Re: [PATCH 4/6] drivers:misc: sources for Init manager module

[Marcel Holtmann <marcel@holtmann.org>]

Hi Pavan,

> > > Which puts me back to square 1,
> > the requirement is TTY device should be open, only when
> > either BT, FM or GPS would want to use it.
> > > [with your patch one of the steps of installing ldisc
> > would be reduced upon opening].
> > 
> > Perhaps but its up to you how you write your low level
> > driver and how the
> > ldisc indicates it wishes to be active (eg speed B0 is used
> > to indicate
> > 'no carrier' in many cases)
> > 
> > > Actually why do I even need a TTY device in this case
> > - right ?
> > > ldisc driver can do the tty->ops->write and
> > tty_read and put it up on different interfaces like
> > eth0/hci0 or /dev/radio0 etc..
> > 
> > You don't need a tty if you are simply demuxing some kind
> > of stream of
> > bytes to/from the hardware and you gain nothing from the
> > tty layer and
> > user space interfaces such as control signal and speed
> > setting.
> > 
> 
> I still want to maintain this as a ldisc driver and the TTY layer is still required to be able to use this with different sort of serial drivers.
> 
> example:
> We do a have a system here where both 8250/omap-serial co-exist creating their own device nodes (ttyS for 8250, ttyO for omap-serial) and I would want to be able to on boot suggest which serial driver to use, so I can open/install ldisc on ttyS or on ttyO.
> 
> 
> > That may be even cleaner in your case as you can then then
> > provide
> > suitable links between your demux and the drivers attached
> > to it
> > indicating when they should be on or off.
> 
> And as Marcel suggested, can't really put in the PM/UART-clk shut off code in ldisc driver because few of these driver provide interfaces and few don't.
> 
> So is there no way this can be accepted with the sysfs entry way of communication ?

to be quite honest, I can't see how that can be accepted right now. It
looks wrong on too many levels.

Regards

Marcel