iso support status.

[Duncan Sands]

Hi Matthieu,

> We plan to release a first version of ueagle-atm, and depreciate 
> eagle-usb for 2.6 kernel.
> 
> But we'll like to have a working usbatm upstream.

Fair enough.

> Are you ok to implement modem quirk (mask EILSEQ status for eagle) ?

Yes.

> How to we should control it : module option, option control per subdriver ?
> I like more the second, has we could have on the same machine 
> sub-drivers that need quirk and others that don't need it.

Well, we could just always mask it.  Otherwise, I suppose adding a
flag to struct usbatm_driver would do the trick.

> Could you find a way to select iso pipe method and have a good default 
> for buffer size ?
> 
> There is Roman method which consist in initialised driver internal 
> before .bind call, then the modem could overwrite it if it want, and 
> then usbatm use it.
> 
> I prefer to have and other option in usbatm_driver, as it will allow 
> static and dynamic allocation (why a modem could not support only iso 
> pipe for example ?).
> Also it avoid that the sub-driver access private data and it much 
> cleaner (don't depend of the order of init in usbatm).

I hope to work on iso support for the speedtouch later this week.
I'll get back to you on this then.

Ciao,

Duncan.