Hi Gilles,

> Yes but most of them have OAM in the TO_DO list

which suggests they are waiting for something - support in the core?

> > > If it is not necessary put this feature in the module, how can this
> users do
> > > a OAM ping ??
> >
> > Using a user-space program?
> >
> > I guess the speedtouch just needs to add AAL0 support.
> >
> If you only accept AAL0 traffic, who will answer when OAM cells will be
> received?

Not the driver!  Either the ATM layer or user-space.  Compare with
standard ping: do device drivers generate the ping replies (yes I
know that this argument is weak)?  Since the ATM layer doesn't do
anything right now, I guess that means either (1) adding support to
the ATM layer, or (2) doing it from user-space.  Having a small
daemon to reply to pings would solve the immediate problem.

> Where will you transmit those cells?

Back out through the modem?

> Do you want to add a library wich will answer specifically to this OAM
> traffic?

Actions should be done at the right level.  What is the right level for
loopback?  The core provides a push_oam method, so it sounds
like it is expecting drivers to identify OAM cells, but not handle them.
The only problem is that this method is always NULL!

> Bewan support OAM both at module and user-mode level
> -the module accept received OAM traffic and answer to OAM cells
> - a user-space program allow to send OAM ping (this is only necessary if you
> want to be able to ping)
> 
> README file say
> "...

That doesn't mean that it is the right thing to do.  It was doubtless easier
for the authors to add support in their module than to add support in the
kernel's ATM layer.  It would be easy for me to write a patch that adds
loopback support to the speedtouch driver.  It would be harder to get it
into the kernel tree, though I could probably sneak it in since the driver
is in the USB subsystem and probably none of the ATM guys would
notice.  In any case, I would like to understand what the right solution
is before I start coding.  The fact that out-of-tree modules have OAM
support, but not the in-tree ones (though the authors seem to want it)
suggests to me that (1) OAM support in an in-tree module will not be
accepted, presumably because it shouldn't be done there, (2) that
out-of-tree modules have it because it seemed easier than trying to
change the kernel ATM layer.

Ciao,

Duncan.

Liste de diffusion modem ALCATEL SpeedTouch USB
Pour se d�sinscrire : mailto:[EMAIL PROTECTED]

        

Reply via email to