Re: new problems with connecting with an LTE modem

[Dan Williams <dcbw redhat com>]

On Fri, 2016-05-20 at 19:43 +0200, Bhima wrote:
> Hey thanks for the quick response!
>
> I'm not able to use the qmi-network helper script anymore because it 
> always fails:
>
> Starting network with 'qmicli -d /dev/cdc-wdm0 
> --wds-start-network=drei.at  --client-no-release-cid'...
> [20 May 2016, 19:28:07] -Warning ** Error reading from istream:
> Resource 
> temporarily unavailable
> error: couldn't create client for the 'wds' service: CID allocation 
> failed in the CTL client: Transaction timed out

If you're running ModemManager in the background, then you want to use
the "proxy" option (-p) for qmicli.  Otherwise ModemManager and qmicli
will fight over the cdc-wdm device.  MM (at least recent versions) will
use the proxy by default to allow for this.

Does just adding "-p" to the qmicli commands in qmi-network make things
work better?

> So I've been using mmcli, like this: "sudo mmcli --verbose 
> --simple-connect="apn=drei.at,ip-type=ipv4" -m 0"
>
> The connection that gets made is pretty bad... like 5% of bandwidth
> I'm 
> used to. The signal strength reported by mmcli status report is "67"
> but 
> I've got no idea what that really means (and it's not like I've
> changed 
> the hardware or even moved it recently).

If you're able to switch back to a previous kernel, and only change the
running kernel, do things work better?  I'm wonder if there's been a
driver change with scheduling USB URBs or something for the network
part, that might be impacting performance.  Something else to try is to
turn off USB runtime power management for the device:

(as root)
echo "0" > /sys/class/usbmisc/cdc-wdm0/device/../power/control

and see if that changes anything.

Dan

> I'm wondering if the modem gets half way through getting configured
> and 
> making the connection with another, perhaps new, mechanism and now 
> trying to go through the process twice is what is making it not work.
>
> On 20/May/16 6:21 pm, Dan Williams wrote:
> > On Fri, 2016-05-20 at 16:10 +0200, Bhima wrote:
> > > Hi Guys!
> > >
> > > I'm having problems with a Sierra MC7710 LTE modem, running in a
> > > Soekris
> > > Net6501, after a recent update to Ubuntu 16.04 to the 4.4.0-22-
> > > generic
> > > kernel.  For the past few years I've been using the qmi-network
> > > helper
> > > script to connect cdc-wdm0, which in turn created a wwan0
> > > port.  Now
> > > that's all gone pear shaped and I've gone around in circles
> > > trying
> > > figure out what's gone wrong.
> > >
> > > looking in dmesg, I see that wwan0 is actually getting renamed to
> > > wwp2s2f3u3i8, which is confirmed by status returned by "mmcli -m
> > > 0"
> > > (ports: 'ttyUSB0 (qcdm), ttyUSB2 (at), cdc-wdm0 (qmi),
> > > wwp2s2f3u3i8
> > > (net)').  I can get a dhcp address on 'wwp2s2f3u3i8' but the
> > > throughput
> > > on the resulting channel is severely constrained.
> > >
> > > So now I'm wondering what exactly the proper way to configure and
> > > use
> > > this modem from the command line these days.
> > Probably the same way you always have; as you've found the
> > interface is
> > now named something different, which is happening because of udev's
> > interface renaming policy in the updated Ubuntu release.  But it's
> > still the same interface as before and should work the same way as
> > before.
> >
> > How much of a drop in throughput have you've found?  Any throughput
> > drop will likely be related to kernel driver changes in qmi_wwan
> > and
> > the kernel USB layer, as opposed to something in ModemManager
> > itself or
> > the qmi-network script.  When the througput is bad, what is the
> > signal
> > strength of LTE connection?
> >
> > Dan
> _______________________________________________
> networkmanager-list mailing list
> networkmanager-list gnome org
> https://mail.gnome.org/mailman/listinfo/networkmanager-list