Point to Point interface over qmi_wwan device

[Dan Williams]

On Tue, 2016-10-04 at 22:49 +0200, Reinhard Speyerer wrote:
> On Tue, Oct 04, 2016 at 08:12:15PM +0000, Tang Nguyen wrote:
> > 
> > 
> > --------------------------------------------
> > On Tue, 10/4/16, Dan Williams <dcbw at redhat.com> wrote:
> > 
> >  Subject: Re: Point to Point interface over qmi_wwan device
> >  To: "Tang Nguyen" <tang_nguyen at yahoo.com>, "Bjørn Mork"
> > <bjorn at mork.no>
> >  Cc: "libqmi (development)" <libqmi-devel at lists.freedesktop.org>
> >  Date: Tuesday, October 4, 2016, 1:17 PM
> >  
> >  On Tue, 2016-10-04 at
> >  16:28 +0000, Tang Nguyen wrote:
> >  > Hi
> >  all,
> >  > I have tested /32 mask on wwan
> >  without running dhclient. but seems
> >  > not
> >  working .  The  packet connection successfully connected
> >  but it
> >  > would be dropped after a
> >  while.  Looks like Sierra firmware waits for
> >  > dhclient discovery request from host, and
> >  after a while it would
> >  > disconnect the
> >  data link.
> >  > My module is MC7354 (ATT)
> >  > What I did as below:
> >  >
> >  1. start network connection.  Verify connection status with
> >  command:
> >  > qmicli  -d /dev/cdc-wdm3
> >  --wds-get-packet-service-status2. Send AT
> >  > command to modem to show assigned IP
> > > 
> > > address3. Using ip command to 
> >  
> >  >Why use the AT command?  Why not:
> >  
> >  >qmicli -d /dev/cdc-wdm3
> >  >--wds-get-current-settings --client-no-release-
> >  >cid --client-cid=<cid from start
> >  >network>
> >  
> >  >I seem to
> >  >recall we've had devices before that didn't fully
> >  >set up the
> > > 
> > > WAN interface until they got
> >  >asked for the IP details, and I wouldn't
> >  >necessarily trust the AT interpreter to do the
> >  >same thing.  But the QMI
> >  >GetCurrentSettings
> >  >request might.
> >  
> > > 
> > > Dan
> > 
> > Using libqmi doesn't help Dan.  I can see IP, assign it to cellular
> > interface but still not able to ping outside
> > 
> > [/dev/cdc-wdm3] Connection status: 'connected'
> > root at rx1500-2:~# qmicli -d /dev/cdc-wdm3 --wds-get-current-
> > settings 
> > [/dev/cdc-wdm3] Current settings retrieved:
> >            IP Family: IPv4
> >         IPv4 address: 10.106.63.25
> >     IPv4 subnet mask: 255.255.255.252
> > IPv4 gateway address: 10.106.63.26
> >     IPv4 primary DNS: 207.219.69.11
> >   IPv4 secondary DNS: 216.218.29.11
> >                  MTU: 1430
> >              Domains: none
> > root at rx1500-2:~# ip add add 10.106.63.25/32 peer 10.106.63.26
> > Not enough information: "dev" argument is required.
> > root at rx1500-2:~# ip add add 10.106.63.25/32 peer 10.106.63.26 dev
> > cel-6-1
> > root at rx1500-2:~# ifconfig cel-6-1
> > cel-6-1   Link encap:Ethernet  HWaddr 0e:9a:50:15:45:49  
> >           inet
> > addr:10.106.105.95  Bcast:0.0.0.0  Mask:255.255.255.255
> >           inet6 addr: fe80::c9a:50ff:fe15:4549/64 Scope:Link
> >           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
> >           RX packets:0 errors:0 dropped:0 overruns:0 frame:0
> >           TX packets:517 errors:29 dropped:0 overruns:0 carrier:0
> >           collisions:0 txqueuelen:1000 
> >           RX bytes:0 (0.0 B)  TX bytes:99778 (97.4 KiB)
> 
> Hi Tang,
> 
> that matches the results from my tests with a MC7304. For MSM9x15
> based devices like the MC7304 the use of DHCP seems to be a trigger
> to
> pass IPv4 traffic over the network interface as long as it is set to
> Ethernet mode. Only using a qmi_wwan version with raw IP support and
> switching the mobile use raw IP with qmicli made this configuration
> work for me without using a DHCP client.

Yeah, that was going to be my next suggestion.  Use rawip mode:

qmicli -d /dev/cdc-wdm0 --wda-set-data-format=raw-ip
qmicli -d /dev/cdc-wdm0 --set-expected-data-format=raw-ip

or, if your device is older and doesn't have the QMI WDA service:

qmicli -d /dev/cdc-wdm0 --device-open-net="net-raw-ip|net-no-qos-header" --dms-get-ids
qmicli -d /dev/cdc-wdm0 --set-expected-data-format=raw-ip

At least, I think those are the right commands.  I'm sure Bjorn will
correct me...

Dan