Re: Interactions with ModemManager

[Colin Helliwell <colin helliwell ln-systems com>]

> On 20 April 2017 at 17:28 Dan Williams <dcbw redhat com> wrote:
>
> On Mon, 2017-04-10 at 14:40 +0100, Colin Helliwell wrote:
>
> > A couple of things I've noticed in integrating NM with ModemManager:
> >
> > 1.  It seems that eth0 doesn't come up - i.e. dhcp client not fired
> > off - until MM has done it's startup operations on the GSM modem. It
> > so happens that my GSM has a long power-on pause, and that is holding
> > up the system's eth network start-up too. Not a huge problem as a
> > hyper-fast startup isn't needed, but not ideal..
>
> Missed this message... This may be due to NM's "startup" function,
> where it waits for all interfaces that exist at startup to settle
> before it tries to do anything with them. This tries to ensure that
> composed devices (eg, bonds, bridges, teams, etc) have all the
> components they need at startup.
>
> So yeah, that could prevent things from happening if the GSM modem
> takes a long time to power on. You might be able to work around this
> by setting the NM "wwan enabled" property to "off" and then having
> scripts or something flip that on later? A dispatcher script could do
> that perhaps.
>
> > 1.  Even after the initial 'query' start-up on the GSM, NM seems to
> > leave it enabled (as per mmcli -m 0), rather than leaving it disabled
> > until called upon (with a nmcli conn up XX)
>
> NM has the 'wwan-enabled' option which controls whether it powers up
> modems on start. If that option is off, NM will leave them in
> 'disabled' state until the 'nmcli con up XXX'.
>
> Try setting 'nmcli radio wwan off', ensuring the modem is disabled in
> MM, then restarting NM. Does that keep the modem disabled?

Just been having a try-out with that control. I see that the state persists across reboot - is there a way to 
force a particular bootup setting?