Re: [PATCH v3 7/9] media: platform: Add c3 ISP driver

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Keke

   a first pass of review without going into details about the
ISP parameters and stats but mostly on architecture.

On Wed, Sep 18, 2024 at 02:07:18PM +0800, Keke Li via B4 Relay wrote:
> From: Keke Li <keke.li@xxxxxxxxxxx>
>
> The C3 ISP supports multi-camera and muti-exposure
> high dynamic range (HDR). It brings together some
> advanced imaging technologies to provide good image quality.
> This driver mainly responsible for driving ISP pipeline
> to process raw image.
>
> Signed-off-by: Keke Li <keke.li@xxxxxxxxxxx>
> ---
>  drivers/media/platform/amlogic/Kconfig             |   1 +
>  drivers/media/platform/amlogic/Makefile            |   1 +
>  drivers/media/platform/amlogic/c3-isp/Kconfig      |  17 +
>  drivers/media/platform/amlogic/c3-isp/Makefile     |  10 +
>  .../media/platform/amlogic/c3-isp/c3-isp-capture.c | 759 ++++++++++++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-common.h  | 327 ++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-core.c    | 675 ++++++++++++++++
>  drivers/media/platform/amlogic/c3-isp/c3-isp-dev.c | 486 ++++++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-params.c  | 857 +++++++++++++++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-regs.h    | 683 ++++++++++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-resizer.c | 768 ++++++++++++++++++
>  .../media/platform/amlogic/c3-isp/c3-isp-stats.c   | 488 ++++++++++++
>  .../amlogic/c3-isp/include/uapi/c3-isp-config.h    | 537 +++++++++++++

This really should be in include/uapi/linux/media/amlogic
so that it is made available to userspace.

>  13 files changed, 5609 insertions(+)
>
> diff --git a/drivers/media/platform/amlogic/Kconfig b/drivers/media/platform/amlogic/Kconfig
> index df09717b28d0..ebda6b7edc2d 100644
> --- a/drivers/media/platform/amlogic/Kconfig
> +++ b/drivers/media/platform/amlogic/Kconfig
> @@ -2,6 +2,7 @@
>
>  comment "Amlogic media platform drivers"
>
> +source "drivers/media/platform/amlogic/c3-isp/Kconfig"
>  source "drivers/media/platform/amlogic/c3-mipi-adapter/Kconfig"
>  source "drivers/media/platform/amlogic/c3-mipi-csi2/Kconfig"

I start wondering if this shouldn't rather be organized in an
drivers/media/platform/amlogic/c3/ folder with an
drivers/media/platform/amlogic/c3/isp/ subfolder

Do you know if any of the csi-2-rx, adapter or ISP can
be used in other SoCs ?

>  source "drivers/media/platform/amlogic/meson-ge2d/Kconfig"
> diff --git a/drivers/media/platform/amlogic/Makefile b/drivers/media/platform/amlogic/Makefile
> index b370154b090c..d0d9363d4d8d 100644
> --- a/drivers/media/platform/amlogic/Makefile
> +++ b/drivers/media/platform/amlogic/Makefile
> @@ -1,5 +1,6 @@
>  # SPDX-License-Identifier: GPL-2.0-only
>
> +obj-y += c3-isp/
>  obj-y += c3-mipi-adapter/
>  obj-y += c3-mipi-csi2/
>  obj-y += meson-ge2d/
> diff --git a/drivers/media/platform/amlogic/c3-isp/Kconfig b/drivers/media/platform/amlogic/c3-isp/Kconfig
> new file mode 100644
> index 000000000000..e317c1e81750
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/Kconfig
> @@ -0,0 +1,17 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +
> +config VIDEO_C3_ISP
> +	tristate "Amlogic C3 Image Signal Processor (ISP) driver"
> +	depends on ARCH_MESON || COMPILE_TEST
> +	depends on VIDEO_DEV
> +	depends on OF
> +	select MEDIA_CONTROLLER
> +	select V4L2_FWNODE
> +	select VIDEO_V4L2_SUBDEV_API
> +	select VIDEOBUF2_DMA_CONTIG
> +	help
> +	  Video4Linux2 driver for Amlogic C3 ISP pipeline.
> +	  C3 ISP pipeline mainly for processing raw image

          The C3 ISP is used for processing raw images
> +	  and output result to memory.

Or something similar, but "C3 ISP pipeline mainly for processing"
doesn't sound right to be

> +
> +	  To compile this driver as a module choose m here.
> diff --git a/drivers/media/platform/amlogic/c3-isp/Makefile b/drivers/media/platform/amlogic/c3-isp/Makefile
> new file mode 100644
> index 000000000000..b1b064170b57
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/Makefile
> @@ -0,0 +1,10 @@
> +# SPDX-License-Identifier: GPL-2.0-only
> +
> +c3-isp-objs := c3-isp-dev.o \
> +	       c3-isp-params.o \
> +	       c3-isp-stats.o \
> +	       c3-isp-capture.o \
> +	       c3-isp-core.o \
> +	       c3-isp-resizer.o
> +
> +obj-$(CONFIG_VIDEO_C3_ISP) += c3-isp.o
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-capture.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-capture.c
> new file mode 100644
> index 000000000000..ee9a7a17a203
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-capture.c
> @@ -0,0 +1,759 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/cleanup.h>
> +#include <linux/pm_runtime.h>
> +
> +#include <media/v4l2-ctrls.h>
> +#include <media/v4l2-event.h>
> +#include <media/v4l2-ioctl.h>
> +#include <media/v4l2-mc.h>
> +#include <media/videobuf2-dma-contig.h>
> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +
> +static const struct c3_isp_capture_format cap_formats[] = {
> +	{
> +		.mbus_code = MEDIA_BUS_FMT_Y8_1X8,
> +		.fourcc = V4L2_PIX_FMT_GREY,
> +		.depth = 8,
> +	},
> +	{
> +		.mbus_code = MEDIA_BUS_FMT_YUYV8_1_5X8,

Does the 5X8 version represents the format on the internal bus between
the resizers and the capture device ?

How does format propagation work from the ISP to the resizers and the
capture devices ? I mean, is there an internal bus where the number of
samples (5X8 vs 2X8) changes depending on the output format ?

> +		.fourcc = V4L2_PIX_FMT_NV12,
> +		.depth = 12,
> +	}, {
> +		.mbus_code = MEDIA_BUS_FMT_YUYV8_1_5X8,
> +		.fourcc = V4L2_PIX_FMT_NV21,
> +		.depth = 12,
> +	}, {
> +		.mbus_code = MEDIA_BUS_FMT_YUYV8_2X8,
> +		.fourcc = V4L2_PIX_FMT_NV16,
> +		.depth = 16,
> +	}, {
> +		.mbus_code = MEDIA_BUS_FMT_YUYV8_2X8,
> +		.fourcc = V4L2_PIX_FMT_NV61,
> +		.depth = 16,
> +	},
> +};
> +
> +/* Hardware configuration */
> +
> +/* Set the address of wrmifx3(write memory interface) */
> +static void c3_isp_cap_wrmifx3_buff(struct c3_isp_capture *cap)
> +{
> +	struct v4l2_pix_format *pix = &cap->vfmt.fmt.pix;
> +	struct c3_isp_vb2_buffer *buff = cap->buff;
> +	u32 offset;
> +
> +	c3_isp_write(cap->isp,
> +		     C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_BADDR, cap->id),
> +		     WRMIFX3_CH0_BADDR(buff->paddr));
> +
> +	if (pix->pixelformat == V4L2_PIX_FMT_NV12 ||
> +	    pix->pixelformat == V4L2_PIX_FMT_NV21 ||
> +	    pix->pixelformat == V4L2_PIX_FMT_NV16 ||
> +	    pix->pixelformat == V4L2_PIX_FMT_NV61) {
> +		offset = pix->width * pix->height;
> +		c3_isp_write(cap->isp,
> +			     C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH1_BADDR, cap->id),
> +			     WRMIFX3_CH1_BADDR(buff->paddr + offset));
> +	}
> +}
> +
> +static void c3_isp_cap_output_size(struct c3_isp_capture *cap)
> +{
> +	struct v4l2_pix_format *pix = &cap->vfmt.fmt.pix;
> +
> +	c3_isp_update_bits(cap->isp, C3_DISP_REG(DISP0_TOP_OUT_SIZE, cap->id),
> +			   DISP_OUT_VSIZE_MASK, pix->height);
> +	c3_isp_update_bits(cap->isp, C3_DISP_REG(DISP0_TOP_OUT_SIZE, cap->id),
> +			   DISP_OUT_HSIZE_MASK, pix->width << DISP_OUT_HSIZE_SHIFT);
> +}
> +
> +static void c3_isp_cap_wrmifx3_grey(struct c3_isp_capture *cap)
> +{
> +	struct v4l2_pix_format *fmt = &cap->vfmt.fmt.pix;
> +	u32 stride;
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_IBITS_MASK, WRMIFX3_FMT_MTX_IBITS_8BIT);
> +
> +	/* Grey has 1 plane*/
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_PLANE_MASK,
> +			   WRMIFX3_FMT_MTX_PLANE_X1 << WRMIFX3_FMT_MTX_PLANE_SHIFT);
> +
> +	/* Set Y only as output format */
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MODE_OUT_MASK,
> +			   WRMIFX3_FMT_MODE_OUT_Y_ONLY << WRMIFX3_FMT_MODE_OUT_SHIFT);
> +
> +	/* The unit of stride is 128 bits */
> +	stride = DIV_ROUND_UP(fmt->width * 8, 128);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL0, cap->id),
> +			   WRMIFX3_CH0_STRIDE_MASK, stride << WRMIFX3_CH0_STRIDE_SHIFT);
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL1, cap->id),
> +			   WRMIFX3_CH0_PIX_BITS_MODE_MASK,
> +			   WRMIFX3_CH0_PIX_BITS_8BITS << WRMIFX3_CH0_PIX_BITS_MODE_SHIFT);
> +}
> +
> +static void c3_isp_cap_wrmifx3_yuv420(struct c3_isp_capture *cap, u32 swap_uv)
> +{
> +	struct v4l2_pix_format *fmt = &cap->vfmt.fmt.pix;
> +	u32 stride;
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_IBITS_MASK, WRMIFX3_FMT_MTX_IBITS_8BIT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_UV_SWAP_MASK,
> +			   swap_uv << WRMIFX3_FMT_MTX_UV_SWAP_SHIFT);
> +
> +	/* NV12 or NV21 has 2 planes*/
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_PLANE_MASK,
> +			   WRMIFX3_FMT_MTX_PLANE_X2 << WRMIFX3_FMT_MTX_PLANE_SHIFT);
> +
> +	/* Set YUV420 as output format */
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MODE_OUT_MASK,
> +			   WRMIFX3_FMT_MODE_OUT_YUV420 << WRMIFX3_FMT_MODE_OUT_SHIFT);
> +
> +	/* The unit of stride is 128 bits */
> +	stride = DIV_ROUND_UP(fmt->width * 8, 128);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL0, cap->id),
> +			   WRMIFX3_CH0_STRIDE_MASK, stride << WRMIFX3_CH0_STRIDE_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH1_CTRL0, cap->id),
> +			   WRMIFX3_CH1_STRIDE_MASK, stride << WRMIFX3_CH1_STRIDE_SHIFT);
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL1, cap->id),
> +			   WRMIFX3_CH0_PIX_BITS_MODE_MASK,
> +			   WRMIFX3_CH0_PIX_BITS_8BITS << WRMIFX3_CH0_PIX_BITS_MODE_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH1_CTRL1, cap->id),
> +			   WRMIFX3_CH1_PIX_BITS_MODE_MASK,
> +			   WRMIFX3_CH1_PIX_BITS_16BITS << WRMIFX3_CH1_PIX_BITS_MODE_SHIFT);
> +}
> +
> +static void c3_isp_cap_wrmifx3_yuv422(struct c3_isp_capture *cap, u32 swap_uv)
> +{
> +	struct v4l2_pix_format *fmt = &cap->vfmt.fmt.pix;
> +	u32 stride;
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_IBITS_MASK, WRMIFX3_FMT_MTX_IBITS_16BIT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_UV_SWAP_MASK,
> +			   swap_uv << WRMIFX3_FMT_MTX_UV_SWAP_SHIFT);
> +
> +	/* NV16 or NV61 has 2 planes*/
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MTX_PLANE_MASK,
> +			   WRMIFX3_FMT_MTX_PLANE_X2 << WRMIFX3_FMT_MTX_PLANE_SHIFT);
> +
> +	/* Set YUV422 as output format */
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_CTRL, cap->id),
> +			   WRMIFX3_FMT_MODE_OUT_MASK,
> +			   WRMIFX3_FMT_MODE_OUT_YUV422 << WRMIFX3_FMT_MODE_OUT_SHIFT);
> +
> +	/* The unit of stride is 128 bits */
> +	stride = DIV_ROUND_UP(fmt->width * 16, 128);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL0, cap->id),
> +			   WRMIFX3_CH0_STRIDE_MASK, stride << WRMIFX3_CH0_STRIDE_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH1_CTRL0, cap->id),
> +			   WRMIFX3_CH1_STRIDE_MASK, stride << WRMIFX3_CH1_STRIDE_SHIFT);
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH0_CTRL1, cap->id),
> +			   WRMIFX3_CH0_PIX_BITS_MODE_MASK,
> +			   WRMIFX3_CH0_PIX_BITS_16BITS << WRMIFX3_CH0_PIX_BITS_MODE_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CH1_CTRL1, cap->id),
> +			   WRMIFX3_CH1_PIX_BITS_MODE_MASK,
> +			   WRMIFX3_CH1_PIX_BITS_32BITS << WRMIFX3_CH1_PIX_BITS_MODE_SHIFT);
> +}
> +
> +static void c3_isp_cap_wrmifx3_size(struct c3_isp_capture *cap)
> +{
> +	struct v4l2_pix_format *fmt = &cap->vfmt.fmt.pix;
> +
> +	if (fmt->pixelformat == V4L2_PIX_FMT_GREY) {
> +		c3_isp_cap_wrmifx3_grey(cap);
> +	} else if (fmt->pixelformat == V4L2_PIX_FMT_NV12) {
> +		c3_isp_cap_wrmifx3_yuv420(cap, WRMIFX3_FMT_MTX_UV_SWAP_UV);
> +	} else if (fmt->pixelformat == V4L2_PIX_FMT_NV21) {
> +		c3_isp_cap_wrmifx3_yuv420(cap, WRMIFX3_FMT_MTX_UV_SWAP_VU);
> +	} else if (fmt->pixelformat == V4L2_PIX_FMT_NV16) {
> +		c3_isp_cap_wrmifx3_yuv422(cap, WRMIFX3_FMT_MTX_UV_SWAP_UV);
> +	} else if (fmt->pixelformat == V4L2_PIX_FMT_NV61) {
> +		c3_isp_cap_wrmifx3_yuv422(cap, WRMIFX3_FMT_MTX_UV_SWAP_VU);
> +	} else {
> +		dev_err(cap->isp->dev, "Invalid pixel format: 0x%x\n", fmt->pixelformat);
> +		return;
> +	}
> +
> +	c3_isp_write(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_FMT_SIZE, cap->id),
> +		     WRMIFX3_FMT_HSIZE(fmt->width) | WRMIFX3_FMT_VSIZE(fmt->height));
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_WIN_LUMA_H, cap->id),
> +			   WRMIFX3_WIN_LUMA_HEND_MASK,
> +			   WRMIFX3_WIN_LUMA_HEND(fmt->width) << WRMIFX3_WIN_LUMA_HEND_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_WIN_LUMA_V, cap->id),
> +			   WRMIFX3_WIN_LUMA_VEND_MASK,
> +			   WRMIFX3_WIN_LUMA_VEND(fmt->height) << WRMIFX3_WIN_LUMA_VEND_SHIFT);
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CRP_HSIZE, cap->id),
> +			   WRMIFX3_CROP_HEND_MASK,
> +			   WRMIFX3_CROP_HEND(fmt->width) << WRMIFX3_CROP_HEND_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_CRP_VSIZE, cap->id),
> +			   WRMIFX3_CROP_VEND_MASK,
> +			   WRMIFX3_CROP_VEND(fmt->height) << WRMIFX3_CROP_VEND_SHIFT);
> +
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_WIN_CHROM_H, cap->id),
> +			   WRMIFX3_WIN_CHROM_HEND_MASK,
> +			   WRMIFX3_WIN_CHROM_HEND(fmt->width) << WRMIFX3_WIN_CHROM_HEND_SHIFT);
> +	c3_isp_update_bits(cap->isp, C3_WRMIFX3_REG(ISP_WRMIFX3_0_WIN_CHROM_V, cap->id),
> +			   WRMIFX3_WIN_CHROM_VEND_MASK,
> +			   WRMIFX3_WIN_CHROM_VEND(fmt->height) << WRMIFX3_WIN_CHROM_VEND_SHIFT);
> +}
> +
> +static void c3_isp_cap_cfg_buff(struct c3_isp_capture *cap)

This function is called from two locations, one with the buff_lock
held, the other one without, if I'm not mistaken.

> +{
> +	cap->buff = list_first_entry_or_null(&cap->pending,
> +					     struct c3_isp_vb2_buffer, list);
> +	if (cap->buff) {
> +		c3_isp_cap_wrmifx3_buff(cap);
> +		list_del(&cap->buff->list);
> +	}
> +}
> +
> +static void c3_isp_cap_start(struct c3_isp_capture *cap)
> +{
> +	c3_isp_cap_cfg_buff(cap);
> +
> +	c3_isp_cap_output_size(cap);
> +	c3_isp_cap_wrmifx3_size(cap);
> +
> +	c3_isp_update_bits(cap->isp, ISP_TOP_PATH_EN,
> +			   TOP_WRMIF_EN(cap->id), TOP_WRMIF_EN(cap->id));
> +
> +	cap->is_streaming = true;
> +}
> +
> +static void c3_isp_cap_stop(struct c3_isp_capture *cap)
> +{
> +	cap->is_streaming = false;
> +
> +	c3_isp_update_bits(cap->isp, ISP_TOP_PATH_EN, TOP_WRMIF_EN(cap->id), 0);
> +}
> +
> +static int c3_isp_cap_done(struct c3_isp_capture *cap)
> +{
> +	struct c3_isp_vb2_buffer *buff = cap->buff;
> +	unsigned long flags;
> +
> +	if (!cap->is_streaming)
> +		return -EINVAL;

Is this an error condition or 0 should be returnes ?

> +
> +	spin_lock_irqsave(&cap->buff_lock, flags);
> +
> +	if (buff) {
> +		buff->vb.sequence = cap->isp->frm_sequence;
> +		buff->vb.vb2_buf.timestamp = ktime_get();
> +		buff->vb.field = V4L2_FIELD_NONE;
> +		vb2_buffer_done(&buff->vb.vb2_buf, VB2_BUF_STATE_DONE);
> +	}
> +
> +	c3_isp_cap_cfg_buff(cap);
> +
> +	spin_unlock_irqrestore(&cap->buff_lock, flags);
> +
> +	return 0;
> +}
> +
> +/* V4L2 video operations */
> +
> +static const struct c3_isp_capture_format
> +*c3_cap_find_fmt(u32 fourcc)
> +{
> +	unsigned int i;
> +
> +	for (i = 0; i < ARRAY_SIZE(cap_formats); i++) {
> +		if (cap_formats[i].fourcc == fourcc)
> +			return &cap_formats[i];
> +	}
> +
> +	return NULL;
> +}
> +
> +static void c3_cap_try_fmt(struct c3_isp_capture *cap,
> +			   struct v4l2_pix_format *pix)
> +{
> +	const struct c3_isp_capture_format *fmt;
> +
> +	fmt = c3_cap_find_fmt(pix->pixelformat);
> +	if (!fmt)
> +		fmt = &cap_formats[0];
> +
> +	pix->width = clamp(pix->width, C3_ISP_MIN_WIDTH, C3_ISP_MAX_WIDTH);
> +	pix->height = clamp(pix->height, C3_ISP_MIN_HEIGHT, C3_ISP_MAX_HEIGHT);
> +	pix->pixelformat = fmt->fourcc;
> +	pix->field = V4L2_FIELD_NONE;
> +	pix->colorspace = V4L2_COLORSPACE_SRGB;
> +	pix->ycbcr_enc = V4L2_YCBCR_ENC_601;
> +	pix->quantization = V4L2_QUANTIZATION_LIM_RANGE;
> +
> +	/* ISP hardware requires 16 bytes alignment */
> +	pix->bytesperline = ALIGN(pix->width, 16);
> +	pix->sizeimage = pix->bytesperline * pix->height * fmt->depth / 8;
> +}
> +
> +static void c3_isp_cap_return_buffers(struct c3_isp_capture *cap,
> +				      enum vb2_buffer_state state)
> +{
> +	unsigned long flags;
> +	struct c3_isp_vb2_buffer *buff;
> +
> +	spin_lock_irqsave(&cap->buff_lock, flags);
> +
> +	if (cap->buff) {
> +		vb2_buffer_done(&cap->buff->vb.vb2_buf, state);
> +		cap->buff = NULL;
> +	}
> +
> +	while (!list_empty(&cap->pending)) {
> +		buff = list_first_entry(&cap->pending,
> +					struct c3_isp_vb2_buffer, list);
> +		list_del(&buff->list);
> +		vb2_buffer_done(&buff->vb.vb2_buf, state);
> +	}
> +
> +	spin_unlock_irqrestore(&cap->buff_lock, flags);
> +}
> +
> +static int c3_isp_cap_querycap(struct file *file, void *fh,
> +			       struct v4l2_capability *cap)
> +{
> +	strscpy(cap->driver, C3_ISP_DRIVER_NAME, sizeof(cap->driver));
> +	strscpy(cap->card, "AML C3 ISP", sizeof(cap->card));
> +
> +	return 0;
> +}
> +
> +static int c3_isp_cap_enum_fmt(struct file *file, void *fh,
> +			       struct v4l2_fmtdesc *f)
> +{
> +	const struct c3_isp_capture_format *fmt;
> +	unsigned int index = 0;
> +	unsigned int i;
> +
> +	if (!f->mbus_code) {
> +		if (f->index >= ARRAY_SIZE(cap_formats))
> +			return -EINVAL;
> +
> +		fmt = &cap_formats[f->index];
> +		f->pixelformat = fmt->fourcc;
> +		return 0;
> +	}
> +
> +	for (i = 0; i < ARRAY_SIZE(cap_formats); i++) {
> +		fmt = &cap_formats[i];
> +		if (f->mbus_code != fmt->mbus_code)
> +			continue;
> +
> +		if (index++ == f->index) {
> +			f->pixelformat = cap_formats[i].fourcc;
> +			return 0;
> +		}
> +	}
> +
> +	return -EINVAL;
> +}
> +
> +static int c3_isp_cap_g_fmt(struct file *file, void *fh,
> +			    struct v4l2_format *f)
> +{
> +	struct c3_isp_capture *cap = video_drvdata(file);
> +
> +	f->fmt.pix = cap->vfmt.fmt.pix;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_cap_s_fmt(struct file *file, void *fh,
> +			    struct v4l2_format *f)
> +{
> +	struct c3_isp_capture *cap = video_drvdata(file);
> +
> +	c3_cap_try_fmt(cap, &f->fmt.pix);
> +	cap->vfmt = *f;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_cap_try_fmt(struct file *file, void *fh,
> +			      struct v4l2_format *f)
> +{
> +	struct c3_isp_capture *cap = video_drvdata(file);
> +
> +	c3_cap_try_fmt(cap, &f->fmt.pix);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_cap_enum_frmsize(struct file *file, void *fh,
> +				   struct v4l2_frmsizeenum *fsize)
> +{
> +	const struct c3_isp_capture_format *fmt;
> +
> +	if (fsize->index)
> +		return -EINVAL;
> +
> +	fmt = c3_cap_find_fmt(fsize->pixel_format);
> +	if (!fmt)
> +		return -EINVAL;
> +
> +	fsize->type = V4L2_FRMSIZE_TYPE_STEPWISE;
> +	fsize->stepwise.min_width = C3_ISP_MIN_WIDTH;
> +	fsize->stepwise.min_height = C3_ISP_MIN_HEIGHT;
> +	fsize->stepwise.max_width = C3_ISP_MAX_WIDTH;
> +	fsize->stepwise.max_height = C3_ISP_MAX_HEIGHT;
> +	fsize->stepwise.step_width = 2;
> +	fsize->stepwise.step_height = 2;
> +
> +	return 0;
> +}
> +
> +static const struct v4l2_ioctl_ops isp_cap_v4l2_ioctl_ops = {
> +	.vidioc_querycap		= c3_isp_cap_querycap,
> +	.vidioc_enum_fmt_vid_cap	= c3_isp_cap_enum_fmt,
> +	.vidioc_g_fmt_vid_cap		= c3_isp_cap_g_fmt,
> +	.vidioc_s_fmt_vid_cap		= c3_isp_cap_s_fmt,
> +	.vidioc_try_fmt_vid_cap		= c3_isp_cap_try_fmt,
> +	.vidioc_reqbufs			= vb2_ioctl_reqbufs,
> +	.vidioc_querybuf		= vb2_ioctl_querybuf,
> +	.vidioc_qbuf			= vb2_ioctl_qbuf,
> +	.vidioc_expbuf			= vb2_ioctl_expbuf,
> +	.vidioc_dqbuf			= vb2_ioctl_dqbuf,
> +	.vidioc_prepare_buf		= vb2_ioctl_prepare_buf,
> +	.vidioc_create_bufs		= vb2_ioctl_create_bufs,
> +	.vidioc_streamon		= vb2_ioctl_streamon,
> +	.vidioc_streamoff		= vb2_ioctl_streamoff,
> +	.vidioc_enum_framesizes         = c3_isp_cap_enum_frmsize,
> +	.vidioc_subscribe_event		= v4l2_ctrl_subscribe_event,
> +	.vidioc_unsubscribe_event	= v4l2_event_unsubscribe,
> +};
> +
> +static const struct v4l2_file_operations isp_cap_v4l2_fops = {
> +	.open = v4l2_fh_open,
> +	.release = vb2_fop_release,
> +	.poll = vb2_fop_poll,
> +	.unlocked_ioctl = video_ioctl2,
> +	.mmap = vb2_fop_mmap,
> +};
> +
> +static int c3_isp_cap_link_validate(struct media_link *link)
> +{
> +	struct video_device *vdev =
> +		media_entity_to_video_device(link->sink->entity);
> +	struct v4l2_subdev *sd =
> +		media_entity_to_v4l2_subdev(link->source->entity);
> +	struct c3_isp_capture *cap = video_get_drvdata(vdev);
> +	struct v4l2_pix_format *pix_fmt = &cap->vfmt.fmt.pix;
> +	struct v4l2_subdev_format src_fmt = {
> +		.which = V4L2_SUBDEV_FORMAT_ACTIVE,
> +		.pad = link->source->index,
> +	};
> +	const struct c3_isp_capture_format *cap_fmt =
> +				c3_cap_find_fmt(pix_fmt->pixelformat);
> +	int ret;
> +
> +	ret = v4l2_subdev_call_state_active(sd, pad, get_fmt, &src_fmt);
> +	if (ret)
> +		return ret;
> +
> +	if (src_fmt.format.width != pix_fmt->width ||
> +	    src_fmt.format.height != pix_fmt->height ||
> +	    src_fmt.format.code != cap_fmt->mbus_code) {
> +		dev_err(cap->isp->dev,
> +			"link %s: %u -> %s: %u not valid: 0x%04x/%ux%u not match 0x%04x/%ux%u\n",
> +			link->source->entity->name, link->source->index,
> +			link->sink->entity->name, link->sink->index,
> +			src_fmt.format.code, src_fmt.format.width,
> +			src_fmt.format.height, cap_fmt->mbus_code,
> +			pix_fmt->width, pix_fmt->height);
> +
> +		return -EPIPE;
> +	};
> +
> +	return 0;
> +}
> +
> +static const struct media_entity_operations isp_cap_entity_ops = {
> +	.link_validate = c3_isp_cap_link_validate,
> +};
> +
> +static int c3_isp_vb2_queue_setup(struct vb2_queue *q,
> +				  unsigned int *num_buffers,
> +				  unsigned int *num_planes,
> +				  unsigned int sizes[],
> +				  struct device *alloc_devs[])
> +{
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(q);
> +	struct v4l2_pix_format *pix = &cap->vfmt.fmt.pix;
> +
> +	if (*num_planes) {
> +		if (*num_planes != 1)
> +			return -EINVAL;
> +
> +		if (sizes[0] < pix->sizeimage)
> +			return -EINVAL;
> +	} else {
> +		*num_planes = 1;
> +		sizes[0] = pix->sizeimage;
> +	}
> +
> +	return 0;
> +}
> +
> +static void c3_isp_vb2_buf_queue(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +			container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&cap->buff_lock, flags);
> +
> +	list_add_tail(&buf->list, &cap->pending);
> +
> +	spin_unlock_irqrestore(&cap->buff_lock, flags);
> +}
> +
> +static int c3_isp_vb2_buf_prepare(struct vb2_buffer *vb)
> +{
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned int size = cap->vfmt.fmt.pix.sizeimage;
> +
> +	if (vb2_plane_size(vb, 0) < size) {
> +		dev_err(cap->isp->dev,
> +			"User buffer too small (%ld < %u)\n",
> +			vb2_plane_size(vb, 0), size);
> +		return -EINVAL;
> +	}
> +
> +	vb2_set_plane_payload(vb, 0, size);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_vb2_buf_init(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +		container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(vb->vb2_queue);
> +
> +	buf->vaddr = vb2_plane_vaddr(vb, 0);
> +	buf->paddr = vb2_dma_contig_plane_dma_addr(vb, 0);
> +
> +	memset(buf->vaddr, 0, cap->vfmt.fmt.pix.sizeimage);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_vb2_start_streaming(struct vb2_queue *q,
> +				      unsigned int count)
> +{
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(q);
> +	int ret;
> +
> +	guard(mutex)(&cap->isp->lock);
> +
> +	ret = pm_runtime_resume_and_get(cap->isp->dev);
> +	if (ret)
> +		return ret;
> +
> +	ret = video_device_pipeline_start(&cap->vdev, &cap->isp->pipe);
> +	if (ret) {
> +		dev_err(cap->isp->dev,
> +			"Failed to start cap%u pipeline: %d\n", cap->id, ret);
> +		goto err_pm_put;
> +	}
> +
> +	if (c3_isp_pipeline_ready(cap->isp)) {
> +		ret = v4l2_subdev_enable_streams(&cap->isp->core.sd,
> +						 C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +						 BIT(0));
> +		if (ret)
> +			goto err_pipeline_stop;
> +	}
> +
> +	c3_isp_rsz_start(cap->rsz);
> +	c3_isp_cap_start(cap);
> +
> +	return 0;
> +
> +err_pipeline_stop:
> +	video_device_pipeline_stop(&cap->vdev);
> +err_pm_put:
> +	pm_runtime_put(cap->isp->dev);
> +	c3_isp_cap_return_buffers(cap, VB2_BUF_STATE_QUEUED);
> +	return ret;
> +}
> +
> +static void c3_isp_vb2_stop_streaming(struct vb2_queue *q)
> +{
> +	struct c3_isp_capture *cap = vb2_get_drv_priv(q);
> +
> +	guard(mutex)(&cap->isp->lock);
> +
> +	c3_isp_cap_stop(cap);
> +	c3_isp_rsz_stop(cap->rsz);
> +	c3_isp_cap_return_buffers(cap, VB2_BUF_STATE_ERROR);
> +
> +	if (cap->isp->pipe.start_count == 1)
> +		v4l2_subdev_disable_streams(&cap->isp->core.sd,
> +					    C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +					    BIT(0));
> +
> +	video_device_pipeline_stop(&cap->vdev);
> +	pm_runtime_put(cap->isp->dev);
> +}
> +
> +static const struct vb2_ops isp_video_vb2_ops = {
> +	.queue_setup = c3_isp_vb2_queue_setup,
> +	.buf_queue = c3_isp_vb2_buf_queue,
> +	.buf_prepare = c3_isp_vb2_buf_prepare,
> +	.buf_init = c3_isp_vb2_buf_init,
> +	.wait_prepare = vb2_ops_wait_prepare,
> +	.wait_finish = vb2_ops_wait_finish,
> +	.start_streaming = c3_isp_vb2_start_streaming,
> +	.stop_streaming = c3_isp_vb2_stop_streaming,
> +};
> +
> +static int c3_isp_register_capture(struct c3_isp_capture *cap)
> +{
> +	struct video_device *vdev = &cap->vdev;
> +	struct vb2_queue *vb2_q = &cap->vb2_q;
> +	int ret;
> +
> +	snprintf(vdev->name, sizeof(vdev->name), "isp-cap%u", cap->id);
> +	vdev->fops = &isp_cap_v4l2_fops;
> +	vdev->ioctl_ops = &isp_cap_v4l2_ioctl_ops;
> +	vdev->v4l2_dev = &cap->isp->v4l2_dev;
> +	vdev->entity.ops = &isp_cap_entity_ops;
> +	vdev->lock = &cap->lock;
> +	vdev->minor = -1;
> +	vdev->queue = vb2_q;
> +	vdev->release = video_device_release_empty;
> +	vdev->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING;
> +	vdev->vfl_dir = VFL_DIR_RX;
> +	video_set_drvdata(vdev, cap);
> +
> +	vb2_q->drv_priv = cap;
> +	vb2_q->mem_ops = &vb2_dma_contig_memops;
> +	vb2_q->ops = &isp_video_vb2_ops;
> +	vb2_q->type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
> +	vb2_q->io_modes = VB2_DMABUF | VB2_MMAP;
> +	vb2_q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
> +	vb2_q->buf_struct_size = sizeof(struct c3_isp_vb2_buffer);
> +	vb2_q->dev = cap->isp->dev;
> +	vb2_q->lock = &cap->lock;
> +	vb2_q->min_queued_buffers = 2;

I'm not sure what your plans regarding libcamera are, but be aware
we're going to get stricter about this value. Ideally, we aim to have
ISP driver set this value to 0 so that the ISP operates (produce
statistics) even if no buffer is queued to the capture devices. The
reason is that we want algorithms to run even if there are no capture
buffers provided.

In order to allow the driver to operate with no buffers, you could
probably allocate a scratch buffer in the driver and use it whenever
you receive a frame completed IRQ and have no buffers available in the
cap->provided queue.

See rkisp1_dummy_buf_create() in the RkISP1 driver and how dummy_buf
is used there.

> +
> +	ret = vb2_queue_init(vb2_q);
> +	if (ret < 0)
> +		goto err_destroy;
> +
> +	cap->pad.flags = MEDIA_PAD_FL_SINK;
> +	ret = media_entity_pads_init(&vdev->entity, 1, &cap->pad);
> +	if (ret < 0)
> +		goto err_queue_release;
> +
> +	ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
> +	if (ret < 0) {
> +		dev_err(cap->isp->dev,
> +			"Failed to register %s: %d\n", vdev->name, ret);
> +		goto err_entity_cleanup;
> +	}
> +
> +	return 0;
> +
> +err_entity_cleanup:
> +	media_entity_cleanup(&vdev->entity);
> +err_queue_release:
> +	vb2_queue_release(vb2_q);
> +err_destroy:
> +	mutex_destroy(&cap->lock);
> +	return ret;
> +}
> +
> +int c3_isp_captures_register(struct c3_isp_device *isp)
> +{
> +	int ret;
> +	unsigned int i;
> +	struct c3_isp_capture *cap;
> +
> +	for (i = C3_ISP_CAP_DEV_0; i < C3_ISP_NUM_CAP_DEVS; i++) {
> +		cap = &isp->caps[i];
> +		memset(cap, 0, sizeof(*cap));
> +
> +		cap->vfmt.fmt.pix.width = C3_ISP_DEFAULT_WIDTH;
> +		cap->vfmt.fmt.pix.height = C3_ISP_DEFAULT_HEIGHT;
> +		cap->vfmt.fmt.pix.field = V4L2_FIELD_NONE;
> +		cap->vfmt.fmt.pix.pixelformat = V4L2_PIX_FMT_NV21;
> +		cap->vfmt.fmt.pix.colorspace = V4L2_COLORSPACE_SRGB;
> +
> +		c3_cap_try_fmt(cap, &cap->vfmt.fmt.pix);
> +
> +		cap->id = i;
> +		if (cap->id == C3_ISP_CAP_DEV_0)
> +			cap->rsz = &isp->resizers[C3_ISP_RSZ_0];
> +		else if (cap->id == C3_ISP_CAP_DEV_1)
> +			cap->rsz = &isp->resizers[C3_ISP_RSZ_1];
> +		else
> +			cap->rsz = &isp->resizers[C3_ISP_RSZ_2];
> +
> +		cap->isp = isp;
> +		INIT_LIST_HEAD(&cap->pending);
> +		spin_lock_init(&cap->buff_lock);
> +		mutex_init(&cap->lock);
> +
> +		ret = c3_isp_register_capture(cap);
> +		if (ret) {
> +			cap->isp = NULL;
> +			mutex_destroy(&cap->lock);
> +			c3_isp_captures_unregister(isp);
> +			return ret;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +void c3_isp_captures_unregister(struct c3_isp_device *isp)
> +{
> +	unsigned int i;
> +	struct c3_isp_capture *cap;
> +
> +	for (i = C3_ISP_CAP_DEV_0; i < C3_ISP_NUM_CAP_DEVS; i++) {
> +		cap = &isp->caps[i];
> +
> +		if (!cap->isp)
> +			continue;
> +		vb2_queue_release(&cap->vb2_q);
> +		media_entity_cleanup(&cap->vdev.entity);
> +		video_unregister_device(&cap->vdev);
> +		mutex_destroy(&cap->lock);
> +	}
> +}
> +
> +void c3_isp_captures_done(struct c3_isp_device *isp)
> +{
> +	c3_isp_cap_done(&isp->caps[C3_ISP_CAP_DEV_0]);
> +	c3_isp_cap_done(&isp->caps[C3_ISP_CAP_DEV_1]);
> +	c3_isp_cap_done(&isp->caps[C3_ISP_CAP_DEV_2]);
> +}
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-common.h b/drivers/media/platform/amlogic/c3-isp/c3-isp-common.h
> new file mode 100644
> index 000000000000..19f2a3bc29c9
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-common.h
> @@ -0,0 +1,327 @@
> +/* SPDX-License-Identifier: (GPL-2.0-only OR MIT) */
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#ifndef __C3_ISP_COMMON_H__
> +#define __C3_ISP_COMMON_H__
> +
> +#include <linux/clk.h>
> +
> +#include <media/media-device.h>
> +#include <media/videobuf2-core.h>
> +#include <media/v4l2-device.h>
> +#include <media/v4l2-subdev.h>
> +#include <media/videobuf2-v4l2.h>
> +
> +#define C3_ISP_DRIVER_NAME            "c3-isp"
> +#define C3_ISP_CLOCK_NUM_MAX          3
> +
> +#define C3_ISP_CORE_DEF_SINK_PAD_FMT  MEDIA_BUS_FMT_SRGGB10_1X10
> +#define C3_ISP_CORE_DEF_SRC_PAD_FMT   MEDIA_BUS_FMT_YUYV8_1_5X8
> +#define C3_ISP_RSZ_DEF_PAD_FMT        MEDIA_BUS_FMT_YUYV8_1_5X8
> +#define C3_ISP_DEFAULT_WIDTH          1920
> +#define C3_ISP_DEFAULT_HEIGHT         1080
> +#define C3_ISP_MAX_WIDTH              2888
> +#define C3_ISP_MAX_HEIGHT             2240
> +#define C3_ISP_MIN_WIDTH              160
> +#define C3_ISP_MIN_HEIGHT             120
> +
> +#define C3_DISP_INTER                 0x400
> +#define C3_DISP_REG(base, id)         ((base) + (id) * C3_DISP_INTER)
> +#define C3_WRMIFX3_INTER              0x100
> +#define C3_WRMIFX3_REG(base, id)      ((base) + (id) * C3_WRMIFX3_INTER)
> +#define C3_PPS_TAP4_S11_H_NUM         33
> +#define C3_PPS_LUT_CTYPE_0            0
> +#define C3_PPS_LUT_CTYPE_2            2
> +#define C3_SCALE_EN                   1
> +#define C3_SCALE_DIS                  0
> +
> +#define C3_ISP_PHASE_OFFSET_0         0
> +#define C3_ISP_PHASE_OFFSET_1         1
> +#define C3_ISP_PHASE_OFFSET_NONE      0xff
> +
> +enum c3_isp_core_pads {
> +	C3_ISP_CORE_PAD_SINK_VIDEO,
> +	C3_ISP_CORE_PAD_SINK_PARAMS,
> +	C3_ISP_CORE_PAD_SOURCE_STATS,
> +	C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +	C3_ISP_CORE_PAD_MAX
> +};
> +
> +enum c3_isp_resizer_ids {
> +	C3_ISP_RSZ_0,
> +	C3_ISP_RSZ_1,
> +	C3_ISP_RSZ_2,
> +	C3_ISP_NUM_RSZ
> +};
> +
> +enum c3_isp_resizer_pads {
> +	C3_ISP_RESIZER_PAD_SINK,
> +	C3_ISP_RESIZER_PAD_SOURCE,
> +	C3_ISP_RESIZER_PAD_MAX
> +};
> +
> +enum c3_isp_cap_devs {
> +	C3_ISP_CAP_DEV_0,
> +	C3_ISP_CAP_DEV_1,
> +	C3_ISP_CAP_DEV_2,
> +	C3_ISP_NUM_CAP_DEVS
> +};
> +
> +/**
> + * struct c3_isp_pps_io_size - isp scaler input and output size
> + *
> + * @thsize: input horizontal size of after preprocessing
> + * @tvsize: input vertical size of after preprocessing
> + * @ohsize: output horizontal size
> + * @ovsize: output vertical size
> + * @ihsize: input horizontal size
> + * @max_hsize: maximum horizontal size
> + */
> +struct c3_isp_pps_io_size {
> +	u32 thsize;
> +	u32 tvsize;
> +	u32 ohsize;
> +	u32 ovsize;
> +	u32 ihsize;
> +	u32 max_hsize;
> +};
> +
> +/**
> + * @mbus_code: the mbus code
> + * @pads: save the pad flag of this mbus_code
> + * @xofst: horizontal phase offset of hardware
> + * @yofst: vertical phase offset of hardware
> + */
> +struct c3_isp_mbus_format_info {
> +	u32 mbus_code;
> +	u32 pads;
> +	u8 xofst;
> +	u8 yofst;
> +};
> +
> +/**
> + * @mbus_code: the mbus code
> + * @fourcc: pixel format
> + * @depth: pixel width
> + */
> +struct c3_isp_capture_format {
> +	u32 mbus_code;
> +	u32 fourcc;
> +	u8 depth;
> +};
> +
> +/**
> + * struct c3_isp_vb2_buffer - A container of vb2 buffer
> + *
> + * @vb: vb2 buffer
> + * @vaddr: buffer virtual address
> + * @paddr: buffer physical address
> + * @list: entry of the buffer in the queue
> + */
> +struct c3_isp_vb2_buffer {
> +	struct vb2_v4l2_buffer vb;
> +	void *vaddr;
> +	dma_addr_t paddr;
> +	struct list_head list;
> +};
> +
> +/**
> + * struct c3_isp_core - ISP core subdev
> + *
> + * @sd: ISP sub-device
> + * @pads: ISP sub-device pads
> + * @src_sd: source sub-device
> + * @isp: pointer to c3_isp_device
> + * @src_sd_pad: source sub-device pad
> + */
> +struct c3_isp_core {
> +	struct v4l2_subdev sd;
> +	struct media_pad pads[C3_ISP_CORE_PAD_MAX];
> +	struct v4l2_subdev *src_sd;
> +	u16 src_sd_pad;
> +	struct c3_isp_device *isp;
> +};
> +
> +/**
> + * struct c3_isp_resizer - ISP resizer subdev
> + *
> + * @id: resizer id
> + * @sd: resizer sub-device
> + * @pads: resizer sub-device pads
> + * @isp: pointer to c3_isp_device
> + * @cap: pointer to c3_isp_capture
> + */
> +struct c3_isp_resizer {
> +	enum c3_isp_resizer_ids id;
> +	struct v4l2_subdev sd;
> +	struct media_pad pads[C3_ISP_RESIZER_PAD_MAX];
> +	struct c3_isp_device *isp;
> +	struct c3_isp_capture *cap;
> +};
> +
> +/**
> + * struct c3_isp_stats - ISP statistics device
> + *
> + * @vb2_q: vb2 buffer queue
> + * @vdev: video node
> + * @vfmt: v4l2_format of the metadata format
> + * @pad: media pad
> + * @lock: protects vb2_q, vdev
> + * @is_streaming: stats status
> + * @isp: pointer to c3_isp_device
> + * @buff: in use buffer
> + * @buff_lock: protects stats buffer
> + * @pending: stats buffer list head
> + */
> +struct c3_isp_stats {
> +	struct vb2_queue vb2_q;
> +	struct video_device vdev;
> +	struct v4l2_format vfmt;
> +	struct media_pad pad;
> +
> +	struct mutex lock; /* Protects vb2_q, vdev */
> +	bool is_streaming;
> +	struct c3_isp_device *isp;
> +
> +	struct c3_isp_vb2_buffer *buff;
> +	spinlock_t buff_lock; /* Protects stream buffer */
> +	struct list_head pending;
> +};
> +
> +/**
> + * struct c3_isp_params - ISP parameters device
> + *
> + * @vb2_q: vb2 buffer queue
> + * @vdev: video node
> + * @vfmt: v4l2_format of the metadata format
> + * @pad: media pad
> + * @lock: protects vb2_q, vdev
> + * @isp: pointer to c3_isp_device
> + * @buff: in use buffer
> + * @buff_lock: protects stats buffer
> + * @pending: stats buffer list head
> + */
> +struct c3_isp_params {
> +	struct vb2_queue vb2_q;
> +	struct video_device vdev;
> +	struct v4l2_format vfmt;
> +	struct media_pad pad;
> +
> +	struct mutex lock; /* Protects vb2_q, vdev */
> +	struct c3_isp_device *isp;
> +
> +	struct c3_isp_vb2_buffer *buff;
> +	spinlock_t buff_lock; /* Protects stream buffer */
> +	struct list_head pending;
> +};
> +
> +/**
> + * struct c3_isp_capture - ISP capture device
> + *
> + * @id: capture device ID
> + * @vb2_q: vb2 buffer queue
> + * @vdev: video node
> + * @vfmt: v4l2_format of the capture format
> + * @pad: media pad
> + * @lock: protects vb2_q, vdev
> + * @is_streaming: capture device status
> + * @isp: pointer to c3_isp_device
> + * @rsz: pointer to c3_isp_resizer
> + * @buff: in use buffer
> + * @buff_lock: protects capture buffer
> + * @pending: capture buffer list head
> + */
> +struct c3_isp_capture {
> +	enum c3_isp_cap_devs id;
> +	struct vb2_queue vb2_q;
> +	struct video_device vdev;
> +	struct v4l2_format vfmt;
> +	struct media_pad pad;
> +
> +	struct mutex lock; /* Protects vb2_q, vdev */
> +	bool is_streaming;
> +	struct c3_isp_device *isp;
> +	struct c3_isp_resizer *rsz;
> +
> +	struct c3_isp_vb2_buffer *buff;
> +	spinlock_t buff_lock; /* Protects stream buffer */
> +	struct list_head pending;
> +};
> +
> +/**
> + * struct c3_isp_info - ISP information
> + *
> + * @clocks: array of ISP clock names
> + * @clock_rates: array of ISP clock rate
> + * @clock_num: actual clock number
> + */
> +struct c3_isp_info {
> +	char *clocks[C3_ISP_CLOCK_NUM_MAX];
> +	u32 clock_rates[C3_ISP_CLOCK_NUM_MAX];
> +	u32 clock_num;
> +};
> +
> +/**
> + * struct c3_isp_device - ISP platform device
> + *
> + * @dev: pointer to the struct device
> + * @base: base register address
> + * @clks: array of clocks
> + * @notifier: notifier to register on the v4l2-async API
> + * @v4l2_dev: v4l2_device variable
> + * @media_dev: media device variable
> + * @pipe: media pipeline
> + * @core: ISP core subdev
> + * @resizer: ISP resizer subdev
> + * @stats: ISP stats device
> + * @params: ISP params device
> + * @caps: array of ISP capture device
> + * @frm_sequence: used to record frame id
> + * @lock: protect ISP device
> + * @info: version-specific ISP information
> + */
> +struct c3_isp_device {
> +	struct device *dev;
> +	void __iomem *base;
> +	struct clk_bulk_data clks[C3_ISP_CLOCK_NUM_MAX];
> +
> +	struct v4l2_async_notifier notifier;
> +	struct v4l2_device v4l2_dev;
> +	struct media_device media_dev;
> +	struct media_pipeline pipe;
> +
> +	struct c3_isp_core core;
> +	struct c3_isp_resizer resizers[C3_ISP_NUM_RSZ];
> +	struct c3_isp_stats stats;
> +	struct c3_isp_params params;
> +	struct c3_isp_capture caps[C3_ISP_NUM_CAP_DEVS];
> +
> +	u32 frm_sequence;
> +	struct mutex lock; /* Protect ISP device */
> +	const struct c3_isp_info *info;
> +};
> +
> +u32 c3_isp_read(struct c3_isp_device *isp, u32 reg);
> +void c3_isp_write(struct c3_isp_device *isp, u32 reg, u32 val);
> +void c3_isp_update_bits(struct c3_isp_device *isp, u32 reg, u32 mask, u32 val);
> +bool c3_isp_pipeline_ready(struct c3_isp_device *isp);
> +
> +int c3_isp_core_register(struct c3_isp_device *isp);
> +void c3_isp_core_unregister(struct c3_isp_device *isp);
> +int c3_isp_resizers_register(struct c3_isp_device *isp);
> +void c3_isp_resizers_unregister(struct c3_isp_device *isp);
> +void c3_isp_rsz_start(struct c3_isp_resizer *rsz);
> +void c3_isp_rsz_stop(struct c3_isp_resizer *rsz);
> +int c3_isp_captures_register(struct c3_isp_device *isp);
> +void c3_isp_captures_unregister(struct c3_isp_device *isp);
> +void c3_isp_captures_done(struct c3_isp_device *isp);
> +int c3_isp_stats_register(struct c3_isp_device *isp);
> +void c3_isp_stats_unregister(struct c3_isp_device *isp);
> +int c3_isp_stats_done(struct c3_isp_device *isp);
> +int c3_isp_params_register(struct c3_isp_device *isp);
> +void c3_isp_params_unregister(struct c3_isp_device *isp);
> +int c3_isp_params_done(struct c3_isp_device *isp);
> +
> +#endif
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-core.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-core.c
> new file mode 100644
> index 000000000000..d3672aff9fd2
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-core.c
> @@ -0,0 +1,675 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/pm_runtime.h>
> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +#include "include/uapi/c3-isp-config.h"
> +
> +#define C3_ISP_CORE_SUBDEV_NAME        "isp-core"
> +
> +static const struct c3_isp_mbus_format_info c3_isp_core_mbus_formats[] = {
> +	/* RAW formats */
> +	{
> +		.mbus_code	= MEDIA_BUS_FMT_SBGGR10_1X10,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_0,
> +		.yofst		= C3_ISP_PHASE_OFFSET_1,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SGBRG10_1X10,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_1,
> +		.yofst		= C3_ISP_PHASE_OFFSET_1,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SGRBG10_1X10,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_0,
> +		.yofst		= C3_ISP_PHASE_OFFSET_0,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SRGGB10_1X10,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_1,
> +		.yofst		= C3_ISP_PHASE_OFFSET_0,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SBGGR12_1X12,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_0,
> +		.yofst		= C3_ISP_PHASE_OFFSET_1,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SGBRG12_1X12,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_1,
> +		.yofst		= C3_ISP_PHASE_OFFSET_1,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SGRBG12_1X12,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_0,
> +		.yofst		= C3_ISP_PHASE_OFFSET_0,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_SRGGB12_1X12,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SINK_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_1,
> +		.yofst		= C3_ISP_PHASE_OFFSET_0,
> +	},
> +	/* YUV formats */
> +	{
> +		.mbus_code	= MEDIA_BUS_FMT_Y8_1X8,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SOURCE_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_YUYV8_1_5X8,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SOURCE_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_YUYV8_2X8,
> +		.pads		= BIT(C3_ISP_CORE_PAD_SOURCE_VIDEO),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	},
> +};
> +
> +static const struct c3_isp_mbus_format_info
> +*core_find_format_by_code(u32 code, u32 pad)
> +{
> +	int i;

unsigned

> +
> +	for (i = 0; i < ARRAY_SIZE(c3_isp_core_mbus_formats); i++) {
> +		const struct c3_isp_mbus_format_info *info =
> +			&c3_isp_core_mbus_formats[i];
> +
> +		if (info->mbus_code == code && info->pads & BIT(pad))
> +			return info;
> +	}
> +
> +	return NULL;
> +}
> +
> +static const struct c3_isp_mbus_format_info
> +*core_find_format_by_index(u32 index, u32 pad)
> +{
> +	int i;

unsigned

> +
> +	for (i = 0; i < ARRAY_SIZE(c3_isp_core_mbus_formats); i++) {
> +		const struct c3_isp_mbus_format_info *info =
> +			&c3_isp_core_mbus_formats[i];
> +
> +		if (!(info->pads & BIT(pad)))
> +			continue;
> +
> +		if (!index)
> +			return info;
> +
> +		index--;
> +	}
> +
> +	return NULL;
> +}
> +
> +static void c3_isp_core_enable(struct c3_isp_device *isp)
> +{
> +	/* Select the line sync signal */
> +	c3_isp_update_bits(isp, ISP_TOP_IRQ_LINE_THRD,
> +			   TOP_IRQ_DIN_HSYNC, TOP_IRQ_DIN_HSYNC);
> +
> +	/* Enable frame done and stats error irq */
> +	c3_isp_update_bits(isp, ISP_TOP_IRQ_EN,
> +			   TOP_IRQ_FRAME_DONE, TOP_IRQ_FRAME_DONE);
> +	c3_isp_update_bits(isp, ISP_TOP_IRQ_EN,
> +			   TOP_IRQ_STATS_ERR, TOP_IRQ_STATS_ERR);
> +
> +	/* Enable image data to ISP core */
> +	c3_isp_update_bits(isp, ISP_TOP_PATH_SEL,
> +			   TOP_DATA_PATH_MASK, TOP_PATH_MIPI_TO_CORE);
> +}
> +
> +static void c3_isp_core_disable(struct c3_isp_device *isp)
> +{
> +	/* Disable image data to ISP core */
> +	c3_isp_update_bits(isp, ISP_TOP_PATH_SEL,
> +			   TOP_DATA_PATH_MASK, 0x0);
> +
> +	/* Disable all irq */
> +	c3_isp_write(isp, ISP_TOP_IRQ_EN, 0x0);
> +}
> +
> +/* Set the phase offset of blc, wb and lns */
> +static void c3_isp_core_lswb_ofst(struct c3_isp_device *isp,
> +				  u8 xofst, u8 yofst)
> +{
> +	c3_isp_update_bits(isp, ISP_LSWB_BLC_PHSOFST,
> +			   LSWB_BLC_XPHS_OFST_MASK,
> +			   xofst << LSWB_BLC_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_LSWB_BLC_PHSOFST,
> +			   LSWB_BLC_YPHS_OFST_MASK, yofst);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_PHSOFST,
> +			   LSWB_WB_XPHS_OFST_MASK,
> +			   xofst << LSWB_WB_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_PHSOFST,
> +			   LSWB_WB_YPHS_OFST_MASK, yofst);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_LNS_PHSOFST,
> +			   LSWB_LNS_XPHS_OFST_MASK,
> +			   xofst << LSWB_LNS_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_LSWB_LNS_PHSOFST,
> +			   LSWB_LNS_YPHS_OFST_MASK, yofst);
> +}
> +
> +/* Set the phase offset of af, ae and awb */
> +static void c3_isp_core_3a_ofst(struct c3_isp_device *isp,
> +				u8 xofst, u8 yofst)
> +{
> +	c3_isp_update_bits(isp, ISP_AF_CTRL, AF_CTRL_XPHS_OFST_MASK,
> +			   xofst << AF_CTRL_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_AF_CTRL, AF_CTRL_YPHS_OFST_MASK,
> +			   yofst << AF_CTRL_YPHS_OFST_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CTRL, AE_CTRL_XPHS_OFST_MASK,
> +			   xofst << AE_CTRL_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_AE_CTRL, AE_CTRL_YPHS_OFST_MASK,
> +			   yofst << AE_CTRL_YPHS_OFST_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_CTRL, AWB_CTRL_XPHS_OFST_MASK,
> +			   xofst << AWB_CTRL_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_AWB_CTRL, AWB_CTRL_YPHS_OFST_MASK, yofst);
> +}
> +
> +/* Set the phase offset of demosaic */
> +static void c3_isp_core_dms_ofst(struct c3_isp_device *isp,
> +				 u8 xofst, u8 yofst)
> +{
> +	c3_isp_update_bits(isp, ISP_DMS_COMMON_PARAM0, DMS_COMMON_XPHS_OFST_MASK,
> +			   xofst << DMS_COMMON_XPHS_OFST_SHIFT);
> +	c3_isp_update_bits(isp, ISP_DMS_COMMON_PARAM0, DMS_COMMON_YPHS_OFST_MASK, yofst);
> +}
> +
> +static void c3_isp_core_cfg_ofst(struct c3_isp_device *isp,
> +				 struct v4l2_mbus_framefmt *fmt)
> +{
> +	const struct c3_isp_mbus_format_info *isp_fmt =
> +			core_find_format_by_code(fmt->code, C3_ISP_CORE_PAD_SINK_VIDEO);
> +
> +	c3_isp_core_lswb_ofst(isp, isp_fmt->xofst, isp_fmt->yofst);
> +	c3_isp_core_3a_ofst(isp, isp_fmt->xofst, isp_fmt->yofst);
> +	c3_isp_core_dms_ofst(isp, isp_fmt->xofst, isp_fmt->yofst);
> +}
> +
> +/* Set format of the hardware control module */
> +static void c3_isp_core_top_fmt(struct c3_isp_device *isp,
> +				struct v4l2_mbus_framefmt *fmt)
> +{
> +	c3_isp_write(isp, ISP_TOP_INPUT_SIZE,
> +		     TOP_INPUT_HSIZE(fmt->width) | TOP_INPUT_VSIZE(fmt->height));
> +
> +	c3_isp_write(isp, ISP_TOP_FRM_SIZE,
> +		     TOP_FRM_CORE_IHSIZE(fmt->width) | TOP_FRM_CORE_IVSIZE(fmt->height));
> +
> +	c3_isp_update_bits(isp, ISP_TOP_HOLD_SIZE, TOP_HOLD_HSIZE_MASK,
> +			   fmt->width << TOP_HOLD_HSIZE_SHIFT);
> +}
> +
> +static void c3_isp_core_af_fmt(struct c3_isp_device *isp,
> +			       struct v4l2_mbus_framefmt *fmt)
> +{
> +	u32 hidx;
> +	u32 vidx;
> +	int i;
> +
> +	c3_isp_update_bits(isp, ISP_AF_HV_BLKNUM, AF_HV_STAT_HBLK_NUM_MASK,
> +			   AF_STAT_BLKH_NUM << AF_HV_STAT_HBLK_NUM_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AF_HV_BLKNUM,
> +			   AF_HV_STAT_VBLK_NUM_MASK, AF_STAT_BLKV_NUM);
> +
> +	c3_isp_write(isp, ISP_AF_HV_SIZE,
> +		     AF_HV_STAT_HSIZE(fmt->width) | AF_HV_STAT_VSIZE(fmt->height));
> +
> +	/* Set the index address to 0 position */
> +	c3_isp_write(isp, ISP_AF_IDX_ADDR, 0);
> +	/*
> +	 * Calculate and set the coordinates of points in the grid.
> +	 * hidx and vidx need to be aligned with 2.
> +	 */
> +	for (i = 0; i <= AF_STAT_BLKH_NUM; i++) {
> +		hidx = i * fmt->width / AF_STAT_BLKH_NUM;
> +		hidx = ALIGN_DOWN(hidx, 2);
> +
> +		vidx = i * fmt->height / AF_STAT_BLKV_NUM;
> +		vidx = min(vidx, fmt->height);
> +		vidx = ALIGN_DOWN(vidx, 2);
> +		c3_isp_write(isp, ISP_AF_IDX_DATA,
> +			     AF_IDX_HIDX_DATA(hidx) | AF_IDX_VIDX_DATA(vidx));
> +	}
> +}
> +
> +static void c3_isp_core_ae_fmt(struct c3_isp_device *isp,
> +			       struct v4l2_mbus_framefmt *fmt)
> +{
> +	u32 hidx;
> +	u32 vidx;
> +	int i;
> +
> +	c3_isp_update_bits(isp, ISP_AE_HV_BLKNUM, AE_HV_STAT_HBLK_NUM_MASK,
> +			   AE_STAT_BLKH_NUM << AE_HV_STAT_HBLK_NUM_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_HV_BLKNUM,
> +			   AE_HV_STAT_VBLK_NUM_MASK, AE_STAT_BLKV_NUM);
> +
> +	c3_isp_write(isp, ISP_AE_HV_SIZE,
> +		     AE_HV_STAT_HSIZE(fmt->width) | AE_HV_STAT_VSIZE(fmt->height));
> +
> +	/* Set the index address to 0 position */
> +	c3_isp_write(isp, ISP_AE_IDX_ADDR, 0);
> +	/*
> +	 * Calculate and set the coordinates of points in the grid.
> +	 * hidx and vidx need to be aligned with 2.
> +	 */
> +	for (i = 0; i <= AE_STAT_BLKH_NUM; i++) {
> +		hidx = i * fmt->width / AE_STAT_BLKH_NUM;
> +		hidx = ALIGN_DOWN(hidx, 2);
> +
> +		vidx = i * fmt->height / AE_STAT_BLKV_NUM;
> +		vidx = min(vidx, fmt->height);
> +		vidx = ALIGN_DOWN(vidx, 2);
> +
> +		c3_isp_write(isp, ISP_AE_IDX_DATA,
> +			     AE_IDX_HIDX_DATA(hidx) | AE_IDX_VIDX_DATA(vidx));
> +	}
> +}
> +
> +static void c3_isp_core_awb_fmt(struct c3_isp_device *isp,
> +				struct v4l2_mbus_framefmt *fmt)
> +{
> +	u32 hidx;
> +	u32 vidx;
> +	int i;
> +
> +	c3_isp_update_bits(isp, ISP_AWB_HV_BLKNUM, AWB_HV_STAT_HBLK_NUM_MASK,
> +			   AWB_STAT_BLKH_NUM << AWB_HV_STAT_HBLK_NUM_SHIFT);
> +	c3_isp_update_bits(isp, ISP_AWB_HV_BLKNUM,
> +			   AWB_HV_STAT_VBLK_NUM_MASK, AWB_STAT_BLKV_NUM);
> +
> +	c3_isp_write(isp, ISP_AWB_HV_SIZE,
> +		     AWB_HV_STAT_HSIZE(fmt->width) | AWB_HV_STAT_VSIZE(fmt->height));
> +
> +	/* Set the index address to 0 position */
> +	c3_isp_write(isp, ISP_AWB_IDX_ADDR, 0);
> +	/*
> +	 * Calculate and set the coordinates of points in the grid.
> +	 * hidx and vidx need to be aligned with 2.
> +	 */
> +	for (i = 0; i <= AWB_STAT_BLKH_NUM; i++) {
> +		hidx = i * fmt->width / AWB_STAT_BLKH_NUM;
> +		hidx = ALIGN_DOWN(hidx, 2);
> +
> +		vidx = i * fmt->height / AWB_STAT_BLKV_NUM;
> +		vidx = min(vidx, fmt->height);
> +		vidx = ALIGN_DOWN(vidx, 2);
> +
> +		c3_isp_write(isp, ISP_AWB_IDX_DATA,
> +			     AWB_IDX_HIDX_DATA(hidx) | AWB_IDX_VIDX_DATA(vidx));
> +	}
> +}
> +
> +static void c3_isp_core_cfg_format(struct c3_isp_device *isp,
> +				   struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_mbus_framefmt *fmt;
> +
> +	fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SINK_VIDEO);
> +
> +	c3_isp_core_cfg_ofst(isp, fmt);
> +	c3_isp_core_top_fmt(isp, fmt);
> +	c3_isp_core_af_fmt(isp, fmt);
> +	c3_isp_core_ae_fmt(isp, fmt);
> +	c3_isp_core_awb_fmt(isp, fmt);
> +}
> +
> +static int c3_isp_core_enable_streams(struct v4l2_subdev *sd,
> +				      struct v4l2_subdev_state *state,
> +				      u32 pad, u64 streams_mask)
> +{
> +	struct c3_isp_core *core = v4l2_get_subdevdata(sd);
> +	u64 sink_streams;
> +	int ret;
> +
> +	core->isp->frm_sequence = 0;
> +	c3_isp_core_cfg_format(core->isp, state);
> +	c3_isp_core_enable(core->isp);
> +
> +	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
> +						       C3_ISP_CORE_PAD_SINK_VIDEO,
> +						       &streams_mask);
> +	ret = v4l2_subdev_enable_streams(core->src_sd,
> +					 core->src_sd_pad, sink_streams);
> +	if (ret) {
> +		c3_isp_core_disable(core->isp);
> +		return ret;
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_core_disable_streams(struct v4l2_subdev *sd,
> +				       struct v4l2_subdev_state *state,
> +				       u32 pad, u64 streams_mask)
> +{
> +	struct c3_isp_core *core = v4l2_get_subdevdata(sd);
> +	u64 sink_streams;
> +	int ret;
> +
> +	sink_streams = v4l2_subdev_state_xlate_streams(state, pad,
> +						       C3_ISP_CORE_PAD_SINK_VIDEO,
> +						       &streams_mask);
> +	ret = v4l2_subdev_disable_streams(core->src_sd,
> +					  core->src_sd_pad, sink_streams);
> +	if (ret)
> +		return ret;
> +
> +	c3_isp_core_disable(core->isp);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_core_cfg_routing(struct v4l2_subdev *sd,
> +				   struct v4l2_subdev_state *state,
> +				   struct v4l2_subdev_krouting *routing)
> +{
> +	static const struct v4l2_mbus_framefmt format = {
> +		.width = C3_ISP_DEFAULT_WIDTH,
> +		.height = C3_ISP_DEFAULT_HEIGHT,
> +		.code = C3_ISP_CORE_DEF_SRC_PAD_FMT,
> +		.field = V4L2_FIELD_NONE,
> +		.colorspace = V4L2_COLORSPACE_SRGB,
> +		.ycbcr_enc = V4L2_YCBCR_ENC_601,
> +		.quantization = V4L2_QUANTIZATION_LIM_RANGE,
> +		.xfer_func = V4L2_XFER_FUNC_SRGB,
> +	};
> +	int ret;
> +
> +	ret = v4l2_subdev_routing_validate(sd, routing,
> +					   V4L2_SUBDEV_ROUTING_NO_SOURCE_STREAM_MIX);
> +	if (ret)
> +		return ret;
> +
> +	ret = v4l2_subdev_set_routing_with_fmt(sd, state, routing, &format);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_core_init_routing(struct v4l2_subdev *sd,
> +				    struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_subdev_route routes[2];
> +	struct v4l2_subdev_krouting routing;
> +
> +	routes[0].sink_pad = C3_ISP_CORE_PAD_SINK_VIDEO;
> +	routes[0].sink_stream = 0;
> +	routes[0].source_pad = C3_ISP_CORE_PAD_SOURCE_STATS;
> +	routes[0].source_stream = 0;
> +	routes[0].flags = V4L2_SUBDEV_ROUTE_FL_ACTIVE;
> +
> +	routes[1].sink_pad = C3_ISP_CORE_PAD_SINK_VIDEO;
> +	routes[1].sink_stream = 0;
> +	routes[1].source_pad = C3_ISP_CORE_PAD_SOURCE_VIDEO;
> +	routes[1].source_stream = 0;
> +	routes[1].flags = V4L2_SUBDEV_ROUTE_FL_ACTIVE;
> +
> +	routing.num_routes = ARRAY_SIZE(routes);
> +	routing.routes = routes;
> +
> +	return c3_isp_core_cfg_routing(sd, state, &routing);
> +}
> +
> +static int c3_isp_core_set_routing(struct v4l2_subdev *sd,
> +				   struct v4l2_subdev_state *state,
> +				   enum v4l2_subdev_format_whence which,
> +				   struct v4l2_subdev_krouting *routing)

I'm not sure I see a reason for the ISP subdev to implement routing.
In my understanding it will only receive a single image stream and
processes it to 3 DMA output nodes, and to control which output is
enabled you use media links.

> +{
> +	bool is_streaming = v4l2_subdev_is_streaming(sd);
> +
> +	if (which == V4L2_SUBDEV_FORMAT_ACTIVE && is_streaming)
> +		return -EBUSY;
> +
> +	return c3_isp_core_cfg_routing(sd, state, routing);
> +}
> +
> +static int c3_isp_core_enum_mbus_code(struct v4l2_subdev *sd,
> +				      struct v4l2_subdev_state *state,
> +				      struct v4l2_subdev_mbus_code_enum *code)
> +{
> +	const struct c3_isp_mbus_format_info *info;
> +	int ret = 0;
> +
> +	switch (code->pad) {
> +	case C3_ISP_CORE_PAD_SINK_VIDEO:
> +	case C3_ISP_CORE_PAD_SOURCE_VIDEO:
> +		info = core_find_format_by_index(code->index, code->pad);
> +		if (!info)
> +			ret = -EINVAL;
> +		else
> +			code->code = info->mbus_code;
> +
> +		break;
> +	case C3_ISP_CORE_PAD_SINK_PARAMS:
> +	case C3_ISP_CORE_PAD_SOURCE_STATS:
> +		if (code->index)
> +			ret = -EINVAL;
> +		else
> +			code->code = MEDIA_BUS_FMT_METADATA_FIXED;
> +
> +		break;
> +	default:
> +		ret = -EINVAL;
> +		break;
> +	}
> +
> +	return ret;
> +}
> +
> +static void c3_isp_core_set_sink_fmt(struct v4l2_subdev_state *state,
> +				     struct v4l2_subdev_format *format)
> +{
> +	struct v4l2_mbus_framefmt *sink_fmt;
> +	const struct c3_isp_mbus_format_info *isp_fmt;
> +
> +	sink_fmt = v4l2_subdev_state_get_format(state, format->pad);
> +
> +	isp_fmt = core_find_format_by_code(format->format.code, format->pad);
> +	if (!isp_fmt)
> +		sink_fmt->code = C3_ISP_CORE_DEF_SINK_PAD_FMT;
> +	else
> +		sink_fmt->code = format->format.code;
> +
> +	sink_fmt->width = clamp_t(u32, format->format.width,
> +				  C3_ISP_MIN_WIDTH, C3_ISP_MAX_WIDTH);
> +	sink_fmt->height = clamp_t(u32, format->format.height,
> +				   C3_ISP_MIN_HEIGHT, C3_ISP_MAX_HEIGHT);
> +
> +	format->format = *sink_fmt;
> +}
> +
> +static void c3_isp_core_set_source_fmt(struct v4l2_subdev_state *state,
> +				       struct v4l2_subdev_format *format)
> +{
> +	const struct c3_isp_mbus_format_info *isp_fmt;
> +	struct v4l2_mbus_framefmt *sink_fmt;
> +	struct v4l2_mbus_framefmt *src_fmt;
> +
> +	sink_fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SINK_VIDEO);
> +	src_fmt = v4l2_subdev_state_get_format(state, format->pad);
> +
> +	isp_fmt = core_find_format_by_code(format->format.code, format->pad);
> +	if (!isp_fmt)
> +		src_fmt->code = C3_ISP_CORE_DEF_SRC_PAD_FMT;
> +	else
> +		src_fmt->code = format->format.code;
> +
> +	/* The source size must be same with the sink size. */
> +	src_fmt->width  = sink_fmt->width;
> +	src_fmt->height = sink_fmt->height;
> +
> +	format->format = *src_fmt;
> +}
> +
> +static int c3_isp_core_set_fmt(struct v4l2_subdev *sd,
> +			       struct v4l2_subdev_state *state,
> +			       struct v4l2_subdev_format *format)
> +{
> +	struct v4l2_mbus_framefmt *fmt;
> +
> +	switch (format->pad) {
> +	case C3_ISP_CORE_PAD_SINK_VIDEO:
> +		c3_isp_core_set_sink_fmt(state, format);
> +		break;
> +	case C3_ISP_CORE_PAD_SINK_PARAMS:
> +	case C3_ISP_CORE_PAD_SOURCE_STATS:
> +		fmt = v4l2_subdev_state_get_format(state, format->pad);
> +		format->format = *fmt;
> +		break;
> +	case C3_ISP_CORE_PAD_SOURCE_VIDEO:
> +		c3_isp_core_set_source_fmt(state, format);
> +		break;
> +	default:

I don't think this can happen. The core validates that format->pad is
correct.

> +		dev_err(sd->dev, "Invalid pad: %u\n", format->pad);
> +		return -ENOTTY;
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_core_init_state(struct v4l2_subdev *sd,
> +				  struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_mbus_framefmt *sink_fmt;
> +	struct v4l2_mbus_framefmt *src_fmt;
> +
> +	/* Video sink pad */
> +	sink_fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SINK_VIDEO);
> +	sink_fmt->width = C3_ISP_DEFAULT_WIDTH;
> +	sink_fmt->height = C3_ISP_DEFAULT_HEIGHT;
> +	sink_fmt->field = V4L2_FIELD_NONE;
> +	sink_fmt->code = C3_ISP_CORE_DEF_SINK_PAD_FMT;
> +	sink_fmt->colorspace = V4L2_COLORSPACE_RAW;
> +	sink_fmt->xfer_func = V4L2_XFER_FUNC_NONE;
> +	sink_fmt->ycbcr_enc = V4L2_YCBCR_ENC_601;
> +	sink_fmt->quantization = V4L2_QUANTIZATION_FULL_RANGE;
> +
> +	/* Video source pad */
> +	src_fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SOURCE_VIDEO);
> +	src_fmt->width = C3_ISP_DEFAULT_WIDTH;
> +	src_fmt->height = C3_ISP_DEFAULT_HEIGHT;
> +	src_fmt->field = V4L2_FIELD_NONE;
> +	src_fmt->code = C3_ISP_CORE_DEF_SRC_PAD_FMT;
> +	src_fmt->colorspace = V4L2_COLORSPACE_SRGB;
> +	src_fmt->xfer_func = V4L2_XFER_FUNC_SRGB;
> +	src_fmt->ycbcr_enc = V4L2_YCBCR_ENC_601;
> +	src_fmt->quantization = V4L2_QUANTIZATION_LIM_RANGE;
> +
> +	/* Parameters pad */
> +	sink_fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SINK_PARAMS);
> +	sink_fmt->width = 0;
> +	sink_fmt->height = 0;
> +	sink_fmt->field = V4L2_FIELD_NONE;
> +	sink_fmt->code = MEDIA_BUS_FMT_METADATA_FIXED;
> +
> +	/* Statistics pad */
> +	src_fmt = v4l2_subdev_state_get_format(state, C3_ISP_CORE_PAD_SOURCE_STATS);
> +	src_fmt->width = 0;
> +	src_fmt->height = 0;
> +	src_fmt->field = V4L2_FIELD_NONE;
> +	src_fmt->code = MEDIA_BUS_FMT_METADATA_FIXED;
> +
> +	return c3_isp_core_init_routing(sd, state);
> +}
> +
> +static const struct v4l2_subdev_pad_ops c3_isp_core_pad_ops = {
> +	.enum_mbus_code = c3_isp_core_enum_mbus_code,
> +	.get_fmt = v4l2_subdev_get_fmt,
> +	.set_fmt = c3_isp_core_set_fmt,
> +	.enable_streams = c3_isp_core_enable_streams,
> +	.disable_streams = c3_isp_core_disable_streams,
> +	.set_routing = c3_isp_core_set_routing,
> +};
> +
> +static const struct v4l2_subdev_ops c3_isp_core_subdev_ops = {
> +	.pad = &c3_isp_core_pad_ops,
> +};
> +
> +static const struct v4l2_subdev_internal_ops c3_isp_core_internal_ops = {
> +	.init_state = c3_isp_core_init_state,
> +};
> +
> +static int c3_isp_core_link_validate(struct media_link *link)
> +{
> +	if (link->sink->index == C3_ISP_CORE_PAD_SINK_PARAMS)
> +		return 0;
> +
> +	return v4l2_subdev_link_validate(link);
> +}
> +
> +/* Media entity operations */
> +static const struct media_entity_operations c3_isp_core_entity_ops = {
> +	.link_validate = c3_isp_core_link_validate,
> +};
> +
> +int c3_isp_core_register(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_core *core = &isp->core;
> +	struct v4l2_subdev *sd = &core->sd;
> +	int ret;
> +
> +	v4l2_subdev_init(sd, &c3_isp_core_subdev_ops);
> +	sd->owner = THIS_MODULE;
> +	sd->flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
> +	sd->internal_ops = &c3_isp_core_internal_ops;
> +	snprintf(sd->name, sizeof(sd->name), "%s", C3_ISP_CORE_SUBDEV_NAME);
> +
> +	sd->entity.function = MEDIA_ENT_F_PROC_VIDEO_PIXEL_FORMATTER;
> +	sd->entity.ops = &c3_isp_core_entity_ops;
> +
> +	core->isp = isp;
> +	sd->dev = isp->dev;
> +	v4l2_set_subdevdata(sd, core);
> +
> +	core->pads[C3_ISP_CORE_PAD_SINK_VIDEO].flags = MEDIA_PAD_FL_SINK;
> +	core->pads[C3_ISP_CORE_PAD_SINK_PARAMS].flags = MEDIA_PAD_FL_SINK;
> +	core->pads[C3_ISP_CORE_PAD_SOURCE_STATS].flags = MEDIA_PAD_FL_SOURCE;
> +	core->pads[C3_ISP_CORE_PAD_SOURCE_VIDEO].flags = MEDIA_PAD_FL_SOURCE;
> +	ret = media_entity_pads_init(&sd->entity, C3_ISP_CORE_PAD_MAX, core->pads);
> +	if (ret)
> +		return ret;
> +
> +	ret = v4l2_subdev_init_finalize(sd);
> +	if (ret)
> +		goto err_entity_cleanup;
> +
> +	ret = v4l2_device_register_subdev(&isp->v4l2_dev, sd);
> +	if (ret)
> +		goto err_subdev_cleanup;
> +
> +	return 0;
> +
> +err_subdev_cleanup:
> +	v4l2_subdev_cleanup(sd);
> +err_entity_cleanup:
> +	media_entity_cleanup(&sd->entity);
> +	return ret;
> +}
> +
> +void c3_isp_core_unregister(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_core *core = &isp->core;
> +	struct v4l2_subdev *sd = &core->sd;
> +
> +	v4l2_device_unregister_subdev(sd);
> +	v4l2_subdev_cleanup(sd);
> +	media_entity_cleanup(&sd->entity);
> +}
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-dev.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-dev.c
> new file mode 100644
> index 000000000000..a57b9f8dbc3c
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-dev.c
> @@ -0,0 +1,486 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/clk.h>
> +#include <linux/device.h>
> +#include <linux/module.h>
> +#include <linux/mutex.h>
> +#include <linux/platform_device.h>
> +#include <linux/pm_runtime.h>
> +
> +#include <media/v4l2-common.h>
> +#include <media/v4l2-device.h>
> +#include <media/v4l2-fwnode.h>
> +#include <media/v4l2-mc.h>
> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +
> +u32 c3_isp_read(struct c3_isp_device *isp, u32 reg)
> +{
> +	return readl(isp->base + reg);
> +}
> +
> +void c3_isp_write(struct c3_isp_device *isp, u32 reg, u32 val)
> +{
> +	writel(val, isp->base + reg);
> +}
> +
> +void c3_isp_update_bits(struct c3_isp_device *isp, u32 reg, u32 mask, u32 val)
> +{
> +	u32 orig, tmp;
> +
> +	orig = c3_isp_read(isp, reg);
> +
> +	tmp = orig & ~mask;
> +	tmp |= val & mask;
> +
> +	if (tmp != orig)
> +		c3_isp_write(isp, reg, tmp);
> +}
> +
> +bool c3_isp_pipeline_ready(struct c3_isp_device *isp)
> +{
> +	struct media_pipeline_entity_iter iter;
> +	unsigned int n_video_devices = 0;
> +	struct media_entity *entity;
> +	int ret;
> +
> +	ret = media_pipeline_entity_iter_init(&isp->pipe, &iter);
> +	if (ret)
> +		return ret;
> +
> +	media_pipeline_for_each_entity(&isp->pipe, &iter, entity) {
> +		if (entity->obj_type == MEDIA_ENTITY_TYPE_VIDEO_DEVICE)
> +			n_video_devices++;
> +	}
> +
> +	media_pipeline_entity_iter_cleanup(&iter);
> +
> +	return n_video_devices == isp->pipe.start_count;
> +}

As suggested in the review of the -params.c module, I suggest to
implement start_streaming on the capture nodes only. From there you
can call enable_streams on the ISP subdevice. The ISP subdevice can
count how many links to resizers are enabled, and actually start its
operation when all the enabled ones have been started.

The idea is to use enabled media links to identify how many capture video
devices are expected to be used, and only start the ISP (and the
downstream subdevices like the CSI-2 Adap and RX and the sensor) when
all of linked ones have been started.

> +
> +/* PM runtime suspend */
> +static int __maybe_unused c3_isp_runtime_suspend(struct device *dev)
> +{
> +	struct c3_isp_device *isp = dev_get_drvdata(dev);
> +
> +	clk_bulk_disable_unprepare(isp->info->clock_num, isp->clks);
> +
> +	return 0;
> +}
> +
> +/* PM runtime resume */
> +static int __maybe_unused c3_isp_runtime_resume(struct device *dev)
> +{
> +	struct c3_isp_device *isp = dev_get_drvdata(dev);
> +
> +	return clk_bulk_prepare_enable(isp->info->clock_num, isp->clks);
> +}
> +
> +static const struct dev_pm_ops c3_isp_pm_ops = {
> +	SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
> +				pm_runtime_force_resume)
> +	SET_RUNTIME_PM_OPS(c3_isp_runtime_suspend,
> +			   c3_isp_runtime_resume, NULL)
> +};
> +
> +/* IRQ handling */
> +static irqreturn_t c3_isp_irq_handler(int irq, void *dev)
> +{
> +	struct c3_isp_device *isp = dev;
> +	u32 status;
> +
> +	/* Get irq status and clear irq status */
> +	status = c3_isp_read(isp, ISP_TOP_RO_IRQ_STAT);
> +	c3_isp_write(isp, ISP_TOP_IRQ_CLR, status);
> +
> +	if (status & TOP_IRQ_FRAME_DONE) {
> +		c3_isp_stats_done(isp);
> +		c3_isp_params_done(isp);
> +		c3_isp_captures_done(isp);
> +		isp->frm_sequence++;
> +	}
> +
> +	if (status & TOP_IRQ_STATS_ERR)
> +		dev_dbg(isp->dev, "ISP IRQ Stats DMA Error\n");
> +
> +	return IRQ_HANDLED;
> +}
> +
> +/* Subdev notifier register */
> +static int c3_isp_notify_bound(struct v4l2_async_notifier *notifier,
> +			       struct v4l2_subdev *sd,
> +			       struct v4l2_async_connection *asc)
> +{
> +	struct c3_isp_device *isp =
> +		container_of(notifier, struct c3_isp_device, notifier);
> +	struct c3_isp_core *core = &isp->core;
> +	struct media_pad *sink = &core->sd.entity.pads[C3_ISP_CORE_PAD_SINK_VIDEO];
> +	int ret;
> +
> +	ret = media_entity_get_fwnode_pad(&sd->entity,
> +					  sd->fwnode, MEDIA_PAD_FL_SOURCE);
> +	if (ret < 0) {
> +		dev_err(isp->dev, "Failed to find pad for %s\n", sd->name);
> +		return ret;
> +	}
> +
> +	core->src_sd = sd;
> +	core->src_sd_pad = ret;
> +
> +	return v4l2_create_fwnode_links_to_pad(sd, sink, MEDIA_LNK_FL_ENABLED |
> +					       MEDIA_LNK_FL_IMMUTABLE);
> +}
> +
> +static int c3_isp_notify_complete(struct v4l2_async_notifier *notifier)
> +{
> +	struct c3_isp_device *isp =
> +		container_of(notifier, struct c3_isp_device, notifier);
> +	int ret;
> +
> +	ret = v4l2_device_register_subdev_nodes(&isp->v4l2_dev);
> +	if (ret < 0) {
> +		dev_err(isp->dev,
> +			"Failed to register subdev nodes: %d\n", ret);
> +		return ret;
> +	}
> +
> +	dev_info(isp->dev, "notify complete\n");
> +
> +	return media_device_register(&isp->media_dev);
> +}
> +
> +static void c3_isp_notify_destroy(struct v4l2_async_connection *asc)
> +{
> +	struct c3_isp_device *isp =
> +		container_of(asc->notifier, struct c3_isp_device, notifier);
> +
> +	media_device_unregister(&isp->media_dev);
> +}
> +
> +static const struct v4l2_async_notifier_operations c3_isp_notify_ops = {
> +	.bound = c3_isp_notify_bound,
> +	.complete = c3_isp_notify_complete,
> +	.destroy = c3_isp_notify_destroy,
> +};
> +
> +static int c3_isp_async_nf_register(struct c3_isp_device *isp)
> +{
> +	struct v4l2_async_connection *asc;
> +	struct fwnode_handle *ep;
> +	int ret;
> +
> +	v4l2_async_nf_init(&isp->notifier, &isp->v4l2_dev);
> +
> +	ep = fwnode_graph_get_endpoint_by_id(dev_fwnode(isp->dev), 0, 0,
> +					     FWNODE_GRAPH_ENDPOINT_NEXT);
> +	if (!ep)
> +		return -ENOTCONN;
> +
> +	asc = v4l2_async_nf_add_fwnode_remote(&isp->notifier, ep,
> +					      struct v4l2_async_connection);
> +	if (IS_ERR(asc)) {
> +		fwnode_handle_put(ep);
> +		return PTR_ERR(asc);
> +	}
> +
> +	fwnode_handle_put(ep);
> +
> +	isp->notifier.ops = &c3_isp_notify_ops;
> +	ret = v4l2_async_nf_register(&isp->notifier);
> +	if (ret)
> +		v4l2_async_nf_cleanup(&isp->notifier);
> +
> +	return ret;
> +}
> +
> +static void c3_isp_async_nf_unregister(struct c3_isp_device *isp)
> +{
> +	v4l2_async_nf_unregister(&isp->notifier);
> +	v4l2_async_nf_cleanup(&isp->notifier);
> +}
> +
> +static int c3_isp_v4l2_register(struct c3_isp_device *isp)
> +{
> +	struct media_device *media_dev = &isp->media_dev;
> +	struct v4l2_device *v4l2_dev = &isp->v4l2_dev;
> +	int ret;
> +
> +	/* Initialize media device */
> +	strscpy(media_dev->model, C3_ISP_DRIVER_NAME,
> +		sizeof(media_dev->model));
> +	media_dev->dev = isp->dev;
> +
> +	media_device_init(media_dev);
> +
> +	/* Initialize v4l2 device */
> +	v4l2_dev->mdev = media_dev;
> +	strscpy(v4l2_dev->name, C3_ISP_DRIVER_NAME,
> +		sizeof(v4l2_dev->name));
> +
> +	ret = v4l2_device_register(isp->dev, v4l2_dev);
> +	if (ret) {
> +		media_device_cleanup(media_dev);
> +		dev_err(isp->dev,
> +			"Failed to register V4L2 device: %d\n", ret);
> +	}
> +
> +	return ret;
> +}
> +
> +static void c3_isp_v4l2_unregister(struct c3_isp_device *isp)
> +{
> +	v4l2_device_unregister(&isp->v4l2_dev);
> +	media_device_cleanup(&isp->media_dev);
> +}
> +
> +static void c3_isp_remove_links(struct c3_isp_device *isp)
> +{
> +	unsigned int i;
> +
> +	media_entity_remove_links(&isp->core.sd.entity);
> +
> +	for (i = 0; i < C3_ISP_NUM_RSZ; i++)
> +		media_entity_remove_links(&isp->resizers[i].sd.entity);
> +
> +	for (i = 0; i < C3_ISP_NUM_CAP_DEVS; i++)
> +		media_entity_remove_links(&isp->caps[i].vdev.entity);
> +}
> +
> +static int c3_isp_create_links(struct c3_isp_device *isp)
> +{
> +	unsigned int i;
> +	int ret;
> +
> +	for (i = 0; i < C3_ISP_NUM_RSZ; i++) {
> +		ret = media_create_pad_link(&isp->resizers[i].sd.entity,
> +					    C3_ISP_RESIZER_PAD_SOURCE,
> +					    &isp->resizers[i].cap->vdev.entity,
> +					    0, MEDIA_LNK_FL_ENABLED);

This could be made IMMUTABLE

> +		if (ret) {
> +			dev_err(isp->dev, "Failed to link resizer %u and capture %u\n", i, i);
> +			goto err_remove_links;
> +		}
> +
> +		ret = media_create_pad_link(&isp->core.sd.entity,
> +					    C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +					    &isp->resizers[i].sd.entity,
> +					    C3_ISP_RESIZER_PAD_SINK,
> +					    MEDIA_LNK_FL_ENABLED);
> +		if (ret) {
> +			dev_err(isp->dev, "Failed to link core and resizer %u\n", i);
> +			goto err_remove_links;
> +		}
> +	}
> +
> +	ret = media_create_pad_link(&isp->core.sd.entity,
> +				    C3_ISP_CORE_PAD_SOURCE_STATS,
> +				    &isp->stats.vdev.entity,
> +				    0, MEDIA_LNK_FL_ENABLED);
> +	if (ret) {
> +		dev_err(isp->dev, "Failed to link core and stats\n");
> +		goto err_remove_links;
> +	}
> +
> +	ret = media_create_pad_link(&isp->params.vdev.entity, 0,
> +				    &isp->core.sd.entity,
> +				    C3_ISP_CORE_PAD_SINK_PARAMS,
> +				    MEDIA_LNK_FL_ENABLED);
> +	if (ret) {
> +		dev_err(isp->dev, "Failed to link params and core\n");
> +		goto err_remove_links;
> +	}
> +
> +	return 0;
> +
> +err_remove_links:
> +	c3_isp_remove_links(isp);
> +	return ret;
> +}
> +
> +static int c3_isp_videos_register(struct c3_isp_device *isp)
> +{
> +	int ret;
> +
> +	ret = c3_isp_captures_register(isp);
> +	if (ret)
> +		return ret;
> +
> +	ret = c3_isp_stats_register(isp);
> +	if (ret)
> +		goto err_captures_unregister;
> +
> +	ret = c3_isp_params_register(isp);
> +	if (ret)
> +		goto err_stats_unregister;
> +
> +	ret = c3_isp_create_links(isp);
> +	if (ret)
> +		goto err_params_unregister;
> +
> +	return 0;
> +
> +err_params_unregister:
> +	c3_isp_params_unregister(isp);
> +err_stats_unregister:
> +	c3_isp_stats_unregister(isp);
> +err_captures_unregister:
> +	c3_isp_captures_unregister(isp);
> +	return ret;
> +}
> +
> +static void c3_isp_videos_unregister(struct c3_isp_device *isp)
> +{
> +	c3_isp_remove_links(isp);
> +	c3_isp_params_unregister(isp);
> +	c3_isp_stats_unregister(isp);
> +	c3_isp_captures_unregister(isp);
> +}
> +
> +static int c3_isp_cfg_clocks(struct c3_isp_device *isp)
> +{
> +	const struct c3_isp_info *info = isp->info;
> +	int ret;
> +	u32 i;
> +
> +	for (i = 0; i < info->clock_num; i++)
> +		isp->clks[i].id = info->clocks[i];
> +
> +	ret = devm_clk_bulk_get(isp->dev, info->clock_num, isp->clks);
> +	if (ret)
> +		return ret;
> +
> +	for (i = 0; i < info->clock_num; i++) {
> +		if (!info->clock_rates[i])
> +			continue;
> +		ret = clk_set_rate(isp->clks[i].clk, info->clock_rates[i]);
> +		if (ret) {
> +			dev_err(isp->dev, "Failed to set %s rate %u\n", info->clocks[i],
> +				info->clock_rates[i]);
> +			return ret;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct c3_isp_device *isp;
> +	int irq;
> +	int ret;
> +
> +	isp = devm_kzalloc(dev, sizeof(*isp), GFP_KERNEL);
> +	if (!isp)
> +		return -ENOMEM;
> +
> +	isp->info = of_device_get_match_data(dev);
> +	isp->dev = dev;
> +
> +	isp->base = devm_platform_ioremap_resource_byname(pdev, "isp");
> +	if (IS_ERR(isp->base))
> +		return dev_err_probe(dev, PTR_ERR(isp->base),
> +				     "Failed to ioremap resource\n");
> +
> +	irq = platform_get_irq(pdev, 0);
> +	if (irq < 0)
> +		return irq;
> +
> +	ret = c3_isp_cfg_clocks(isp);
> +	if (ret)
> +		return dev_err_probe(dev, ret, "Failed to configure clocks\n");
> +
> +	platform_set_drvdata(pdev, isp);
> +
> +	pm_runtime_enable(dev);
> +
> +	ret = c3_isp_v4l2_register(isp);
> +	if (ret)
> +		goto err_runtime_disable;
> +
> +	ret = c3_isp_core_register(isp);
> +	if (ret)
> +		goto err_v4l2_unregister;
> +
> +	ret = c3_isp_resizers_register(isp);
> +	if (ret)
> +		goto err_core_unregister;
> +
> +	ret = c3_isp_async_nf_register(isp);
> +	if (ret)
> +		goto err_resizers_unregister;
> +
> +	ret = c3_isp_videos_register(isp);
> +	if (ret)
> +		goto err_nf_unregister;
> +
> +	ret = devm_request_irq(dev, irq,
> +			       c3_isp_irq_handler, IRQF_SHARED,
> +			       dev_driver_string(dev), isp);
> +	if (ret)
> +		goto err_streams_unregister;

I would request the IRQ before registering devices to userspace.

> +
> +	mutex_init(&isp->lock);
> +
> +	return 0;
> +
> +err_streams_unregister:
> +	c3_isp_videos_unregister(isp);
> +err_nf_unregister:
> +	c3_isp_async_nf_unregister(isp);
> +err_resizers_unregister:
> +	c3_isp_resizers_unregister(isp);
> +err_core_unregister:
> +	c3_isp_core_unregister(isp);
> +err_v4l2_unregister:
> +	c3_isp_v4l2_unregister(isp);
> +err_runtime_disable:
> +	pm_runtime_disable(dev);
> +	return ret;
> +};
> +
> +static void c3_isp_remove(struct platform_device *pdev)
> +{
> +	struct c3_isp_device *isp = platform_get_drvdata(pdev);
> +
> +	mutex_destroy(&isp->lock);
> +	c3_isp_videos_unregister(isp);
> +	c3_isp_async_nf_unregister(isp);
> +	c3_isp_core_unregister(isp);
> +	c3_isp_resizers_unregister(isp);
> +	c3_isp_v4l2_unregister(isp);
> +	pm_runtime_disable(isp->dev);
> +};
> +
> +static const struct c3_isp_info isp_info = {
> +	.clocks = {"vapb", "isp0"},
> +	.clock_rates = {0, 400000000},
> +	.clock_num = 2
> +};
> +
> +static const struct of_device_id c3_isp_of_match[] = {
> +	{ .compatible = "amlogic,c3-isp",
> +	  .data = &isp_info },
> +	{ },
> +};
> +MODULE_DEVICE_TABLE(of, c3_isp_of_match);
> +
> +static struct platform_driver c3_isp_driver = {
> +	.probe = c3_isp_probe,
> +	.remove = c3_isp_remove,
> +	.driver = {
> +		.name = "c3-isp",
> +		.of_match_table = c3_isp_of_match,
> +		.pm = &c3_isp_pm_ops,
> +	},
> +};
> +
> +module_platform_driver(c3_isp_driver);
> +
> +MODULE_AUTHOR("Keke Li <keke.li@xxxxxxxxxxx>");
> +MODULE_DESCRIPTION("Amlogic C3 ISP pipeline");
> +MODULE_LICENSE("GPL");
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-params.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-params.c
> new file mode 100644
> index 000000000000..8a6b7ce86eaf
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-params.c
> @@ -0,0 +1,857 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/pm_runtime.h>
> +
> +#include <media/v4l2-ctrls.h>
> +#include <media/v4l2-event.h>

Do you need these ?

> +#include <media/v4l2-ioctl.h>
> +#include <media/v4l2-mc.h>
> +#include <media/videobuf2-dma-contig.h>

See below about using the dma-contig vb2 ops

> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +#include "include/uapi/c3-isp-config.h"
> +
> +typedef void (*block_handler)(struct c3_isp_device *isp,
> +			      struct c3_isp_param_block_header *block);
> +
> +struct c3_isp_block_handler {
> +	size_t size;
> +	block_handler handler;
> +};
> +
> +/* Hardware configuration */
> +
> +static void c3_isp_params_cfg_wb_change(struct c3_isp_device *isp,
> +					struct c3_isp_param_block_header *block)
> +{
> +	struct wb_change_cfg *wb = (struct wb_change_cfg *)block;
> +
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BEO_CTRL,
> +				   TOP_BEO_CTRL_WB_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BEO_CTRL,
> +			   TOP_BEO_CTRL_WB_EN, TOP_BEO_CTRL_WB_EN);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_GAIN0, LSWB_WB_GAIN0_MASK,
> +			   wb->wb_gain[0] << LSWB_WB_GAIN0_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_GAIN0,
> +			   LSWB_WB_GAIN1_MASK, wb->wb_gain[1]);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_GAIN1, LSWB_WB_GAIN2_MASK,
> +			   wb->wb_gain[2] << LSWB_WB_GAIN2_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_GAIN1,
> +			   LSWB_WB_GAIN3_MASK, wb->wb_gain[3]);
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_GAIN2,
> +			   LSWB_WB_GAIN4_MASK, wb->wb_gain[4]);
> +
> +	c3_isp_write(isp, ISP_LSWB_WB_LIMIT0,
> +		     LSWB_WB_LIMIT0(wb->wb_limit[0]) | LSWB_WB_LIMIT1(wb->wb_limit[1]));
> +
> +	c3_isp_write(isp, ISP_LSWB_WB_LIMIT1,
> +		     LSWB_WB_LIMIT2(wb->wb_limit[2]) | LSWB_WB_LIMIT3(wb->wb_limit[3]));
> +
> +	c3_isp_update_bits(isp, ISP_LSWB_WB_LIMIT2, LSWB_WB_LIMIT4_MASK,
> +			   wb->wb_limit[4]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_0, AE_CRTL2_BL12_GRBGI0_MASK,
> +			   wb->ae_bl12_grbgi[0]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_0, AE_CRTL2_GAIN_GRBGI0_MASK,
> +			   wb->ae_gain_grbgi[0] << AE_CRTL2_GAIN_GRBGI0_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_1, AE_CRTL2_BL12_GRBGI1_MASK,
> +			   wb->ae_bl12_grbgi[1]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_1, AE_CRTL2_GAIN_GRBGI1_MASK,
> +			   wb->ae_gain_grbgi[1] << AE_CRTL2_GAIN_GRBGI1_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_2, AE_CRTL2_BL12_GRBGI2_MASK,
> +			   wb->ae_bl12_grbgi[2]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_2, AE_CRTL2_GAIN_GRBGI2_MASK,
> +			   wb->ae_gain_grbgi[2] << AE_CRTL2_GAIN_GRBGI2_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_3, AE_CRTL2_BL12_GRBGI3_MASK,
> +			   wb->ae_bl12_grbgi[3]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_3, AE_CRTL2_GAIN_GRBGI3_MASK,
> +			   wb->ae_gain_grbgi[3] << AE_CRTL2_GAIN_GRBGI3_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_4, AE_CRTL2_BL12_GRBGI4_MASK,
> +			   wb->ae_bl12_grbgi[4]);
> +
> +	c3_isp_update_bits(isp, ISP_AE_CRTL2_4, AE_CRTL2_GAIN_GRBGI4_MASK,
> +			   wb->ae_gain_grbgi[4] << AE_CRTL2_GAIN_GRBGI4_SHIFT);
> +}
> +
> +static void c3_isp_params_cfg_wb_luma(struct c3_isp_device *isp,
> +				      struct c3_isp_param_block_header *block)
> +{
> +	struct wb_luma_cfg *wb_luma = (struct wb_luma_cfg *)block;
> +
> +	if (!block->enabled)
> +		return;
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BLC20_0,
> +			   AWB_STAT_BLC20_GR_MASK, wb_luma->awb_stat_blc20[0]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BLC20_1,
> +			   AWB_STAT_BLC20_R_MASK, wb_luma->awb_stat_blc20[1]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BLC20_2,
> +			   AWB_STAT_BLC20_B_MASK, wb_luma->awb_stat_blc20[2]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BLC20_3,
> +			   AWB_STAT_BLC20_GB_MASK, wb_luma->awb_stat_blc20[3]);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_GAIN10_0,
> +			   AWB_STAT_GAIN10_GR_MASK, wb_luma->awb_stat_gain10[0]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_GAIN10_1,
> +			   AWB_STAT_GAIN10_R_MASK, wb_luma->awb_stat_gain10[1]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_GAIN10_2,
> +			   AWB_STAT_GAIN10_B_MASK, wb_luma->awb_stat_gain10[2]);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_GAIN10_3,
> +			   AWB_STAT_GAIN10_GB_MASK, wb_luma->awb_stat_gain10[3]);
> +
> +	c3_isp_write(isp, ISP_AWB_STAT_SATUR_CTRL,
> +		     AWB_STAT_SATUR_LOW(wb_luma->awb_stat_satur_low) |
> +		     AWB_STAT_SATUR_HIGH(wb_luma->awb_stat_satur_high));
> +}
> +
> +static void c3_isp_params_cfg_wb_triangle(struct c3_isp_device *isp,
> +					  struct c3_isp_param_block_header *block)
> +{
> +	struct wb_triangle_cfg *wb = (struct wb_triangle_cfg *)block;
> +
> +	if (!block->enabled)
> +		return;
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_CTRL2, AWB_STAT_SATUR_VALID_MASK,
> +			   wb->awb_stat_satur_vald);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_RG, AWB_STAT_RG_MIN_MASK,
> +			   wb->awb_stat_rg_min);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_RG, AWB_STAT_RG_MAX_MASK,
> +			   wb->awb_stat_rg_max << AWB_STAT_RG_MAX_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BG, AWB_STAT_BG_MIN_MASK,
> +			   wb->awb_stat_bg_min);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BG, AWB_STAT_BG_MAX_MASK,
> +			   wb->awb_stat_bg_max << AWB_STAT_BG_MAX_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_RG_HL, AWB_STAT_RG_LOW_MASK,
> +			   wb->awb_stat_rg_low);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_RG_HL, AWB_STAT_RG_HIGH_MASK,
> +			   wb->awb_stat_rg_high << AWB_STAT_RG_HIGH_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BG_HL, AWB_STAT_BG_LOW_MASK,
> +			   wb->awb_stat_bg_low);
> +	c3_isp_update_bits(isp, ISP_AWB_STAT_BG_HL, AWB_STAT_BG_HIGH_MASK,
> +			   wb->awb_stat_bg_high << AWB_STAT_BG_HIGH_SHIFT);
> +}
> +
> +static void c3_isp_params_cfg_awb_stats(struct c3_isp_device *isp,
> +					struct c3_isp_param_block_header *block)
> +{
> +	struct awb_stats_cfg *awb_stats = (struct awb_stats_cfg *)block;
> +	u32 *weight = awb_stats->awb_stat_blk_weight;
> +	int idx_base;
> +	int group;
> +	int i;
> +
> +	if (!block->enabled)
> +		return;
> +
> +	c3_isp_update_bits(isp, ISP_TOP_3A_STAT_CRTL, TOP_3A_AWB_STAT_SWITCH_MASK,
> +			   awb_stats->awb_stat_switch << TOP_3A_AWB_STAT_SWITCH_SHIFT);
> +
> +	/* Calculate the group number */
> +	group = AWB_BLOCK_WT_NUM / AWB_BLK_WT_DATA_NUM_OF_GROUP;
> +
> +	/* Set the weight address to 0 position */
> +	c3_isp_write(isp, ISP_AWB_BLK_WT_ADDR, 0);
> +	for (i = 0; i < group; i++) {

you can now use
        for (unsigned int i = 0; ...)

if 'i' is not needed outside of the loop

> +		idx_base = i * AWB_BLK_WT_DATA_NUM_OF_GROUP;
> +		c3_isp_write(isp, ISP_AWB_BLK_WT_DATA,
> +			     AWB_BLK_WT_DATA0(weight[idx_base + 0]) |
> +			     AWB_BLK_WT_DATA1(weight[idx_base + 1]) |
> +			     AWB_BLK_WT_DATA2(weight[idx_base + 2]) |
> +			     AWB_BLK_WT_DATA3(weight[idx_base + 3]) |
> +			     AWB_BLK_WT_DATA4(weight[idx_base + 4]) |
> +			     AWB_BLK_WT_DATA5(weight[idx_base + 5]) |
> +			     AWB_BLK_WT_DATA6(weight[idx_base + 6]) |
> +			     AWB_BLK_WT_DATA7(weight[idx_base + 7]));
> +	}
> +}
> +
> +static void c3_isp_params_cfg_ae_stats(struct c3_isp_device *isp,
> +				       struct c3_isp_param_block_header *block)
> +{
> +	struct ae_stats_cfg *ae_stats = (struct ae_stats_cfg *)block;
> +	u32 *weight = ae_stats->ae_stat_blk_weight;
> +	int idx_base;
> +	int group;
> +	int i;
> +
> +	if (!block->enabled)
> +		return;
> +
> +	c3_isp_update_bits(isp, ISP_TOP_3A_STAT_CRTL, TOP_3A_AE_STAT_SWITCH_MASK,
> +			   ae_stats->ae_stat_switch << TOP_3A_AE_STAT_SWITCH_SHIFT);
> +
> +	/* Calculate the group number */
> +	group = AE_BLOCK_WT_NUM / AE_BLK_WT_DATA_NUM_OF_GROUP;
> +
> +	/* Set the weight address to 0 position */
> +	c3_isp_write(isp, ISP_AE_BLK_WT_ADDR, 0);
> +	for (i = 0; i < group; i++) {
> +		idx_base = i * AE_BLK_WT_DATA_NUM_OF_GROUP;
> +		c3_isp_write(isp, ISP_AE_BLK_WT_DATA,
> +			     AE_BLK_WT_DATA0(weight[idx_base + 0]) |
> +			     AE_BLK_WT_DATA1(weight[idx_base + 1]) |
> +			     AE_BLK_WT_DATA2(weight[idx_base + 2]) |
> +			     AE_BLK_WT_DATA3(weight[idx_base + 3]) |
> +			     AE_BLK_WT_DATA4(weight[idx_base + 4]) |
> +			     AE_BLK_WT_DATA5(weight[idx_base + 5]) |
> +			     AE_BLK_WT_DATA6(weight[idx_base + 6]) |
> +			     AE_BLK_WT_DATA7(weight[idx_base + 7]));
> +	}
> +
> +	/* Write the last weight data */
> +	idx_base = i * AE_BLK_WT_DATA_NUM_OF_GROUP;
> +	c3_isp_write(isp, ISP_AE_BLK_WT_DATA,
> +		     AE_BLK_WT_DATA0(weight[idx_base + 0]) |
> +		     AE_BLK_WT_DATA1(weight[idx_base + 1]) |
> +		     AE_BLK_WT_DATA2(weight[idx_base + 2]) |
> +		     AE_BLK_WT_DATA3(weight[idx_base + 3]) |
> +		     AE_BLK_WT_DATA4(weight[idx_base + 4]) |
> +		     AE_BLK_WT_DATA5(weight[idx_base + 5]) |
> +		     AE_BLK_WT_DATA6(weight[idx_base + 6]));
> +}
> +
> +static void c3_isp_params_cfg_af_stats(struct c3_isp_device *isp,
> +				       struct c3_isp_param_block_header *block)
> +{
> +	struct af_stats_cfg *af_stats = (struct af_stats_cfg *)block;
> +
> +	if (!block->enabled)
> +		return;
> +
> +	c3_isp_update_bits(isp, ISP_TOP_3A_STAT_CRTL, TOP_3A_AF_STAT_SWITCH_MASK,
> +			   af_stats->af_stat_switch << TOP_3A_AF_STAT_SWITCH_SHIFT);
> +}
> +
> +static void c3_isp_params_cfg_pst_gamma(struct c3_isp_device *isp,
> +					struct c3_isp_param_block_header *block)
> +{
> +	struct pst_gamma_cfg *gamma = (struct pst_gamma_cfg *)block;
> +	int idx_base;
> +	int i, j;
> +
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_GAMMA_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_GAMMA_EN, TOP_BED_GAMMA_EN);
> +
> +	for (j = 0; j < GAMMA_LUT_GROUP_NUM; j++) {
> +		c3_isp_write(isp, ISP_PST_GAMMA_LUT_ADDR, PST_GAMMA_LUT_ADDR(j));
> +
> +		/* Calculate the block number */
> +		for (i = 0; i < GAMMA_LUT_POINT_NUM / GAMMA_LUT_DATA_NUM_OF_GROUP; i++) {
> +			idx_base = i * GAMMA_LUT_DATA_NUM_OF_GROUP;
> +			c3_isp_write(isp, ISP_PST_GAMMA_LUT_DATA,
> +				     PST_GAMMA_LUT_DATA0(gamma->pst_gamma_lut[j][idx_base]) |
> +				     PST_GAMMA_LUT_DATA1(gamma->pst_gamma_lut[j][idx_base + 1]));
> +		}
> +
> +		/* Write the last one lut data of group j */
> +		idx_base = i * GAMMA_LUT_DATA_NUM_OF_GROUP;
> +		c3_isp_write(isp, ISP_PST_GAMMA_LUT_DATA,
> +			     PST_GAMMA_LUT_DATA0(gamma->pst_gamma_lut[j][idx_base]));
> +	}
> +}
> +
> +static void c3_isp_params_cfg_dmsc(struct c3_isp_device *isp,
> +				   struct c3_isp_param_block_header *block)
> +{
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_DMSC_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_DMSC_EN, TOP_BED_DMSC_EN);
> +}
> +
> +/* Configure 4 x 3 ccm matrix */
> +static void c3_isp_params_cfg_ccm(struct c3_isp_device *isp,
> +				  struct c3_isp_param_block_header *block)
> +{
> +	struct ccm_cfg *ccm = (struct ccm_cfg *)block;
> +
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_CCM_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_CCM_EN, TOP_BED_CCM_EN);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_00_01, CCM_MTX_00_MASK,
> +			   ccm->ccm_4x3matrix[0][0]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_00_01, CCM_MTX_01_MASK,
> +			   ccm->ccm_4x3matrix[0][1] << CCM_MTX_01_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_02_03, CCM_MTX_02_MASK,
> +			   ccm->ccm_4x3matrix[0][2]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_02_03, CCM_MTX_03_MASK,
> +			   ccm->ccm_4x3matrix[0][3] << CCM_MTX_03_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_10_11, CCM_MTX_10_MASK,
> +			   ccm->ccm_4x3matrix[1][0]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_10_11, CCM_MTX_11_MASK,
> +			   ccm->ccm_4x3matrix[1][1] << CCM_MTX_11_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_12_13, CCM_MTX_12_MASK,
> +			   ccm->ccm_4x3matrix[1][2]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_12_13, CCM_MTX_13_MASK,
> +			   ccm->ccm_4x3matrix[1][3] << CCM_MTX_13_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_20_21, CCM_MTX_20_MASK,
> +			   ccm->ccm_4x3matrix[2][0]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_20_21, CCM_MTX_21_MASK,
> +			   ccm->ccm_4x3matrix[2][1] << CCM_MTX_21_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_22_23_RS, CCM_MTX_22_MASK,
> +			   ccm->ccm_4x3matrix[2][2]);
> +	c3_isp_update_bits(isp, ISP_CCM_MTX_22_23_RS, CCM_MTX_23_MASK,
> +			   ccm->ccm_4x3matrix[2][3] << CCM_MTX_23_SHIFT);
> +}
> +
> +/* Configure color space conversion matrix parameters */
> +static void c3_isp_params_cfg_csc(struct c3_isp_device *isp,
> +				  struct c3_isp_param_block_header *block)
> +{
> +	struct csc_cfg *csc = (struct csc_cfg *)block;
> +
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_CM0_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BED_CTRL, TOP_BED_CM0_EN, TOP_BED_CM0_EN);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_INP_OFST01, CM0_INP_OFST0_MASK,
> +			   csc->cm0_offset_inp[0]);
> +	c3_isp_update_bits(isp, ISP_CM0_INP_OFST01, CM0_INP_OFST1_MASK,
> +			   csc->cm0_offset_inp[1] << CM0_INP_OFST1_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_INP_OFST2, CM0_INP_OFST2_MASK,
> +			   csc->cm0_offset_inp[2]);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_COEF00_01, CM0_MTX_00_MASK,
> +			   csc->cm0_3x3matrix[0][0]);
> +	c3_isp_update_bits(isp, ISP_CM0_COEF00_01, CM0_MTX_01_MASK,
> +			   csc->cm0_3x3matrix[0][1] << CM0_MTX_01_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_COEF02_10, CM0_MTX_02_MASK,
> +			   csc->cm0_3x3matrix[0][2]);
> +	c3_isp_update_bits(isp, ISP_CM0_COEF02_10, CM0_MTX_10_MASK,
> +			   csc->cm0_3x3matrix[1][0] << CM0_MTX_10_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_COEF11_12, CM0_MTX_11_MASK,
> +			   csc->cm0_3x3matrix[1][1]);
> +	c3_isp_update_bits(isp, ISP_CM0_COEF11_12, CM0_MTX_12_MASK,
> +			   csc->cm0_3x3matrix[1][2] << CM0_MTX_12_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_COEF20_21, CM0_MTX_20_MASK,
> +			   csc->cm0_3x3matrix[2][0]);
> +	c3_isp_update_bits(isp, ISP_CM0_COEF20_21, CM0_MTX_21_MASK,
> +			   csc->cm0_3x3matrix[2][1] << CM0_MTX_21_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_COEF22_OUP_OFST0, CM0_MTX_22_MASK,
> +			   csc->cm0_3x3matrix[2][2]);
> +	c3_isp_update_bits(isp, ISP_CM0_COEF22_OUP_OFST0, CM0_OFST_OUP0_MASK,
> +			   csc->cm0_offset_oup[0] << CM0_OFST_OUP0_SHIFT);
> +
> +	c3_isp_update_bits(isp, ISP_CM0_OUP_OFST12_RS, CM0_OFST_OUP1_MASK,
> +			   csc->cm0_offset_oup[1]);
> +	c3_isp_update_bits(isp, ISP_CM0_OUP_OFST12_RS, CM0_OFST_OUP2_MASK,
> +			   csc->cm0_offset_oup[2] << CM0_OFST_OUP2_SHIFT);
> +	c3_isp_update_bits(isp, ISP_CM0_OUP_OFST12_RS, CM0_MTX_RS_MASK,
> +			   csc->cm0_3x3mtrx_rs << CM0_MTX_RS_SHIFT);
> +}
> +
> +/* Set blc offset of each color channel */
> +static void c3_isp_params_cfg_blc(struct c3_isp_device *isp,
> +				  struct c3_isp_param_block_header *block)
> +{
> +	struct blc_cfg *blc = (struct blc_cfg *)block;
> +
> +	if (!block->enabled) {
> +		c3_isp_update_bits(isp, ISP_TOP_BEO_CTRL, TOP_BEO_CTRL_BLC_EN, false);
> +		return;
> +	}
> +
> +	c3_isp_update_bits(isp, ISP_TOP_BEO_CTRL, TOP_BEO_CTRL_BLC_EN,
> +			   TOP_BEO_CTRL_BLC_EN);
> +
> +	c3_isp_write(isp, ISP_FED_BL_OFST_GR, FED_BL_GR_OFST(blc->fe_bl_ofst[0]));
> +	c3_isp_write(isp, ISP_FED_BL_OFST_R, FED_BL_R_OFST(blc->fe_bl_ofst[1]));
> +	c3_isp_write(isp, ISP_FED_BL_OFST_B, FED_BL_B_OFST(blc->fe_bl_ofst[2]));
> +	c3_isp_write(isp, ISP_FED_BL_OFST_GB, FED_BL_GB_OFST(blc->fe_bl_ofst[3]));
> +	c3_isp_write(isp, ISP_FED_BL_OFST_IR, FED_BL_IR_OFST(blc->fe_bl_ofst[4]));
> +
> +	c3_isp_write(isp, ISP_LSWB_BLC_OFST0,
> +		     LSWB_BLC_OFST0(blc->blc_ofst[0]) | LSWB_BLC_OFST1(blc->blc_ofst[1]));
> +	c3_isp_write(isp, ISP_LSWB_BLC_OFST1,
> +		     LSWB_BLC_OFST2(blc->blc_ofst[2]) | LSWB_BLC_OFST3(blc->blc_ofst[3]));
> +	c3_isp_write(isp, ISP_LSWB_BLC_OFST2, LSWB_BLC_OFST4(blc->blc_ofst[4]));
> +}
> +
> +static const struct c3_isp_block_handler c3_isp_block_handlers[] = {
> +	[C3_ISP_PARAM_BLOCK_WB_CHANGE] = {
> +		.size = sizeof(struct wb_change_cfg),
> +		.handler = c3_isp_params_cfg_wb_change,
> +	},
> +	[C3_ISP_PARAM_BLOCK_WB_LUMA] = {
> +		.size = sizeof(struct wb_luma_cfg),
> +		.handler = c3_isp_params_cfg_wb_luma,
> +	},
> +	[C3_ISP_PARAM_BLOCK_WB_TRIANGLE] = {
> +		.size = sizeof(struct wb_triangle_cfg),
> +		.handler = c3_isp_params_cfg_wb_triangle,
> +	},
> +	[C3_ISP_PARAM_BLOCK_AWB_STATS] = {
> +		.size = sizeof(struct awb_stats_cfg),
> +		.handler = c3_isp_params_cfg_awb_stats,
> +	},
> +	[C3_ISP_PARAM_BLOCK_AE_STATS] = {
> +		.size = sizeof(struct ae_stats_cfg),
> +		.handler = c3_isp_params_cfg_ae_stats,
> +	},
> +	[C3_ISP_PARAM_BLOCK_AF_STATS] = {
> +		.size = sizeof(struct af_stats_cfg),
> +		.handler = c3_isp_params_cfg_af_stats,
> +	},
> +	[C3_ISP_PARAM_BLOCK_PST_GAMMA] = {
> +		.size = sizeof(struct pst_gamma_cfg),
> +		.handler = c3_isp_params_cfg_pst_gamma,
> +	},
> +	[C3_ISP_PARAM_BLOCK_DMSC] = {
> +		.size = sizeof(struct dmsc_cfg),
> +		.handler = c3_isp_params_cfg_dmsc,
> +	},
> +	[C3_ISP_PARAM_BLOCK_CCM] = {
> +		.size = sizeof(struct ccm_cfg),
> +		.handler = c3_isp_params_cfg_ccm,
> +	},
> +	[C3_ISP_PARAM_BLOCK_CSC] = {
> +		.size = sizeof(struct csc_cfg),
> +		.handler = c3_isp_params_cfg_csc,
> +	},
> +	[C3_ISP_PARAM_BLOCK_BLC] = {
> +		.size = sizeof(struct blc_cfg),
> +		.handler = c3_isp_params_cfg_blc,
> +	},
> +};
> +
> +static enum vb2_buffer_state
> +c3_isp_params_cfg_blocks(struct c3_isp_params *params)
> +{
> +	struct c3_isp_params_buffer *config = params->buff->vaddr;
> +	enum vb2_buffer_state state = VB2_BUF_STATE_DONE;
> +	size_t block_offset = 0;
> +	size_t max_offset = 0;
> +
> +	if (config->total_size > C3_ISP_PARAMS_MAX_SIZE) {
> +		dev_dbg(params->isp->dev, "Invalid parameters buffer size %lu\n",
> +			config->total_size);
> +		state = VB2_BUF_STATE_ERROR;
> +		goto err_return_state;
> +	}

I suggest to move validation of the parameters buffer to .buf_prepare
time.

This function is called in irq context, and it's better to do all
validation check at buffer queuing time instead.

You can have a look at the rkisp1-params.c module, where
rkisp1_params_prepare_ext_params does the validation in the
.buf_prepare call path

> +
> +	/* Ensure config->data has a full struct c3_isp_param_block_header */
> +	max_offset = config->total_size - sizeof(struct c3_isp_param_block_header);
> +
> +	while (block_offset <= max_offset) {
> +		const struct c3_isp_block_handler *block_handler;
> +		struct c3_isp_param_block_header *block;
> +
> +		block = (struct c3_isp_param_block_header *)
> +			 &config->data[block_offset];
> +
> +		if (block->type >= C3_ISP_PARAM_BLOCK_SENTINEL) {
> +			dev_dbg(params->isp->dev, "Invalid parameters block type\n");
> +			state = VB2_BUF_STATE_ERROR;
> +			goto err_return_state;
> +		}
> +
> +		block_handler = &c3_isp_block_handlers[block->type];
> +		if (block->size != block_handler->size) {
> +			dev_dbg(params->isp->dev, "Invalid parameters block size\n");
> +			state = VB2_BUF_STATE_ERROR;
> +			goto err_return_state;
> +		}
> +
> +		block_handler->handler(params->isp, block);
> +
> +		block_offset += block->size;
> +	}
> +
> +err_return_state:
> +	return state;
> +}
> +
> +/* Initialize ISP pipeline */
> +static int c3_isp_params_start(struct c3_isp_params *params)
> +{
> +	enum vb2_buffer_state state;
> +	unsigned long flags;
> +
> +	/* Reset these controllers */
> +	c3_isp_write(params->isp, ISP_TOP_FEO_CTRL0, TOP_FEO_CTRL0_ALL_DIS);
> +	c3_isp_write(params->isp, ISP_TOP_FEO_CTRL1_0, TOP_FEO_CTRL1_0_ALL_DIS);
> +	c3_isp_write(params->isp, ISP_TOP_FEO_CTRL1_1, TOP_FEO_CTRL1_1_ALL_DIS);
> +	c3_isp_write(params->isp, ISP_TOP_FED_CTRL, TOP_FED_CTRL_ALL_DIS);
> +	c3_isp_write(params->isp, ISP_TOP_BEO_CTRL, TOP_BEO_CTRL_ALL_DIS);
> +	c3_isp_write(params->isp, ISP_TOP_BED_CTRL, TOP_BED_CTRL_ALL_DIS);
> +
> +	spin_lock_irqsave(&params->buff_lock, flags);
> +
> +	/* Only use the first buffer to initialize ISP */
> +	params->buff = list_first_entry_or_null(&params->pending,
> +						struct c3_isp_vb2_buffer, list);
> +	if (!params->buff) {
> +		spin_unlock_irqrestore(&params->buff_lock, flags);
> +		return -EINVAL;
> +	}
> +
> +	state = c3_isp_params_cfg_blocks(params);
> +
> +	spin_unlock_irqrestore(&params->buff_lock, flags);
> +
> +	return 0;
> +}
> +
> +/* V4L2 video operations */
> +
> +static void c3_isp_params_return_buffers(struct c3_isp_params *params,
> +					 enum vb2_buffer_state state)
> +{
> +	unsigned long flags;
> +	struct c3_isp_vb2_buffer *buff;
> +
> +	spin_lock_irqsave(&params->buff_lock, flags);
> +
> +	while (!list_empty(&params->pending)) {
> +		buff = list_first_entry(&params->pending,
> +					struct c3_isp_vb2_buffer, list);
> +		list_del(&buff->list);
> +		vb2_buffer_done(&buff->vb.vb2_buf, state);
> +	}
> +
> +	spin_unlock_irqrestore(&params->buff_lock, flags);
> +}
> +
> +static int c3_isp_params_querycap(struct file *file, void *fh,
> +				  struct v4l2_capability *cap)
> +{
> +	strscpy(cap->driver, C3_ISP_DRIVER_NAME, sizeof(cap->driver));
> +	strscpy(cap->card, "AML C3 ISP", sizeof(cap->card));
> +
> +	return 0;
> +}
> +
> +static int c3_isp_params_enum_fmt(struct file *file, void *fh,
> +				  struct v4l2_fmtdesc *f)
> +{
> +	if (f->index)
> +		return -EINVAL;
> +
> +	f->pixelformat = V4L2_META_FMT_C3ISP_PARAMS;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_params_g_fmt(struct file *file, void *fh,
> +			       struct v4l2_format *f)
> +{
> +	struct c3_isp_params *params = video_drvdata(file);
> +
> +	f->fmt.meta = params->vfmt.fmt.meta;
> +
> +	return 0;
> +}
> +
> +static const struct v4l2_ioctl_ops isp_params_v4l2_ioctl_ops = {
> +	.vidioc_querycap                = c3_isp_params_querycap,
> +	.vidioc_enum_fmt_meta_out       = c3_isp_params_enum_fmt,
> +	.vidioc_g_fmt_meta_out          = c3_isp_params_g_fmt,
> +	.vidioc_s_fmt_meta_out          = c3_isp_params_g_fmt,
> +	.vidioc_try_fmt_meta_out        = c3_isp_params_g_fmt,
> +	.vidioc_reqbufs                 = vb2_ioctl_reqbufs,
> +	.vidioc_querybuf                = vb2_ioctl_querybuf,
> +	.vidioc_qbuf                    = vb2_ioctl_qbuf,
> +	.vidioc_expbuf                  = vb2_ioctl_expbuf,
> +	.vidioc_dqbuf                   = vb2_ioctl_dqbuf,
> +	.vidioc_prepare_buf             = vb2_ioctl_prepare_buf,
> +	.vidioc_create_bufs             = vb2_ioctl_create_bufs,
> +	.vidioc_streamon                = vb2_ioctl_streamon,
> +	.vidioc_streamoff               = vb2_ioctl_streamoff,
> +	.vidioc_subscribe_event         = v4l2_ctrl_subscribe_event,
> +	.vidioc_unsubscribe_event       = v4l2_event_unsubscribe,
> +};
> +
> +static const struct v4l2_file_operations isp_params_v4l2_fops = {
> +	.open = v4l2_fh_open,
> +	.release = vb2_fop_release,
> +	.poll = vb2_fop_poll,
> +	.unlocked_ioctl = video_ioctl2,
> +	.mmap = vb2_fop_mmap,
> +};
> +
> +static int c3_isp_params_vb2_queue_setup(struct vb2_queue *q,
> +					 unsigned int *num_buffers,
> +					 unsigned int *num_planes,
> +					 unsigned int sizes[],
> +					 struct device *alloc_devs[])
> +{
> +	if (*num_planes) {
> +		if (*num_planes != 1)
> +			return -EINVAL;
> +
> +		if (sizes[0] < sizeof(struct c3_isp_params_buffer))
> +			return -EINVAL;
> +	} else {
> +		*num_planes = 1;
> +		sizes[0] = sizeof(struct c3_isp_params_buffer);
> +	}
> +
> +	return 0;
> +}
> +
> +static void c3_isp_params_vb2_buf_queue(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +			container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_params *params = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&params->buff_lock, flags);
> +
> +	list_add_tail(&buf->list, &params->pending);
> +
> +	spin_unlock_irqrestore(&params->buff_lock, flags);
> +}
> +
> +static int c3_isp_params_vb2_buf_prepare(struct vb2_buffer *vb)
> +{
> +	struct c3_isp_params *params = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned int size = params->vfmt.fmt.meta.buffersize;
> +
> +	if (vb2_plane_size(vb, 0) < size) {

How does this work ?

'size' is
	params->vfmt.fmt.meta.buffersize = sizeof(struct c3_isp_params_buffer);

if I got this right

Now, as you're using an extensible parameters implementation,
userspace is allowed to submit buffers of a smaller size, with only
the "interesting" blocks there.

This check instead makes sure that userspace always fill the
paramteres buffer with all blocks, am I wrong ?

This defeates the purpose of extensible formats, where instead
userspace is allowed to only fill the buffers with a subset of the
configuration blocks.

Have I missed something ?

> +		dev_err(params->isp->dev,
> +			"User buffer too small (%ld < %u)\n",
> +			vb2_plane_size(vb, 0), size);
> +		return -EINVAL;
> +	}
> +
> +	vb2_set_plane_payload(vb, 0, size);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_params_vb2_buf_init(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +		container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_params *params = vb2_get_drv_priv(vb->vb2_queue);
> +
> +	buf->vaddr = vb2_plane_vaddr(vb, 0);
> +	buf->paddr = vb2_dma_contig_plane_dma_addr(vb, 0);

This is not used

> +
> +	memset(buf->vaddr, 0, params->vfmt.fmt.meta.buffersize);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_params_vb2_start_streaming(struct vb2_queue *q,
> +					     unsigned int count)
> +{
> +	struct c3_isp_params *params = vb2_get_drv_priv(q);
> +	int ret;
> +
> +	guard(mutex)(&params->isp->lock);
> +
> +	ret = pm_runtime_resume_and_get(params->isp->dev);
> +	if (ret)
> +		return ret;
> +
> +	ret = video_device_pipeline_start(&params->vdev, &params->isp->pipe);
> +	if (ret) {
> +		dev_err(params->isp->dev,
> +			"Failed to start params pipeline: %d\n", ret);
> +		goto err_pm_put;
> +	}
> +
> +	if (c3_isp_pipeline_ready(params->isp)) {

I understand this counts how many video devices have been started...


> +		ret = v4l2_subdev_enable_streams(&params->isp->core.sd,
> +						 C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +						 BIT(0));

... but does it need to be called by this module ? it's not like there is a
parameters stream to be enabled in the ISP (see also the suggestion to
drop routing support from there).


> +		if (ret)
> +			goto err_pipeline_stop;
> +	}
> +
> +	c3_isp_params_start(params);

Does this perform initialization and apply the first parameters from a
queued buffer ?

I'm wondering if it wouldn't be better to:

- Implement a start_streaming operation on capture nodes only
- Call the ISP's enable_streams
- The ISP driver counts how many enabled links to resizers are there
- If the number of enable_streams calls matches the number of enabled
  links:
  - call c3_isp_params_start() and any function that does the stats or
    params setup
  - actually start the ISP
  - propagate the enable_streams call to the downstream subdevs (Adap,
    then csi2-rx then sensor)

> +
> +	return 0;
> +
> +err_pipeline_stop:
> +	video_device_pipeline_stop(&params->vdev);
> +err_pm_put:
> +	pm_runtime_put(params->isp->dev);
> +	c3_isp_params_return_buffers(params, VB2_BUF_STATE_QUEUED);
> +	return ret;
> +}
> +
> +static void c3_isp_params_vb2_stop_streaming(struct vb2_queue *q)
> +{
> +	struct c3_isp_params *params = vb2_get_drv_priv(q);
> +
> +	guard(mutex)(&params->isp->lock);
> +
> +	c3_isp_params_return_buffers(params, VB2_BUF_STATE_ERROR);

Even if you don't implement start_streaming as I've suggested, you
will need this one to return buffers to userspace.

> +
> +	if (params->isp->pipe.start_count == 1)
> +		v4l2_subdev_disable_streams(&params->isp->core.sd,
> +					    C3_ISP_CORE_PAD_SOURCE_VIDEO,
> +					    BIT(0));
> +
> +	video_device_pipeline_stop(&params->vdev);
> +	pm_runtime_put(params->isp->dev);
> +}
> +
> +static const struct vb2_ops isp_params_vb2_ops = {
> +	.queue_setup = c3_isp_params_vb2_queue_setup,
> +	.buf_queue = c3_isp_params_vb2_buf_queue,
> +	.buf_prepare = c3_isp_params_vb2_buf_prepare,
> +	.buf_init = c3_isp_params_vb2_buf_init,
> +	.wait_prepare = vb2_ops_wait_prepare,
> +	.wait_finish = vb2_ops_wait_finish,
> +	.start_streaming = c3_isp_params_vb2_start_streaming,
> +	.stop_streaming = c3_isp_params_vb2_stop_streaming,
> +};
> +
> +int c3_isp_params_register(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_params *params = &isp->params;
> +	struct video_device *vdev = &params->vdev;
> +	struct vb2_queue *vb2_q = &params->vb2_q;
> +	int ret;
> +
> +	memset(params, 0, sizeof(*params));
> +	params->vfmt.fmt.meta.dataformat = V4L2_META_FMT_C3ISP_PARAMS;
> +	params->vfmt.fmt.meta.buffersize = sizeof(struct c3_isp_params_buffer);
> +	params->isp = isp;
> +	INIT_LIST_HEAD(&params->pending);
> +	spin_lock_init(&params->buff_lock);
> +	mutex_init(&params->lock);
> +
> +	snprintf(vdev->name, sizeof(vdev->name), "isp-params");

Here and in all other names, I would prefix them with "c3-"

> +	vdev->fops = &isp_params_v4l2_fops;
> +	vdev->ioctl_ops = &isp_params_v4l2_ioctl_ops;
> +	vdev->v4l2_dev = &isp->v4l2_dev;
> +	vdev->lock = &params->lock;
> +	vdev->minor = -1;
> +	vdev->queue = vb2_q;
> +	vdev->release = video_device_release_empty;
> +	vdev->device_caps = V4L2_CAP_META_OUTPUT | V4L2_CAP_STREAMING;
> +	vdev->vfl_dir = VFL_DIR_TX;
> +	video_set_drvdata(vdev, params);
> +
> +	vb2_q->drv_priv = params;
> +	vb2_q->mem_ops = &vb2_dma_contig_memops;

Do you need to use DMA ?

This implementation seems to rather receive a buffer of parameters,
inspect its content and write registers according to values there.

I don't see direct DMA transfers to a memory mapped register area, but
instead single writes to registers.

Should you use vb2_vmalloc_memops instead ?
Make sure to change the headers inclusions and Kconfig dependencies
accordingly.

> +	vb2_q->ops = &isp_params_vb2_ops;
> +	vb2_q->type = V4L2_BUF_TYPE_META_OUTPUT;
> +	vb2_q->io_modes = VB2_DMABUF | VB2_MMAP;
> +	vb2_q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
> +	vb2_q->buf_struct_size = sizeof(struct c3_isp_vb2_buffer);
> +	vb2_q->dev = isp->dev;
> +	vb2_q->lock = &params->lock;
> +	vb2_q->min_queued_buffers = 1;
> +
> +	ret = vb2_queue_init(vb2_q);
> +	if (ret)
> +		goto err_detroy;
> +
> +	params->pad.flags = MEDIA_PAD_FL_SOURCE;
> +	ret = media_entity_pads_init(&vdev->entity, 1, &params->pad);
> +	if (ret)
> +		goto err_queue_release;
> +
> +	ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
> +	if (ret < 0) {
> +		dev_err(isp->dev,
> +			"Failed to register %s: %d\n", vdev->name, ret);
> +		goto err_entity_cleanup;
> +	}
> +
> +	return 0;
> +
> +err_entity_cleanup:
> +	media_entity_cleanup(&vdev->entity);
> +err_queue_release:
> +	vb2_queue_release(vb2_q);
> +err_detroy:
> +	mutex_destroy(&params->lock);
> +	return ret;
> +}
> +
> +void c3_isp_params_unregister(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_params *params = &isp->params;
> +
> +	vb2_queue_release(&params->vb2_q);
> +	media_entity_cleanup(&params->vdev.entity);
> +	video_unregister_device(&params->vdev);
> +	mutex_destroy(&params->lock);
> +}
> +
> +int c3_isp_params_done(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_params *params = &isp->params;
> +	enum vb2_buffer_state state;
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&params->buff_lock, flags);
> +
> +	params->buff = list_first_entry_or_null(&params->pending,
> +						struct c3_isp_vb2_buffer, list);
> +	if (!params->buff) {
> +		spin_unlock_irqrestore(&params->buff_lock, flags);
> +		return -EINVAL;
> +	}
> +
> +	list_del(&params->buff->list);
> +
> +	state = c3_isp_params_cfg_blocks(params);
> +
> +	params->buff->vb.sequence = params->isp->frm_sequence;
> +	params->buff->vb.vb2_buf.timestamp = ktime_get();
> +	params->buff->vb.field = V4L2_FIELD_NONE;
> +	vb2_buffer_done(&params->buff->vb.vb2_buf, state);
> +
> +	spin_unlock_irqrestore(&params->buff_lock, flags);
> +
> +	return 0;
> +}
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-regs.h b/drivers/media/platform/amlogic/c3-isp/c3-isp-regs.h
> new file mode 100644
> index 000000000000..de1938f7c354
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-regs.h
> @@ -0,0 +1,683 @@
> +/* SPDX-License-Identifier: (GPL-2.0-only OR MIT) */
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#ifndef __C3_ISP_REGS_H__
> +#define __C3_ISP_REGS_H__
> +
> +#define ISP_TOP_INPUT_SIZE                       0x0000
> +#define TOP_INPUT_HSIZE(x)                       (((x) & 0xffff) << 16)
> +#define TOP_INPUT_VSIZE(x)                       ((x) & 0xffff)
> +
> +#define ISP_TOP_FRM_SIZE                         0x0004
> +#define TOP_FRM_CORE_IHSIZE(x)                   (((x) & 0xffff) << 16)
> +#define TOP_FRM_CORE_IVSIZE(x)                   ((x) & 0xffff)
> +
> +#define ISP_TOP_HOLD_SIZE                        0x0008
> +#define TOP_HOLD_HSIZE_MASK                      GENMASK(31, 16)
> +#define TOP_HOLD_HSIZE_SHIFT                     16
> +
> +#define ISP_TOP_PATH_EN                          0x0010
> +#define TOP_DISP_EN(x)                           BIT((x))
> +#define TOP_WRMIF_EN(x)                          BIT((x) + 8)
> +
> +#define ISP_TOP_PATH_SEL                         0x0014
> +#define TOP_DATA_PATH_MASK                       GENMASK(18, 16)
> +#define TOP_PATH_MIPI_TO_CORE                    BIT(16)
> +
> +#define ISP_TOP_IRQ_EN                           0x0080
> +#define TOP_IRQ_FRAME_DONE                       BIT(0)
> +#define TOP_IRQ_STATS_ERR                        BIT(5)
> +
> +#define ISP_TOP_IRQ_CLR                          0x0084
> +#define ISP_TOP_IRQ_LINE_THRD                    0x0088
> +#define TOP_IRQ_DIN_HSYNC                        BIT(16)
> +
> +#define ISP_TOP_RO_IRQ_STAT                      0x01c4
> +#define ISP_TOP_MODE_CTRL                        0x0400
> +#define ISP_TOP_FEO_CTRL0                        0x040c
> +#define TOP_FEO_CTRL0_ALL_DIS                    0
> +#define TOP_FEO_INPUT_FMT_EN                     BIT(8)
> +
> +#define ISP_TOP_FEO_CTRL1_0                      0x0410
> +#define TOP_FEO_CTRL1_0_ALL_DIS                  0
> +
> +#define ISP_TOP_FEO_CTRL1_1                      0x0414
> +#define TOP_FEO_CTRL1_1_ALL_DIS                  0
> +
> +#define ISP_TOP_FED_CTRL                         0x0418
> +#define TOP_FED_CTRL_ALL_DIS                     0
> +
> +#define ISP_TOP_BEO_CTRL                         0x041c
> +#define TOP_BEO_CTRL_ALL_DIS                     0
> +#define TOP_BEO_CTRL_GTM_EN                      BIT(2)
> +#define TOP_BEO_CTRL_WB_EN                       BIT(6)
> +#define TOP_BEO_CTRL_BLC_EN                      BIT(7)
> +#define TOP_BEO_CTRL_IDGAIN_EN                   BIT(8)
> +#define TOP_BEO_CTRL_EOTF_EN                     BIT(9)
> +
> +#define ISP_TOP_BED_CTRL                         0x0420
> +#define TOP_BED_CTRL_ALL_DIS                     0
> +#define TOP_BED_CM0_EN                           BIT(14)
> +#define TOP_BED_GAMMA_EN                         BIT(16)
> +#define TOP_BED_CCM_EN                           BIT(18)
> +#define TOP_BED_DMSC_EN                          BIT(19)
> +
> +#define ISP_TOP_3A_STAT_CRTL                     0x0424
> +#define TOP_3A_AE_STAT_EN                        BIT(0)
> +#define TOP_3A_AWB_STAT_EN                       BIT(1)
> +#define TOP_3A_AF_STAT_EN                        BIT(2)
> +#define TOP_3A_AWB_STAT_SWITCH_MASK              GENMASK(6, 4)
> +#define TOP_3A_AWB_STAT_SWITCH_SHIFT             4
> +#define TOP_3A_AWB_STAT_SWITCH_BEFORE_WB         2
> +
> +#define TOP_3A_AE_STAT_SWITCH_MASK               GENMASK(9, 8)
> +#define TOP_3A_AE_STAT_SWITCH_SHIFT              8
> +#define TOP_3A_AE_STAT_SWITCH_FROM_MLS           1
> +#define TOP_3A_AF_STAT_SWITCH_MASK               GENMASK(13, 12)
> +#define TOP_3A_AF_STAT_SWITCH_SHIFT              12
> +#define TOP_3A_AF_STAT_SWITCH_FROM_SNR            0
> +
> +#define ISP_FED_BL_OFST_GR                       0x2018
> +#define FED_BL_GR_OFST(x)                        ((x) & 0x1fffff)
> +
> +#define ISP_FED_BL_OFST_R                        0x201c
> +#define FED_BL_R_OFST(x)                         ((x) & 0x1fffff)
> +
> +#define ISP_FED_BL_OFST_B                        0x2020
> +#define FED_BL_B_OFST(x)                         ((x) & 0x1fffff)
> +
> +#define ISP_FED_BL_OFST_GB                       0x2024
> +#define FED_BL_GB_OFST(x)                        ((x) & 0x1fffff)
> +
> +#define ISP_FED_BL_OFST_IR                       0x2028
> +#define FED_BL_IR_OFST(x)                        ((x) & 0x1fffff)
> +
> +#define ISP_LSWB_BLC_OFST0                       0x4028
> +#define LSWB_BLC_OFST1_MASK                      GENMASK(15, 0)
> +#define LSWB_BLC_OFST1(x)                        ((x) & 0xffff)
> +#define LSWB_BLC_OFST1_EXT(x)                    ((x) & 0xffff)
> +#define LSWB_BLC_OFST0_MASK                      GENMASK(31, 16)
> +#define LSWB_BLC_OFST0(x)                        (((x) & 0xffff) << 16)
> +#define LSWB_BLC_OFST0_EXT(x)                    (((x) >> 16) & 0xffff)
> +
> +#define ISP_LSWB_BLC_OFST1                       0x402c
> +#define LSWB_BLC_OFST3_MASK                      GENMASK(15, 0)
> +#define LSWB_BLC_OFST3(x)                        ((x) & 0xffff)
> +#define LSWB_BLC_OFST3_EXT(x)                    ((x) & 0xffff)
> +#define LSWB_BLC_OFST2_MASK                      GENMASK(31, 16)
> +#define LSWB_BLC_OFST2(x)                        (((x) & 0xffff) << 16)
> +#define LSWB_BLC_OFST2_EXT(x)                    (((x) >> 16) & 0xffff)
> +
> +#define ISP_LSWB_BLC_OFST2                       0x4030
> +#define LSWB_BLC_OFST4_MASK                      GENMASK(15, 0)
> +#define LSWB_BLC_OFST4(x)                        ((x) & 0xffff)
> +#define LSWB_BLC_OFST4_EXT(x)                    ((x) & 0xffff)
> +
> +#define ISP_LSWB_BLC_PHSOFST                     0x4034
> +#define LSWB_BLC_YPHS_OFST_MASK                  GENMASK(1, 0)
> +#define LSWB_BLC_XPHS_OFST_MASK                  GENMASK(3, 2)
> +#define LSWB_BLC_XPHS_OFST_SHIFT                 2
> +
> +#define ISP_LSWB_WB_GAIN0                        0x4038
> +#define LSWB_WB_GAIN1_MASK                       GENMASK(11, 0)
> +#define LSWB_WB_GAIN0_MASK                       GENMASK(27, 16)
> +#define LSWB_WB_GAIN0_SHIFT                      16
> +
> +#define ISP_LSWB_WB_GAIN1                        0x403c
> +#define LSWB_WB_GAIN3_MASK                       GENMASK(11, 0)
> +#define LSWB_WB_GAIN2_MASK                       GENMASK(27, 16)
> +#define LSWB_WB_GAIN2_SHIFT                      16
> +
> +#define ISP_LSWB_WB_GAIN2                        0x4040
> +#define LSWB_WB_GAIN4_MASK                       GENMASK(11, 0)
> +
> +#define ISP_LSWB_WB_LIMIT0                       0x4044
> +#define LSWB_WB_LIMIT1(x)                        ((x) & 0xffff)
> +#define LSWB_WB_LIMIT0(x)                        (((x) & 0xffff) << 16)
> +
> +#define ISP_LSWB_WB_LIMIT1                       0x4048
> +#define LSWB_WB_LIMIT3(x)                        ((x) & 0xffff)
> +#define LSWB_WB_LIMIT2(x)                        (((x) & 0xffff) << 16)
> +
> +#define ISP_LSWB_WB_LIMIT2                       0x404c
> +#define LSWB_WB_LIMIT4_MASK                      GENMASK(15, 0)
> +
> +#define ISP_LSWB_WB_PHSOFST                      0x4050
> +#define LSWB_WB_YPHS_OFST_MASK                   GENMASK(1, 0)
> +#define LSWB_WB_XPHS_OFST_MASK                   GENMASK(3, 2)
> +#define LSWB_WB_XPHS_OFST_SHIFT                  2
> +
> +#define ISP_LSWB_LNS_PHSOFST                     0x4054
> +#define LSWB_LNS_YPHS_OFST_MASK                  GENMASK(1, 0)
> +#define LSWB_LNS_XPHS_OFST_MASK                  GENMASK(3, 2)
> +#define LSWB_LNS_XPHS_OFST_SHIFT                 2
> +
> +#define ISP_DMS_COMMON_PARAM0                    0x5000
> +#define DMS_COMMON_YPHS_OFST_MASK                GENMASK(1, 0)
> +#define DMS_COMMON_XPHS_OFST_MASK                GENMASK(3, 2)
> +#define DMS_COMMON_XPHS_OFST_SHIFT               2
> +
> +#define ISP_CM0_INP_OFST01                       0x6040
> +#define CM0_INP_OFST0_MASK                       GENMASK(12, 0)
> +#define CM0_INP_OFST1_MASK                       GENMASK(28, 16)
> +#define CM0_INP_OFST1_SHIFT                      16
> +
> +#define ISP_CM0_INP_OFST2                        0x6044
> +#define CM0_INP_OFST2_MASK                       GENMASK(12, 0)
> +
> +#define ISP_CM0_COEF00_01                        0x6048
> +#define CM0_MTX_00_MASK                          GENMASK(12, 0)
> +#define CM0_MTX_01_MASK                          GENMASK(28, 16)
> +#define CM0_MTX_01_SHIFT                         16
> +
> +#define ISP_CM0_COEF02_10                        0x604c
> +#define CM0_MTX_02_MASK                          GENMASK(12, 0)
> +#define CM0_MTX_10_MASK                          GENMASK(28, 16)
> +#define CM0_MTX_10_SHIFT                         16
> +
> +#define ISP_CM0_COEF11_12                        0x6050
> +#define CM0_MTX_11_MASK                          GENMASK(12, 0)
> +#define CM0_MTX_12_MASK                          GENMASK(28, 16)
> +#define CM0_MTX_12_SHIFT                         16
> +
> +#define ISP_CM0_COEF20_21                        0x6054
> +#define CM0_MTX_20_MASK                          GENMASK(12, 0)
> +#define CM0_MTX_21_MASK                          GENMASK(28, 16)
> +#define CM0_MTX_21_SHIFT                         16
> +
> +#define ISP_CM0_COEF22_OUP_OFST0                 0x6058
> +#define CM0_MTX_22_MASK                          GENMASK(12, 0)
> +#define CM0_OFST_OUP0_MASK                       GENMASK(28, 16)
> +#define CM0_OFST_OUP0_SHIFT                      16
> +
> +#define ISP_CM0_OUP_OFST12_RS                    0x605c
> +#define CM0_OFST_OUP1_MASK                       GENMASK(12, 0)
> +#define CM0_OFST_OUP2_MASK                       GENMASK(28, 16)
> +#define CM0_OFST_OUP2_SHIFT                      16
> +#define CM0_MTX_RS_MASK                          GENMASK(31, 30)
> +#define CM0_MTX_RS_SHIFT                         30
> +
> +#define ISP_CCM_MTX_00_01                        0x6098
> +#define CCM_MTX_00_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_01_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_01_SHIFT                         16
> +
> +#define ISP_CCM_MTX_02_03                        0x609c
> +#define CCM_MTX_02_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_03_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_03_SHIFT                         16
> +
> +#define ISP_CCM_MTX_10_11                        0x60A0
> +#define CCM_MTX_10_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_11_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_11_SHIFT                         16
> +
> +#define ISP_CCM_MTX_12_13                        0x60A4
> +#define CCM_MTX_12_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_13_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_13_SHIFT                         16
> +
> +#define ISP_CCM_MTX_20_21                        0x60A8
> +#define CCM_MTX_20_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_21_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_21_SHIFT                         16
> +
> +#define ISP_CCM_MTX_22_23_RS                     0x60Ac
> +#define CCM_MTX_22_MASK                          GENMASK(12, 0)
> +#define CCM_MTX_23_MASK                          GENMASK(28, 16)
> +#define CCM_MTX_23_SHIFT                         16
> +
> +#define ISP_PST_GAMMA_MODE                       0x60C0
> +#define PST_GAMMA_MODE                           BIT(0)
> +#define PST_GAMMA_MODE_EXT(x)                    ((x) & 0x1)
> +
> +#define ISP_PST_GAMMA_LUT_ADDR                   0x60cc
> +#define PST_GAMMA_LUT_ADDR(x)                    ((x) << 7)
> +
> +#define ISP_PST_GAMMA_LUT_DATA                   0x60d0
> +#define PST_GAMMA_LUT_DATA0(x)                   ((x) & 0xffff)
> +#define PST_GAMMA_LUT_DATA1(x)                   (((x) & 0xffff) << 16)
> +
> +#define DISP0_TOP_TOP_CTRL                       0x8000
> +#define DISP_CRP2_EN                             BIT(5)
> +
> +#define DISP0_TOP_CRP2_START                     0x8004
> +#define DISP_CRP2_VSTART_MASK                    GENMASK(15, 0)
> +#define DISP_CRP2_VSTART(x)                      ((x) & 0xffff)
> +#define DISP_CRP2_HSTART_MASK                    GENMASK(31, 16)
> +#define DISP_CRP2_HSTART(x)                      (((x) & 0xffff) << 16)
> +
> +#define DISP0_TOP_CRP2_SIZE                      0x8008
> +#define DISP_CRP2_VSIZE_MASK                     GENMASK(15, 0)
> +#define DISP_CRP2_VSIZE(x)                       ((x) & 0xffff)
> +#define DISP_CRP2_HSIZE_MASK                     GENMASK(31, 16)
> +#define DISP_CRP2_HSIZE(x)                       (((x) & 0xffff) << 16)
> +
> +#define DISP0_TOP_OUT_SIZE                       0x800c
> +#define DISP_OUT_VSIZE_MASK                      GENMASK(12, 0)
> +#define DISP_OUT_HSIZE_MASK                      GENMASK(28, 16)
> +#define DISP_OUT_HSIZE_SHIFT                     16
> +
> +#define ISP_DISP0_TOP_IN_SIZE                    0x804c
> +#define DISP_TOP_IN_VSIZE(x)                     ((x) & 0x1fff)
> +#define DISP_TOP_IN_HSIZE(x)                     (((x) & 0x1fff) << 16)
> +
> +#define DISP0_PPS_SCALE_EN                       0x8200
> +#define PPS_VSC_TAP_NUM_MASK                     GENMASK(3, 0)
> +#define PPS_HSC_TAP_NUM_MASK                     GENMASK(7, 4)
> +#define PPS_HSC_TAP_NUM_SHIFT                    4
> +#define PPS_HSC_TAP_NUM_INIT                     4
> +#define PPS_PREVSC_FLT_NUM_MASK                  GENMASK(11, 8)
> +#define PPS_PREVSC_FLT_NUM_SHIFT                 8
> +#define PPS_PREHSC_FLT_NUM_MASK                  GENMASK(15, 12)
> +#define PPS_PREHSC_FLT_NUM_SHIFT                 12
> +#define PPS_PREHSC_FLT_NUM_INIT                  8
> +#define PPS_PREVSC_RATE_MASK                     GENMASK(17, 16)
> +#define PPS_PREVSC_RATE_SHIFT                    16
> +#define PPS_PREHSC_RATE_MASK                     GENMASK(19, 18)
> +#define PPS_PREHSC_RATE_SHIFT                    18
> +#define PPS_HSC_EN_MASK                          BIT(20)
> +#define PPS_HSC_EN_SHIFT                         20
> +#define PPS_VSC_EN_MASK                          BIT(21)
> +#define PPS_VSC_EN_SHIFT                         21
> +#define PPS_PREVSC_EN_MASK                       BIT(22)
> +#define PPS_PREVSC_EN_SHIFT                      22
> +#define PPS_PREHSC_EN_MASK                       BIT(23)
> +#define PPS_PREHSC_EN_SHIFT                      23
> +#define PPS_HSC_NOR_RS_BITS_MASK                 GENMASK(27, 24)
> +#define PPS_HSC_NOR_RS_BITS_SHIFT                24
> +#define PPS_HSC_NOR_RS_BITS_INIT                 9
> +#define PPS_VSC_NOR_RS_BITS_MASK                 GENMASK(31, 28)
> +#define PPS_VSC_NOR_RS_BITS_SHIFT                28
> +#define PPS_VSC_NOR_RS_BITS_INIT                 9
> +
> +#define DISP0_PPS_PRE_HSCALE_COEF_0              0x8204
> +#define PPS_PREHSC_LUMA_COEF0_MASK               GENMASK(9, 0)
> +#define PPS_PREHSC_LUMA_COEF0_INIT               128
> +#define PPS_PREHSC_LUMA_COEF1_MASK               GENMASK(25, 16)
> +#define PPS_PREHSC_LUMA_COEF1_SHIFT              16
> +#define PPS_PREHSC_LUMA_COEF1_INIT               128
> +
> +#define DISP0_PPS_PRE_HSCALE_COEF_1              0x8208
> +#define PPS_PREHSC_LUMA_COEF2_MASK               GENMASK(9, 0)
> +#define PPS_PREHSC_LUMA_COEF2_INIT               32
> +#define PPS_PREHSC_LUMA_COEF3_MASK               GENMASK(25, 16)
> +#define PPS_PREHSC_LUMA_COEF3_SHIFT              16
> +#define PPS_PREHSC_LUMA_COEF3_INIT               32
> +
> +#define DISP0_PPS_VSC_START_PHASE_STEP           0x8224
> +#define PPS_VSC_FRACTION_PART_MASK               GENMASK(23, 0)
> +#define PPS_VSC_INTEGER_PART_MASK                GENMASK(27, 24)
> +#define PPS_VSC_INTEGER_PART_SHIFT               24
> +
> +#define DISP0_PPS_HSC_START_PHASE_STEP           0x8230
> +#define PPS_HSC_FRACTION_PART_MASK               GENMASK(23, 0)
> +#define PPS_HSC_INTEGER_PART_MASK                GENMASK(27, 24)
> +#define PPS_HSC_INTEGER_PART_SHIFT               24
> +
> +#define DISP0_PPS_444TO422                       0x823c
> +#define PPS_444TO422_EN_MASK                     BIT(0)
> +
> +#define ISP_SCALE0_COEF_IDX_LUMA                 0x8240
> +#define SCALE_LUMA_COEF_S11_MODE                 BIT(9)
> +#define SCALE_LUMA_CTYPE(x)                      (((x) & 0x7) << 10)
> +
> +#define ISP_SCALE0_COEF_LUMA                     0x8244
> +#define SCALE_COEF_LUMA_DATA1(x)                 ((x) & 0x7ff)
> +#define SCALE_COEF_LUMA_DATA0(x)                 (((x) & 0x7ff) << 16)
> +
> +#define ISP_SCALE0_COEF_IDX_CHRO                 0x8248
> +#define SCALE_CHRO_COEF_S11_MODE                 BIT(9)
> +#define SCALE_CHRO_CTYPE(x)                      (((x) & 0x7) << 10)
> +
> +#define ISP_SCALE0_COEF_CHRO                     0x824c
> +#define SCALE_COEF_CHRO_DATA1(x)                 ((x) & 0x7ff)
> +#define SCALE_COEF_CHRO_DATA0(x)                 (((x) & 0x7ff) << 16)
> +
> +#define ISP_AF_ROI0_WIN01                        0xa00c
> +#define AF_ROI_XYWIN_00_INIT                     (50 & 0xffff)
> +#define AF_ROI_XYWIN_01_INIT                     ((50 & 0xffff) << 16)
> +
> +#define ISP_AF_ROI1_WIN01                        0xa010
> +#define AF_ROI_XYWIN_10_INIT                     (100 & 0xffff)
> +#define AF_ROI_XYWIN_11_INIT                     ((50 & 0xffff) << 16)
> +
> +#define ISP_AF_ROI0_WIN23                        0xa014
> +#define AF_ROI_XYWIN_02_INIT                     (80 & 0xffff)
> +#define AF_ROI_XYWIN_03_INIT                     ((40 & 0xffff) << 16)
> +
> +#define ISP_AF_ROI1_WIN23                        0xa018
> +#define AF_ROI_XYWIN_12_INIT                     (80 & 0xffff)
> +#define AF_ROI_XYWIN_13_INIT                     ((40 & 0xffff) << 16)
> +
> +#define ISP_AF_CTRL                              0xa044
> +#define AF_CTRL_YPHS_OFST_MASK                   GENMASK(15, 14)
> +#define AF_CTRL_YPHS_OFST_SHIFT                  14
> +#define AF_CTRL_XPHS_OFST_MASK                   GENMASK(17, 16)
> +#define AF_CTRL_XPHS_OFST_SHIFT                  16
> +
> +#define ISP_AF_HV_SIZE                           0xa04c
> +#define AF_HV_STAT_VSIZE(x)                      ((x) & 0xffff)
> +#define AF_HV_STAT_HSIZE(x)                      (((x) & 0xffff) << 16)
> +
> +#define ISP_AF_HV_BLKNUM                         0xa050
> +#define AF_HV_STAT_VBLK_NUM_MASK                 GENMASK(5, 0)
> +#define AF_HV_STAT_VBLK_NUM_EXT(x)               ((x) & 0x3f)
> +#define AF_HV_STAT_HBLK_NUM_MASK                 GENMASK(21, 16)
> +#define AF_HV_STAT_HBLK_NUM_SHIFT                16
> +
> +#define ISP_AF_EN_CTRL                           0xa054
> +#define AF_STAT_SELECT                           BIT(21)
> +#define AF_STAT_SELECT_SHIFT                     21
> +
> +#define ISP_RO_AF_GLB_STAT_PCK0                  0xa0e0
> +#define ISP_RO_AF_GLB_STAT_PCK1                  0xa0e4
> +#define ISP_RO_AF_GLB_STAT_PCK2                  0xa0e8
> +#define ISP_RO_AF_GLB_STAT_PCK3                  0xa0eC
> +#define ISP_AF_IDX_ADDR                          0xa1c0
> +#define ISP_AF_IDX_DATA                          0xa1c4
> +#define AF_IDX_VIDX_DATA(x)                      ((x) & 0xffff)
> +#define AF_IDX_HIDX_DATA(x)                      (((x) & 0xffff) << 16)
> +
> +#define ISP_AE_ROI0_WIN01                        0xa40c
> +#define AE_ROI_XYWIN_00_INIT                     (50 & 0xffff)
> +#define AE_ROI_XYWIN_01_INIT                     ((50 & 0xffff) << 16)
> +
> +#define ISP_AE_ROI1_WIN01                        0xa410
> +#define AE_ROI_XYWIN_10_INIT                     (100 & 0xffff)
> +#define AE_ROI_XYWIN_11_INIT                     ((50 & 0xffff) << 16)
> +
> +#define ISP_AE_ROI0_WIN23                        0xa414
> +#define AE_ROI_XYWIN_02_INIT                     (80 & 0xffff)
> +#define AE_ROI_XYWIN_03_INIT                     ((40 & 0xffff) << 16)
> +
> +#define ISP_AE_ROI1_WIN23                        0xa418
> +#define AE_ROI_XYWIN_12_INIT                     (80 & 0xffff)
> +#define AE_ROI_XYWIN_13_INIT                     ((40 & 0xffff) << 16)
> +
> +#define ISP_RO_AE_ROI_STAT_PCK0_0                0xa424
> +#define ISP_RO_AE_ROI_STAT_PCK1_0                0xa428
> +#define ISP_RO_AE_ROI_STAT_PCK0_1                0xa42c
> +#define ISP_RO_AE_ROI_STAT_PCK1_1                0xa430
> +#define ISP_AE_CTRL                              0xa448
> +#define AE_CTRL_AE_STAT_LOCAL_MODE               GENMASK(3, 2)
> +#define AE_CTRL_AE_STAT_LOCAL_MODE_EXT(x)        (((x) >> 2) & 0x3)
> +#define AE_CTRL_INPUT_2LINE_TOGETHER             BIT(7)
> +#define AE_CTRL_LUMA_MODE_MASK                   GENMASK(9, 8)
> +#define AE_CTRL_LUMA_MODE_SHIFT                  8
> +#define AE_CTRL_LUMA_MODE_FILTER                 2
> +#define AE_CTRL_YPHS_OFST_MASK                   GENMASK(25, 24)
> +#define AE_CTRL_YPHS_OFST_SHIFT                  24
> +#define AE_CTRL_XPHS_OFST_MASK                   GENMASK(27, 26)
> +#define AE_CTRL_XPHS_OFST_SHIFT                  26
> +
> +#define ISP_AE_CRTL2_0                           0xa44c
> +#define AE_CRTL2_BL12_GRBGI0_MASK                GENMASK(11, 0)
> +#define AE_CRTL2_GAIN_GRBGI0_MASK                GENMASK(23, 16)
> +#define AE_CRTL2_GAIN_GRBGI0_SHIFT               16
> +
> +#define ISP_AE_CRTL2_1                           0xa450
> +#define AE_CRTL2_BL12_GRBGI1_MASK                GENMASK(11, 0)
> +#define AE_CRTL2_GAIN_GRBGI1_MASK                GENMASK(23, 16)
> +#define AE_CRTL2_GAIN_GRBGI1_SHIFT               16
> +
> +#define ISP_AE_CRTL2_2                           0xa454
> +#define AE_CRTL2_BL12_GRBGI2_MASK                GENMASK(11, 0)
> +#define AE_CRTL2_GAIN_GRBGI2_MASK                GENMASK(23, 16)
> +#define AE_CRTL2_GAIN_GRBGI2_SHIFT               16
> +
> +#define ISP_AE_CRTL2_3                           0xa458
> +#define AE_CRTL2_BL12_GRBGI3_MASK                GENMASK(11, 0)
> +#define AE_CRTL2_GAIN_GRBGI3_MASK                GENMASK(23, 16)
> +#define AE_CRTL2_GAIN_GRBGI3_SHIFT               16
> +
> +#define ISP_AE_CRTL2_4                           0xa45C
> +#define AE_CRTL2_BL12_GRBGI4_MASK                GENMASK(11, 0)
> +#define AE_CRTL2_GAIN_GRBGI4_MASK                GENMASK(23, 16)
> +#define AE_CRTL2_GAIN_GRBGI4_SHIFT               16
> +
> +#define ISP_AE_HV_SIZE                           0xa464
> +#define AE_HV_STAT_VSIZE(x)                      ((x) & 0xffff)
> +#define AE_HV_STAT_HSIZE(x)                      (((x) & 0xffff) << 16)
> +
> +#define ISP_AE_HV_BLKNUM                         0xa468
> +#define AE_HV_STAT_VBLK_NUM_MASK                 GENMASK(6, 0)
> +#define AE_HV_STAT_VBLK_NUM_EXT(x)               ((x) & 0x7f)
> +#define AE_HV_STAT_HBLK_NUM_MASK                 GENMASK(22, 16)
> +#define AE_HV_STAT_HBLK_NUM_SHIFT                16
> +
> +#define ISP_AE_STAT_THD01                        0xa46c
> +#define AE_STAT_THRD0_INIT                       (13107 & 0xffff)
> +#define AE_STAT_THRD1_INIT                       ((26214 & 0xffff) << 16)
> +
> +#define ISP_AE_STAT_THD23                        0xa470
> +#define AE_STAT_THRD2_INIT                       (39321 & 0xffff)
> +#define AE_STAT_THRD3_INIT                       ((52428 & 0xffff) << 16)
> +
> +#define ISP_RO_AE_STAT_GPNUM                     0xa49c
> +#define ISP_AE_IDX_ADDR                          0xa600
> +#define ISP_AE_IDX_DATA                          0xa604
> +#define AE_IDX_VIDX_DATA(x)                      ((x) & 0xffff)
> +#define AE_IDX_HIDX_DATA(x)                      (((x) & 0xffff) << 16)
> +
> +#define ISP_AE_BLK_WT_ADDR                       0xa608
> +#define ISP_AE_BLK_WT_DATA                       0xa60c
> +#define AE_BLK_WT_DATA0(x)                       ((x) & 0xf)
> +#define AE_BLK_WT_DATA1(x)                       (((x) & 0xf) << 4)
> +#define AE_BLK_WT_DATA2(x)                       (((x) & 0xf) << 8)
> +#define AE_BLK_WT_DATA3(x)                       (((x) & 0xf) << 12)
> +#define AE_BLK_WT_DATA4(x)                       (((x) & 0xf) << 16)
> +#define AE_BLK_WT_DATA5(x)                       (((x) & 0xf) << 20)
> +#define AE_BLK_WT_DATA6(x)                       (((x) & 0xf) << 24)
> +#define AE_BLK_WT_DATA7(x)                       (((x) & 0xf) << 28)
> +
> +#define ISP_AWB_CTRL                             0xa834
> +#define AWB_CTRL_YPHS_OFST_MASK                  GENMASK(1, 0)
> +#define AWB_CTRL_XPHS_OFST_MASK                  GENMASK(3, 2)
> +#define AWB_CTRL_XPHS_OFST_SHIFT                 2
> +#define AWB_CTRL_STAT_RATIO_MODE                 BIT(4)
> +
> +#define ISP_AWB_HV_SIZE                          0xa83c
> +#define AWB_HV_STAT_VSIZE(x)                     ((x) & 0xffff)
> +#define AWB_HV_STAT_HSIZE(x)                     (((x) & 0xffff) << 16)
> +
> +#define ISP_AWB_STAT_ROI0_WIN01                  0xa80c
> +#define AWB_ROI_XYWIN_00_INIT                    (32 & 0xffff)
> +#define AWB_ROI_XYWIN_01_INIT                    ((62 & 0xffff) << 16)
> +
> +#define ISP_AWB_STAT_ROI1_WIN01                  0xa810
> +#define AWB_ROI_XYWIN_10_INIT                    (32 & 0xffff)
> +#define AWB_ROI_XYWIN_11_INIT                    ((82 & 0xffff) << 16)
> +
> +#define ISP_AWB_STAT_ROI0_WIN23                  0xa814
> +#define AWB_ROI_XYWIN_02_INIT                    (32 & 0xffff)
> +#define AWB_ROI_XYWIN_03_INIT                    ((42 & 0xffff) << 16)
> +
> +#define ISP_AWB_STAT_ROI1_WIN23                  0xa818
> +#define AWB_ROI_XYWIN_12_INIT                    (64 & 0xffff)
> +#define AWB_ROI_XYWIN_13_INIT                    ((52 & 0xffff) << 16)
> +
> +#define ISP_AWB_HV_BLKNUM                        0xa840
> +#define AWB_HV_STAT_VBLK_NUM_MASK                GENMASK(5, 0)
> +#define AWB_HV_STAT_VBLK_NUM_EXT(x)              ((x) & 0x3f)
> +#define AWB_HV_STAT_HBLK_NUM_MASK                GENMASK(21, 16)
> +#define AWB_HV_STAT_HBLK_NUM_SHIFT               16
> +
> +#define ISP_AWB_STAT_RG                          0xa848
> +#define AWB_STAT_RG_MIN_MASK                     GENMASK(11, 0)
> +#define AWB_STAT_RG_MAX_MASK                     GENMASK(27, 16)
> +#define AWB_STAT_RG_MAX_SHIFT                    16
> +
> +#define ISP_AWB_STAT_BG                          0xa84c
> +#define AWB_STAT_BG_MIN_MASK                     GENMASK(11, 0)
> +#define AWB_STAT_BG_MAX_MASK                     GENMASK(27, 16)
> +#define AWB_STAT_BG_MAX_SHIFT                    16
> +
> +#define ISP_AWB_STAT_RG_HL                       0xa850
> +#define AWB_STAT_RG_LOW_MASK                     GENMASK(11, 0)
> +#define AWB_STAT_RG_HIGH_MASK                    GENMASK(27, 16)
> +#define AWB_STAT_RG_HIGH_SHIFT                   16
> +
> +#define ISP_AWB_STAT_BG_HL                       0xa854
> +#define AWB_STAT_BG_LOW_MASK                     GENMASK(11, 0)
> +#define AWB_STAT_BG_HIGH_MASK                    GENMASK(27, 16)
> +#define AWB_STAT_BG_HIGH_SHIFT                   16
> +
> +#define ISP_AWB_STAT_CTRL2                       0xa858
> +#define AWB_STAT_SATUR_VALID_MASK                BIT(0)
> +#define AWB_STAT_LOCAL_MODE                      BIT(2)
> +#define AWB_STAT_LOCAL_MODE_EXT(x)               (((x) >> 2) & 0x1)
> +#define AWB_STAT_LUMA_DIV_MODE                   GENMASK(4, 3)
> +#define AWB_STAT_LUMA_DIV_MODE_EXT(x)            (((x) >> 3) & 0x3)
> +
> +#define ISP_AWB_STAT_BLC20_0                     0xa85c
> +#define AWB_STAT_BLC20_GR_MASK                   GENMASK(19, 0)
> +
> +#define ISP_AWB_STAT_BLC20_1                     0xa860
> +#define AWB_STAT_BLC20_R_MASK                    GENMASK(19, 0)
> +
> +#define ISP_AWB_STAT_BLC20_2                     0xa864
> +#define AWB_STAT_BLC20_B_MASK                    GENMASK(19, 0)
> +
> +#define ISP_AWB_STAT_BLC20_3                     0xa868
> +#define AWB_STAT_BLC20_GB_MASK                   GENMASK(19, 0)
> +
> +#define ISP_AWB_STAT_GAIN10_0                    0xa86c
> +#define AWB_STAT_GAIN10_GR_MASK                  GENMASK(9, 0)
> +
> +#define ISP_AWB_STAT_GAIN10_1                    0xa870
> +#define AWB_STAT_GAIN10_R_MASK                   GENMASK(9, 0)
> +
> +#define ISP_AWB_STAT_GAIN10_2                    0xa874
> +#define AWB_STAT_GAIN10_B_MASK                   GENMASK(9, 0)
> +
> +#define ISP_AWB_STAT_GAIN10_3                    0xa878
> +#define AWB_STAT_GAIN10_GB_MASK                  GENMASK(9, 0)
> +
> +#define ISP_AWB_STAT_SATUR_CTRL                  0xa884
> +#define AWB_STAT_SATUR_LOW(x)                    ((x) & 0xffff)
> +#define AWB_STAT_SATUR_HIGH(x)                   (((x) & 0xffff) << 16)
> +
> +#define ISP_AWB_IDX_ADDR                         0xaa00
> +#define ISP_AWB_IDX_DATA                         0xaa04
> +#define AWB_IDX_VIDX_DATA(x)                     ((x) & 0xffff)
> +#define AWB_IDX_HIDX_DATA(x)                     (((x) & 0xffff) << 16)
> +
> +#define ISP_AWB_BLK_WT_ADDR                      0xaa08
> +#define ISP_AWB_BLK_WT_DATA                      0xaa0c
> +#define AWB_BLK_WT_DATA0(x)                      ((x) & 0xf)
> +#define AWB_BLK_WT_DATA1(x)                      (((x) & 0xf) << 4)
> +#define AWB_BLK_WT_DATA2(x)                      (((x) & 0xf) << 8)
> +#define AWB_BLK_WT_DATA3(x)                      (((x) & 0xf) << 12)
> +#define AWB_BLK_WT_DATA4(x)                      (((x) & 0xf) << 16)
> +#define AWB_BLK_WT_DATA5(x)                      (((x) & 0xf) << 20)
> +#define AWB_BLK_WT_DATA6(x)                      (((x) & 0xf) << 24)
> +#define AWB_BLK_WT_DATA7(x)                      (((x) & 0xf) << 28)
> +
> +#define ISP_WRMIFX3_0_CH0_CTRL0                  0xc400
> +#define WRMIFX3_CH0_STRIDE_MASK                  GENMASK(28, 16)
> +#define WRMIFX3_CH0_STRIDE_SHIFT                 16
> +
> +#define ISP_WRMIFX3_0_CH0_CTRL1                  0xc404
> +#define WRMIFX3_CH0_PIX_BITS_MODE_MASK           GENMASK(30, 27)
> +#define WRMIFX3_CH0_PIX_BITS_MODE_SHIFT          27
> +#define WRMIFX3_CH0_PIX_BITS_8BITS               1
> +#define WRMIFX3_CH0_PIX_BITS_16BITS              2
> +
> +#define ISP_WRMIFX3_0_CH1_CTRL0                  0xc408
> +#define WRMIFX3_CH1_STRIDE_MASK                  GENMASK(28, 16)
> +#define WRMIFX3_CH1_STRIDE_SHIFT                 16
> +
> +#define ISP_WRMIFX3_0_CH1_CTRL1                  0xc40c
> +#define WRMIFX3_CH1_PIX_BITS_MODE_MASK           GENMASK(30, 27)
> +#define WRMIFX3_CH1_PIX_BITS_MODE_SHIFT          27
> +#define WRMIFX3_CH1_PIX_BITS_16BITS              2
> +#define WRMIFX3_CH1_PIX_BITS_32BITS              3
> +
> +#define ISP_WRMIFX3_0_CH0_BADDR                  0xc440
> +/* WRMIF base address need 16 bits alignment */
> +#define WRMIFX3_CH0_BADDR(x)                     (((x) >> 4) & 0xffffffff)
> +
> +#define ISP_WRMIFX3_0_CH1_BADDR                  0xc444
> +#define WRMIFX3_CH1_BADDR(x)                     (((x) >> 4) & 0xffffffff)
> +
> +#define ISP_WRMIFX3_0_FMT_SIZE                   0xc464
> +#define WRMIFX3_FMT_HSIZE(x)                     ((x) & 0xffff)
> +#define WRMIFX3_FMT_VSIZE(x)                     (((x) & 0xffff) << 16)
> +
> +#define ISP_WRMIFX3_0_FMT_CTRL                   0xc468
> +#define WRMIFX3_FMT_MTX_IBITS_MASK               GENMASK(1, 0)
> +#define WRMIFX3_FMT_MTX_IBITS_8BIT               0
> +#define WRMIFX3_FMT_MTX_IBITS_10BIT              1
> +#define WRMIFX3_FMT_MTX_IBITS_12BIT              2
> +#define WRMIFX3_FMT_MTX_IBITS_16BIT              3
> +#define WRMIFX3_FMT_MTX_UV_SWAP_MASK             BIT(2)
> +#define WRMIFX3_FMT_MTX_UV_SWAP_SHIFT            2
> +#define WRMIFX3_FMT_MTX_UV_SWAP_VU               0
> +#define WRMIFX3_FMT_MTX_UV_SWAP_UV               1
> +#define WRMIFX3_FMT_MTX_PLANE_MASK               GENMASK(5, 4)
> +#define WRMIFX3_FMT_MTX_PLANE_SHIFT              4
> +#define WRMIFX3_FMT_MTX_PLANE_X1                 0
> +#define WRMIFX3_FMT_MTX_PLANE_X2                 1
> +#define WRMIFX3_FMT_MODE_OUT_MASK                GENMASK(18, 16)
> +#define WRMIFX3_FMT_MODE_OUT_SHIFT               16
> +#define WRMIFX3_FMT_MODE_OUT_YUV422              1
> +#define WRMIFX3_FMT_MODE_OUT_YUV420              2
> +#define WRMIFX3_FMT_MODE_OUT_Y_ONLY              3
> +
> +#define ISP_WRMIFX3_0_WIN_LUMA_H                 0xc420
> +#define WRMIFX3_WIN_LUMA_HEND_MASK               GENMASK(28, 16)
> +#define WRMIFX3_WIN_LUMA_HEND_SHIFT              16
> +#define WRMIFX3_WIN_LUMA_HEND(x)                 ((x) - 1)
> +
> +#define ISP_WRMIFX3_0_WIN_LUMA_V                 0xc424
> +#define WRMIFX3_WIN_LUMA_VEND_MASK               GENMASK(28, 16)
> +#define WRMIFX3_WIN_LUMA_VEND_SHIFT              16
> +#define WRMIFX3_WIN_LUMA_VEND(x)                 ((x) - 1)
> +
> +#define ISP_WRMIFX3_0_WIN_CHROM_H                0xc428
> +#define WRMIFX3_WIN_CHROM_HEND_MASK              GENMASK(28, 16)
> +#define WRMIFX3_WIN_CHROM_HEND_SHIFT             16
> +#define WRMIFX3_WIN_CHROM_HEND(x)                (((x) >> 1) - 1)
> +
> +#define ISP_WRMIFX3_0_WIN_CHROM_V                0xc42c
> +#define WRMIFX3_WIN_CHROM_VEND_MASK              GENMASK(28, 16)
> +#define WRMIFX3_WIN_CHROM_VEND_SHIFT             16
> +#define WRMIFX3_WIN_CHROM_VEND(x)                (((x) >> 1) - 1)
> +
> +#define ISP_WRMIFX3_0_CRP_HSIZE                  0xc48c
> +#define WRMIFX3_CROP_HEND_MASK                   GENMASK(31, 16)
> +#define WRMIFX3_CROP_HEND_SHIFT                  16
> +#define WRMIFX3_CROP_HEND(x)                     ((x) - 1)
> +
> +#define ISP_WRMIFX3_0_CRP_VSIZE                  0xc490
> +#define WRMIFX3_CROP_VEND_MASK                   GENMASK(31, 16)
> +#define WRMIFX3_CROP_VEND_SHIFT                  16
> +#define WRMIFX3_CROP_VEND(x)                     ((x) - 1)
> +
> +#define VIU_DMAWR_BADDR0                         0xc840
> +#define VIU_DMAWR_AF_BADDR_MASK                  GENMASK(27, 0)
> +/* AF base address need 16 bits alignment */
> +#define VIU_DMAWR_AF_BADDR(x)                    ((x) >> 4)
> +
> +#define VIU_DMAWR_BADDR1                         0xc844
> +#define VIU_DMAWR_AWB_BADDR_MASK                  GENMASK(27, 0)
> +/* AWB base address need 16 bits alignment */
> +#define VIU_DMAWR_AWB_BADDR(x)                    ((x) >> 4)
> +
> +#define VIU_DMAWR_BADDR2                         0xc848
> +#define VIU_DMAWR_AE_BADDR_MASK                  GENMASK(27, 0)
> +/* AE base address need 16 bits alignment */
> +#define VIU_DMAWR_AE_BADDR(x)                    ((x) >> 4)
> +
> +#define VIU_DMAWR_SIZE0                          0xc854
> +#define VIU_DMAWR_SIZE_AF_MASK                   GENMASK(15, 0)
> +#define VIU_DMAWR_SIZE_AWB_MASK                  GENMASK(31, 16)
> +#define VIU_DMAWR_SIZE_AWB_SHIFT                 16
> +
> +#define VIU_DMAWR_SIZE1                          0xc858
> +#define VIU_DMAWR_SIZE_AE_MASK                   GENMASK(15, 0)
> +
> +#endif
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-resizer.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-resizer.c
> new file mode 100644
> index 000000000000..01d99b66cb32
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-resizer.c
> @@ -0,0 +1,768 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/pm_runtime.h>
> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +
> +static const struct c3_isp_mbus_format_info c3_isp_rsz_mbus_formats[] = {
> +	/* YUV formats */
> +	{
> +		.mbus_code	= MEDIA_BUS_FMT_Y8_1X8,
> +		.pads		= BIT(C3_ISP_RESIZER_PAD_SINK)
> +				| BIT(C3_ISP_RESIZER_PAD_SOURCE),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_YUYV8_1_5X8,
> +		.pads		= BIT(C3_ISP_RESIZER_PAD_SINK)
> +				| BIT(C3_ISP_RESIZER_PAD_SOURCE),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	}, {
> +		.mbus_code	= MEDIA_BUS_FMT_YUYV8_2X8,

These mbus_codes come from the ISP, I presume after debayering.
Is the different samples number (5X8, 2X8) a representation of the
format on the internal bus between the ISP and the resizers ?

> +		.pads		= BIT(C3_ISP_RESIZER_PAD_SINK)
> +				| BIT(C3_ISP_RESIZER_PAD_SOURCE),
> +		.xofst		= C3_ISP_PHASE_OFFSET_NONE,
> +		.yofst		= C3_ISP_PHASE_OFFSET_NONE,
> +	},
> +};
> +
> +/* The normal parameters of pps module */
> +static const int pps_lut_tap4_s11[C3_PPS_TAP4_S11_H_NUM][4] =  {
> +	{  0, 511,   0,   0}, { -5, 511,   5,   0}, {-10, 511,  11,   0},
> +	{-14, 510,  17,  -1}, {-18, 508,  23,  -1}, {-22, 506,  29,  -1},
> +	{-25, 503,  36,  -2}, {-28, 500,  43,  -3}, {-32, 496,  51,  -3},
> +	{-34, 491,  59,  -4}, {-37, 487,  67,  -5}, {-39, 482,  75,  -6},
> +	{-41, 476,  84,  -7}, {-42, 470,  92,  -8}, {-44, 463, 102,  -9},
> +	{-45, 456, 111, -10}, {-45, 449, 120, -12}, {-47, 442, 130, -13},
> +	{-47, 434, 140, -15}, {-47, 425, 151, -17}, {-47, 416, 161, -18},
> +	{-47, 407, 172, -20}, {-47, 398, 182, -21}, {-47, 389, 193, -23},
> +	{-46, 379, 204, -25}, {-45, 369, 215, -27}, {-44, 358, 226, -28},
> +	{-43, 348, 237, -30}, {-43, 337, 249, -31}, {-41, 326, 260, -33},
> +	{-40, 316, 271, -35}, {-39, 305, 282, -36}, {-37, 293, 293, -37}
> +};
> +
> +static const struct c3_isp_mbus_format_info
> +*rsz_find_format_by_code(u32 code, u32 pad)
> +{
> +	int i;

unsigned

> +
> +	for (i = 0; i < ARRAY_SIZE(c3_isp_rsz_mbus_formats); i++) {
> +		const struct c3_isp_mbus_format_info *info =
> +			&c3_isp_rsz_mbus_formats[i];
> +
> +		if (info->mbus_code == code && info->pads & BIT(pad))
> +			return info;
> +	}
> +
> +	return NULL;
> +}
> +
> +static void c3_isp_rsz_cfg_fmt(struct c3_isp_resizer *rsz,
> +			       struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_mbus_framefmt *fmt;
> +
> +	fmt = v4l2_subdev_state_get_format(state, C3_ISP_RESIZER_PAD_SINK);
> +
> +	c3_isp_write(rsz->isp, C3_DISP_REG(ISP_DISP0_TOP_IN_SIZE, rsz->id),
> +		     DISP_TOP_IN_HSIZE(fmt->width) | DISP_TOP_IN_VSIZE(fmt->height));
> +}
> +
> +static void c3_isp_rsz_crop_enable(struct c3_isp_resizer *rsz,
> +				   struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_rect *crop;
> +
> +	crop = v4l2_subdev_state_get_crop(state, C3_ISP_RESIZER_PAD_SINK);
> +
> +	c3_isp_write(rsz->isp, C3_DISP_REG(DISP0_TOP_CRP2_START, rsz->id),
> +		     DISP_CRP2_VSTART(crop->top) | DISP_CRP2_HSTART(crop->left));
> +	c3_isp_write(rsz->isp, C3_DISP_REG(DISP0_TOP_CRP2_SIZE, rsz->id),
> +		     DISP_CRP2_VSIZE(crop->height) | DISP_CRP2_HSIZE(crop->width));
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_TOP_TOP_CTRL, rsz->id),
> +			   DISP_CRP2_EN, DISP_CRP2_EN);
> +}
> +
> +static void c3_isp_rsz_pps_size(struct c3_isp_resizer *rsz,
> +				struct c3_isp_pps_io_size *io_size)
> +{
> +	int thsize = io_size->thsize;
> +	int tvsize = io_size->tvsize;
> +	u32 ohsize = io_size->ohsize;
> +	u32 ovsize = io_size->ovsize;
> +	u32 ihsize = io_size->ihsize;
> +	u32 max_hsize = io_size->max_hsize;
> +	int step_h_integer, step_v_integer;
> +	int step_h_fraction, step_v_fraction;
> +	int yuv444to422_en;
> +
> +	/* Calculate the integer part of horizonal scaler step */
> +	step_h_integer = thsize / ohsize;
> +
> +	/* Calculate the vertical part of horizonal scaler step */
> +	step_v_integer = tvsize / ovsize;
> +
> +	/*
> +	 * Calculate the fraction part of horizonal scaler step.
> +	 * step_h_fraction = (source / dest) * 2^24,
> +	 * so step_h_fraction = ((source << 12) / dest) << 12.
> +	 */
> +	step_h_fraction = ((thsize << 12) / ohsize) << 12;
> +
> +	/*
> +	 * Calculate the fraction part of vertical scaler step
> +	 * step_v_fraction = (source / dest) * 2^24,
> +	 * so step_v_fraction = ((source << 12) / dest) << 12.
> +	 */
> +	step_v_fraction = ((tvsize << 12) / ovsize) << 12;
> +
> +	yuv444to422_en = ihsize > (max_hsize / 2) ? 1 : 0;
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_444TO422, rsz->id),
> +			   PPS_444TO422_EN_MASK, yuv444to422_en);
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_VSC_START_PHASE_STEP, rsz->id),
> +			   PPS_VSC_FRACTION_PART_MASK, step_v_fraction);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_VSC_START_PHASE_STEP, rsz->id),
> +			   PPS_VSC_INTEGER_PART_MASK,
> +			   step_v_integer << PPS_VSC_INTEGER_PART_SHIFT);
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_HSC_START_PHASE_STEP, rsz->id),
> +			   PPS_HSC_FRACTION_PART_MASK, step_h_fraction);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_HSC_START_PHASE_STEP, rsz->id),
> +			   PPS_HSC_INTEGER_PART_MASK,
> +			   step_h_integer << PPS_HSC_INTEGER_PART_SHIFT);
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_PRE_HSCALE_COEF_0, rsz->id),
> +			   PPS_PREHSC_LUMA_COEF0_MASK, PPS_PREHSC_LUMA_COEF0_INIT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_PRE_HSCALE_COEF_0, rsz->id),
> +			   PPS_PREHSC_LUMA_COEF1_MASK,
> +			   PPS_PREHSC_LUMA_COEF1_INIT << PPS_PREHSC_LUMA_COEF1_SHIFT);
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_PRE_HSCALE_COEF_1, rsz->id),
> +			   PPS_PREHSC_LUMA_COEF2_MASK, PPS_PREHSC_LUMA_COEF2_INIT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_PRE_HSCALE_COEF_1, rsz->id),
> +			   PPS_PREHSC_LUMA_COEF3_MASK,
> +			   PPS_PREHSC_LUMA_COEF3_INIT << PPS_PREHSC_LUMA_COEF3_SHIFT);
> +}
> +
> +static void c3_isp_rsz_pps_lut(struct c3_isp_resizer *rsz, u32 ctype)
> +{
> +	int i;

unsigned

> +
> +	/*
> +	 * Default value of this register is 0,
> +	 * so only need to set SCALE_LUMA_COEF_S11_MODE
> +	 * and SCALE_LUMA_CTYPE.

You can fit this in 2 lines ?

> +	 * This register needs to be written in one time.
> +	 */
> +	c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_IDX_LUMA, rsz->id),
> +		     SCALE_LUMA_COEF_S11_MODE | SCALE_LUMA_CTYPE(ctype));
> +
> +	for (i = 0; i < C3_PPS_TAP4_S11_H_NUM; i++) {
> +		c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_LUMA, rsz->id),
> +			     SCALE_COEF_LUMA_DATA0(pps_lut_tap4_s11[i][0]) |
> +			     SCALE_COEF_LUMA_DATA1(pps_lut_tap4_s11[i][1]));
> +		c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_LUMA, rsz->id),
> +			     SCALE_COEF_LUMA_DATA0(pps_lut_tap4_s11[i][2]) |
> +			     SCALE_COEF_LUMA_DATA1(pps_lut_tap4_s11[i][3]));
> +	}
> +
> +	/*
> +	 * Default value of this register is 0,
> +	 * so only need to set SCALE_CHRO_COEF_S11_MODE
> +	 * and SCALE_CHRO_CTYPE.

same

> +	 * This register needs to be written in one time.
> +	 */
> +	c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_IDX_CHRO, rsz->id),
> +		     SCALE_CHRO_COEF_S11_MODE | SCALE_CHRO_CTYPE(ctype));
> +
> +	for (i = 0; i < C3_PPS_TAP4_S11_H_NUM; i++) {
> +		c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_CHRO, rsz->id),
> +			     SCALE_COEF_CHRO_DATA0(pps_lut_tap4_s11[i][0]) |
> +			     SCALE_COEF_CHRO_DATA1(pps_lut_tap4_s11[i][1]));
> +		c3_isp_write(rsz->isp, C3_DISP_REG(ISP_SCALE0_COEF_CHRO, rsz->id),
> +			     SCALE_COEF_CHRO_DATA0(pps_lut_tap4_s11[i][2]) |
> +			     SCALE_COEF_CHRO_DATA1(pps_lut_tap4_s11[i][3]));
> +	}
> +}
> +
> +static void c3_isp_rsz_pps_disable(struct c3_isp_resizer *rsz)
> +{
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_HSC_EN_MASK, 0);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_VSC_EN_MASK, 0);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREVSC_EN_MASK, 0);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREHSC_EN_MASK, 0);
> +}
> +
> +static int c3_isp_rsz_pps_enable(struct c3_isp_resizer *rsz,
> +				 struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_rect *crop;
> +	struct v4l2_rect *cmps;
> +	int max_hsize;
> +	int hsc_en, vsc_en;
> +	int preh_en, prev_en;
> +	u32 reg_prehsc_rate;
> +	u32 reg_prevsc_flt_num;
> +	int pre_vscale_max_hsize;
> +	u32 ihsize_after_pre_hsc;
> +	u32 ihsize_after_pre_hsc_alt;
> +	u32 reg_vsc_tap_num_alt;
> +	u32 ihsize;
> +	u32 ivsize;
> +	struct c3_isp_pps_io_size io_size;
> +
> +	crop = v4l2_subdev_state_get_crop(state, C3_ISP_RESIZER_PAD_SINK);
> +	cmps = v4l2_subdev_state_get_compose(state, C3_ISP_RESIZER_PAD_SINK);
> +
> +	ihsize = crop->width;
> +	ivsize = crop->height;
> +
> +	hsc_en = (ihsize == cmps->width) ? C3_SCALE_DIS : C3_SCALE_EN;
> +	vsc_en = (ivsize == cmps->height) ? C3_SCALE_DIS : C3_SCALE_EN;
> +
> +	/* Disable pps when there no need to use pps */
> +	if (!hsc_en && !vsc_en) {
> +		c3_isp_rsz_pps_disable(rsz);
> +		return 0;
> +	}
> +
> +	/*
> +	 * Pre-scale needs to be enable
> +	 * if the down scaling factor exceeds 4.
> +	 */
> +	preh_en = (ihsize > cmps->width * 4) ? C3_SCALE_EN : C3_SCALE_DIS;
> +	prev_en = (ivsize > cmps->height * 4) ? C3_SCALE_EN : C3_SCALE_DIS;
> +
> +	if (rsz->id == C3_ISP_RSZ_2) {
> +		max_hsize = C3_ISP_MAX_WIDTH;
> +		/*
> +		 * Set vertical tap number and
> +		 * the max hsize of pre-vertical scale.
> +		 */
> +		reg_prevsc_flt_num = 4;
> +		pre_vscale_max_hsize = max_hsize / 2;
> +	} else {
> +		max_hsize = C3_ISP_DEFAULT_WIDTH;
> +		preh_en  = (ihsize > max_hsize) ? C3_SCALE_EN : C3_SCALE_DIS;
> +		/*
> +		 * Set vertical tap number and
> +		 * the max hsize of pre-vertical scale.
> +		 */
> +		if (ihsize > (max_hsize / 2) &&
> +		    ihsize <= max_hsize && prev_en) {
> +			reg_prevsc_flt_num = 2;
> +			pre_vscale_max_hsize = max_hsize;
> +		} else {
> +			reg_prevsc_flt_num = 4;
> +			pre_vscale_max_hsize = max_hsize / 2;
> +		}
> +	}
> +
> +	/*
> +	 * Set pre-horizonal scale rate and
> +	 * the hsize of after pre-horizonal scale.
> +	 */
> +	if (preh_en) {
> +		reg_prehsc_rate = 1;
> +		ihsize_after_pre_hsc = DIV_ROUND_UP(ihsize, 2);
> +	} else {
> +		reg_prehsc_rate = 0;
> +		ihsize_after_pre_hsc = ihsize;
> +	}
> +
> +	/* Change pre-horizonal scale rate */
> +	if (prev_en && ihsize_after_pre_hsc >= pre_vscale_max_hsize)
> +		reg_prehsc_rate += 1;
> +
> +	/* Set the actual hsize of after pre-horizonal scale */
> +	if (preh_en)
> +		ihsize_after_pre_hsc_alt =
> +			DIV_ROUND_UP(ihsize, 1 << reg_prehsc_rate);
> +	else
> +		ihsize_after_pre_hsc_alt = ihsize;
> +
> +	/* Set vertical scaler bank length */
> +	if (ihsize_after_pre_hsc_alt <= (max_hsize / 2))
> +		reg_vsc_tap_num_alt = 4;
> +	else if (ihsize_after_pre_hsc_alt <= max_hsize)
> +		reg_vsc_tap_num_alt = prev_en ? 2 : 4;
> +	else
> +		reg_vsc_tap_num_alt = prev_en ? 4 : 2;
> +
> +	io_size.thsize = ihsize_after_pre_hsc_alt;
> +	io_size.tvsize = prev_en ? DIV_ROUND_UP(ivsize, 2) : ivsize;
> +	io_size.ohsize = cmps->width;
> +	io_size.ovsize = cmps->height;
> +	io_size.ihsize = ihsize;
> +	io_size.max_hsize = max_hsize;
> +
> +	c3_isp_rsz_pps_size(rsz, &io_size);
> +	c3_isp_rsz_pps_lut(rsz, C3_PPS_LUT_CTYPE_0);
> +	c3_isp_rsz_pps_lut(rsz, C3_PPS_LUT_CTYPE_2);
> +
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_VSC_TAP_NUM_MASK, reg_vsc_tap_num_alt);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_HSC_TAP_NUM_MASK,
> +			   PPS_HSC_TAP_NUM_INIT << PPS_HSC_TAP_NUM_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREVSC_FLT_NUM_MASK,
> +			   reg_prevsc_flt_num << PPS_PREVSC_FLT_NUM_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREHSC_FLT_NUM_MASK,
> +			   PPS_PREHSC_FLT_NUM_INIT << PPS_PREHSC_FLT_NUM_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREVSC_RATE_MASK, prev_en << PPS_PREVSC_RATE_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREHSC_RATE_MASK, reg_prehsc_rate << PPS_PREHSC_RATE_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_HSC_EN_MASK, hsc_en << PPS_HSC_EN_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_VSC_EN_MASK, vsc_en << PPS_VSC_EN_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREVSC_EN_MASK, prev_en << PPS_PREVSC_EN_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_PREHSC_EN_MASK, preh_en << PPS_PREHSC_EN_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_HSC_NOR_RS_BITS_MASK,
> +			   PPS_HSC_NOR_RS_BITS_INIT << PPS_HSC_NOR_RS_BITS_SHIFT);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_PPS_SCALE_EN, rsz->id),
> +			   PPS_VSC_NOR_RS_BITS_MASK,
> +			   PPS_VSC_NOR_RS_BITS_INIT << PPS_VSC_NOR_RS_BITS_SHIFT);
> +
> +	return 0;
> +}
> +
> +void c3_isp_rsz_start(struct c3_isp_resizer *rsz)
> +{
> +	struct v4l2_subdev_state *state;
> +
> +	state = v4l2_subdev_lock_and_get_active_state(&rsz->sd);
> +
> +	c3_isp_rsz_cfg_fmt(rsz, state);
> +	c3_isp_rsz_crop_enable(rsz, state);
> +	c3_isp_rsz_pps_enable(rsz, state);
> +
> +	c3_isp_update_bits(rsz->isp, ISP_TOP_PATH_EN,
> +			   TOP_DISP_EN(rsz->id), TOP_DISP_EN(rsz->id));
> +
> +	v4l2_subdev_unlock_state(state);
> +}
> +
> +void c3_isp_rsz_stop(struct c3_isp_resizer *rsz)
> +{
> +	c3_isp_update_bits(rsz->isp, ISP_TOP_PATH_EN, TOP_DISP_EN(rsz->id), 0);
> +	c3_isp_update_bits(rsz->isp, C3_DISP_REG(DISP0_TOP_TOP_CTRL, rsz->id),
> +			   DISP_CRP2_EN, 0x0);
> +
> +	c3_isp_rsz_pps_disable(rsz);
> +}
> +
> +static int c3_isp_rsz_cfg_routing(struct v4l2_subdev *sd,
> +				  struct v4l2_subdev_state *state,
> +				  struct v4l2_subdev_krouting *routing)
> +{
> +	static const struct v4l2_mbus_framefmt format = {
> +		.width = C3_ISP_DEFAULT_WIDTH,
> +		.height = C3_ISP_DEFAULT_HEIGHT,
> +		.code = C3_ISP_RSZ_DEF_PAD_FMT,
> +		.field = V4L2_FIELD_NONE,
> +		.colorspace = V4L2_COLORSPACE_SRGB,
> +		.ycbcr_enc = V4L2_YCBCR_ENC_601,
> +		.quantization = V4L2_QUANTIZATION_LIM_RANGE,
> +		.xfer_func = V4L2_XFER_FUNC_SRGB,
> +	};
> +	int ret;
> +
> +	ret = v4l2_subdev_routing_validate(sd, routing, V4L2_SUBDEV_ROUTING_ONLY_1_TO_1);
> +	if (ret)
> +		return ret;
> +
> +	ret = v4l2_subdev_set_routing_with_fmt(sd, state, routing, &format);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_rsz_init_routing(struct v4l2_subdev *sd,
> +				   struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_subdev_route routes;
> +	struct v4l2_subdev_krouting routing;
> +
> +	routes.sink_pad = C3_ISP_RESIZER_PAD_SINK;
> +	routes.sink_stream = 0;
> +	routes.source_pad = C3_ISP_RESIZER_PAD_SOURCE;
> +	routes.source_stream = 0;
> +	routes.flags = V4L2_SUBDEV_ROUTE_FL_ACTIVE;
> +
> +	routing.num_routes = 1;
> +	routing.routes = &routes;
> +
> +	return c3_isp_rsz_cfg_routing(sd, state, &routing);
> +}
> +
> +static int c3_isp_rsz_set_routing(struct v4l2_subdev *sd,
> +				  struct v4l2_subdev_state *state,
> +				  enum v4l2_subdev_format_whence which,
> +				  struct v4l2_subdev_krouting *routing)
> +{
> +	bool is_streaming = v4l2_subdev_is_streaming(sd);
> +
> +	if (which == V4L2_SUBDEV_FORMAT_ACTIVE && is_streaming)
> +		return -EBUSY;
> +
> +	return c3_isp_rsz_cfg_routing(sd, state, routing);
> +}

Unless there are reasons I missed, I would drop routing support from
the resizers

> +
> +static int c3_isp_rsz_enum_mbus_code(struct v4l2_subdev *sd,
> +				     struct v4l2_subdev_state *state,
> +				     struct v4l2_subdev_mbus_code_enum *code)
> +{
> +	if (code->index >= ARRAY_SIZE(c3_isp_rsz_mbus_formats))
> +		return -EINVAL;
> +
> +	code->code = c3_isp_rsz_mbus_formats[code->index].mbus_code;
> +
> +	return 0;
> +}
> +
> +static void c3_isp_rsz_set_sink_fmt(struct v4l2_subdev_state *state,
> +				    struct v4l2_subdev_format *format)
> +{
> +	struct v4l2_mbus_framefmt *sink_fmt;
> +	const struct c3_isp_mbus_format_info *isp_fmt;
> +	struct v4l2_rect *sink_crop;
> +	struct v4l2_rect *sink_cmps;
> +
> +	sink_fmt = v4l2_subdev_state_get_format(state, format->pad);
> +	sink_crop = v4l2_subdev_state_get_crop(state, format->pad);
> +	sink_cmps = v4l2_subdev_state_get_compose(state, format->pad);
> +
> +	isp_fmt = rsz_find_format_by_code(format->format.code, format->pad);
> +	if (!isp_fmt)
> +		sink_fmt->code = C3_ISP_RSZ_DEF_PAD_FMT;
> +	else
> +		sink_fmt->code = format->format.code;
> +
> +	sink_fmt->width = clamp_t(u32, format->format.width,
> +				  C3_ISP_MIN_WIDTH, C3_ISP_MAX_WIDTH);
> +	sink_fmt->height = clamp_t(u32, format->format.height,
> +				   C3_ISP_MIN_HEIGHT, C3_ISP_MAX_HEIGHT);
> +
> +	sink_crop->width = sink_fmt->width;
> +	sink_crop->height = sink_fmt->height;
> +	sink_crop->left = 0;
> +	sink_crop->top = 0;
> +
> +	sink_cmps->width = sink_crop->width;
> +	sink_cmps->height = sink_crop->height;
> +	sink_cmps->left = 0;
> +	sink_cmps->top = 0;
> +
> +	format->format = *sink_fmt;
> +}
> +
> +static void c3_isp_rsz_set_source_fmt(struct v4l2_subdev_state *state,
> +				      struct v4l2_subdev_format *format)
> +{
> +	const struct c3_isp_mbus_format_info *rsz_fmt;
> +	struct v4l2_mbus_framefmt *src_fmt;
> +	struct v4l2_rect *sink_crop;
> +	struct v4l2_rect *sink_cmps;
> +
> +	src_fmt = v4l2_subdev_state_get_format(state, format->pad);
> +	sink_cmps = v4l2_subdev_state_get_compose(state, C3_ISP_RESIZER_PAD_SINK);
> +	sink_crop = v4l2_subdev_state_get_crop(state, C3_ISP_RESIZER_PAD_SINK);
> +
> +	rsz_fmt = rsz_find_format_by_code(format->format.code, format->pad);
> +	if (!rsz_fmt)
> +		src_fmt->code = C3_ISP_RSZ_DEF_PAD_FMT;
> +	else
> +		src_fmt->code = format->format.code;
> +
> +	src_fmt->width = clamp_t(u32, format->format.width,
> +				 C3_ISP_MIN_WIDTH, sink_crop->width);
> +	src_fmt->height = clamp_t(u32, format->format.height,
> +				  C3_ISP_MIN_HEIGHT, sink_crop->height);
> +
> +	/* The sink compose size must be same with the source size. */
> +	sink_cmps->width = src_fmt->width;
> +	sink_cmps->height = src_fmt->height;

Shouldn't it be the other way around ? The source sizes should always
match the sink compose rectangle sizes ?

> +
> +	format->format = *src_fmt;
> +}
> +
> +static int c3_isp_rsz_set_fmt(struct v4l2_subdev *sd,
> +			      struct v4l2_subdev_state *state,
> +			      struct v4l2_subdev_format *format)
> +{
> +	if (format->pad == C3_ISP_RESIZER_PAD_SINK) {
> +		c3_isp_rsz_set_sink_fmt(state, format);
> +	} else if (format->pad == C3_ISP_RESIZER_PAD_SOURCE) {
> +		c3_isp_rsz_set_source_fmt(state, format);
> +	} else {

This can't happen

> +		dev_err(sd->dev, "Invalid pad: %u\n", format->pad);
> +		return -ENOTTY;
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_rsz_get_selection(struct v4l2_subdev *sd,
> +				    struct v4l2_subdev_state *state,
> +				    struct v4l2_subdev_selection *sel)
> +{
> +	struct v4l2_mbus_framefmt *fmt;
> +	struct v4l2_rect *crop;
> +	struct v4l2_rect *cmps;
> +
> +	if (sel->pad == C3_ISP_RESIZER_PAD_SOURCE)
> +		return -EINVAL;
> +
> +	switch (sel->target) {
> +	case V4L2_SEL_TGT_CROP_BOUNDS:
> +		fmt = v4l2_subdev_state_get_format(state, sel->pad);
> +		sel->r.width = fmt->width;
> +		sel->r.height = fmt->height;
> +		sel->r.left = 0;
> +		sel->r.top = 0;
> +		break;
> +	case V4L2_SEL_TGT_COMPOSE_BOUNDS:
> +		crop = v4l2_subdev_state_get_crop(state, sel->pad);
> +		sel->r.width = crop->width;
> +		sel->r.height = crop->height;
> +		sel->r.left = 0;
> +		sel->r.top = 0;
> +		break;
> +	case V4L2_SEL_TGT_CROP:
> +		crop = v4l2_subdev_state_get_crop(state, sel->pad);
> +		sel->r = *crop;
> +		break;
> +	case V4L2_SEL_TGT_COMPOSE:
> +		cmps = v4l2_subdev_state_get_compose(state, sel->pad);
> +		sel->r = *cmps;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_rsz_set_selection(struct v4l2_subdev *sd,
> +				    struct v4l2_subdev_state *state,
> +				    struct v4l2_subdev_selection *sel)
> +{
> +	struct v4l2_mbus_framefmt *fmt;
> +	struct v4l2_rect *crop;
> +	struct v4l2_rect *cmps;
> +
> +	if (sel->pad == C3_ISP_RESIZER_PAD_SOURCE)
> +		return -EINVAL;
> +
> +	switch (sel->target) {
> +	case V4L2_SEL_TGT_CROP:
> +		fmt = v4l2_subdev_state_get_format(state, sel->pad);
> +		crop = v4l2_subdev_state_get_crop(state, sel->pad);
> +
> +		sel->r.left = clamp_t(s32, sel->r.left, 0, fmt->width - 1);
> +		sel->r.top = clamp_t(s32, sel->r.top, 0, fmt->height - 1);
> +		sel->r.width = clamp(sel->r.width, C3_ISP_MIN_WIDTH,
> +				     fmt->width - sel->r.left);
> +		sel->r.height = clamp(sel->r.height, C3_ISP_MIN_HEIGHT,
> +				      fmt->height - sel->r.top);
> +
> +		crop->width = ALIGN(sel->r.width, 2);
> +		crop->height = ALIGN(sel->r.height, 2);
> +		crop->left = sel->r.left;
> +		crop->top = sel->r.top;
> +
> +		sel->r = *crop;
> +		break;
> +	case V4L2_SEL_TGT_COMPOSE:
> +		crop = v4l2_subdev_state_get_crop(state, sel->pad);
> +		cmps = v4l2_subdev_state_get_compose(state, sel->pad);
> +
> +		sel->r.left = 0;
> +		sel->r.top = 0;
> +		sel->r.width = clamp(sel->r.width, C3_ISP_MIN_WIDTH, crop->width);
> +		sel->r.height = clamp(sel->r.height, C3_ISP_MIN_HEIGHT, crop->height);
> +
> +		cmps->width = ALIGN(sel->r.width, 2);
> +		cmps->height = ALIGN(sel->r.height, 2);
> +		cmps->left = sel->r.left;
> +		cmps->top = sel->r.top;
> +
> +		sel->r = *cmps;
> +
> +		fmt = v4l2_subdev_state_get_format(state, C3_ISP_RESIZER_PAD_SOURCE);
> +		fmt->width = cmps->width;
> +		fmt->height = cmps->height;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
> +static int c3_isp_rsz_init_state(struct v4l2_subdev *sd,
> +				 struct v4l2_subdev_state *state)
> +{
> +	struct v4l2_mbus_framefmt *sink_fmt;
> +	struct v4l2_mbus_framefmt *src_fmt;
> +	struct v4l2_rect *crop;
> +	struct v4l2_rect *cmps;
> +
> +	sink_fmt = v4l2_subdev_state_get_format(state, C3_ISP_RESIZER_PAD_SINK);
> +	sink_fmt->width = C3_ISP_DEFAULT_WIDTH;
> +	sink_fmt->height = C3_ISP_DEFAULT_HEIGHT;
> +	sink_fmt->field = V4L2_FIELD_NONE;
> +	sink_fmt->code = C3_ISP_RSZ_DEF_PAD_FMT;
> +	sink_fmt->colorspace = V4L2_COLORSPACE_SRGB;
> +	sink_fmt->xfer_func = V4L2_XFER_FUNC_SRGB;
> +	sink_fmt->ycbcr_enc = V4L2_YCBCR_ENC_601;
> +	sink_fmt->quantization = V4L2_QUANTIZATION_LIM_RANGE;
> +
> +	crop = v4l2_subdev_state_get_crop(state, C3_ISP_RESIZER_PAD_SINK);
> +	crop->width = C3_ISP_DEFAULT_WIDTH;
> +	crop->height = C3_ISP_DEFAULT_HEIGHT;
> +	crop->left = 0;
> +	crop->top = 0;
> +
> +	cmps = v4l2_subdev_state_get_compose(state, C3_ISP_RESIZER_PAD_SINK);
> +	cmps->width = C3_ISP_DEFAULT_WIDTH;
> +	cmps->height = C3_ISP_DEFAULT_HEIGHT;
> +	cmps->left = 0;
> +	cmps->top = 0;
> +
> +	src_fmt = v4l2_subdev_state_get_format(state, C3_ISP_RESIZER_PAD_SOURCE);
> +	*src_fmt = *sink_fmt;
> +
> +	return c3_isp_rsz_init_routing(sd, state);
> +}
> +
> +static const struct v4l2_subdev_pad_ops c3_isp_rsz_pad_ops = {
> +	.enum_mbus_code = c3_isp_rsz_enum_mbus_code,
> +	.get_fmt = v4l2_subdev_get_fmt,
> +	.set_fmt = c3_isp_rsz_set_fmt,
> +	.get_selection = c3_isp_rsz_get_selection,
> +	.set_selection = c3_isp_rsz_set_selection,
> +	.set_routing = c3_isp_rsz_set_routing,
> +};
> +
> +static const struct v4l2_subdev_ops c3_isp_rsz_subdev_ops = {
> +	.pad = &c3_isp_rsz_pad_ops,
> +};
> +
> +static const struct v4l2_subdev_internal_ops c3_isp_rsz_internal_ops = {
> +	.init_state = c3_isp_rsz_init_state,
> +};
> +
> +/* Media entity operations */
> +static const struct media_entity_operations c3_isp_rsz_entity_ops = {
> +	.link_validate = v4l2_subdev_link_validate,
> +};
> +
> +static int c3_isp_rsz_register(struct c3_isp_resizer *rsz)
> +{
> +	struct v4l2_subdev *sd = &rsz->sd;
> +	int ret;
> +
> +	v4l2_subdev_init(sd, &c3_isp_rsz_subdev_ops);
> +	sd->owner = THIS_MODULE;
> +	sd->flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
> +	sd->internal_ops = &c3_isp_rsz_internal_ops;
> +	snprintf(sd->name, sizeof(sd->name), "isp-resizer%u", rsz->id);

maybe "c3-isp_resizer%u"

> +
> +	sd->entity.function = MEDIA_ENT_F_PROC_VIDEO_SCALER;
> +	sd->entity.ops = &c3_isp_rsz_entity_ops;
> +
> +	sd->dev = rsz->isp->dev;
> +	v4l2_set_subdevdata(sd, rsz);
> +
> +	rsz->pads[C3_ISP_RESIZER_PAD_SINK].flags = MEDIA_PAD_FL_SINK;
> +	rsz->pads[C3_ISP_RESIZER_PAD_SOURCE].flags = MEDIA_PAD_FL_SOURCE;
> +	ret = media_entity_pads_init(&sd->entity, C3_ISP_RESIZER_PAD_MAX, rsz->pads);
> +	if (ret)
> +		return ret;
> +
> +	ret = v4l2_subdev_init_finalize(sd);
> +	if (ret)
> +		goto err_entity_cleanup;
> +
> +	ret = v4l2_device_register_subdev(&rsz->isp->v4l2_dev, sd);
> +	if (ret)
> +		goto err_subdev_cleanup;
> +
> +	return 0;
> +
> +err_subdev_cleanup:
> +	v4l2_subdev_cleanup(sd);
> +err_entity_cleanup:
> +	media_entity_cleanup(&sd->entity);
> +	return ret;
> +}
> +
> +static void c3_isp_rsz_unregister(struct c3_isp_resizer *rsz)
> +{
> +	struct v4l2_subdev *sd = &rsz->sd;
> +
> +	v4l2_device_unregister_subdev(sd);
> +	v4l2_subdev_cleanup(sd);
> +	media_entity_cleanup(&sd->entity);
> +}
> +
> +int c3_isp_resizers_register(struct c3_isp_device *isp)
> +{
> +	u32 i;
> +	int ret;
> +
> +	for (i = C3_ISP_RSZ_0; i < C3_ISP_NUM_RSZ; i++) {
> +		struct c3_isp_resizer *rsz = &isp->resizers[i];
> +
> +		rsz->id = i;
> +		rsz->isp = isp;
> +
> +		if (rsz->id == C3_ISP_RSZ_0)
> +			rsz->cap = &isp->caps[C3_ISP_CAP_DEV_0];
> +		else if (rsz->id == C3_ISP_RSZ_1)
> +			rsz->cap = &isp->caps[C3_ISP_CAP_DEV_1];
> +		else
> +			rsz->cap = &isp->caps[C3_ISP_CAP_DEV_2];
> +
> +		ret = c3_isp_rsz_register(rsz);
> +		if (ret) {
> +			rsz->isp = NULL;
> +			c3_isp_resizers_unregister(isp);
> +			return ret;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +void c3_isp_resizers_unregister(struct c3_isp_device *isp)
> +{
> +	u32 i;
> +
> +	for (i = C3_ISP_RSZ_0; i < C3_ISP_NUM_RSZ; i++) {
> +		struct c3_isp_resizer *rsz = &isp->resizers[i];
> +
> +		if (rsz->isp)
> +			c3_isp_rsz_unregister(rsz);
> +	};
> +}
> diff --git a/drivers/media/platform/amlogic/c3-isp/c3-isp-stats.c b/drivers/media/platform/amlogic/c3-isp/c3-isp-stats.c
> new file mode 100644
> index 000000000000..72024442d48f
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/c3-isp-stats.c
> @@ -0,0 +1,488 @@
> +// SPDX-License-Identifier: (GPL-2.0-only OR MIT)
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#include <linux/cleanup.h>
> +#include <linux/pm_runtime.h>
> +
> +#include <media/v4l2-ctrls.h>
> +#include <media/v4l2-event.h>
> +#include <media/v4l2-ioctl.h>
> +#include <media/v4l2-mc.h>
> +#include <media/videobuf2-dma-contig.h>
> +
> +#include "c3-isp-common.h"
> +#include "c3-isp-regs.h"
> +#include "include/uapi/c3-isp-config.h"
> +
> +/* Hardware configuration */
> +
> +static void c3_isp_stats_af_init(struct c3_isp_stats *stats)
> +{
> +	c3_isp_write(stats->isp, ISP_AF_ROI0_WIN01,
> +		     AF_ROI_XYWIN_00_INIT | AF_ROI_XYWIN_01_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AF_ROI0_WIN23,
> +		     AF_ROI_XYWIN_02_INIT | AF_ROI_XYWIN_03_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AF_ROI1_WIN01,
> +		     AF_ROI_XYWIN_10_INIT | AF_ROI_XYWIN_11_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AF_ROI1_WIN23,
> +		     AF_ROI_XYWIN_12_INIT | AF_ROI_XYWIN_13_INIT);
> +
> +	/* 0: old statistics output, 1: new statistics output. */
> +	c3_isp_update_bits(stats->isp, ISP_AF_EN_CTRL,
> +			   AF_STAT_SELECT, 0x1 << AF_STAT_SELECT_SHIFT);
> +}
> +
> +static void c3_isp_stats_ae_init(struct c3_isp_stats *stats)
> +{
> +	c3_isp_write(stats->isp, ISP_AE_ROI0_WIN01,
> +		     AE_ROI_XYWIN_00_INIT | AE_ROI_XYWIN_01_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AE_ROI0_WIN23,
> +		     AE_ROI_XYWIN_02_INIT | AE_ROI_XYWIN_03_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AE_ROI1_WIN01,
> +		     AE_ROI_XYWIN_10_INIT | AE_ROI_XYWIN_11_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AE_ROI1_WIN23,
> +		     AE_ROI_XYWIN_12_INIT | AE_ROI_XYWIN_13_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AE_STAT_THD01,
> +		     AE_STAT_THRD0_INIT | AE_STAT_THRD1_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AE_STAT_THD23,
> +		     AE_STAT_THRD2_INIT | AE_STAT_THRD3_INIT);
> +
> +	/* Set 0 when ae_stat_switch is not 0 */
> +	c3_isp_update_bits(stats->isp, ISP_AE_CTRL,
> +			   AE_CTRL_INPUT_2LINE_TOGETHER, 0);
> +
> +	/* Configure ae luma mode */
> +	c3_isp_update_bits(stats->isp, ISP_AE_CTRL, AE_CTRL_LUMA_MODE_MASK,
> +			   AE_CTRL_LUMA_MODE_FILTER << AE_CTRL_LUMA_MODE_SHIFT);
> +}
> +
> +static void c3_isp_stats_awb_init(struct c3_isp_stats *stats)
> +{
> +	/* Initialize the awb statistics rectangle of image */
> +	c3_isp_write(stats->isp, ISP_AWB_STAT_ROI0_WIN01,
> +		     AWB_ROI_XYWIN_00_INIT | AWB_ROI_XYWIN_01_INIT);
> +	c3_isp_write(stats->isp, ISP_AWB_STAT_ROI0_WIN23,
> +		     AWB_ROI_XYWIN_02_INIT | AWB_ROI_XYWIN_03_INIT);
> +
> +	c3_isp_write(stats->isp, ISP_AWB_STAT_ROI1_WIN01,
> +		     AWB_ROI_XYWIN_10_INIT | AWB_ROI_XYWIN_11_INIT);
> +	c3_isp_write(stats->isp, ISP_AWB_STAT_ROI1_WIN23,
> +		     AWB_ROI_XYWIN_12_INIT | AWB_ROI_XYWIN_13_INIT);
> +}
> +
> +static void c3_isp_stats_cfg_dmawr_addr(struct c3_isp_stats *stats)
> +{
> +	struct c3_isp_device *isp = stats->isp;
> +	struct c3_isp_stats_info *stats_info = stats->buff->vaddr;
> +	u32 awb_dma_size = sizeof(stats_info->awb_stats);
> +	u32 ae_dma_size = sizeof(stats_info->ae_stats);
> +	u32 awb_dma_addr = stats->buff->paddr;
> +	u32 af_dma_addr;
> +	u32 ae_dma_addr;
> +
> +	ae_dma_addr = awb_dma_addr + awb_dma_size;
> +	af_dma_addr = ae_dma_addr + ae_dma_size;
> +
> +	c3_isp_update_bits(isp, VIU_DMAWR_BADDR0, VIU_DMAWR_AF_BADDR_MASK,
> +			   VIU_DMAWR_AF_BADDR(af_dma_addr));
> +	c3_isp_update_bits(isp, VIU_DMAWR_BADDR1, VIU_DMAWR_AWB_BADDR_MASK,
> +			   VIU_DMAWR_AWB_BADDR(awb_dma_addr));
> +	c3_isp_update_bits(isp, VIU_DMAWR_BADDR2, VIU_DMAWR_AE_BADDR_MASK,
> +			   VIU_DMAWR_AE_BADDR(ae_dma_addr));
> +}
> +
> +static void c3_isp_stats_enable(struct c3_isp_stats *stats)
> +{
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AE_STAT_EN, TOP_3A_AE_STAT_EN);
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AWB_STAT_EN, TOP_3A_AWB_STAT_EN);
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AF_STAT_EN, TOP_3A_AF_STAT_EN);
> +}
> +
> +static void c3_isp_stats_disable(struct c3_isp_stats *stats)
> +{
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AE_STAT_EN, 0);
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AWB_STAT_EN, 0);
> +	c3_isp_update_bits(stats->isp, ISP_TOP_3A_STAT_CRTL,
> +			   TOP_3A_AF_STAT_EN, 0);
> +}
> +
> +/* The unit of dma_size is 16 bytes */
> +static void c3_isp_stats_cfg_dmawr_size(struct c3_isp_stats *stats)
> +{
> +	u32 dma_size;
> +
> +	dma_size = sizeof(struct af_stats_info) / 16;
> +	c3_isp_update_bits(stats->isp, VIU_DMAWR_SIZE0,
> +			   VIU_DMAWR_SIZE_AF_MASK, dma_size);
> +
> +	dma_size = sizeof(struct awb_stats_info) / 16;
> +	c3_isp_update_bits(stats->isp, VIU_DMAWR_SIZE0, VIU_DMAWR_SIZE_AWB_MASK,
> +			   dma_size << VIU_DMAWR_SIZE_AWB_SHIFT);
> +
> +	dma_size = sizeof(struct ae_stats_info) / 16;
> +	c3_isp_update_bits(stats->isp, VIU_DMAWR_SIZE1,
> +			   VIU_DMAWR_SIZE_AE_MASK, dma_size);
> +}
> +
> +static void c3_isp_stats_cfg_buff(struct c3_isp_stats *stats)
> +{
> +	stats->buff = list_first_entry_or_null(&stats->pending,
> +					       struct c3_isp_vb2_buffer, list);
> +	if (stats->buff) {
> +		c3_isp_stats_cfg_dmawr_addr(stats);
> +		list_del(&stats->buff->list);
> +	}
> +}
> +
> +static void c3_isp_stats_start(struct c3_isp_stats *stats)
> +{
> +	c3_isp_stats_af_init(stats);
> +	c3_isp_stats_ae_init(stats);
> +	c3_isp_stats_awb_init(stats);
> +
> +	c3_isp_stats_cfg_dmawr_size(stats);
> +	c3_isp_stats_cfg_buff(stats);
> +	c3_isp_stats_enable(stats);
> +
> +	stats->is_streaming = true;
> +}
> +
> +static void c3_isp_stats_stop(struct c3_isp_stats *stats)
> +{
> +	stats->is_streaming = false;
> +
> +	c3_isp_stats_disable(stats);
> +}
> +
> +static void c3_isp_stats_return_buffers(struct c3_isp_stats *stats,
> +					enum vb2_buffer_state state)
> +{
> +	unsigned long flags;
> +	struct c3_isp_vb2_buffer *buff;
> +
> +	spin_lock_irqsave(&stats->buff_lock, flags);
> +
> +	if (stats->buff) {
> +		vb2_buffer_done(&stats->buff->vb.vb2_buf, state);
> +		stats->buff = NULL;
> +	}
> +
> +	while (!list_empty(&stats->pending)) {
> +		buff = list_first_entry(&stats->pending,
> +					struct c3_isp_vb2_buffer, list);
> +		list_del(&buff->list);
> +		vb2_buffer_done(&buff->vb.vb2_buf, state);
> +	}
> +
> +	spin_unlock_irqrestore(&stats->buff_lock, flags);
> +}
> +
> +static int c3_isp_stats_querycap(struct file *file, void *fh,
> +				 struct v4l2_capability *cap)
> +{
> +	strscpy(cap->driver, C3_ISP_DRIVER_NAME, sizeof(cap->driver));
> +	strscpy(cap->card, "AML C3 ISP", sizeof(cap->card));
> +
> +	return 0;
> +}
> +
> +static int c3_isp_stats_enum_fmt(struct file *file, void *fh,
> +				 struct v4l2_fmtdesc *f)
> +{
> +	struct c3_isp_stats *stats = video_drvdata(file);
> +
> +	if (f->index > 0 || f->type != stats->vb2_q.type)
> +		return -EINVAL;
> +
> +	f->pixelformat = V4L2_META_FMT_C3ISP_STATS;
> +
> +	return 0;
> +}
> +
> +static int c3_isp_stats_g_fmt(struct file *file, void *fh,
> +			      struct v4l2_format *f)
> +{
> +	struct c3_isp_stats *stats = video_drvdata(file);
> +
> +	f->fmt.meta = stats->vfmt.fmt.meta;
> +
> +	return 0;
> +}
> +
> +static const struct v4l2_ioctl_ops isp_stats_v4l2_ioctl_ops = {
> +	.vidioc_querycap                = c3_isp_stats_querycap,
> +	.vidioc_enum_fmt_meta_cap       = c3_isp_stats_enum_fmt,
> +	.vidioc_g_fmt_meta_cap          = c3_isp_stats_g_fmt,
> +	.vidioc_s_fmt_meta_cap          = c3_isp_stats_g_fmt,
> +	.vidioc_try_fmt_meta_cap        = c3_isp_stats_g_fmt,
> +	.vidioc_reqbufs	                = vb2_ioctl_reqbufs,
> +	.vidioc_querybuf                = vb2_ioctl_querybuf,
> +	.vidioc_qbuf                    = vb2_ioctl_qbuf,
> +	.vidioc_expbuf                  = vb2_ioctl_expbuf,
> +	.vidioc_dqbuf                   = vb2_ioctl_dqbuf,
> +	.vidioc_prepare_buf             = vb2_ioctl_prepare_buf,
> +	.vidioc_create_bufs             = vb2_ioctl_create_bufs,
> +	.vidioc_streamon                = vb2_ioctl_streamon,
> +	.vidioc_streamoff               = vb2_ioctl_streamoff,
> +	.vidioc_subscribe_event         = v4l2_ctrl_subscribe_event,
> +	.vidioc_unsubscribe_event       = v4l2_event_unsubscribe,
> +};
> +
> +static const struct v4l2_file_operations isp_stats_v4l2_fops = {
> +	.open = v4l2_fh_open,
> +	.release = vb2_fop_release,
> +	.poll = vb2_fop_poll,
> +	.unlocked_ioctl = video_ioctl2,
> +	.mmap = vb2_fop_mmap,
> +};
> +
> +static int c3_isp_stats_vb2_queue_setup(struct vb2_queue *q,
> +					unsigned int *num_buffers,
> +					unsigned int *num_planes,
> +					unsigned int sizes[],
> +					struct device *alloc_devs[])
> +{
> +	if (*num_planes) {
> +		if (*num_planes != 1)
> +			return -EINVAL;
> +
> +		if (sizes[0] < sizeof(struct c3_isp_stats_info))
> +			return -EINVAL;
> +	} else {
> +		*num_planes = 1;
> +		sizes[0] = sizeof(struct c3_isp_stats_info);
> +	}
> +
> +	return 0;
> +}
> +
> +static void c3_isp_stats_vb2_buf_queue(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +			container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_stats *stats = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned long flags;
> +
> +	spin_lock_irqsave(&stats->buff_lock, flags);
> +
> +	list_add_tail(&buf->list, &stats->pending);
> +
> +	spin_unlock_irqrestore(&stats->buff_lock, flags);
> +}
> +
> +static int c3_isp_stats_vb2_buf_prepare(struct vb2_buffer *vb)
> +{
> +	struct c3_isp_stats *stats = vb2_get_drv_priv(vb->vb2_queue);
> +	unsigned int size = stats->vfmt.fmt.meta.buffersize;
> +
> +	if (vb2_plane_size(vb, 0) < size) {
> +		dev_err(stats->isp->dev,
> +			"User buffer too small (%ld < %u)\n",
> +			vb2_plane_size(vb, 0), size);
> +		return -EINVAL;
> +	}
> +
> +	vb2_set_plane_payload(vb, 0, size);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_stats_vb2_buf_init(struct vb2_buffer *vb)
> +{
> +	struct vb2_v4l2_buffer *v4l2_buf = to_vb2_v4l2_buffer(vb);
> +	struct c3_isp_vb2_buffer *buf =
> +			container_of(v4l2_buf, struct c3_isp_vb2_buffer, vb);
> +	struct c3_isp_stats *stats = vb2_get_drv_priv(vb->vb2_queue);
> +
> +	buf->vaddr = vb2_plane_vaddr(vb, 0);
> +	buf->paddr = vb2_dma_contig_plane_dma_addr(vb, 0);
> +
> +	memset(buf->vaddr, 0, stats->vfmt.fmt.meta.buffersize);
> +
> +	return 0;
> +}
> +
> +static int c3_isp_stats_vb2_start_streaming(struct vb2_queue *q,
> +					    unsigned int count)
> +{
> +	struct c3_isp_stats *stats = vb2_get_drv_priv(q);
> +	int ret;
> +
> +	guard(mutex)(&stats->isp->lock);
> +
> +	ret = pm_runtime_resume_and_get(stats->isp->dev);
> +	if (ret)
> +		return ret;
> +
> +	ret = video_device_pipeline_start(&stats->vdev, &stats->isp->pipe);
> +	if (ret) {
> +		dev_err(stats->isp->dev,
> +			"Failed to start stats pipeline: %d\n", ret);
> +		goto err_pm_put;
> +	}
> +
> +	if (c3_isp_pipeline_ready(stats->isp)) {
> +		ret = v4l2_subdev_enable_streams(&stats->isp->core.sd,
> +						 C3_ISP_CORE_PAD_SOURCE_STATS,
> +						 BIT(0));
> +		if (ret)
> +			goto err_pipeline_stop;
> +	}
> +
> +	c3_isp_stats_start(stats);
> +
> +	return 0;
> +
> +err_pipeline_stop:
> +	video_device_pipeline_stop(&stats->vdev);
> +err_pm_put:
> +	pm_runtime_put(stats->isp->dev);
> +	c3_isp_stats_return_buffers(stats, VB2_BUF_STATE_QUEUED);
> +	return ret;
> +}
> +
> +static void c3_isp_stats_vb2_stop_streaming(struct vb2_queue *q)
> +{
> +	struct c3_isp_stats *stats = vb2_get_drv_priv(q);
> +
> +	guard(mutex)(&stats->isp->lock);
> +
> +	c3_isp_stats_stop(stats);
> +	c3_isp_stats_return_buffers(stats, VB2_BUF_STATE_ERROR);
> +
> +	if (stats->isp->pipe.start_count == 1)
> +		v4l2_subdev_disable_streams(&stats->isp->core.sd,
> +					    C3_ISP_CORE_PAD_SOURCE_STATS,
> +					    BIT(0));
> +
> +	video_device_pipeline_stop(&stats->vdev);
> +	pm_runtime_put(stats->isp->dev);
> +}
> +
> +static const struct vb2_ops isp_stats_vb2_ops = {
> +	.queue_setup = c3_isp_stats_vb2_queue_setup,
> +	.buf_queue = c3_isp_stats_vb2_buf_queue,
> +	.buf_prepare = c3_isp_stats_vb2_buf_prepare,
> +	.buf_init = c3_isp_stats_vb2_buf_init,
> +	.wait_prepare = vb2_ops_wait_prepare,
> +	.wait_finish = vb2_ops_wait_finish,
> +	.start_streaming = c3_isp_stats_vb2_start_streaming,
> +	.stop_streaming = c3_isp_stats_vb2_stop_streaming,
> +};
> +
> +int c3_isp_stats_register(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_stats *stats = &isp->stats;
> +	struct video_device *vdev = &stats->vdev;
> +	struct vb2_queue *vb2_q = &stats->vb2_q;
> +	int ret;
> +
> +	memset(stats, 0, sizeof(*stats));
> +	stats->vfmt.fmt.meta.dataformat = V4L2_META_FMT_C3ISP_STATS;
> +	stats->vfmt.fmt.meta.buffersize = sizeof(struct c3_isp_stats_info);
> +	stats->isp = isp;
> +	INIT_LIST_HEAD(&stats->pending);
> +	spin_lock_init(&stats->buff_lock);
> +
> +	mutex_init(&stats->lock);
> +
> +	snprintf(vdev->name, sizeof(vdev->name), "isp-stats");
> +	vdev->fops = &isp_stats_v4l2_fops;
> +	vdev->ioctl_ops = &isp_stats_v4l2_ioctl_ops;
> +	vdev->v4l2_dev = &isp->v4l2_dev;
> +	vdev->lock = &stats->lock;
> +	vdev->minor = -1;
> +	vdev->queue = vb2_q;
> +	vdev->release = video_device_release_empty;
> +	vdev->device_caps = V4L2_CAP_META_CAPTURE | V4L2_CAP_STREAMING;
> +	vdev->vfl_dir = VFL_DIR_RX;
> +	video_set_drvdata(vdev, stats);
> +
> +	vb2_q->drv_priv = stats;
> +	vb2_q->mem_ops = &vb2_dma_contig_memops;
> +	vb2_q->ops = &isp_stats_vb2_ops;
> +	vb2_q->type = V4L2_BUF_TYPE_META_CAPTURE;
> +	vb2_q->io_modes = VB2_DMABUF | VB2_MMAP;
> +	vb2_q->timestamp_flags = V4L2_BUF_FLAG_TIMESTAMP_MONOTONIC;
> +	vb2_q->buf_struct_size = sizeof(struct c3_isp_vb2_buffer);
> +	vb2_q->dev = isp->dev;
> +	vb2_q->lock = &stats->lock;
> +	vb2_q->min_queued_buffers = 2;
> +
> +	ret = vb2_queue_init(vb2_q);
> +	if (ret)
> +		goto err_destroy;
> +
> +	stats->pad.flags = MEDIA_PAD_FL_SINK;
> +	ret = media_entity_pads_init(&vdev->entity, 1, &stats->pad);
> +	if (ret)
> +		goto err_queue_release;
> +
> +	ret = video_register_device(vdev, VFL_TYPE_VIDEO, -1);
> +	if (ret) {
> +		dev_err(isp->dev,
> +			"Failed to register %s: %d\n", vdev->name, ret);
> +		goto err_entity_cleanup;
> +	}
> +
> +	return 0;
> +
> +err_entity_cleanup:
> +	media_entity_cleanup(&vdev->entity);
> +err_queue_release:
> +	vb2_queue_release(vb2_q);
> +err_destroy:
> +	mutex_destroy(&stats->lock);
> +	return ret;
> +}
> +
> +void c3_isp_stats_unregister(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_stats *stats = &isp->stats;
> +
> +	vb2_queue_release(&stats->vb2_q);
> +	media_entity_cleanup(&stats->vdev.entity);
> +	video_unregister_device(&stats->vdev);
> +	mutex_destroy(&stats->lock);
> +}
> +
> +int c3_isp_stats_done(struct c3_isp_device *isp)
> +{
> +	struct c3_isp_stats *stats = &isp->stats;
> +	struct c3_isp_vb2_buffer *buff = stats->buff;
> +	unsigned long flags;
> +
> +	if (!stats->is_streaming)
> +		return -EINVAL;
> +
> +	spin_lock_irqsave(&stats->buff_lock, flags);
> +
> +	if (buff) {
> +		buff->vb.sequence = stats->isp->frm_sequence;
> +		buff->vb.vb2_buf.timestamp = ktime_get();
> +		buff->vb.field = V4L2_FIELD_NONE;
> +		vb2_buffer_done(&buff->vb.vb2_buf, VB2_BUF_STATE_DONE);
> +	}
> +
> +	c3_isp_stats_cfg_buff(stats);
> +
> +	spin_unlock_irqrestore(&stats->buff_lock, flags);
> +
> +	return 0;
> +}
> diff --git a/drivers/media/platform/amlogic/c3-isp/include/uapi/c3-isp-config.h b/drivers/media/platform/amlogic/c3-isp/include/uapi/c3-isp-config.h
> new file mode 100644
> index 000000000000..84ff5741357a
> --- /dev/null
> +++ b/drivers/media/platform/amlogic/c3-isp/include/uapi/c3-isp-config.h
> @@ -0,0 +1,537 @@
> +/* SPDX-License-Identifier: (GPL-2.0-only OR MIT) */
> +/*
> + * Copyright (C) 2024 Amlogic, Inc. All rights reserved
> + */
> +
> +#ifndef __C3_ISP_CONFIG_H__
> +#define __C3_ISP_CONFIG_H__

When moving to include/uapi/linux/media/amlogic prefix this with
_UAPI_

> +
> +#define AF_STAT_BLKH_NUM             17
> +#define AF_STAT_BLKV_NUM             15
> +#define AF_STAT_BLK_NUM              (AF_STAT_BLKH_NUM * AF_STAT_BLKV_NUM)
> +/* AF stats block size need to be aligned with 2 */
> +#define AF_STAT_BLK_SIZE             ALIGN(AF_STAT_BLK_NUM, 2)
> +#define AE_HISTOGRAM_SIZE	     1024
> +#define AE_STAT_BLKH_NUM             17
> +#define AE_STAT_BLKV_NUM             15
> +#define AE_STAT_BLK_NUM              (AE_STAT_BLKH_NUM * AE_STAT_BLKV_NUM)
> +/* AE stats block size need to be aligned with 2 */
> +#define AE_STAT_BLK_SIZE             ALIGN(AE_STAT_BLK_NUM, 2)
> +#define AE_BLOCK_WT_NUM              255
> +#define AE_BLK_WT_DATA_NUM_OF_GROUP  8
> +#define AWB_STAT_BLKH_NUM            32
> +#define AWB_STAT_BLKV_NUM            24
> +#define AWB_STAT_BLK_NUM             (AWB_STAT_BLKH_NUM * AWB_STAT_BLKV_NUM)
> +/* AWB stats block size need to be aligned with 2 */
> +#define AWB_STAT_BLK_SIZE            ALIGN(AWB_STAT_BLK_NUM, 2)
> +#define AWB_BLOCK_WT_NUM             768
> +#define AWB_BLK_WT_DATA_NUM_OF_GROUP 8
> +#define AWB_STAT_BLC20_NUM           4
> +#define AWB_STAT_GAIN10_NUM          4
> +#define BLC_OFFSET_NUM               5
> +#define GAMMA_LUT_GROUP_NUM          4
> +#define GAMMA_LUT_POINT_NUM          129
> +#define GAMMA_LUT_DATA_NUM_OF_GROUP  2
> +
> +/**
> + * struct awb_zone_stats - AWB statistics of a block
> + *
> + * AWB zone stats is aligned with 8 bytes
> + *
> + * @rg: the ratio of R / G in a zone
> + * @bg: the ratio of B / G in a zone
> + * @pixel_sum: the total number of pixels used in a zone
> + */
> +struct awb_zone_stats {
> +	u16 rg;
> +	u16 bg;
> +	u32 pixel_sum;
> +};
> +
> +/**
> + * struct awb_stats_info - Auto white balance statistics information.
> + *
> + * AWB statistical information of all blocks.
> + *
> + * @awb_stats: array of auto white balance statistics
> + */
> +struct awb_stats_info {
> +	struct awb_zone_stats awb_stats[AWB_STAT_BLK_SIZE];
> +};
> +
> +/**
> + * struct ae_zone_stats - AE statistics of a block
> + *
> + * AE zone stats is aligned with 8 bytes.
> + * This is a 5-bin histogram and the total sum is
> + * normalized to 0xffff.
> + * So hist2 = 0xffff - (hist0 + hist1 + hist3 + hist4)
> + *
> + * @hist0: the global normalized pixel count for bin 0
> + * @hist1: the global normalized pixel count for bin 1
> + * @hist3: the global normalized pixel count for bin 3
> + * @hist4: the global normalized pixel count for bin 4
> + */
> +struct ae_zone_stats {
> +	u16 hist0;

Should you include  <linux/types.h> and use types prefixed with __
(__u16) ?

> +	u16 hist1;
> +	u16 hist3;
> +	u16 hist4;
> +};
> +
> +/**
> + * struct ae_stats_info - Exposure statistics information
> + *
> + * AE statistical information consists of
> + * all blocks information and a 1024-bin histogram.
> + *
> + * @ae_stats: array of auto exposure block statistics
> + * @hist: a 1024-bin histogram for the entire image
> + */
> +struct ae_stats_info {
> +	struct ae_zone_stats ae_stats[AE_STAT_BLK_SIZE];
> +	u32 hist[AE_HISTOGRAM_SIZE];
> +};
> +
> +/**
> + * struct af_zone_stats - AF statistics of a block
> + *
> + * AF block stats is aligned with 8 bytes.
> + * The zonal accumulated contrast metrics are stored
> + * in floating point format with 16 bits mantissa and
> + * 5 or 6 bits exponent.
> + * Apart from contrast metrics we accumulate squared image and
> + * quartic image data over the zone.
> + *
> + * @i2_mat: the mantissa of zonal squared image pixel sum
> + * @i4_mat: the mantissa of zonal quartic image pixel sum
> + * @e4_mat: the mantissa of zonal multi-directional quartic edge sum
> + * @i2_exp: the exponent of zonal squared image pixel sum
> + * @i4_exp: the exponent of zonal quartic image pixel sum
> + * @e4_exp: the exponent of zonal multi-directional quartic edge sum
> + */
> +struct af_zone_stats {
> +	u16 i2_mat;
> +	u16 i4_mat;
> +	u16 e4_mat;
> +	u16 i2_exp: 5;
> +	u16 i4_exp: 6;
> +	u16 e4_exp: 5;
> +};
> +
> +/**
> + * struct af_stats_info - Auto Focus statistics information
> + *
> + * AF statistical information of each block
> + *
> + * @af_stats: array of auto focus block statistics
> + */
> +struct af_stats_info {
> +	struct af_zone_stats af_stats[AF_STAT_BLK_SIZE];
> +};
> +
> +/**
> + * struct c3_isp_stats_info - V4L2_META_FMT_C3ISP_STATS
> + *
> + * Contains ISP statistics
> + *
> + * @awb_stats: auto white balance stats
> + * @ae_stats: auto exposure stats
> + * @af_stats: auto focus stats
> + */
> +struct c3_isp_stats_info {
> +	struct awb_stats_info awb_stats;
> +	struct ae_stats_info ae_stats;
> +	struct af_stats_info af_stats;
> +};
> +
> +/**
> + * enum c3_isp_param_buffer_version -  C3 ISP parameters block versioning
> + *
> + * @C3_ISP_PARAM_BUFFER_V0: First version of C3 ISP parameters block
> + */
> +enum c3_isp_param_buffer_version {
> +	C3_ISP_PARAM_BUFFER_V0,
> +};
> +
> +/**
> + * enum c3_isp_param_block_type - Enumeration of C3 ISP parameter blocks
> + *
> + * Each block configures a specific processing block of the C3 ISP.
> + * The block type allows the driver to correctly interpret
> + * the parameters block data.
> + *
> + * @C3_ISP_PARAM_BLOCK_WB_CHANGE: White balance change parameters
> + * @C3_ISP_PARAM_BLOCK_WB_LUMA: White balance Luma-based parameters
> + * @C3_ISP_PARAM_BLOCK_WB_TRIANGLE: White balance triangle parameters
> + * @C3_ISP_PARAM_BLOCK_AWB_STATS: AWB statistics parameters
> + * @C3_ISP_PARAM_BLOCK_AE_STATS: AE statistics parameters
> + * @C3_ISP_PARAM_BLOCK_AF_STATS: AF statistics parameters
> + * @C3_ISP_PARAM_BLOCK_PST_GAMMA: post gamma parameters
> + * @C3_ISP_PARAM_BLOCK_DMSC: demosaic parameters
> + * @C3_ISP_PARAM_BLOCK_CCM: Color correction matrix parameters
> + * @C3_ISP_PARAM_BLOCK_CSC: Color space conversion parameters
> + * @C3_ISP_PARAM_BLOCK_BLC: Black level correction parameters
> + * @C3_ISP_PARAM_BLOCK_SENTINEL: First non-valid block index
> + */
> +enum c3_isp_param_block_type {
> +	C3_ISP_PARAM_BLOCK_WB_CHANGE,
> +	C3_ISP_PARAM_BLOCK_WB_LUMA,
> +	C3_ISP_PARAM_BLOCK_WB_TRIANGLE,
> +	C3_ISP_PARAM_BLOCK_AWB_STATS,
> +	C3_ISP_PARAM_BLOCK_AE_STATS,
> +	C3_ISP_PARAM_BLOCK_AF_STATS,
> +	C3_ISP_PARAM_BLOCK_PST_GAMMA,
> +	C3_ISP_PARAM_BLOCK_DMSC,
> +	C3_ISP_PARAM_BLOCK_CCM,
> +	C3_ISP_PARAM_BLOCK_CSC,
> +	C3_ISP_PARAM_BLOCK_BLC,
> +	C3_ISP_PARAM_BLOCK_SENTINEL
> +};
> +
> +/**
> + * struct c3_isp_param_block_header - C3 ISP parameter block header
> + *
> + * This structure represents the common part of all the ISP configuration
> + * blocks. Each parameters block shall embed an instance of this structure type
> + * as its first member, followed by the block-specific configuration data. The
> + * driver inspects this common header to discern the block type and its size and
> + * properly handle the block content by casting it to the correct block-specific
> + * type.
> + *
> + * @type: The parameters block type (enum c3_isp_param_block_type)
> + * @enabled: Block enabled/disabled flag
> + * @size: Size (in bytes) of parameters block
> + */
> +
> +struct c3_isp_param_block_header {
> +	enum c3_isp_param_block_type type;
> +	bool enabled;
> +	size_t size;
> +};

What is the size of this structure ? Is it aligned or does the
compiler inserts padding bytes ? In general, I would try to align
everything to 8 bytes to avoid the compiler inserting padding bytes.

A tool that can help you identifies holes is pahole. Just write a
small userspace program that includes types from this header and
declare a variable of each type of defined in this header. Pass the
executable to pahole and it will show the memory layout of each
member.

> +
> +/**
> + * struct wb_change_cfg - White Balance configuration
> + *
> + * @header: The C3 ISP parameters block header
> + * @wb_gain: white balance gain of each color
> + *	wb_gain[0]: Gr gain, range 0~0xfff
> + *	wb_gain[1]: R gain, range 0~0xfff
> + *	wb_gain[2]: B gain, range 0~0xfff
> + *	wb_gain[3]: Gb gain, range 0~0xfff
> + *	wb_gain[4]: Ir gain, range 0~0xfff
> + * @wb_limit: white balance limit of each color
> + *	wb_limit[0]: Gr limit, 16 bits float
> + *	wb_limit[1]: R limit, 16 bits float
> + *	wb_limit[2]: B limit, 16 bits float
> + *	wb_limit[3]: Gb limit, 16 bits float
> + *	wb_limit[4]: Ir limit, 16 bits float
> + * @ae_gain_grbgi: Gain of each color before blending to luma
> + *	ae_gain_grbgi[0]: Gr gain, range 0~255
> + *	ae_gain_grbgi[1]: R gain, range 0~255
> + *	ae_gain_grbgi[2]: B gain, range 0~255
> + *	ae_gain_grbgi[3]: Gb gain, range 0~255
> + *	ae_gain_grbgi[4]: Ir gain, range 0~255
> + * @ae_bl12_grbgi: Offset of each color before ae_gain_grbgi
> + *	ae_bl12_grbgi[0]: Gr offset, range 0~4095
> + *	ae_bl12_grbgi[1]: R offset, range 0~4095
> + *	ae_bl12_grbgi[2]: B offset, range 0~4095
> + *	ae_bl12_grbgi[3]: Gb offset, range 0~4095
> + *	ae_bl12_grbgi[4]: Ir offset, range 0~4095
> + */
> +struct wb_change_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 wb_gain[5];
> +	u32 wb_limit[5];
> +	u32 ae_gain_grbgi[5];
> +	u32 ae_bl12_grbgi[5];
> +};
> +
> +/**
> + * struct wb_luma_cfg - White Balance Luma-based configuration
> + *
> + * @header: The C3 ISP parameters block header
> + * @awb_stat_blc20: BLC in AWB statistic
> + *	awb_stat_blc20[0]: Gr blc, range 0~0xfffff
> + *	awb_stat_blc20[1]: R blc, range 0~0xfffff
> + *	awb_stat_blc20[2]: B blc, range 0~0xfffff
> + *	awb_stat_blc20[3]: Gb blc, range 0~0xfffff
> + * @awb_stat_gain10: Gain in AWB statistic
> + *	awb_stat_gain10[0]: Gr gain, range 0~1023
> + *	awb_stat_gain10[1]: R gain, range 0~1023
> + *	awb_stat_gain10[2]: B gain, range 0~1023
> + *	awb_stat_gain10[3]: Gb gain, range 0~1023
> + * @awb_stat_satur_low: AWB statistic under-saturation threshold
> + *	value: range 0~65535
> + * @awb_stat_satur_high: AWB statistic over-saturation threshold
> + *	value: range 0~65535
> + */
> +struct wb_luma_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 awb_stat_blc20[AWB_STAT_BLC20_NUM];
> +	u32 awb_stat_gain10[AWB_STAT_GAIN10_NUM];
> +	u32 awb_stat_satur_low;
> +	u32 awb_stat_satur_high;
> +};
> +
> +/**
> + * struct wb_triangle_cfg - White Balance Triangle
> + *
> + * @header: The C3 ISP parameters block header
> + * @awb_stat_satur_vald: AWB statistic over saturation control
> + *	value: 0: disable, 1: enable
> + * @awb_stat_rg_min: min value of r/g
> + *	value: 0~4095
> + * @awb_stat_rg_max: max value of r/g
> + *	value: 0~4095
> + * @awb_stat_bg_min: min value of b/g
> + *	value: 0~4095
> + * @awb_stat_bg_max: max value of b/g
> + *	value: 0~4095
> + * @awb_stat_rg_low: low value of r/g
> + *	value: 0~4095
> + * @awb_stat_rg_high: high value of r/g
> + *	value: 0~4095
> + * @awb_stat_bg_low: low value of b/g
> + *	value: 0~4095
> + * @awb_stat_bg_high: high value of b/g
> + *	value: 0~4095
> + */
> +struct wb_triangle_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 awb_stat_satur_vald;
> +	u32 awb_stat_rg_min;
> +	u32 awb_stat_rg_max;
> +	u32 awb_stat_bg_min;
> +	u32 awb_stat_bg_max;
> +	u32 awb_stat_rg_low;
> +	u32 awb_stat_rg_high;
> +	u32 awb_stat_bg_low;
> +	u32 awb_stat_bg_high;
> +};
> +
> +/**
> + * struct awb_stats_cfg - AWB statistics configuration
> + *
> + * This structure contains AWB statistics control information.
> + *
> + * @header: The C3 ISP parameters block header
> + * @awb_stat_switch: the switch of AWB statistics
> + *	value: 0~7
> + * @awb_stat_blk_weight: Array of weights for AWB statistics blocks
> + *	value: 0~15
> + */
> +struct awb_stats_cfg {
> +	struct c3_isp_param_block_header header;
> +	u8 awb_stat_switch;
> +	u32 awb_stat_blk_weight[AWB_BLOCK_WT_NUM];
> +};
> +
> +/**
> + * struct ae_stats_cfg - AE statistics configuration
> + *
> + * This structure contains AE statistics control information.
> + *
> + * @header: The C3 ISP parameters block header
> + * @ae_stat_switch: the switch of AE statistics
> + *	value: 0~3
> + * @ae_stat_blk_weight: Array of weights for AE statistics blocks
> + *	value: 0~15
> + */
> +struct ae_stats_cfg {
> +	struct c3_isp_param_block_header header;
> +	u8 ae_stat_switch;
> +	u32 ae_stat_blk_weight[AE_BLOCK_WT_NUM];
> +};
> +
> +/**
> + * struct af_stats_cfg - AF statistics configuration
> + *
> + * This structure contains AF statistics control information.
> + *
> + * @header: The C3 ISP parameters block header
> + * @af_stat_switch: the switch of AF statistics
> + *	value: 0~3
> + */
> +struct af_stats_cfg {
> +	struct c3_isp_param_block_header header;
> +	u8 af_stat_switch;
> +};
> +
> +/**
> + * struct pst_gamma_cfg - Post gamma configuration
> + *
> + * This structure contains post gamma parameters
> + *
> + * @header: The C3 ISP parameters block header
> + * @pst_gamma_lut: LUT for P-Stitch gamma
> + *	value: 0~65535
> + */
> +struct pst_gamma_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 pst_gamma_lut[GAMMA_LUT_GROUP_NUM][GAMMA_LUT_POINT_NUM];
> +};
> +
> +/**
> + * struct dmsc_cfg - Demosaic configuration
> + *
> + * This structure contains demosaic parameters
> + *
> + * @header: The C3 ISP parameters block header
> + */
> +struct dmsc_cfg {
> +	struct c3_isp_param_block_header header;
> +};
> +
> +/**
> + * struct ccm_cfg - ISP CCM configuration
> + *
> + * This structure holds the parameters for configuring the CCM,
> + * which is used for color correction.
> + *
> + * @header: The C3 ISP parameters block header
> + * @ccm_4x3matrix: A 3x4 matrix used for color correction
> + *	value: 0~8191
> + */
> +struct ccm_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 ccm_4x3matrix[3][4];
> +};
> +
> +/**
> + * struct csc_cfg - ISP Color Space Conversion configuration
> + *
> + * This structure contains settings for color space conversion.
> + *
> + * @header: The C3 ISP parameters block header
> + * @cm0_offset_inp: Input offset values for the 0-order color matrix
> + *	value: 0~8191
> + * @cm0_offset_oup: Output offset values for the 0-order color matrix
> + *	value: 0~8191
> + * @cm0_3x3mtrx_rs: matrix right shift for cm0
> + *	value: 0~3
> + * @cm0_3x3matrix: A 3x3 matrix used for the color matrix
> + *	value: 0~8191
> + */
> +struct csc_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 cm0_offset_inp[3];
> +	u32 cm0_offset_oup[3];
> +	u32 cm0_3x3mtrx_rs;
> +	u32 cm0_3x3matrix[3][3];
> +};
> +
> +/**
> + * struct blc_cfg - ISP Black Level Correction (BLC) configuration
> + *
> + * This structure holds the parameters for BLC in image processing.
> + *
> + * @header: The C3 ISP parameters block header
> + * @fe_bl_ofst: Array of front-end BLC offsets for each color each channels
> + *	fe_bl_ofst[0]: Gr blc offset, range 0~0x1fffff
> + *	fe_bl_ofst[1]: R blc offset, range 0~0x1fffff
> + *	fe_bl_ofst[2]: B blc offset, range 0~0x1fffff
> + *	fe_bl_ofst[3]: Gb blc offset, range 0~0x1fffff
> + *	fe_bl_ofst[4]: Ir blc offset, range 0~0x1fffff
> + * @blc_ofst: Array of LSWB BLC offsets
> + *	blc_ofst[0]: Gr blc offset, 16 bits float
> + *	blc_ofst[1]: R blc offset, 16 bits float
> + *	blc_ofst[2]: B blc offset, 16 bits float
> + *	blc_ofst[3]: Gb blc offset, 16 bits float
> + *	blc_ofst[4]: Ir blc offset, 16 bits float
> + */
> +struct blc_cfg {
> +	struct c3_isp_param_block_header header;
> +	u32 fe_bl_ofst[BLC_OFFSET_NUM];
> +	u32 blc_ofst[BLC_OFFSET_NUM];
> +};
> +
> +/**
> + * define C3_ISP_PARAMS_MAX_SIZE - Maximum size of all C3 ISP Parameters
> + *
> + * Though the parameters for the C3 ISP are passed as optional blocks, the
> + * driver still needs to know the absolute maximum size so that it can allocate
> + * a buffer sized appropriately to accommodate userspace attempting to set all
> + * possible parameters in a single frame.
> + */
> +#define C3_ISP_PARAMS_MAX_SIZE                 \
> +	(sizeof(struct wb_change_cfg) +        \
> +	sizeof(struct wb_luma_cfg)   +         \
> +	sizeof(struct wb_triangle_cfg) +       \
> +	sizeof(struct awb_stats_cfg) +          \
> +	sizeof(struct ae_stats_cfg) +           \
> +	sizeof(struct af_stats_cfg) +           \
> +	sizeof(struct pst_gamma_cfg) +         \
> +	sizeof(struct dmsc_cfg) +              \
> +	sizeof(struct ccm_cfg) +               \
> +	sizeof(struct csc_cfg) +               \
> +	sizeof(struct blc_cfg))
> +
> +/**
> + * struct c3_isp_params_buffer - C3 ISP configuration parameters
> + *
> + * This struct contains the configuration parameters of the C3 ISP
> + * algorithms, serialized by userspace into an opaque data buffer. Each
> + * configuration parameter block is represented by a block-specific structure
> + * which contains a :c:type:`c3_isp_param_block_header` entry as first
> + * member. Userspace populates the @data buffer with configuration parameters
> + * for the blocks that it intends to configure. As a consequence, the data
> + * buffer effective size changes according to the number of ISP blocks that
> + * userspace intends to configure.
> + *
> + * The parameters buffer is versioned by the @version field to allow modifying
> + * and extending its definition. Userspace should populate the @version field to
> + * inform the driver about the version it intends to use. The driver will parse
> + * and handle the @data buffer according to the data layout specific to the
> + * indicated revision and return an error if the desired revision is not
> + * supported.
> + *
> + * For each ISP block that userspace wants to configure, a block-specific
> + * structure is appended to the @data buffer, one after the other without gaps
> + * in between nor overlaps. Userspace shall populate the @total_size field with
> + * the effective size, in bytes, of the @data buffer.
> + *
> + * The expected memory layout of the parameters buffer is::
> + *
> + *	+-------------------- struct c3_isp_params_buffer ------------------+
> + *	| version = C3_ISP_PARAM_BUFFER_V0;                                   |
> + *	| total_size = sizeof(sizeof(struct wb_change_cfg))                   |
> + *	|              sizeof(sizeof(struct wb_luma_cfg));                    |
> + *	| +------------------------- data  ---------------------------------+ |
> + *	| | +------------------ struct wb_change_cfg) --------------------+ | |
> + *	| | | +---------  struct c3_isp_param_block_header header  -----+ | | |
> + *	| | | | type = C3_ISP_PARAM_BLOCK_WB_CHANGE;                    | | | |
> + *	| | | | enabled = true;                                            | | | |
> + *	| | | | size =                                                  | | | |
> + *	| | | |    sizeof(struct c3_isp_param_block_header header);     | | | |
> + *	| | | +---------------------------------------------------------+ | | |
> + *	| | | wb_gain[5] = ...;                                           | | |
> + *	| | | wb_limit[5] = ...;                                          | | |
> + *	| | | ae_gain_grbgi[5] = ...;                                     | | |
> + *	| | | ae_bl12_grbgi[5] = ...;                                     | | |
> + *	| | +------------------ struct wb_luma_cfg -----------------------+ | |
> + *	| | | +---------- struct c3_isp_param_block_header header ------+ | | |
> + *	| | | | type = C3_ISP_PARAM_BLOCK_WB_LUMA;                      | | | |
> + *	| | | | enabled = true;                                            | | | |
> + *	| | | | size = sizeof(struct wb_luma_cfg);                      | | | |
> + *	| | | +---------------------------------------------------------+ | | |
> + *	| | | awb_stat_blc20[4] = ...;                                    | | |
> + *	| | | awb_stat_gain10[4] = ...;                                   | | |
> + *	| | | awb_stat_satur_low = ...;                                   | | |
> + *	| | | awb_stat_satur_high = ...;                                  | | |
> + *	| | +-------------------------------------------------------------+ | |
> + *	| +-----------------------------------------------------------------+ |
> + *	+---------------------------------------------------------------------+
> + *
> + * @version: The C3 ISP parameters buffer version
> + * @total_size: The C3 ISP configuration data effective size,
> + *		excluding this header
> + * @data: The C3 ISP configuration blocks data
> + */
> +struct c3_isp_params_buffer {
> +	enum c3_isp_param_buffer_version version;
> +	size_t total_size;
> +	u8 data[C3_ISP_PARAMS_MAX_SIZE];
> +};
> +
> +#endif
>
> --
> 2.46.1
>
>
>




[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux