
-		.features = VSP1_HAS_BRU | VSP1_HAS_LUT,
+		.features = VSP1_HAS_BRU | VSP1_HAS_CLU | VSP1_HAS_LUT
+			  | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
 		.wpf_count = 1,
 		.num_bru_inputs = 5,
@@ -619,7 +626,7 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 	}, {
 		.version = VI6_IP_VERSION_MODEL_VSPD_GEN3,
 		.gen = 3,
-		.features = VSP1_HAS_BRU | VSP1_HAS_LIF,
+		.features = VSP1_HAS_BRU | VSP1_HAS_LIF | VSP1_HAS_WPF_VFLIP,
 		.rpf_count = 5,
 		.wpf_count = 2,
 		.num_bru_inputs = 5,
@@ -629,6 +636,7 @@ static const struct vsp1_device_info vsp1_device_infos[] = {
 static int vsp1_probe(struct platform_device *pdev)
 {
 	struct vsp1_device *vsp1;
+	struct device_node *fcp_node;
 	struct resource *irq;
 	struct resource *io;
 	unsigned int i;
@@ -640,22 +648,17 @@ static int vsp1_probe(struct platform_device *pdev)
 		return -ENOMEM;
 
 	vsp1->dev = &pdev->dev;
-	mutex_init(&vsp1->lock);
 	INIT_LIST_HEAD(&vsp1->entities);
 	INIT_LIST_HEAD(&vsp1->videos);
 
-	/* I/O, IRQ and clock resources */
+	platform_set_drvdata(pdev, vsp1);
+
+	/* I/O and IRQ resources (clock managed by the clock PM domain) */
 	io = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	vsp1->mmio = devm_ioremap_resource(&pdev->dev, io);
 	if (IS_ERR(vsp1->mmio))
 		return PTR_ERR(vsp1->mmio);
 
-	vsp1->clock = devm_clk_get(&pdev->dev, NULL);
-	if (IS_ERR(vsp1->clock)) {
-		dev_err(&pdev->dev, "failed to get clock\n");
-		return PTR_ERR(vsp1->clock);
-	}
-
 	irq = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
 	if (!irq) {
 		dev_err(&pdev->dev, "missing IRQ\n");
@@ -669,13 +672,27 @@ static int vsp1_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	/* FCP (optional) */
+	fcp_node = of_parse_phandle(pdev->dev.of_node, "renesas,fcp", 0);
+	if (fcp_node) {
+		vsp1->fcp = rcar_fcp_get(fcp_node);
+		of_node_put(fcp_node);
+		if (IS_ERR(vsp1->fcp)) {
+			dev_dbg(&pdev->dev, "FCP not found (%ld)\n",
+				PTR_ERR(vsp1->fcp));
+			return PTR_ERR(vsp1->fcp);
+		}
+	}
+
 	/* Configure device parameters based on the version register. */
-	ret = clk_prepare_enable(vsp1->clock);
+	pm_runtime_enable(&pdev->dev);
+
+	ret = pm_runtime_get_sync(&pdev->dev);
 	if (ret < 0)
-		return ret;
+		goto done;
 
 	version = vsp1_read(vsp1, VI6_IP_VERSION);
-	clk_disable_unprepare(vsp1->clock);
+	pm_runtime_put_sync(&pdev->dev);
 
 	for (i = 0; i < ARRAY_SIZE(vsp1_device_infos); ++i) {
 		if ((version & VI6_IP_VERSION_MODEL_MASK) ==
@@ -687,7 +704,8 @@ static int vsp1_probe(struct platform_device *pdev)
 
 	if (!vsp1->info) {
 		dev_err(&pdev->dev, "unsupported IP version 0x%08x\n", version);
-		return -ENXIO;
+		ret = -ENXIO;
+		goto done;
 	}
 
 	dev_dbg(&pdev->dev, "IP version 0x%08x\n", version);
@@ -696,12 +714,14 @@ static int vsp1_probe(struct platform_device *pdev)
 	ret = vsp1_create_entities(vsp1);
 	if (ret < 0) {
 		dev_err(&pdev->dev, "failed to create entities\n");
-		return ret;
+		goto done;
 	}
 
-	platform_set_drvdata(pdev, vsp1);
+done:
+	if (ret)
+		pm_runtime_disable(&pdev->dev);
 
-	return 0;
+	return ret;
 }
 
 static int vsp1_remove(struct platform_device *pdev)
@@ -709,6 +729,9 @@ static int vsp1_remove(struct platform_device *pdev)
 	struct vsp1_device *vsp1 = platform_get_drvdata(pdev);
 
 	vsp1_destroy_entities(vsp1);
+	rcar_fcp_put(vsp1->fcp);
+
+	pm_runtime_disable(&pdev->dev);
 
 	return 0;
 }
diff --git a/drivers/media/platform/vsp1/vsp1_entity.c b/drivers/media/platform/vsp1/vsp1_entity.c
index 3d070bcc6053..4cf6cc719c00 100644
--- a/drivers/media/platform/vsp1/vsp1_entity.c
+++ b/drivers/media/platform/vsp1/vsp1_entity.c
@@ -22,6 +22,12 @@
 #include "vsp1_dl.h"
 #include "vsp1_entity.h"
 
+static inline struct vsp1_entity *
+media_entity_to_vsp1_entity(struct media_entity *entity)
+{
+	return container_of(entity, struct vsp1_entity, subdev.entity);
+}
+
 void vsp1_entity_route_setup(struct vsp1_entity *source,
 			     struct vsp1_dl_list *dl)
 {
@@ -30,7 +36,7 @@ void vsp1_entity_route_setup(struct vsp1_entity *source,
 	if (source->route->reg == 0)
 		return;
 
-	sink = container_of(source->sink, struct vsp1_entity, subdev.entity);
+	sink = media_entity_to_vsp1_entity(source->sink);
 	vsp1_dl_list_write(dl, source->route->reg,
 			   sink->route->inputs[source->sink_pad]);
 }
@@ -81,12 +87,30 @@ vsp1_entity_get_pad_format(struct vsp1_entity *entity,
 	return v4l2_subdev_get_try_format(&entity->subdev, cfg, pad);
 }
 
+/**
+ * vsp1_entity_get_pad_selection - Get a pad selection from storage for entity
+ * @entity: the entity
+ * @cfg: the configuration storage
+ * @pad: the pad number
+ * @target: the selection target
+ *
+ * Return the selection rectangle stored in the given configuration for an
+ * entity's pad. The configuration can be an ACTIVE or TRY configuration. The
+ * selection target can be COMPOSE or CROP.
+ */
 struct v4l2_rect *
-vsp1_entity_get_pad_compose(struct vsp1_entity *entity,
-			    struct v4l2_subdev_pad_config *cfg,
-			    unsigned int pad)
+vsp1_entity_get_pad_selection(struct vsp1_entity *entity,
+			      struct v4l2_subdev_pad_config *cfg,
+			      unsigned int pad, unsigned int target)
 {
-	return v4l2_subdev_get_try_compose(&entity->subdev, cfg, pad);
+	switch (target) {
+	case V4L2_SEL_TGT_COMPOSE:
+		return v4l2_subdev_get_try_compose(&entity->subdev, cfg, pad);
+	case V4L2_SEL_TGT_CROP:
+		return v4l2_subdev_get_try_crop(&entity->subdev, cfg, pad);
+	default:
+		return NULL;
+	}
 }
 
 /*
@@ -252,7 +276,7 @@ int vsp1_entity_link_setup(struct media_entity *entity,
 	if (!(local->flags & MEDIA_PAD_FL_SOURCE))
 		return 0;
 
-	source = container_of(local->entity, struct vsp1_entity, subdev.entity);
+	source = media_entity_to_vsp1_entity(local->entity);
 
 	if (!source->route)
 		return 0;
@@ -274,33 +298,50 @@ int vsp1_entity_link_setup(struct media_entity *entity,
  * Initialization
  */
 
+#define VSP1_ENTITY_ROUTE(ent)						\
+	{ VSP1_ENTITY_##ent, 0, VI6_DPR_##ent##_ROUTE,			\
+	  { VI6_DPR_NODE_##ent }, VI6_DPR_NODE_##ent }
+
+#define VSP1_ENTITY_ROUTE_RPF(idx)					\
+	{ VSP1_ENTITY_RPF, idx, VI6_DPR_RPF_ROUTE(idx),			\
+	  { 0, }, VI6_DPR_NODE_RPF(idx) }
+
+#define VSP1_ENTITY_ROUTE_UDS(idx)					\
+	{ VSP1_ENTITY_UDS, idx, VI6_DPR_UDS_ROUTE(idx),			\
+	  { VI6_DPR_NODE_UDS(idx) }, VI6_DPR_NODE_UDS(idx) }
+
+#define VSP1_ENTITY_ROUTE_WPF(idx)					\
+	{ VSP1_ENTITY_WPF, idx, 0,					\
+	  { VI6_DPR_NODE_WPF(idx) }, VI6_DPR_NODE_WPF(idx) }
+
 static const struct vsp1_route vsp1_routes[] = {
 	{ VSP1_ENTITY_BRU, 0, VI6_DPR_BRU_ROUTE,
 	  { VI6_DPR_NODE_BRU_IN(0), VI6_DPR_NODE_BRU_IN(1),
 	    VI6_DPR_NODE_BRU_IN(2), VI6_DPR_NODE_BRU_IN(3),
-	    VI6_DPR_NODE_BRU_IN(4) } },
-	{ VSP1_ENTITY_HSI, 0, VI6_DPR_HSI_ROUTE, { VI6_DPR_NODE_HSI, } },
-	{ VSP1_ENTITY_HST, 0, VI6_DPR_HST_ROUTE, { VI6_DPR_NODE_HST, } },
-	{ VSP1_ENTITY_LIF, 0, 0, { VI6_DPR_NODE_LIF, } },
-	{ VSP1_ENTITY_LUT, 0, VI6_DPR_LUT_ROUTE, { VI6_DPR_NODE_LUT, } },
-	{ VSP1_ENTITY_RPF, 0, VI6_DPR_RPF_ROUTE(0), { 0, } },
-	{ VSP1_ENTITY_RPF, 1, VI6_DPR_RPF_ROUTE(1), { 0, } },
-	{ VSP1_ENTITY_RPF, 2, VI6_DPR_RPF_ROUTE(2), { 0, } },
-	{ VSP1_ENTITY_RPF, 3, VI6_DPR_RPF_ROUTE(3), { 0, } },
-	{ VSP1_ENTITY_RPF, 4, VI6_DPR_RPF_ROUTE(4), { 0, } },
-	{ VSP1_ENTITY_SRU, 0, VI6_DPR_SRU_ROUTE, { VI6_DPR_NODE_SRU, } },
-	{ VSP1_ENTITY_UDS, 0, VI6_DPR_UDS_ROUTE(0), { VI6_DPR_NODE_UDS(0), } },
-	{ VSP1_ENTITY_UDS, 1, VI6_DPR_UDS_ROUTE(1), { VI6_DPR_NODE_UDS(1), } },
-	{ VSP1_ENTITY_UDS, 2, VI6_DPR_UDS_ROUTE(2), { VI6_DPR_NODE_UDS(2), } },
-	{ VSP1_ENTITY_WPF, 0, 0, { VI6_DPR_NODE_WPF(0), } },
-	{ VSP1_ENTITY_WPF, 1, 0, { VI6_DPR_NODE_WPF(1), } },
-	{ VSP1_ENTITY_WPF, 2, 0, { VI6_DPR_NODE_WPF(2), } },
-	{ VSP1_ENTITY_WPF, 3, 0, { VI6_DPR_NODE_WPF(3), } },
+	    VI6_DPR_NODE_BRU_IN(4) }, VI6_DPR_NODE_BRU_OUT },
+	VSP1_ENTITY_ROUTE(CLU),
+	VSP1_ENTITY_ROUTE(HSI),
+	VSP1_ENTITY_ROUTE(HST),
+	{ VSP1_ENTITY_LIF, 0, 0, { VI6_DPR_NODE_LIF, }, VI6_DPR_NODE_LIF },
+	VSP1_ENTITY_ROUTE(LUT),
+	VSP1_ENTITY_ROUTE_RPF(0),
+	VSP1_ENTITY_ROUTE_RPF(1),
+	VSP1_ENTITY_ROUTE_RPF(2),
+	VSP1_ENTITY_ROUTE_RPF(3),
+	VSP1_ENTITY_ROUTE_RPF(4),
+	VSP1_ENTITY_ROUTE(SRU),
+	VSP1_ENTITY_ROUTE_UDS(0),
+	VSP1_ENTITY_ROUTE_UDS(1),
+	VSP1_ENTITY_ROUTE_UDS(2),
+	VSP1_ENTITY_ROUTE_WPF(0),
+	VSP1_ENTITY_ROUTE_WPF(1),
+	VSP1_ENTITY_ROUTE_WPF(2),
+	VSP1_ENTITY_ROUTE_WPF(3),
 };
 
 int vsp1_entity_init(struct vsp1_device *vsp1, struct vsp1_entity *entity,
 		     const char *name, unsigned int num_pads,
-		     const struct v4l2_subdev_ops *ops)
+		     const struct v4l2_subdev_ops *ops, u32 function)
 {
 	struct v4l2_subdev *subdev;
 	unsigned int i;
@@ -341,6 +382,7 @@ int vsp1_entity_init(struct vsp1_device *vsp1, struct vsp1_entity *entity,
 	subdev = &entity->subdev;
 	v4l2_subdev_init(subdev, ops);
 
+	subdev->entity.function = function;
 	subdev->entity.ops = &vsp1->media_ops;
 	subdev->flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
 
diff --git a/drivers/media/platform/vsp1/vsp1_entity.h b/drivers/media/platform/vsp1/vsp1_entity.h
index 69eff4e17350..b43457fd2c43 100644
--- a/drivers/media/platform/vsp1/vsp1_entity.h
+++ b/drivers/media/platform/vsp1/vsp1_entity.h
@@ -24,6 +24,7 @@ struct vsp1_pipeline;
 
 enum vsp1_entity_type {
 	VSP1_ENTITY_BRU,
+	VSP1_ENTITY_CLU,
 	VSP1_ENTITY_HSI,
 	VSP1_ENTITY_HST,
 	VSP1_ENTITY_LIF,
@@ -42,17 +43,21 @@ enum vsp1_entity_type {
  * @index: Entity index this routing entry is associated with
  * @reg: Output routing configuration register
  * @inputs: Target node value for each input
+ * @output: Target node value for entity output
  *
  * Each $vsp1_route entry describes routing configuration for the entity
  * specified by the entry's @type and @index. @reg indicates the register that
  * holds output routing configuration for the entity, and the @inputs array
- * store the target node value for each input of the entity.
+ * store the target node value for each input of the entity. The @output field
+ * stores the target node value of the entity output when used as a source for
+ * histogram generation.
  */
 struct vsp1_route {
 	enum vsp1_entity_type type;
 	unsigned int index;
 	unsigned int reg;
 	unsigned int inputs[VSP1_ENTITY_MAX_INPUTS];
+	unsigned int output;
 };
 
 /**
@@ -68,7 +73,7 @@ struct vsp1_entity_operations {
 	void (*destroy)(struct vsp1_entity *);
 	void (*set_memory)(struct vsp1_entity *, struct vsp1_dl_list *dl);
 	void (*configure)(struct vsp1_entity *, struct vsp1_pipeline *,
-			  struct vsp1_dl_list *);
+			  struct vsp1_dl_list *, bool);
 };
 
 struct vsp1_entity {
@@ -100,7 +105,7 @@ static inline struct vsp1_entity *to_vsp1_entity(struct v4l2_subdev *subdev)
 
 int vsp1_entity_init(struct vsp1_device *vsp1, struct vsp1_entity *entity,
 		     const char *name, unsigned int num_pads,
-		     const struct v4l2_subdev_ops *ops);
+		     const struct v4l2_subdev_ops *ops, u32 function);
 void vsp1_entity_destroy(struct vsp1_entity *entity);
 
 extern const struct v4l2_subdev_internal_ops vsp1_subdev_internal_ops;
@@ -118,9 +123,9 @@ vsp1_entity_get_pad_format(struct vsp1_entity *entity,
 			   struct v4l2_subdev_pad_config *cfg,
 			   unsigned int pad);
 struct v4l2_rect *
-vsp1_entity_get_pad_compose(struct vsp1_entity *entity,
-			    struct v4l2_subdev_pad_config *cfg,
-			    unsigned int pad);
+vsp1_entity_get_pad_selection(struct vsp1_entity *entity,
+			      struct v4l2_subdev_pad_config *cfg,
+			      unsigned int pad, unsigned int target);
 int vsp1_entity_init_cfg(struct v4l2_subdev *subdev,
 			 struct v4l2_subdev_pad_config *cfg);
 
diff --git a/drivers/media/platform/vsp1/vsp1_hsit.c b/drivers/media/platform/vsp1/vsp1_hsit.c
index 68b8567b374d..6e5077beb38c 100644
--- a/drivers/media/platform/vsp1/vsp1_hsit.c
+++ b/drivers/media/platform/vsp1/vsp1_hsit.c
@@ -107,7 +107,7 @@ static int hsit_set_format(struct v4l2_subdev *subdev,
 	return 0;
 }
 
-static struct v4l2_subdev_pad_ops hsit_pad_ops = {
+static const struct v4l2_subdev_pad_ops hsit_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = hsit_enum_mbus_code,
 	.enum_frame_size = hsit_enum_frame_size,
@@ -115,7 +115,7 @@ static struct v4l2_subdev_pad_ops hsit_pad_ops = {
 	.set_fmt = hsit_set_format,
 };
 
-static struct v4l2_subdev_ops hsit_ops = {
+static const struct v4l2_subdev_ops hsit_ops = {
 	.pad    = &hsit_pad_ops,
 };
 
@@ -125,10 +125,13 @@ static struct v4l2_subdev_ops hsit_ops = {
 
 static void hsit_configure(struct vsp1_entity *entity,
 			   struct vsp1_pipeline *pipe,
-			   struct vsp1_dl_list *dl)
+			   struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_hsit *hsit = to_hsit(&entity->subdev);
 
+	if (!full)
+		return;
+
 	if (hsit->inverse)
 		vsp1_hsit_write(hsit, dl, VI6_HSI_CTRL, VI6_HSI_CTRL_EN);
 	else
@@ -161,8 +164,9 @@ struct vsp1_hsit *vsp1_hsit_create(struct vsp1_device *vsp1, bool inverse)
 	else
 		hsit->entity.type = VSP1_ENTITY_HST;
 
-	ret = vsp1_entity_init(vsp1, &hsit->entity, inverse ? "hsi" : "hst", 2,
-			       &hsit_ops);
+	ret = vsp1_entity_init(vsp1, &hsit->entity, inverse ? "hsi" : "hst",
+			       2, &hsit_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_ENC_CONV);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/drivers/media/platform/vsp1/vsp1_lif.c b/drivers/media/platform/vsp1/vsp1_lif.c
index 0217393f22df..a720063f38c5 100644
--- a/drivers/media/platform/vsp1/vsp1_lif.c
+++ b/drivers/media/platform/vsp1/vsp1_lif.c
@@ -104,7 +104,7 @@ static int lif_set_format(struct v4l2_subdev *subdev,
 	return 0;
 }
 
-static struct v4l2_subdev_pad_ops lif_pad_ops = {
+static const struct v4l2_subdev_pad_ops lif_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = lif_enum_mbus_code,
 	.enum_frame_size = lif_enum_frame_size,
@@ -112,7 +112,7 @@ static struct v4l2_subdev_pad_ops lif_pad_ops = {
 	.set_fmt = lif_set_format,
 };
 
-static struct v4l2_subdev_ops lif_ops = {
+static const struct v4l2_subdev_ops lif_ops = {
 	.pad    = &lif_pad_ops,
 };
 
@@ -122,7 +122,7 @@ static struct v4l2_subdev_ops lif_ops = {
 
 static void lif_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	const struct v4l2_mbus_framefmt *format;
 	struct vsp1_lif *lif = to_lif(&entity->subdev);
@@ -130,6 +130,9 @@ static void lif_configure(struct vsp1_entity *entity,
 	unsigned int obth = 400;
 	unsigned int lbth = 200;
 
+	if (!full)
+		return;
+
 	format = vsp1_entity_get_pad_format(&lif->entity, lif->entity.config,
 					    LIF_PAD_SOURCE);
 
@@ -165,7 +168,12 @@ struct vsp1_lif *vsp1_lif_create(struct vsp1_device *vsp1)
 	lif->entity.ops = &lif_entity_ops;
 	lif->entity.type = VSP1_ENTITY_LIF;
 
-	ret = vsp1_entity_init(vsp1, &lif->entity, "lif", 2, &lif_ops);
+	/* The LIF is never exposed to userspace, but media entity registration
+	 * requires a function to be set. Use PROC_VIDEO_PIXEL_FORMATTER just to
+	 * avoid triggering a WARN_ON(), the value won't be seen anywhere.
+	 */
+	ret = vsp1_entity_init(vsp1, &lif->entity, "lif", 2, &lif_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/drivers/media/platform/vsp1/vsp1_lut.c b/drivers/media/platform/vsp1/vsp1_lut.c
index aa09e59f0ab8..dc31de9602ba 100644
--- a/drivers/media/platform/vsp1/vsp1_lut.c
+++ b/drivers/media/platform/vsp1/vsp1_lut.c
@@ -13,7 +13,6 @@
 
 #include <linux/device.h>
 #include <linux/gfp.h>
-#include <linux/vsp1.h>
 
 #include <media/v4l2-subdev.h>
 
@@ -35,43 +34,62 @@ static inline void vsp1_lut_write(struct vsp1_lut *lut, struct vsp1_dl_list *dl,
 }
 
 /* -----------------------------------------------------------------------------
- * V4L2 Subdevice Core Operations
+ * Controls
  */
 
-static int lut_set_table(struct vsp1_lut *lut, struct vsp1_lut_config *config)
+#define V4L2_CID_VSP1_LUT_TABLE			(V4L2_CID_USER_BASE | 0x1001)
+
+static int lut_set_table(struct vsp1_lut *lut, struct v4l2_ctrl *ctrl)
 {
 	struct vsp1_dl_body *dlb;
 	unsigned int i;
 
-	dlb = vsp1_dl_fragment_alloc(lut->entity.vsp1, ARRAY_SIZE(config->lut));
+	dlb = vsp1_dl_fragment_alloc(lut->entity.vsp1, 256);
 	if (!dlb)
 		return -ENOMEM;
 
-	for (i = 0; i < ARRAY_SIZE(config->lut); ++i)
+	for (i = 0; i < 256; ++i)
 		vsp1_dl_fragment_write(dlb, VI6_LUT_TABLE + 4 * i,
-				       config->lut[i]);
+				       ctrl->p_new.p_u32[i]);
 
-	mutex_lock(&lut->lock);
+	spin_lock_irq(&lut->lock);
 	swap(lut->lut, dlb);
-	mutex_unlock(&lut->lock);
+	spin_unlock_irq(&lut->lock);
 
 	vsp1_dl_fragment_free(dlb);
 	return 0;
 }
 
-static long lut_ioctl(struct v4l2_subdev *subdev, unsigned int cmd, void *arg)
+static int lut_s_ctrl(struct v4l2_ctrl *ctrl)
 {
-	struct vsp1_lut *lut = to_lut(subdev);
-
-	switch (cmd) {
-	case VIDIOC_VSP1_LUT_CONFIG:
-		return lut_set_table(lut, arg);
+	struct vsp1_lut *lut =
+		container_of(ctrl->handler, struct vsp1_lut, ctrls);
 
-	default:
-		return -ENOIOCTLCMD;
+	switch (ctrl->id) {
+	case V4L2_CID_VSP1_LUT_TABLE:
+		lut_set_table(lut, ctrl);
+		break;
 	}
+
+	return 0;
 }
 
+static const struct v4l2_ctrl_ops lut_ctrl_ops = {
+	.s_ctrl = lut_s_ctrl,
+};
+
+static const struct v4l2_ctrl_config lut_table_control = {
+	.ops = &lut_ctrl_ops,
+	.id = V4L2_CID_VSP1_LUT_TABLE,
+	.name = "Look-Up Table",
+	.type = V4L2_CTRL_TYPE_U32,
+	.min = 0x00000000,
+	.max = 0x00ffffff,
+	.step = 1,
+	.def = 0,
+	.dims = { 256},
+};
+
 /* -----------------------------------------------------------------------------
  * V4L2 Subdevice Pad Operations
  */
@@ -147,11 +165,7 @@ static int lut_set_format(struct v4l2_subdev *subdev,
  * V4L2 Subdevice Operations
  */
 
-static struct v4l2_subdev_core_ops lut_core_ops = {
-	.ioctl = lut_ioctl,
-};
-
-static struct v4l2_subdev_pad_ops lut_pad_ops = {
+static const struct v4l2_subdev_pad_ops lut_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = lut_enum_mbus_code,
 	.enum_frame_size = lut_enum_frame_size,
@@ -159,8 +173,7 @@ static struct v4l2_subdev_pad_ops lut_pad_ops = {
 	.set_fmt = lut_set_format,
 };
 
-static struct v4l2_subdev_ops lut_ops = {
-	.core	= &lut_core_ops,
+static const struct v4l2_subdev_ops lut_ops = {
 	.pad    = &lut_pad_ops,
 };
 
@@ -170,18 +183,24 @@ static struct v4l2_subdev_ops lut_ops = {
 
 static void lut_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_lut *lut = to_lut(&entity->subdev);
+	struct vsp1_dl_body *dlb;
+	unsigned long flags;
 
-	vsp1_lut_write(lut, dl, VI6_LUT_CTRL, VI6_LUT_CTRL_EN);
-
-	mutex_lock(&lut->lock);
-	if (lut->lut) {
-		vsp1_dl_list_add_fragment(dl, lut->lut);
-		lut->lut = NULL;
+	if (full) {
+		vsp1_lut_write(lut, dl, VI6_LUT_CTRL, VI6_LUT_CTRL_EN);
+		return;
 	}
-	mutex_unlock(&lut->lock);
+
+	spin_lock_irqsave(&lut->lock, flags);
+	dlb = lut->lut;
+	lut->lut = NULL;
+	spin_unlock_irqrestore(&lut->lock, flags);
+
+	if (dlb)
+		vsp1_dl_list_add_fragment(dl, dlb);
 }
 
 static const struct vsp1_entity_operations lut_entity_ops = {
@@ -201,12 +220,30 @@ struct vsp1_lut *vsp1_lut_create(struct vsp1_device *vsp1)
 	if (lut == NULL)
 		return ERR_PTR(-ENOMEM);
 
+	spin_lock_init(&lut->lock);
+
 	lut->entity.ops = &lut_entity_ops;
 	lut->entity.type = VSP1_ENTITY_LUT;
 
-	ret = vsp1_entity_init(vsp1, &lut->entity, "lut", 2, &lut_ops);
+	ret = vsp1_entity_init(vsp1, &lut->entity, "lut", 2, &lut_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_LUT);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
+	/* Initialize the control handler. */
+	v4l2_ctrl_handler_init(&lut->ctrls, 1);
+	v4l2_ctrl_new_custom(&lut->ctrls, &lut_table_control, NULL);
+
+	lut->entity.subdev.ctrl_handler = &lut->ctrls;
+
+	if (lut->ctrls.error) {
+		dev_err(vsp1->dev, "lut: failed to initialize controls\n");
+		ret = lut->ctrls.error;
+		vsp1_entity_destroy(&lut->entity);
+		return ERR_PTR(ret);
+	}
+
+	v4l2_ctrl_handler_setup(&lut->ctrls);
+
 	return lut;
 }
diff --git a/drivers/media/platform/vsp1/vsp1_lut.h b/drivers/media/platform/vsp1/vsp1_lut.h
index cef874f22b6a..f8c4e8f0a79d 100644
--- a/drivers/media/platform/vsp1/vsp1_lut.h
+++ b/drivers/media/platform/vsp1/vsp1_lut.h
@@ -13,9 +13,10 @@
 #ifndef __VSP1_LUT_H__
 #define __VSP1_LUT_H__
 
-#include <linux/mutex.h>
+#include <linux/spinlock.h>
 
 #include <media/media-entity.h>
+#include <media/v4l2-ctrls.h>
 #include <media/v4l2-subdev.h>
 
 #include "vsp1_entity.h"
@@ -28,7 +29,9 @@ struct vsp1_device;
 struct vsp1_lut {
 	struct vsp1_entity entity;
 
-	struct mutex lock;
+	struct v4l2_ctrl_handler ctrls;
+
+	spinlock_t lock;
 	struct vsp1_dl_body *lut;
 };
 
diff --git a/drivers/media/platform/vsp1/vsp1_pipe.c b/drivers/media/platform/vsp1/vsp1_pipe.c
index 4f3b4a1d028a..3e75fb3fcace 100644
--- a/drivers/media/platform/vsp1/vsp1_pipe.c
+++ b/drivers/media/platform/vsp1/vsp1_pipe.c
@@ -172,13 +172,17 @@ void vsp1_pipeline_reset(struct vsp1_pipeline *pipe)
 			bru->inputs[i].rpf = NULL;
 	}
 
-	for (i = 0; i < pipe->num_inputs; ++i) {
-		pipe->inputs[i]->pipe = NULL;
-		pipe->inputs[i] = NULL;
+	for (i = 0; i < ARRAY_SIZE(pipe->inputs); ++i) {
+		if (pipe->inputs[i]) {
+			pipe->inputs[i]->pipe = NULL;
+			pipe->inputs[i] = NULL;
+		}
 	}
 
-	pipe->output->pipe = NULL;
-	pipe->output = NULL;
+	if (pipe->output) {
+		pipe->output->pipe = NULL;
+		pipe->output = NULL;
+	}
 
 	INIT_LIST_HEAD(&pipe->entities);
 	pipe->state = VSP1_PIPELINE_STOPPED;
@@ -286,6 +290,8 @@ void vsp1_pipeline_frame_end(struct vsp1_pipeline *pipe)
 
 	if (pipe->frame_end)
 		pipe->frame_end(pipe);
+
+	pipe->sequence++;
 }
 
 /*
@@ -295,42 +301,20 @@ void vsp1_pipeline_frame_end(struct vsp1_pipeline *pipe)
  * to be scaled, we disable alpha scaling when the UDS input has a fixed alpha
  * value. The UDS then outputs a fixed alpha value which needs to be programmed
  * from the input RPF alpha.
- *
- * This function can only be called from a subdev s_stream handler as it
- * requires a valid display list context.
  */
 void vsp1_pipeline_propagate_alpha(struct vsp1_pipeline *pipe,
-				   struct vsp1_entity *input,
-				   struct vsp1_dl_list *dl,
-				   unsigned int alpha)
+				   struct vsp1_dl_list *dl, unsigned int alpha)
 {
-	struct vsp1_entity *entity;
-	struct media_pad *pad;
-
-	pad = media_entity_remote_pad(&input->pads[RWPF_PAD_SOURCE]);
-
-	while (pad) {
-		if (!is_media_entity_v4l2_subdev(pad->entity))
-			break;
-
-		entity = to_vsp1_entity(media_entity_to_v4l2_subdev(pad->entity));
-
-		/* The BRU background color has a fixed alpha value set to 255,
-		 * the output alpha value is thus always equal to 255.
-		 */
-		if (entity->type == VSP1_ENTITY_BRU)
-			alpha = 255;
-
-		if (entity->type == VSP1_ENTITY_UDS) {
-			struct vsp1_uds *uds = to_uds(&entity->subdev);
+	if (!pipe->uds)
+		return;
 
-			vsp1_uds_set_alpha(uds, dl, alpha);
-			break;
-		}
+	/* The BRU background color has a fixed alpha value set to 255, the
+	 * output alpha value is thus always equal to 255.
+	 */
+	if (pipe->uds_input->type == VSP1_ENTITY_BRU)
+		alpha = 255;
 
-		pad = &entity->pads[entity->source_pad];
-		pad = media_entity_remote_pad(pad);
-	}
+	vsp1_uds_set_alpha(pipe->uds, dl, alpha);
 }
 
 void vsp1_pipelines_suspend(struct vsp1_device *vsp1)
@@ -383,7 +367,7 @@ void vsp1_pipelines_resume(struct vsp1_device *vsp1)
 {
 	unsigned int i;
 
-	/* Resume pipeline all running pipelines. */
+	/* Resume all running pipelines. */
 	for (i = 0; i < vsp1->info->wpf_count; ++i) {
 		struct vsp1_rwpf *wpf = vsp1->wpf[i];
 		struct vsp1_pipeline *pipe;
diff --git a/drivers/media/platform/vsp1/vsp1_pipe.h b/drivers/media/platform/vsp1/vsp1_pipe.h
index 7b56113511dd..d20d997b1fda 100644
--- a/drivers/media/platform/vsp1/vsp1_pipe.h
+++ b/drivers/media/platform/vsp1/vsp1_pipe.h
@@ -61,12 +61,13 @@ enum vsp1_pipeline_state {
  * @pipe: the media pipeline
  * @irqlock: protects the pipeline state
  * @state: current state
- * @wq: work queue to wait for state change completion
+ * @wq: wait queue to wait for state change completion
  * @frame_end: frame end interrupt handler
  * @lock: protects the pipeline use count and stream count
  * @kref: pipeline reference count
  * @stream_count: number of streaming video nodes
  * @buffers_ready: bitmask of RPFs and WPFs with at least one buffer available
+ * @sequence: frame sequence number
  * @num_inputs: number of RPFs
  * @inputs: array of RPFs in the pipeline (indexed by RPF index)
  * @output: WPF at the output of the pipeline
@@ -90,6 +91,7 @@ struct vsp1_pipeline {
 	struct kref kref;
 	unsigned int stream_count;
 	unsigned int buffers_ready;
+	unsigned int sequence;
 
 	unsigned int num_inputs;
 	struct vsp1_rwpf *inputs[VSP1_MAX_RPF];
@@ -115,9 +117,7 @@ bool vsp1_pipeline_ready(struct vsp1_pipeline *pipe);
 void vsp1_pipeline_frame_end(struct vsp1_pipeline *pipe);
 
 void vsp1_pipeline_propagate_alpha(struct vsp1_pipeline *pipe,
-				   struct vsp1_entity *input,
-				   struct vsp1_dl_list *dl,
-				   unsigned int alpha);
+				   struct vsp1_dl_list *dl, unsigned int alpha);
 
 void vsp1_pipelines_suspend(struct vsp1_device *vsp1);
 void vsp1_pipelines_resume(struct vsp1_device *vsp1);
diff --git a/drivers/media/platform/vsp1/vsp1_regs.h b/drivers/media/platform/vsp1/vsp1_regs.h
index 927b5fb94c48..3b03007ba625 100644
--- a/drivers/media/platform/vsp1/vsp1_regs.h
+++ b/drivers/media/platform/vsp1/vsp1_regs.h
@@ -154,10 +154,10 @@
 #define VI6_RPF_ALPH_SEL_AEXT_EXT	(1 << 18)
 #define VI6_RPF_ALPH_SEL_AEXT_ONE	(2 << 18)
 #define VI6_RPF_ALPH_SEL_AEXT_MASK	(3 << 18)
-#define VI6_RPF_ALPH_SEL_ALPHA0_MASK	(0xff << 8)
-#define VI6_RPF_ALPH_SEL_ALPHA0_SHIFT	8
-#define VI6_RPF_ALPH_SEL_ALPHA1_MASK	(0xff << 0)
-#define VI6_RPF_ALPH_SEL_ALPHA1_SHIFT	0
+#define VI6_RPF_ALPH_SEL_ALPHA1_MASK	(0xff << 8)
+#define VI6_RPF_ALPH_SEL_ALPHA1_SHIFT	8
+#define VI6_RPF_ALPH_SEL_ALPHA0_MASK	(0xff << 0)
+#define VI6_RPF_ALPH_SEL_ALPHA0_SHIFT	0
 
 #define VI6_RPF_VRTCOL_SET		0x0318
 #define VI6_RPF_VRTCOL_SET_LAYA_MASK	(0xff << 24)
@@ -255,6 +255,8 @@
 #define VI6_WPF_OUTFMT_PDV_MASK		(0xff << 24)
 #define VI6_WPF_OUTFMT_PDV_SHIFT	24
 #define VI6_WPF_OUTFMT_PXA		(1 << 23)
+#define VI6_WPF_OUTFMT_ROT		(1 << 18)
+#define VI6_WPF_OUTFMT_HFLP		(1 << 17)
 #define VI6_WPF_OUTFMT_FLP		(1 << 16)
 #define VI6_WPF_OUTFMT_SPYCS		(1 << 15)
 #define VI6_WPF_OUTFMT_SPUVS		(1 << 14)
@@ -289,6 +291,11 @@
 #define VI6_WPF_RNDCTRL_CLMD_EXT	(2 << 12)
 #define VI6_WPF_RNDCTRL_CLMD_MASK	(3 << 12)
 
+#define VI6_WPF_ROT_CTRL		0x1018
+#define VI6_WPF_ROT_CTRL_LN16		(1 << 17)
+#define VI6_WPF_ROT_CTRL_LMEM_WD_MASK	(0x1fff << 0)
+#define VI6_WPF_ROT_CTRL_LMEM_WD_SHIFT	0
+
 #define VI6_WPF_DSTM_STRIDE_Y		0x101c
 #define VI6_WPF_DSTM_STRIDE_C		0x1020
 #define VI6_WPF_DSTM_ADDR_Y		0x1024
@@ -444,6 +451,15 @@
  */
 
 #define VI6_CLU_CTRL			0x2900
+#define VI6_CLU_CTRL_AAI		(1 << 28)
+#define VI6_CLU_CTRL_MVS		(1 << 24)
+#define VI6_CLU_CTRL_AX1I_2D		(3 << 14)
+#define VI6_CLU_CTRL_AX2I_2D		(1 << 12)
+#define VI6_CLU_CTRL_OS0_2D		(3 << 8)
+#define VI6_CLU_CTRL_OS1_2D		(1 << 6)
+#define VI6_CLU_CTRL_OS2_2D		(3 << 4)
+#define VI6_CLU_CTRL_M2D		(1 << 1)
+#define VI6_CLU_CTRL_EN			(1 << 0)
 
 /* -----------------------------------------------------------------------------
  * HST Control Registers
diff --git a/drivers/media/platform/vsp1/vsp1_rpf.c b/drivers/media/platform/vsp1/vsp1_rpf.c
index 49168db3f529..388838913205 100644
--- a/drivers/media/platform/vsp1/vsp1_rpf.c
+++ b/drivers/media/platform/vsp1/vsp1_rpf.c
@@ -38,7 +38,7 @@ static inline void vsp1_rpf_write(struct vsp1_rwpf *rpf,
  * V4L2 Subdevice Operations
  */
 
-static struct v4l2_subdev_ops rpf_ops = {
+static const struct v4l2_subdev_ops rpf_ops = {
 	.pad    = &vsp1_rwpf_pad_ops,
 };
 
@@ -60,7 +60,7 @@ static void rpf_set_memory(struct vsp1_entity *entity, struct vsp1_dl_list *dl)
 
 static void rpf_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_rwpf *rpf = to_rwpf(&entity->subdev);
 	const struct vsp1_format_info *fmtinfo = rpf->fmtinfo;
@@ -73,6 +73,16 @@ static void rpf_configure(struct vsp1_entity *entity,
 	u32 pstride;
 	u32 infmt;
 
+	if (!full) {
+		vsp1_rpf_write(rpf, dl, VI6_RPF_VRTCOL_SET,
+			       rpf->alpha << VI6_RPF_VRTCOL_SET_LAYA_SHIFT);
+		vsp1_rpf_write(rpf, dl, VI6_RPF_MULT_ALPHA, rpf->mult_alpha |
+			       (rpf->alpha << VI6_RPF_MULT_ALPHA_RATIO_SHIFT));
+
+		vsp1_pipeline_propagate_alpha(pipe, dl, rpf->alpha);
+		return;
+	}
+
 	/* Source size, stride and crop offsets.
 	 *
 	 * The crop offsets correspond to the location of the crop rectangle top
@@ -130,9 +140,10 @@ static void rpf_configure(struct vsp1_entity *entity,
 	if (pipe->bru) {
 		const struct v4l2_rect *compose;
 
-		compose = vsp1_entity_get_pad_compose(pipe->bru,
-						      pipe->bru->config,
-						      rpf->bru_input);
+		compose = vsp1_entity_get_pad_selection(pipe->bru,
+							pipe->bru->config,
+							rpf->bru_input,
+							V4L2_SEL_TGT_COMPOSE);
 		left = compose->left;
 		top = compose->top;
 	}
@@ -167,9 +178,6 @@ static void rpf_configure(struct vsp1_entity *entity,
 		       (fmtinfo->alpha ? VI6_RPF_ALPH_SEL_ASEL_PACKED
 				       : VI6_RPF_ALPH_SEL_ASEL_FIXED));
 
-	vsp1_rpf_write(rpf, dl, VI6_RPF_VRTCOL_SET,
-		       rpf->alpha << VI6_RPF_VRTCOL_SET_LAYA_SHIFT);
-
 	if (entity->vsp1->info->gen == 3) {
 		u32 mult;
 
@@ -187,8 +195,7 @@ static void rpf_configure(struct vsp1_entity *entity,
 			mult = VI6_RPF_MULT_ALPHA_A_MMD_RATIO
 			     | (premultiplied ?
 				VI6_RPF_MULT_ALPHA_P_MMD_RATIO :
-				VI6_RPF_MULT_ALPHA_P_MMD_NONE)
-			     | (rpf->alpha << VI6_RPF_MULT_ALPHA_RATIO_SHIFT);
+				VI6_RPF_MULT_ALPHA_P_MMD_NONE);
 		} else {
 			/* When the input doesn't contain an alpha channel the
 			 * global alpha value is applied in the unpacking unit,
@@ -199,11 +206,9 @@ static void rpf_configure(struct vsp1_entity *entity,
 			     | VI6_RPF_MULT_ALPHA_P_MMD_NONE;
 		}
 
-		vsp1_rpf_write(rpf, dl, VI6_RPF_MULT_ALPHA, mult);
+		rpf->mult_alpha = mult;
 	}
 
-	vsp1_pipeline_propagate_alpha(pipe, &rpf->entity, dl, rpf->alpha);
-
 	vsp1_rpf_write(rpf, dl, VI6_RPF_MSK_CTRL, 0);
 	vsp1_rpf_write(rpf, dl, VI6_RPF_CKEY_CTRL, 0);
 
@@ -236,18 +241,21 @@ struct vsp1_rwpf *vsp1_rpf_create(struct vsp1_device *vsp1, unsigned int index)
 	rpf->entity.index = index;
 
 	sprintf(name, "rpf.%u", index);
-	ret = vsp1_entity_init(vsp1, &rpf->entity, name, 2, &rpf_ops);
+	ret = vsp1_entity_init(vsp1, &rpf->entity, name, 2, &rpf_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
 	/* Initialize the control handler. */
-	ret = vsp1_rwpf_init_ctrls(rpf);
+	ret = vsp1_rwpf_init_ctrls(rpf, 0);
 	if (ret < 0) {
 		dev_err(vsp1->dev, "rpf%u: failed to initialize controls\n",
 			index);
 		goto error;
 	}
 
+	v4l2_ctrl_handler_setup(&rpf->ctrls);
+
 	return rpf;
 
 error:
diff --git a/drivers/media/platform/vsp1/vsp1_rwpf.c b/drivers/media/platform/vsp1/vsp1_rwpf.c
index 3b6e032e7806..8d461b375e91 100644
--- a/drivers/media/platform/vsp1/vsp1_rwpf.c
+++ b/drivers/media/platform/vsp1/vsp1_rwpf.c
@@ -241,11 +241,9 @@ static const struct v4l2_ctrl_ops vsp1_rwpf_ctrl_ops = {
 	.s_ctrl = vsp1_rwpf_s_ctrl,
 };
 
-int vsp1_rwpf_init_ctrls(struct vsp1_rwpf *rwpf)
+int vsp1_rwpf_init_ctrls(struct vsp1_rwpf *rwpf, unsigned int ncontrols)
 {
-	rwpf->alpha = 255;
-
-	v4l2_ctrl_handler_init(&rwpf->ctrls, 1);
+	v4l2_ctrl_handler_init(&rwpf->ctrls, ncontrols + 1);
 	v4l2_ctrl_new_std(&rwpf->ctrls, &vsp1_rwpf_ctrl_ops,
 			  V4L2_CID_ALPHA_COMPONENT, 0, 255, 1, 255);
 
diff --git a/drivers/media/platform/vsp1/vsp1_rwpf.h b/drivers/media/platform/vsp1/vsp1_rwpf.h
index 9ff7c78f239e..cb20484e80da 100644
--- a/drivers/media/platform/vsp1/vsp1_rwpf.h
+++ b/drivers/media/platform/vsp1/vsp1_rwpf.h
@@ -13,6 +13,8 @@
 #ifndef __VSP1_RWPF_H__
 #define __VSP1_RWPF_H__
 
+#include <linux/spinlock.h>
+
 #include <media/media-entity.h>
 #include <media/v4l2-ctrls.h>
 #include <media/v4l2-subdev.h>
@@ -49,6 +51,16 @@ struct vsp1_rwpf {
 
 	unsigned int alpha;
 
+	u32 mult_alpha;
+	u32 outfmt;
+
+	struct {
+		spinlock_t lock;
+		struct v4l2_ctrl *ctrls[2];
+		unsigned int pending;
+		unsigned int active;
+	} flip;
+
 	unsigned int offsets[2];
 	struct vsp1_rwpf_memory mem;
 
@@ -68,7 +80,7 @@ static inline struct vsp1_rwpf *entity_to_rwpf(struct vsp1_entity *entity)
 struct vsp1_rwpf *vsp1_rpf_create(struct vsp1_device *vsp1, unsigned int index);
 struct vsp1_rwpf *vsp1_wpf_create(struct vsp1_device *vsp1, unsigned int index);
 
-int vsp1_rwpf_init_ctrls(struct vsp1_rwpf *rwpf);
+int vsp1_rwpf_init_ctrls(struct vsp1_rwpf *rwpf, unsigned int ncontrols);
 
 extern const struct v4l2_subdev_pad_ops vsp1_rwpf_pad_ops;
 
diff --git a/drivers/media/platform/vsp1/vsp1_sru.c b/drivers/media/platform/vsp1/vsp1_sru.c
index 97ef997ae735..47f5e0cea2ce 100644
--- a/drivers/media/platform/vsp1/vsp1_sru.c
+++ b/drivers/media/platform/vsp1/vsp1_sru.c
@@ -37,7 +37,7 @@ static inline void vsp1_sru_write(struct vsp1_sru *sru, struct vsp1_dl_list *dl,
  * Controls
  */
 
-#define V4L2_CID_VSP1_SRU_INTENSITY		(V4L2_CID_USER_BASE + 1)
+#define V4L2_CID_VSP1_SRU_INTENSITY		(V4L2_CID_USER_BASE | 0x1001)
 
 struct vsp1_sru_param {
 	u32 ctrl0;
@@ -239,7 +239,7 @@ static int sru_set_format(struct v4l2_subdev *subdev,
 	return 0;
 }
 
-static struct v4l2_subdev_pad_ops sru_pad_ops = {
+static const struct v4l2_subdev_pad_ops sru_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = sru_enum_mbus_code,
 	.enum_frame_size = sru_enum_frame_size,
@@ -247,7 +247,7 @@ static struct v4l2_subdev_pad_ops sru_pad_ops = {
 	.set_fmt = sru_set_format,
 };
 
-static struct v4l2_subdev_ops sru_ops = {
+static const struct v4l2_subdev_ops sru_ops = {
 	.pad    = &sru_pad_ops,
 };
 
@@ -257,7 +257,7 @@ static struct v4l2_subdev_ops sru_ops = {
 
 static void sru_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	const struct vsp1_sru_param *param;
 	struct vsp1_sru *sru = to_sru(&entity->subdev);
@@ -265,6 +265,9 @@ static void sru_configure(struct vsp1_entity *entity,
 	struct v4l2_mbus_framefmt *output;
 	u32 ctrl0;
 
+	if (!full)
+		return;
+
 	input = vsp1_entity_get_pad_format(&sru->entity, sru->entity.config,
 					   SRU_PAD_SINK);
 	output = vsp1_entity_get_pad_format(&sru->entity, sru->entity.config,
@@ -308,7 +311,8 @@ struct vsp1_sru *vsp1_sru_create(struct vsp1_device *vsp1)
 	sru->entity.ops = &sru_entity_ops;
 	sru->entity.type = VSP1_ENTITY_SRU;
 
-	ret = vsp1_entity_init(vsp1, &sru->entity, "sru", 2, &sru_ops);
+	ret = vsp1_entity_init(vsp1, &sru->entity, "sru", 2, &sru_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_SCALER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/drivers/media/platform/vsp1/vsp1_uds.c b/drivers/media/platform/vsp1/vsp1_uds.c
index 1875e29da184..652dcd895022 100644
--- a/drivers/media/platform/vsp1/vsp1_uds.c
+++ b/drivers/media/platform/vsp1/vsp1_uds.c
@@ -40,9 +40,11 @@ static inline void vsp1_uds_write(struct vsp1_uds *uds, struct vsp1_dl_list *dl,
  * Scaling Computation
  */
 
-void vsp1_uds_set_alpha(struct vsp1_uds *uds, struct vsp1_dl_list *dl,
+void vsp1_uds_set_alpha(struct vsp1_entity *entity, struct vsp1_dl_list *dl,
 			unsigned int alpha)
 {
+	struct vsp1_uds *uds = to_uds(&entity->subdev);
+
 	vsp1_uds_write(uds, dl, VI6_UDS_ALPVAL,
 		       alpha << VI6_UDS_ALPVAL_VAL0_SHIFT);
 }
@@ -226,7 +228,7 @@ static int uds_set_format(struct v4l2_subdev *subdev,
  * V4L2 Subdevice Operations
  */
 
-static struct v4l2_subdev_pad_ops uds_pad_ops = {
+static const struct v4l2_subdev_pad_ops uds_pad_ops = {
 	.init_cfg = vsp1_entity_init_cfg,
 	.enum_mbus_code = uds_enum_mbus_code,
 	.enum_frame_size = uds_enum_frame_size,
@@ -234,7 +236,7 @@ static struct v4l2_subdev_pad_ops uds_pad_ops = {
 	.set_fmt = uds_set_format,
 };
 
-static struct v4l2_subdev_ops uds_ops = {
+static const struct v4l2_subdev_ops uds_ops = {
 	.pad    = &uds_pad_ops,
 };
 
@@ -244,7 +246,7 @@ static struct v4l2_subdev_ops uds_ops = {
 
 static void uds_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_uds *uds = to_uds(&entity->subdev);
 	const struct v4l2_mbus_framefmt *output;
@@ -253,6 +255,9 @@ static void uds_configure(struct vsp1_entity *entity,
 	unsigned int vscale;
 	bool multitap;
 
+	if (!full)
+		return;
+
 	input = vsp1_entity_get_pad_format(&uds->entity, uds->entity.config,
 					   UDS_PAD_SINK);
 	output = vsp1_entity_get_pad_format(&uds->entity, uds->entity.config,
@@ -314,7 +319,8 @@ struct vsp1_uds *vsp1_uds_create(struct vsp1_device *vsp1, unsigned int index)
 	uds->entity.index = index;
 
 	sprintf(name, "uds.%u", index);
-	ret = vsp1_entity_init(vsp1, &uds->entity, name, 2, &uds_ops);
+	ret = vsp1_entity_init(vsp1, &uds->entity, name, 2, &uds_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_SCALER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/drivers/media/platform/vsp1/vsp1_uds.h b/drivers/media/platform/vsp1/vsp1_uds.h
index 5c8cbfcad4cc..7bf3cdcffc65 100644
--- a/drivers/media/platform/vsp1/vsp1_uds.h
+++ b/drivers/media/platform/vsp1/vsp1_uds.h
@@ -35,7 +35,7 @@ static inline struct vsp1_uds *to_uds(struct v4l2_subdev *subdev)
 
 struct vsp1_uds *vsp1_uds_create(struct vsp1_device *vsp1, unsigned int index);
 
-void vsp1_uds_set_alpha(struct vsp1_uds *uds, struct vsp1_dl_list *dl,
+void vsp1_uds_set_alpha(struct vsp1_entity *uds, struct vsp1_dl_list *dl,
 			unsigned int alpha);
 
 #endif /* __VSP1_UDS_H__ */
diff --git a/drivers/media/platform/vsp1/vsp1_video.c b/drivers/media/platform/vsp1/vsp1_video.c
index 424a54122190..9fb4fc26a359 100644
--- a/drivers/media/platform/vsp1/vsp1_video.c
+++ b/drivers/media/platform/vsp1/vsp1_video.c
@@ -219,7 +219,7 @@ vsp1_video_complete_buffer(struct vsp1_video *video)
 
 	spin_unlock_irqrestore(&video->irqlock, flags);
 
-	done->buf.sequence = video->sequence++;
+	done->buf.sequence = pipe->sequence;
 	done->buf.vb2_buf.timestamp = ktime_get_ns();
 	for (i = 0; i < done->buf.vb2_buf.num_planes; ++i)
 		vb2_set_plane_payload(&done->buf.vb2_buf, i,
@@ -251,11 +251,17 @@ static void vsp1_video_frame_end(struct vsp1_pipeline *pipe,
 static void vsp1_video_pipeline_run(struct vsp1_pipeline *pipe)
 {
 	struct vsp1_device *vsp1 = pipe->output->entity.vsp1;
+	struct vsp1_entity *entity;
 	unsigned int i;
 
 	if (!pipe->dl)
 		pipe->dl = vsp1_dl_list_get(pipe->output->dlm);
 
+	list_for_each_entry(entity, &pipe->entities, list_pipe) {
+		if (entity->ops->configure)
+			entity->ops->configure(entity, pipe, pipe->dl, false);
+	}
+
 	for (i = 0; i < vsp1->info->rpf_count; ++i) {
 		struct vsp1_rwpf *rwpf = pipe->inputs[i];
 
@@ -628,7 +634,7 @@ static int vsp1_video_setup_pipeline(struct vsp1_pipeline *pipe)
 		vsp1_entity_route_setup(entity, pipe->dl);
 
 		if (entity->ops->configure)
-			entity->ops->configure(entity, pipe, pipe->dl);
+			entity->ops->configure(entity, pipe, pipe->dl, true);
 	}
 
 	return 0;
@@ -670,7 +676,7 @@ static void vsp1_video_stop_streaming(struct vb2_queue *vq)
 	int ret;
 
 	mutex_lock(&pipe->lock);
-	if (--pipe->stream_count == 0) {
+	if (--pipe->stream_count == pipe->num_inputs) {
 		/* Stop the pipeline. */
 		ret = vsp1_pipeline_stop(pipe);
 		if (ret == -ETIMEDOUT)
@@ -692,7 +698,7 @@ static void vsp1_video_stop_streaming(struct vb2_queue *vq)
 	spin_unlock_irqrestore(&video->irqlock, flags);
 }
 
-static struct vb2_ops vsp1_video_queue_qops = {
+static const struct vb2_ops vsp1_video_queue_qops = {
 	.queue_setup = vsp1_video_queue_setup,
 	.buf_prepare = vsp1_video_buffer_prepare,
 	.buf_queue = vsp1_video_buffer_queue,
@@ -801,8 +807,6 @@ vsp1_video_streamon(struct file *file, void *fh, enum v4l2_buf_type type)
 	if (video->queue.owner && video->queue.owner != file->private_data)
 		return -EBUSY;
 
-	video->sequence = 0;
-
 	/* Get a pipeline for the video node and start streaming on it. No link
 	 * touching an entity in the pipeline can be activated or deactivated
 	 * once streaming is started.
@@ -911,7 +915,7 @@ static int vsp1_video_release(struct file *file)
 	return 0;
 }
 
-static struct v4l2_file_operations vsp1_video_fops = {
+static const struct v4l2_file_operations vsp1_video_fops = {
 	.owner = THIS_MODULE,
 	.unlocked_ioctl = video_ioctl2,
 	.open = vsp1_video_open,
diff --git a/drivers/media/platform/vsp1/vsp1_video.h b/drivers/media/platform/vsp1/vsp1_video.h
index 4487dc89f25d..50ea7f02205f 100644
--- a/drivers/media/platform/vsp1/vsp1_video.h
+++ b/drivers/media/platform/vsp1/vsp1_video.h
@@ -48,7 +48,6 @@ struct vsp1_video {
 	struct vb2_queue queue;
 	spinlock_t irqlock;
 	struct list_head irqqueue;
-	unsigned int sequence;
 };
 
 static inline struct vsp1_video *to_vsp1_video(struct video_device *vdev)
diff --git a/drivers/media/platform/vsp1/vsp1_wpf.c b/drivers/media/platform/vsp1/vsp1_wpf.c
index 6c91eaa35e75..31983169c24a 100644
--- a/drivers/media/platform/vsp1/vsp1_wpf.c
+++ b/drivers/media/platform/vsp1/vsp1_wpf.c
@@ -37,6 +37,97 @@ static inline void vsp1_wpf_write(struct vsp1_rwpf *wpf,
 }
 
 /* -----------------------------------------------------------------------------
+ * Controls
+ */
+
+enum wpf_flip_ctrl {
+	WPF_CTRL_VFLIP = 0,
+	WPF_CTRL_HFLIP = 1,
+	WPF_CTRL_MAX,
+};
+
+static int vsp1_wpf_s_ctrl(struct v4l2_ctrl *ctrl)
+{
+	struct vsp1_rwpf *wpf =
+		container_of(ctrl->handler, struct vsp1_rwpf, ctrls);
+	unsigned int i;
+	u32 flip = 0;
+
+	switch (ctrl->id) {
+	case V4L2_CID_HFLIP:
+	case V4L2_CID_VFLIP:
+		for (i = 0; i < WPF_CTRL_MAX; ++i) {
+			if (wpf->flip.ctrls[i])
+				flip |= wpf->flip.ctrls[i]->val ? BIT(i) : 0;
+		}
+
+		spin_lock_irq(&wpf->flip.lock);
+		wpf->flip.pending = flip;
+		spin_unlock_irq(&wpf->flip.lock);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static const struct v4l2_ctrl_ops vsp1_wpf_ctrl_ops = {
+	.s_ctrl = vsp1_wpf_s_ctrl,
+};
+
+static int wpf_init_controls(struct vsp1_rwpf *wpf)
+{
+	struct vsp1_device *vsp1 = wpf->entity.vsp1;
+	unsigned int num_flip_ctrls;
+
+	spin_lock_init(&wpf->flip.lock);
+
+	if (wpf->entity.index != 0) {
+		/* Only WPF0 supports flipping. */
+		num_flip_ctrls = 0;
+	} else if (vsp1->info->features & VSP1_HAS_WPF_HFLIP) {
+		/* When horizontal flip is supported the WPF implements two
+		 * controls (horizontal flip and vertical flip).
+		 */
+		num_flip_ctrls = 2;
+	} else if (vsp1->info->features & VSP1_HAS_WPF_VFLIP) {
+		/* When only vertical flip is supported the WPF implements a
+		 * single control (vertical flip).
+		 */
+		num_flip_ctrls = 1;
+	} else {
+		/* Otherwise flipping is not supported. */
+		num_flip_ctrls = 0;
+	}
+
+	vsp1_rwpf_init_ctrls(wpf, num_flip_ctrls);
+
+	if (num_flip_ctrls >= 1) {
+		wpf->flip.ctrls[WPF_CTRL_VFLIP] =
+			v4l2_ctrl_new_std(&wpf->ctrls, &vsp1_wpf_ctrl_ops,
+					  V4L2_CID_VFLIP, 0, 1, 1, 0);
+	}
+
+	if (num_flip_ctrls == 2) {
+		wpf->flip.ctrls[WPF_CTRL_HFLIP] =
+			v4l2_ctrl_new_std(&wpf->ctrls, &vsp1_wpf_ctrl_ops,
+					  V4L2_CID_HFLIP, 0, 1, 1, 0);
+
+		v4l2_ctrl_cluster(2, wpf->flip.ctrls);
+	}
+
+	if (wpf->ctrls.error) {
+		dev_err(vsp1->dev, "wpf%u: failed to initialize controls\n",
+			wpf->entity.index);
+		return wpf->ctrls.error;
+	}
+
+	return 0;
+}
+
+/* -----------------------------------------------------------------------------
  * V4L2 Subdevice Core Operations
  */
 
@@ -62,11 +153,11 @@ static int wpf_s_stream(struct v4l2_subdev *subdev, int enable)
  * V4L2 Subdevice Operations
  */
 
-static struct v4l2_subdev_video_ops wpf_video_ops = {
+static const struct v4l2_subdev_video_ops wpf_video_ops = {
 	.s_stream = wpf_s_stream,
 };
 
-static struct v4l2_subdev_ops wpf_ops = {
+static const struct v4l2_subdev_ops wpf_ops = {
 	.video	= &wpf_video_ops,
 	.pad    = &vsp1_rwpf_pad_ops,
 };
@@ -85,15 +176,37 @@ static void vsp1_wpf_destroy(struct vsp1_entity *entity)
 static void wpf_set_memory(struct vsp1_entity *entity, struct vsp1_dl_list *dl)
 {
 	struct vsp1_rwpf *wpf = entity_to_rwpf(entity);
+	const struct v4l2_pix_format_mplane *format = &wpf->format;
+	struct vsp1_rwpf_memory mem = wpf->mem;
+	unsigned int flip = wpf->flip.active;
+	unsigned int offset;
+
+	/* Update the memory offsets based on flipping configuration. The
+	 * destination addresses point to the locations where the VSP starts
+	 * writing to memory, which can be different corners of the image
+	 * depending on vertical flipping. Horizontal flipping is handled
+	 * through a line buffer and doesn't modify the start address.
+	 */
+	if (flip & BIT(WPF_CTRL_VFLIP)) {
+		mem.addr[0] += (format->height - 1)
+			     * format->plane_fmt[0].bytesperline;
+
+		if (format->num_planes > 1) {
+			offset = (format->height / wpf->fmtinfo->vsub - 1)
+			       * format->plane_fmt[1].bytesperline;
+			mem.addr[1] += offset;
+			mem.addr[2] += offset;
+		}
+	}
 
-	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_Y, wpf->mem.addr[0]);
-	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_C0, wpf->mem.addr[1]);
-	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_C1, wpf->mem.addr[2]);
+	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_Y, mem.addr[0]);
+	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_C0, mem.addr[1]);
+	vsp1_wpf_write(wpf, dl, VI6_WPF_DSTM_ADDR_C1, mem.addr[2]);
 }
 
 static void wpf_configure(struct vsp1_entity *entity,
 			  struct vsp1_pipeline *pipe,
-			  struct vsp1_dl_list *dl)
+			  struct vsp1_dl_list *dl, bool full)
 {
 	struct vsp1_rwpf *wpf = to_rwpf(&entity->subdev);
 	struct vsp1_device *vsp1 = wpf->entity.vsp1;
@@ -104,6 +217,26 @@ static void wpf_configure(struct vsp1_entity *entity,
 	u32 outfmt = 0;
 	u32 srcrpf = 0;
 
+	if (!full) {
+		const unsigned int mask = BIT(WPF_CTRL_VFLIP)
+					| BIT(WPF_CTRL_HFLIP);
+
+		spin_lock(&wpf->flip.lock);
+		wpf->flip.active = (wpf->flip.active & ~mask)
+				 | (wpf->flip.pending & mask);
+		spin_unlock(&wpf->flip.lock);
+
+		outfmt = (wpf->alpha << VI6_WPF_OUTFMT_PDV_SHIFT) | wpf->outfmt;
+
+		if (wpf->flip.active & BIT(WPF_CTRL_VFLIP))
+			outfmt |= VI6_WPF_OUTFMT_FLP;
+		if (wpf->flip.active & BIT(WPF_CTRL_HFLIP))
+			outfmt |= VI6_WPF_OUTFMT_HFLP;
+
+		vsp1_wpf_write(wpf, dl, VI6_WPF_OUTFMT, outfmt);
+		return;
+	}
+
 	/* Cropping */
 	crop = vsp1_rwpf_get_crop(wpf, wpf->entity.config);
 
@@ -143,13 +276,18 @@ static void wpf_configure(struct vsp1_entity *entity,
 				       format->plane_fmt[1].bytesperline);
 
 		vsp1_wpf_write(wpf, dl, VI6_WPF_DSWAP, fmtinfo->swap);
+
+		if (vsp1->info->features & VSP1_HAS_WPF_HFLIP &&
+		    wpf->entity.index == 0)
+			vsp1_wpf_write(wpf, dl, VI6_WPF_ROT_CTRL,
+				       VI6_WPF_ROT_CTRL_LN16 |
+				       (256 << VI6_WPF_ROT_CTRL_LMEM_WD_SHIFT));
 	}
 
 	if (sink_format->code != source_format->code)
 		outfmt |= VI6_WPF_OUTFMT_CSC;
 
-	outfmt |= wpf->alpha << VI6_WPF_OUTFMT_PDV_SHIFT;
-	vsp1_wpf_write(wpf, dl, VI6_WPF_OUTFMT, outfmt);
+	wpf->outfmt = outfmt;
 
 	vsp1_dl_list_write(dl, VI6_DPR_WPF_FPORCH(wpf->entity.index),
 			   VI6_DPR_WPF_FPORCH_FP_WPFN);
@@ -216,7 +354,8 @@ struct vsp1_rwpf *vsp1_wpf_create(struct vsp1_device *vsp1, unsigned int index)
 	wpf->entity.index = index;
 
 	sprintf(name, "wpf.%u", index);
-	ret = vsp1_entity_init(vsp1, &wpf->entity, name, 2, &wpf_ops);
+	ret = vsp1_entity_init(vsp1, &wpf->entity, name, 2, &wpf_ops,
+			       MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
@@ -228,13 +367,15 @@ struct vsp1_rwpf *vsp1_wpf_create(struct vsp1_device *vsp1, unsigned int index)
 	}
 
 	/* Initialize the control handler. */
-	ret = vsp1_rwpf_init_ctrls(wpf);
+	ret = wpf_init_controls(wpf);
 	if (ret < 0) {
 		dev_err(vsp1->dev, "wpf%u: failed to initialize controls\n",
 			index);
 		goto error;
 	}
 
+	v4l2_ctrl_handler_setup(&wpf->ctrls);
+
 	return wpf;
 
 error: