[PATCH 0/4] Improvements for OmniVision OV2685 driver

[PATCH 0/4] Improvements for OmniVision OV2685 driver

[Luca Weiss]

The first two patches make the sensor work in my setup, and adds a
missing error print that can happen when the dts is wrong.

The last two patches add extra API support for orientation/rotation and
get_selection that is wanted by libcamera to function correctly.

Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
---
Luca Weiss (4):
      media: i2c: ov2685: Make reset gpio optional
      media: i2c: ov2685: Add print for power on write failed
      media: i2c: ov2685: Add controls from fwnode
      media: i2c: ov2685: Add .get_selection() support

 drivers/media/i2c/ov2685.c | 78 ++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 76 insertions(+), 2 deletions(-)
---
base-commit: b4c97ca9bb2381580a132a6c71b0e93ac6a4524a
change-id: 20230129-ov2685-improvements-b03bdcf1c290

Best regards,
-- 
Luca Weiss <luca@z3ntu.xyz>

[PATCH 1/4] media: i2c: ov2685: Make reset gpio optional

[Luca Weiss]

In some setups XSHUTDOWN is connected to DOVDD when it's unused,
therefore treat the reset gpio as optional.

Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
---
 drivers/media/i2c/ov2685.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
index a3b524f15d89..a422f4c8a2eb 100644
--- a/drivers/media/i2c/ov2685.c
+++ b/drivers/media/i2c/ov2685.c
@@ -734,7 +734,7 @@ static int ov2685_probe(struct i2c_client *client,
 	if (clk_get_rate(ov2685->xvclk) != OV2685_XVCLK_FREQ)
 		dev_warn(dev, "xvclk mismatched, modes are based on 24MHz\n");
 
-	ov2685->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
+	ov2685->reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
 	if (IS_ERR(ov2685->reset_gpio)) {
 		dev_err(dev, "Failed to get reset-gpios\n");
 		return -EINVAL;

-- 
2.39.1

Re: [PATCH 1/4] media: i2c: ov2685: Make reset gpio optional

[Jacopo Mondi]

Hi Luca

On Sun, Jan 29, 2023 at 10:42:35AM +0100, Luca Weiss wrote:
> In some setups XSHUTDOWN is connected to DOVDD when it's unused,
> therefore treat the reset gpio as optional.

I don't have a datasheet for this sensor, but OV sensors usually have
to gpio lines to control powerdown and reset. Datasheets usually
suggest to hook one of the 2 to DOVDD and control the other from the
SoC. How is the sensor hooked up in your design ? No gpio lines is
controlled by the SoC ?

Another question is if we need to software-reset the sensor if no gpio
line is hooked up to XSHUTDOWN.

>
> Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
> ---
>  drivers/media/i2c/ov2685.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> index a3b524f15d89..a422f4c8a2eb 100644
> --- a/drivers/media/i2c/ov2685.c
> +++ b/drivers/media/i2c/ov2685.c
> @@ -734,7 +734,7 @@ static int ov2685_probe(struct i2c_client *client,
>  	if (clk_get_rate(ov2685->xvclk) != OV2685_XVCLK_FREQ)
>  		dev_warn(dev, "xvclk mismatched, modes are based on 24MHz\n");
>
> -	ov2685->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
> +	ov2685->reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_LOW);
>  	if (IS_ERR(ov2685->reset_gpio)) {
>  		dev_err(dev, "Failed to get reset-gpios\n");
>  		return -EINVAL;
>
> --
> 2.39.1
>

Re: [PATCH 1/4] media: i2c: ov2685: Make reset gpio optional

[Luca Weiss]

On Sonntag, 29. Jänner 2023 12:22:49 CET Jacopo Mondi wrote:
> Hi Luca
> 
> On Sun, Jan 29, 2023 at 10:42:35AM +0100, Luca Weiss wrote:
> > In some setups XSHUTDOWN is connected to DOVDD when it's unused,
> > therefore treat the reset gpio as optional.
> 
> I don't have a datasheet for this sensor, but OV sensors usually have
> to gpio lines to control powerdown and reset. Datasheets usually
> suggest to hook one of the 2 to DOVDD and control the other from the
> SoC. How is the sensor hooked up in your design ? No gpio lines is
> controlled by the SoC ?

It looks like this sensor only has XSHUTDOWN pin and no extra reset pin.

In my setup there's the normal I2C & CSI & mclk hookups, but the supply lines 
and shutdown line are all just connected to regulator-fixed, so gpio-
controlled on/off regulators.

> 
> Another question is if we need to software-reset the sensor if no gpio
> line is hooked up to XSHUTDOWN.

The datasheet mentions it resets itself during power up (so when the supplies 
are turned on), so I don't think we need to add anything.

Regards
Luca

> 
> > Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
> > ---
> > 
> >  drivers/media/i2c/ov2685.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> > index a3b524f15d89..a422f4c8a2eb 100644
> > --- a/drivers/media/i2c/ov2685.c
> > +++ b/drivers/media/i2c/ov2685.c
> > @@ -734,7 +734,7 @@ static int ov2685_probe(struct i2c_client *client,
> > 
> >  	if (clk_get_rate(ov2685->xvclk) != OV2685_XVCLK_FREQ)
> >  	
> >  		dev_warn(dev, "xvclk mismatched, modes are based on 
24MHz\n");
> > 
> > -	ov2685->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
> > +	ov2685->reset_gpio = devm_gpiod_get_optional(dev, "reset",
> > GPIOD_OUT_LOW);> 
> >  	if (IS_ERR(ov2685->reset_gpio)) {
> >  	
> >  		dev_err(dev, "Failed to get reset-gpios\n");
> >  		return -EINVAL;
> > 
> > --
> > 2.39.1

Re: [PATCH 1/4] media: i2c: ov2685: Make reset gpio optional

[Jacopo Mondi]

Hi Luca

On Sun, Jan 29, 2023 at 12:49:03PM +0100, Luca Weiss wrote:
> On Sonntag, 29. Jänner 2023 12:22:49 CET Jacopo Mondi wrote:
> > Hi Luca
> >
> > On Sun, Jan 29, 2023 at 10:42:35AM +0100, Luca Weiss wrote:
> > > In some setups XSHUTDOWN is connected to DOVDD when it's unused,
> > > therefore treat the reset gpio as optional.
> >
> > I don't have a datasheet for this sensor, but OV sensors usually have
> > to gpio lines to control powerdown and reset. Datasheets usually
> > suggest to hook one of the 2 to DOVDD and control the other from the
> > SoC. How is the sensor hooked up in your design ? No gpio lines is
> > controlled by the SoC ?
>
> It looks like this sensor only has XSHUTDOWN pin and no extra reset pin.
>

Ack, I see the same for OV2680 (for which I have a datasheet)

> In my setup there's the normal I2C & CSI & mclk hookups, but the supply lines
> and shutdown line are all just connected to regulator-fixed, so gpio-
> controlled on/off regulators.
>
> >
> > Another question is if we need to software-reset the sensor if no gpio
> > line is hooked up to XSHUTDOWN.
>
> The datasheet mentions it resets itself during power up (so when the supplies
> are turned on), so I don't think we need to add anything.
>

Thanks for the clarification!

Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>

Thanks
   j


> Regards
> Luca
>
> >
> > > Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
> > > ---
> > >
> > >  drivers/media/i2c/ov2685.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> > > index a3b524f15d89..a422f4c8a2eb 100644
> > > --- a/drivers/media/i2c/ov2685.c
> > > +++ b/drivers/media/i2c/ov2685.c
> > > @@ -734,7 +734,7 @@ static int ov2685_probe(struct i2c_client *client,
> > >
> > >  	if (clk_get_rate(ov2685->xvclk) != OV2685_XVCLK_FREQ)
> > >
> > >  		dev_warn(dev, "xvclk mismatched, modes are based on
> 24MHz\n");
> > >
> > > -	ov2685->reset_gpio = devm_gpiod_get(dev, "reset", GPIOD_OUT_LOW);
> > > +	ov2685->reset_gpio = devm_gpiod_get_optional(dev, "reset",
> > > GPIOD_OUT_LOW);>
> > >  	if (IS_ERR(ov2685->reset_gpio)) {
> > >
> > >  		dev_err(dev, "Failed to get reset-gpios\n");
> > >  		return -EINVAL;
> > >
> > > --
> > > 2.39.1
>
>
>
>

[PATCH 2/4] media: i2c: ov2685: Add print for power on write failed

[Luca Weiss]

If the sensor doens't power up correctly, for example due to incorrect
devicetree description, the power up i2c writes will fail.

Add an error print for this situation.

Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
---
 drivers/media/i2c/ov2685.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
index a422f4c8a2eb..844a91dbc8e5 100644
--- a/drivers/media/i2c/ov2685.c
+++ b/drivers/media/i2c/ov2685.c
@@ -419,8 +419,10 @@ static int __ov2685_power_on(struct ov2685 *ov2685)
 	 * writing register before .s_stream() as a workaround
 	 */
 	ret = ov2685_write_array(ov2685->client, ov2685->cur_mode->reg_list);
-	if (ret)
+	if (ret) {
+		dev_err(dev, "Failed to set regs for power on\n");
 		goto disable_supplies;
+	}
 
 	return 0;
 

-- 
2.39.1

Re: [PATCH 2/4] media: i2c: ov2685: Add print for power on write failed

[Jacopo Mondi]

Hi Luca

On Sun, Jan 29, 2023 at 10:42:36AM +0100, Luca Weiss wrote:
> If the sensor doens't power up correctly, for example due to incorrect
> devicetree description, the power up i2c writes will fail.
>
> Add an error print for this situation.
>
> Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
> ---
>  drivers/media/i2c/ov2685.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> index a422f4c8a2eb..844a91dbc8e5 100644
> --- a/drivers/media/i2c/ov2685.c
> +++ b/drivers/media/i2c/ov2685.c
> @@ -419,8 +419,10 @@ static int __ov2685_power_on(struct ov2685 *ov2685)
>  	 * writing register before .s_stream() as a workaround
>  	 */
>  	ret = ov2685_write_array(ov2685->client, ov2685->cur_mode->reg_list);
> -	if (ret)
> +	if (ret) {
> +		dev_err(dev, "Failed to set regs for power on\n");
>  		goto disable_supplies;
> +	}

This is fine. I would also consider if it's worth to fail loud in
ov2685_write_reg().

Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
>
>  	return 0;
>
>
> --
> 2.39.1
>

[PATCH 3/4] media: i2c: ov2685: Add controls from fwnode

[Luca Weiss]

Add V4L2_CID_CAMERA_ORIENTATION and V4L2_CID_CAMERA_SENSOR_ROTATION
controls to the ov2685 driver by attempting to parse them from firmware.

Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
---
 drivers/media/i2c/ov2685.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
index 844a91dbc8e5..bfced11b178b 100644
--- a/drivers/media/i2c/ov2685.c
+++ b/drivers/media/i2c/ov2685.c
@@ -17,6 +17,7 @@
 #include <media/media-entity.h>
 #include <media/v4l2-async.h>
 #include <media/v4l2-ctrls.h>
+#include <media/v4l2-fwnode.h>
 #include <media/v4l2-subdev.h>
 
 #define CHIP_ID				0x2685
@@ -613,6 +614,7 @@ static int ov2685_initialize_controls(struct ov2685 *ov2685)
 	const struct ov2685_mode *mode;
 	struct v4l2_ctrl_handler *handler;
 	struct v4l2_ctrl *ctrl;
+	struct v4l2_fwnode_device_properties props;
 	u64 exposure_max;
 	u32 pixel_rate, h_blank;
 	int ret;
@@ -661,6 +663,15 @@ static int ov2685_initialize_controls(struct ov2685 *ov2685)
 				ARRAY_SIZE(ov2685_test_pattern_menu) - 1,
 				0, 0, ov2685_test_pattern_menu);
 
+	/* set properties from fwnode (e.g. rotation, orientation) */
+	ret = v4l2_fwnode_device_parse(&ov2685->client->dev, &props);
+	if (ret)
+		goto err_free_handler;
+
+	ret = v4l2_ctrl_new_fwnode_properties(handler, &ov2685_ctrl_ops, &props);
+	if (ret)
+		goto err_free_handler;
+
 	if (handler->error) {
 		ret = handler->error;
 		dev_err(&ov2685->client->dev,

-- 
2.39.1

Re: [PATCH 3/4] media: i2c: ov2685: Add controls from fwnode

[Jacopo Mondi]

Hi Luca,

On Sun, Jan 29, 2023 at 10:42:37AM +0100, Luca Weiss wrote:
> Add V4L2_CID_CAMERA_ORIENTATION and V4L2_CID_CAMERA_SENSOR_ROTATION
> controls to the ov2685 driver by attempting to parse them from firmware.
>
> Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
> ---
>  drivers/media/i2c/ov2685.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
>
> diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> index 844a91dbc8e5..bfced11b178b 100644
> --- a/drivers/media/i2c/ov2685.c
> +++ b/drivers/media/i2c/ov2685.c
> @@ -17,6 +17,7 @@
>  #include <media/media-entity.h>
>  #include <media/v4l2-async.h>
>  #include <media/v4l2-ctrls.h>
> +#include <media/v4l2-fwnode.h>
>  #include <media/v4l2-subdev.h>
>
>  #define CHIP_ID				0x2685
> @@ -613,6 +614,7 @@ static int ov2685_initialize_controls(struct ov2685 *ov2685)
>  	const struct ov2685_mode *mode;
>  	struct v4l2_ctrl_handler *handler;
>  	struct v4l2_ctrl *ctrl;
> +	struct v4l2_fwnode_device_properties props;
>  	u64 exposure_max;
>  	u32 pixel_rate, h_blank;
>  	int ret;
> @@ -661,6 +663,15 @@ static int ov2685_initialize_controls(struct ov2685 *ov2685)
>  				ARRAY_SIZE(ov2685_test_pattern_menu) - 1,
>  				0, 0, ov2685_test_pattern_menu);
>

As the below function can register up to 2 controls, you should also
reserve space for them when initializing the control handler to avoid
relocations

-       ret = v4l2_ctrl_handler_init(handler, 8);
+       ret = v4l2_ctrl_handler_init(handler, 10);


> +	/* set properties from fwnode (e.g. rotation, orientation) */
> +	ret = v4l2_fwnode_device_parse(&ov2685->client->dev, &props);
> +	if (ret)
> +		goto err_free_handler;
> +
> +	ret = v4l2_ctrl_new_fwnode_properties(handler, &ov2685_ctrl_ops, &props);
> +	if (ret)
> +		goto err_free_handler;
> +
>  	if (handler->error) {
>  		ret = handler->error;
>  		dev_err(&ov2685->client->dev,
>
> --
> 2.39.1
>

[PATCH 4/4] media: i2c: ov2685: Add .get_selection() support

[Luca Weiss]

Add support for the .get_selection() pad operation to the ov2685 sensor
driver.

Report the native sensor size (pixel array), the crop bounds (readable
pixel array area) and the current and default analog crop rectangles.

Signed-off-by: Luca Weiss <luca@z3ntu.xyz>
---
 drivers/media/i2c/ov2685.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 61 insertions(+)

diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
index bfced11b178b..7ebf36d1a8cc 100644
--- a/drivers/media/i2c/ov2685.c
+++ b/drivers/media/i2c/ov2685.c
@@ -56,6 +56,9 @@
 #define OV2685_REG_VALUE_16BIT		2
 #define OV2685_REG_VALUE_24BIT		3
 
+#define OV2685_NATIVE_WIDTH		1616
+#define OV2685_NATIVE_HEIGHT		1216
+
 #define OV2685_LANES			1
 #define OV2685_BITS_PER_SAMPLE		10
 
@@ -78,6 +81,7 @@ struct ov2685_mode {
 	u32 exp_def;
 	u32 hts_def;
 	u32 vts_def;
+	const struct v4l2_rect *analog_crop;
 	const struct regval *reg_list;
 };
 
@@ -231,6 +235,13 @@ static const int ov2685_test_pattern_val[] = {
 	OV2685_TEST_PATTERN_COLOR_SQUARE,
 };
 
+static const struct v4l2_rect ov2685_analog_crop = {
+	.left	= 8,
+	.top	= 8,
+	.width	= 1600,
+	.height	= 1200,
+};
+
 static const struct ov2685_mode supported_modes[] = {
 	{
 		.width = 1600,
@@ -238,6 +249,7 @@ static const struct ov2685_mode supported_modes[] = {
 		.exp_def = 0x04ee,
 		.hts_def = 0x06a4,
 		.vts_def = 0x050e,
+		.analog_crop = &ov2685_analog_crop,
 		.reg_list = ov2685_1600x1200_regs,
 	},
 };
@@ -384,6 +396,53 @@ static int ov2685_enum_frame_sizes(struct v4l2_subdev *sd,
 	return 0;
 }
 
+static const struct v4l2_rect *
+__ov2685_get_pad_crop(struct ov2685 *ov2685,
+		      struct v4l2_subdev_state *state, unsigned int pad,
+		      enum v4l2_subdev_format_whence which)
+{
+	const struct ov2685_mode *mode = ov2685->cur_mode;
+
+	switch (which) {
+	case V4L2_SUBDEV_FORMAT_TRY:
+		return v4l2_subdev_get_try_crop(&ov2685->subdev, state, pad);
+	case V4L2_SUBDEV_FORMAT_ACTIVE:
+		return mode->analog_crop;
+	}
+
+	return NULL;
+}
+
+static int ov2685_get_selection(struct v4l2_subdev *sd,
+				struct v4l2_subdev_state *sd_state,
+				struct v4l2_subdev_selection *sel)
+{
+	struct ov2685 *ov2685 = to_ov2685(sd);
+
+	switch (sel->target) {
+	case V4L2_SEL_TGT_CROP:
+		mutex_lock(&ov2685->mutex);
+		sel->r = *__ov2685_get_pad_crop(ov2685, sd_state, sel->pad,
+				sel->which);
+		mutex_unlock(&ov2685->mutex);
+		break;
+	case V4L2_SEL_TGT_NATIVE_SIZE:
+	case V4L2_SEL_TGT_CROP_BOUNDS:
+		sel->r.top = 0;
+		sel->r.left = 0;
+		sel->r.width = OV2685_NATIVE_WIDTH;
+		sel->r.height = OV2685_NATIVE_HEIGHT;
+		break;
+	case V4L2_SEL_TGT_CROP_DEFAULT:
+		sel->r = ov2685_analog_crop;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 /* Calculate the delay in us by clock rate and clock cycles */
 static inline u32 ov2685_cal_delay(u32 cycles)
 {
@@ -592,6 +651,8 @@ static const struct v4l2_subdev_pad_ops ov2685_pad_ops = {
 	.enum_frame_size = ov2685_enum_frame_sizes,
 	.get_fmt = ov2685_get_fmt,
 	.set_fmt = ov2685_set_fmt,
+	.get_selection = ov2685_get_selection,
+	.set_selection = ov2685_get_selection,
 };
 
 static const struct v4l2_subdev_ops ov2685_subdev_ops = {

-- 
2.39.1

Re: [PATCH 4/4] media: i2c: ov2685: Add .get_selection() support

[Jacopo Mondi]

Hi Luca

On Sun, Jan 29, 2023 at 10:42:38AM +0100, Luca Weiss wrote:
> Add support for the .get_selection() pad operation to the ov2685 sensor
> driver.
>
> Report the native sensor size (pixel array), the crop bounds (readable
> pixel array area) and the current and default analog crop rectangles.
>
> Signed-off-by: Luca Weiss <luca@z3ntu.xyz>

As the driver supports a single mode you could have hard-coded
the rectangle sizes in get_selection(), but this way is much better as
it prepares for adding more modes to the driver eventually.

Thanks
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>

> ---
>  drivers/media/i2c/ov2685.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 61 insertions(+)
>
> diff --git a/drivers/media/i2c/ov2685.c b/drivers/media/i2c/ov2685.c
> index bfced11b178b..7ebf36d1a8cc 100644
> --- a/drivers/media/i2c/ov2685.c
> +++ b/drivers/media/i2c/ov2685.c
> @@ -56,6 +56,9 @@
>  #define OV2685_REG_VALUE_16BIT		2
>  #define OV2685_REG_VALUE_24BIT		3
>
> +#define OV2685_NATIVE_WIDTH		1616
> +#define OV2685_NATIVE_HEIGHT		1216
> +
>  #define OV2685_LANES			1
>  #define OV2685_BITS_PER_SAMPLE		10
>
> @@ -78,6 +81,7 @@ struct ov2685_mode {
>  	u32 exp_def;
>  	u32 hts_def;
>  	u32 vts_def;
> +	const struct v4l2_rect *analog_crop;
>  	const struct regval *reg_list;
>  };
>
> @@ -231,6 +235,13 @@ static const int ov2685_test_pattern_val[] = {
>  	OV2685_TEST_PATTERN_COLOR_SQUARE,
>  };
>
> +static const struct v4l2_rect ov2685_analog_crop = {
> +	.left	= 8,
> +	.top	= 8,
> +	.width	= 1600,
> +	.height	= 1200,
> +};
> +
>  static const struct ov2685_mode supported_modes[] = {
>  	{
>  		.width = 1600,
> @@ -238,6 +249,7 @@ static const struct ov2685_mode supported_modes[] = {
>  		.exp_def = 0x04ee,
>  		.hts_def = 0x06a4,
>  		.vts_def = 0x050e,
> +		.analog_crop = &ov2685_analog_crop,
>  		.reg_list = ov2685_1600x1200_regs,
>  	},
>  };
> @@ -384,6 +396,53 @@ static int ov2685_enum_frame_sizes(struct v4l2_subdev *sd,
>  	return 0;
>  }
>
> +static const struct v4l2_rect *
> +__ov2685_get_pad_crop(struct ov2685 *ov2685,
> +		      struct v4l2_subdev_state *state, unsigned int pad,
> +		      enum v4l2_subdev_format_whence which)
> +{
> +	const struct ov2685_mode *mode = ov2685->cur_mode;
> +
> +	switch (which) {
> +	case V4L2_SUBDEV_FORMAT_TRY:
> +		return v4l2_subdev_get_try_crop(&ov2685->subdev, state, pad);
> +	case V4L2_SUBDEV_FORMAT_ACTIVE:
> +		return mode->analog_crop;
> +	}
> +
> +	return NULL;
> +}
> +
> +static int ov2685_get_selection(struct v4l2_subdev *sd,
> +				struct v4l2_subdev_state *sd_state,
> +				struct v4l2_subdev_selection *sel)
> +{
> +	struct ov2685 *ov2685 = to_ov2685(sd);
> +
> +	switch (sel->target) {
> +	case V4L2_SEL_TGT_CROP:
> +		mutex_lock(&ov2685->mutex);
> +		sel->r = *__ov2685_get_pad_crop(ov2685, sd_state, sel->pad,
> +				sel->which);
> +		mutex_unlock(&ov2685->mutex);
> +		break;
> +	case V4L2_SEL_TGT_NATIVE_SIZE:
> +	case V4L2_SEL_TGT_CROP_BOUNDS:
> +		sel->r.top = 0;
> +		sel->r.left = 0;
> +		sel->r.width = OV2685_NATIVE_WIDTH;
> +		sel->r.height = OV2685_NATIVE_HEIGHT;
> +		break;
> +	case V4L2_SEL_TGT_CROP_DEFAULT:
> +		sel->r = ov2685_analog_crop;
> +		break;
> +	default:
> +		return -EINVAL;
> +	}
> +
> +	return 0;
> +}
> +
>  /* Calculate the delay in us by clock rate and clock cycles */
>  static inline u32 ov2685_cal_delay(u32 cycles)
>  {
> @@ -592,6 +651,8 @@ static const struct v4l2_subdev_pad_ops ov2685_pad_ops = {
>  	.enum_frame_size = ov2685_enum_frame_sizes,
>  	.get_fmt = ov2685_get_fmt,
>  	.set_fmt = ov2685_set_fmt,
> +	.get_selection = ov2685_get_selection,
> +	.set_selection = ov2685_get_selection,
>  };
>
>  static const struct v4l2_subdev_ops ov2685_subdev_ops = {
>
> --
> 2.39.1
>