Re: [MM] grabbing QMI/WDM ports

[Aleksander Morgado <aleksander lanedo com>]

> > For anyone interested, there's some work ('qmi-wdm' branch in the
> > upstream MM tree) to be able to grab qmi/wdm ports, currently just
> > exposed as raw MMPort objects. Of course, you'll need a device with
> > such ports (got a Pantech UML290 here) and a recent enough kernel with
> > Björn's qmi_wwan driver. Any review greatly appreciated, specially
> > because I used just 1 device for testing it.
>
> Great!  I'll see if I can get around to playing with it.  Testing these
> things tend to take me some time and effort as I'm not used to using NM
> and MM ;-)

Great, thanks!

> I noticed one issue though, without actual testing: I believe you want
> to remove the DRIVERS match from the udev rule:
>
>    KERNEL=="cdc-wdm*", SUBSYSTEM=="usb", DRIVERS=="qmi_wwan", ENV{ID_MM_CANDIDATE}="1"
>
>
> That line will be correct for your Pantech device and for lots of Gobi
> devices where the qmi_wwan driver uses cdc-wdm as a "subdriver".  But
> the subdriver design may make different drivers own cdc-wdm devices.
> Currently the only ones are cdc-wdm and qmi_wwan, but other drivers may
> show up.  E.g. if a device needs a driver supporting a combined
> QMI/CDC-NCM USB interface.

Noted. Will change that later.

> In short, I believe all cdc-wdm devices will be useful to ModemManager
> regardless of driver.  It doesn't even matter which protocol they use,
> as long as ModemManager has some way to detect it (by device id or
> probing).
>
> The classical example is the Ericsson MBM modems used in Lenovo and Dell
> laptops, which present AT command cdc-wdm interfaces.  ModemManager
> should collect these and use them if necessary.

So those cdc-wdm interfaces are pure AT ports, no QMI involved in there? 
Then we can include AT probing for these ports as well, and if they are 
not AT assume they are QMI; or even setup new QMI probing as well.


--
Aleksander