> (perhaps i2c_verify_client() checking dev was not such a great idea, as
>  callers need to act on dev && !verified anyway?)

Can be argued. I will have a second thought about it.