> Don't do this hard-coded form of extensiblity. ioctl()s are inherently
> extensible because they encode the size. Instead of switching on the
> full ioctl, switch on the ioctl number. See for example fs/pidfs:
>
> /* Extensible IOCTL. */
> if (_IOC_NR(cmd) == _IOC_NR(PIDFD_GET_INFO))
> return pidfd_info(file, cmd, arg);
>
> static long pidfd_info(struct file *file, unsigned int cmd, unsigned long arg)
> {
> struct pidfd_info __user *uinfo = (struct pidfd_info __user *)arg;
> <snip>
> size_t usize = _IOC_SIZE(cmd);
> struct pidfd_info kinfo = {};
>
> if (!uinfo)
> return -EINVAL;
> if (usize < PIDFD_INFO_SIZE_VER0)
> return -EINVAL; /* First version, no smaller struct possible */
>
> pidfs uses a mask field to allow request-response modification:
Thanks for the detailed feedback — very helpful.
For now, I'll keep it simple and skip adding a mask field since all
fields in the struct are always returned.
> (Only requirement is that a zero value means "no info", i.e., can't be a
> valid value. If you want zero to be a valid value then a mask member
> might be helpful where the info that was available is raised.)
To clarify on the zero values: the fields in this struct are
capability fields, where a zero value indicates that the hardware
doesn’t support the corresponding feature. None of the fields have zero
as a valid value when the feature is supported, so a mask isn’t
necessary.
I sent another version [1] with your feedback applied. Please see if it
aligns with what you had in mind.
[1] https://lore.kernel.org/linux-block/20250618055153.48823-1-anuj20.g@xxxxxxxxxxx/
--
Anuj Gupta
