Hi!

> +/* External clock frequency is 24.0M */
> +#define IMX219_XCLK_FREQ		24000000

MHz?

> +/*Frame Length Line*/

Make this usual comment style.

> +/*
> + * Initialisation delay between XCLR low->high and the moment when the sensor
> + * can start capture (i.e. can leave software stanby) must be not less than:
> + *   t4 + max(t5, t6 + <time to initialize the sensor register over I2C>)
> + * where
> + *   t4 is fixed, and is max 200uS,
> + *   t5 is fixed, and is 6000uS,
...
> + * 1185 to 5333 uS, and is always less than t5.
...
> + * For this reason this is always safe to wait (t4 + t5) = 6200 uS, then

I believe this should be "us" or "usec".

> +	/*
> +	 * Mutex for serialized access:
> +	 * Protect sensor module set pad format and start/stop streaming safely.
> +	 */

This is not really english.

> +/* Read registers up to 2 at a time */
> +static int imx219_read_reg(struct imx219 *imx219, u16 reg, u32 len, u32 *val)
> +{

> +	msgs[1].buf = &data_buf[4 - len];
> +
> +	ret = i2c_transfer(client->adapter, msgs, ARRAY_SIZE(msgs));
> +	if (ret != ARRAY_SIZE(msgs))
> +		return -EIO;
> +
> +	*val = get_unaligned_be32(data_buf);

That's really interesting piece of code. len is in bytes, but return
value is in 32 bit values, magic with be32...

> +static int imx219_write_reg(struct imx219 *imx219, u16 reg, u32 len, u32 val)
> +{
> +	struct i2c_client *client = v4l2_get_subdevdata(&imx219->sd);
> +	u8 buf[6];
> +
> +	if (len > 4)
> +		return -EINVAL;
> +
> +	put_unaligned_be16(reg, buf);
> +	put_unaligned_be32(val << (8 * (4 - len)), buf + 2);

Wow. When I thought above was interesting.... this is even more so.


> +static int imx219_set_ctrl(struct v4l2_ctrl *ctrl)
> +{
...
> +	/*
> +	 * Applying V4L2 control value only happens
> +	 * when power is up for streaming
> +	 */
> +	if (pm_runtime_get_if_in_use(&client->dev) == 0)
> +		return 0;

Is this usual for other cameras? Will not it cause problems to
applications doing autogain?

> +	switch (ctrl->id) {
...
> +	case V4L2_CID_HFLIP:
> +	case V4L2_CID_VFLIP:
> +		ret = imx219_write_reg(imx219, IMX219_REG_ORIENTATION, 1,
> +				       imx219->hflip->val |
> +				       imx219->vflip->val << 1);
> +		break;

ctrl->val is unused in this case?

> +		break;
> +	default:
> +		dev_info(&client->dev,
> +			 "ctrl(id:0x%x,val:0x%x) is not handled\n",
> +			 ctrl->id, ctrl->val);
> +		ret = -EINVAL;
> +		break;
> +	}

Rate limit this as user can trigger these?

> +static int imx219_set_stream(struct v4l2_subdev *sd, int enable)
> +{
> +	struct imx219 *imx219 = to_imx219(sd);
> +	struct i2c_client *client = v4l2_get_subdevdata(sd);
> +	int ret = 0;
> +
> +	mutex_lock(&imx219->mutex);
> +	if (imx219->streaming == enable) {
> +		mutex_unlock(&imx219->mutex);
> +		return 0;
> +	}
> +
> +	if (enable) {
> +		ret = pm_runtime_get_sync(&client->dev);
> +		if (ret < 0) {
> +			pm_runtime_put_noidle(&client->dev);
> +			goto err_unlock;
> +		}
> +
> +		/*
> +		 * Apply default & customized values
> +		 * and then start streaming.
> +		 */
> +		ret = imx219_start_streaming(imx219);
> +		if (ret)
> +			goto err_rpm_put;

Is it correct/vital that one error path uses put_noidle() and second
uses plain put()?

> +static int __maybe_unused imx219_resume(struct device *dev)
> +{
> +	struct i2c_client *client = to_i2c_client(dev);
> +	struct v4l2_subdev *sd = i2c_get_clientdata(client);
> +	struct imx219 *imx219 = to_imx219(sd);
> +	int ret;
> +
> +	if (imx219->streaming) {
> +		ret = imx219_start_streaming(imx219);
> +		if (ret)
> +			goto error;
> +	}
> +
> +	return 0;
> +
> +error:
> +	imx219_stop_streaming(imx219);
> +	imx219->streaming = 0;

This error path will leave driver in inconsistent state. hflip/vflip
controls will be locked and pm_runtime_put() is not executed.

> +static int imx219_check_hwcfg(struct device *dev)
> +{
> +	struct v4l2_fwnode_endpoint *ep_cfg;
> +	struct device_node *ep;
> +	int ret = -EINVAL;
> +
> +	ep = of_graph_get_next_endpoint(dev->of_node, NULL);
> +	if (!ep) {
> +		dev_err(dev, "missing endpoint node\n");
> +		return -EINVAL;
> +	}
> +
> +	ep_cfg = v4l2_fwnode_endpoint_alloc_parse(of_fwnode_handle(ep));
> +	if (IS_ERR(ep_cfg)) {
> +		dev_err(dev, "could not parse endpoint\n");
> +		goto error_out;
> +	}
...
> +error_out:
> +	v4l2_fwnode_endpoint_free(ep_cfg);

Is it correct to call endpoint_free on error pointer?

> +MODULE_AUTHOR("Dave Stevenson <dave.stevenson@raspberrypi.com");

Missing ">" at end of address.

Best regards,
								Pavel
-- 
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany