Re: QMI support in network-manager-applet

[Aleksander Morgado <aleksander lanedo com>]

On 07/05/13 17:47, Dan Williams wrote:
> > > > > > > Ok here's the output of modem manager in debug, I've removed all the
> > > > > > > > > > > > > > > > > > > > > novatel drivers and compiled them(Modem Manager, Network Manager,
> > > > > > > > > > > > > > > > > > > > > nm-applet) in the order that Dan recommended. The output is a bit
> > > > > > > > > > > > > > > > > > > > > lengthy so I've attached it as a text document.
> > > > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > > > Are you able to test with the new "aleksander/novatel-fixes" branch from
> > > > > > > > > > > > > > > > > git? The main problem was that the Novatel plugin didn't allow QMI
> > > > > > > > > > > > > > > > > modems yet, it should allow them now.
> > > > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > > > Dan, quick review of the branch?
> > > > > > > > > > > > >
> > > > > > > > > > > > > Do we want those driven by Novatel or by Generic though?
> > > > > > > > >
> > > > > > > > > If we can grab them by the Novatel plugin, I think we should. We don't
> > > > > > > > > want all QMI modems to be handled by the Generic plugin, as that would
> > > > > > > > > mean handling all QMI modems with QMI... and I don't think we should do
> > > > > > > > > that e.g. for ZTE QMI modems (which are not handled with QMI currently,
> > > > > > > > > as they suck).
> > > > > The branch works and is fine, but the problem with using the Novatel
> > > > > plugin is that it takes a lot longer to probe than the Generic one, due
> > > > > to the $NWDMAT=1 stuff that the plugin does, which is completely
> > > > > unecessary on the QMI-capable devices.  I'm not sure there's a good way
> > > > > to suppress that given that we're not sure if the modem has a QMI port
> > > > > at the time we send the DMAT=1 stuff.
> > >
> > > But the generic plugin also does AT probing... or is it that NWDMAT
> > > takes longer to reply?
> The NWDMAT stuff just takes longer in my experience, at least 3 or 4
> retries on the ports.  Plus, the QMI devices just have more ports (eg 4
> or so) than the non-QMI devices that NWDMAT is useful on, which
> typically have 2 or 3.
>
> > > Instead of calling NWDMAT as a custom AT probing method, we could set it
> > > up in a custom_init() instead, and only run it if there is no QMI port
> > > in the device?
> Yeah, that sounds better.

Implemented that logic in the branch now. Are you able to test it?

-- 
Aleksander