Re: [PATCH 1/3] media: i2c: gc0310: fix probe error handling and unwind resources properly

[Hans de Goede <hansg@kernel.org>]

Hi,

On 1-Apr-26 20:16, Sanjay Chitroda wrote:
> From: Sanjay Chitroda <sanjayembeddedse@gmail.com>
> 
> The GC0310 probe path currently performs error cleanup by jumping to a
> common label that mirrors the driver's remove() callback. This is unsafe,
> as remove() assumes that the subdevice has been fully registered with
> the V4L2 framework, media and control resources have been initialized.

That is simply not true, all functions called in remove() internally
check if their init counter-part has succeeded and if not are a no-op.

If you're aware of any specific calls in remove() where this is not
the case, please explicitly describe these cases and describe an
example exit-error path from probe() where things actually go wrong.

> Calling remove() from probe can result in unregistering or cleaning up
> subdevice, leading to resource leaks and subtle lifecycle bugs.
> 
> Rewrite the probe() error handling to unwind resources explicitly, using
> fine‑grained goto labels along with appropriate error logs. Each failure
> path now frees only successfully acquired resource, without remove().
> 
> This aligns the driver with standard V4L2 sensor lifecycle expectations
> and avoids incorrect teardown during probe failures.

The rest of this reads very much like this was AI generated.

Did you use AI to generate these patches ? If so please:

Make sure you actually understand what the patch is doing and
very yourself that it actually is correct, which in this case
I believe it is not.

Regards,

Hans


> 
> Fixes: 1e72afb5146a ("media: Move gc0310 sensor drivers to drivers/media/i2c/")
> Signed-off-by: Sanjay Chitroda <sanjayembeddedse@gmail.com>
> ---
>  drivers/media/i2c/gc0310.c | 39 ++++++++++++++++++++++++++++----------
>  1 file changed, 29 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/media/i2c/gc0310.c b/drivers/media/i2c/gc0310.c
> index 7af4d66f42a0..72a82ad4118a 100644
> --- a/drivers/media/i2c/gc0310.c
> +++ b/drivers/media/i2c/gc0310.c
> @@ -721,7 +721,7 @@ static int gc0310_probe(struct i2c_client *client)
>  
>  	ret = gc0310_detect(sensor);
>  	if (ret)
> -		goto err_power_down;
> +		goto error_power_off;
>  
>  	sensor->sd.internal_ops = &gc0310_internal_ops;
>  	sensor->sd.flags |= V4L2_SUBDEV_FL_HAS_DEVNODE;
> @@ -730,20 +730,27 @@ static int gc0310_probe(struct i2c_client *client)
>  
>  	ret = gc0310_init_controls(sensor);
>  	if (ret)
> -		goto err_power_down;
> +		goto error_power_off;
>  
>  	ret = media_entity_pads_init(&sensor->sd.entity, 1, &sensor->pad);
> -	if (ret)
> -		goto err_power_down;
> +	if (ret) {
> +		dev_err_probe(&client->dev, ret, "failed to init entity pads\n");
> +		goto error_handler_free;
> +	}
>  
>  	sensor->sd.state_lock = sensor->ctrls.handler.lock;
>  	ret = v4l2_subdev_init_finalize(&sensor->sd);
> -	if (ret)
> -		goto err_power_down;
> +	if (ret) {
> +		dev_err_probe(&client->dev, ret, "subdev init error\n");
> +		goto error_media_entity;
> +	}
>  
>  	ret = v4l2_async_register_subdev_sensor(&sensor->sd);
> -	if (ret)
> -		goto err_power_down;
> +	if (ret) {
> +		dev_err_probe(&client->dev, ret,
> +			      "failed to register gc0310 sub-device\n");
> +		goto error_subdev_cleanup;
> +	}
>  
>  	pm_runtime_set_autosuspend_delay(&client->dev, 1000);
>  	pm_runtime_use_autosuspend(&client->dev);
> @@ -751,9 +758,21 @@ static int gc0310_probe(struct i2c_client *client)
>  
>  	return 0;
>  
> -err_power_down:
> +error_subdev_cleanup:
> +	v4l2_subdev_cleanup(&sensor->sd);
> +	pm_runtime_disable(&client->dev);
> +	pm_runtime_set_suspended(&client->dev);
> +
> +error_media_entity:
> +	media_entity_cleanup(&sensor->sd.entity);
> +
> +error_handler_free:
> +	v4l2_ctrl_handler_free(&sensor->ctrls.handler);
> +
> +error_power_off:
>  	pm_runtime_put_noidle(&client->dev);
> -	gc0310_remove(client);
> +	gc0310_power_off(&client->dev);
> +
>  	return ret;
>  }
>