Re: [v6 08/14] iio: imu: add Bosch Sensortec BNO055 core driver

[Andrea Merello <Andrea.Merello@iit.it>]

    >
> ...
>
>> >> +static void bno055_debugfs_init(struct iio_dev *iio_dev)
>> >> +{
>> >> +       struct bno055_priv *priv = iio_priv(iio_dev);
>> >> +
>> >> +       priv->debugfs = debugfs_create_file("firmware_version", 0400,
>> >> +                                           iio_get_debugfs_dentry(iio_dev),
>> >> +                                           priv, &bno055_fw_version_ops);
>> >
>> >> +       devm_add_action_or_reset(priv->dev, bno055_debugfs_remove, priv->debugfs);
>> >
>> >Shouldn't we report the potential error here? It's not directly
>> >related to debugfs, but something which is not directly related.
>>
>> The error eventually comes out from something that has nothing to do with debugs per se (i.e. the devm stuff), but it will only affect debugfs indeed.
>>
>> Assuming that we don't want to make the whole driver fail in case debugfs stuff fails (see last part of the comment above debugfs_create_file() implementation), and given that the devm_add_action_or_reset(), should indeed "reset" in case of failure (i.e.  we should be in a clean situation anyway), I would say it should be OK not to propagate the error and let things go on.
>
>As I said, it's not directly related to debugfs. Here is the resource
>leak possible or bad things happen if you probe the driver, that fails
>to add this call for removal, remove it, and try to insert again, in
>such case the debugfs will be stale.

Hum, I would say this shouldn't ever happen: AFAICS devm_add_action_or_reset() is a wrapper around devm_add_action() and it's purpose is exactly to add a check for failure; devm_add_action_or_reset() immediately invokes the action handler in case devm_add_action() fails. IOW in case of failure to add the devm stuff, the debugfs file is removed immediately and it shouldn't cause any mess with next times probe()s; just the driver will go on without the debugfs file being here.

I think this is the point of using devm_add_action_or_reset() instead of dev_add_action()  indeed, or am I missing something?

>> However we can add a dev_warn() to report what happened.
>
>Not sure if it would suffice, I leave it to Jonathan.
>
>-- 
>With Best Regards,
>Andy Shevchenko