Re: [RTnet-developers] RTDM devices class and subclass

[Jan K. <jan...@we...>]

Marcelo Coelho wrote:
>> Definitely not. I was referring to the "rtnet" part - RTnet is only on=
e
>> RTDM user.
>=20
> Yes, but you're the only developer i can contact, and this is the less
> active mailling list...
>=20

I would prefer to see such a generic discussion on a generic mailing
list, like xenomai-core or rtai. There might be more people hanging
around, willing to contribute comments and ideas.

>=20
> Now about the driver, i searched in rtdm for something that could be
> related to the TTY associated structs and functions. Neither in the
> serial device nor in rtdm api, these functions are used. Is it possible=

> that you (the RTDM team) change that and start giving support for those=

> kind of devices in the same way the kernel does?

If this would be only about wrapping a *few* IOCTLs, I see no big
problems (except that the driver code may increase). What interfaces do
you have in mind?

Well, the termios interface as-is appeared to me as not well suited for
typical serial usage scenarios in hard-RT applications. Specifically,
the RTDM profile introduced an event concept and per-character
timestamping. As there were several extensions required anyway, I
consolidated the whole interface, making it more regular and targeted
with respect to UART devices - that's at least my feeling about it.

If your driver reveals some need to move common code into a high-level
serial layer anyway, adding a compat layer for certain IOCTLs the driver
does not need to know about might be an options. But note that, even
then, using tc*() functions will not work as they map to the Linux
ioctl() invocation, not the RTDM one.

Jan