Issues with Netgear 340U latest firmware

[Bjørn Mork]

Aleksander Morgado <aleksander at aleksander.es> writes:
> On Wed, Aug 13, 2014 at 9:19 PM, Dan Williams <dcbw at redhat.com> wrote:
>> Bjorn: I see that the Sierra drivers expect the following layout
>>
>> Serial:
>> 1199:68A2 - blacklisted: 8, 10, 11, 19, 20
>> 1199:68C0 - blacklisted: 8, 10, 11
>> 1199:9057 - blacklisted: 0, 1, 5, 8, 10, 11 ("Netgear AC341U IPT2 mode")
>>
>> Net:
>> 1199:68A2 - 8, 10, 19 ("MDM9x15 PDNs")
>> 1199:68C0 - 8, 10, 19
>> 1199:9057 - 8, 10, 11
>>
>> Which means that qmi_wwan is missing:
>>
>> 1199:68a2: 10
>> 1199:68c0: 19
>> 1199:9057: 10, 11

Yes. These are the USB interface numbers used for network functions on
Sierra QMI devices, but only a subset is available on any given PID:

5  MBIM
6  MBIM DATA
8  QMI/RMNET1
10 RMNET2
11 RMNET3
12 MBIM
13 MBIM DATA
19 RMNET2
20 RMNET3

(RMNETx are QMI functions)

I have so far considered the 10 addition to the 68a2 PID as merely a
macro convenience thing in the Sierra GobiNet driver - it's just simpler
to say that 8, 10, 19 are QMI interfaces regardless of PID using a
common macro. The same applies to the 19 interface for the 68c0 PID.  Is
that wrong?  Has anyone seen device configurations with these PID and
interface number combos?

This is the reason why qmi_wwan is missing the first two. The last one
should maybe have intf #10 added. Is really intf #11 enabled for
1199:9057 in the Sierra driver?  That's a bit unexpected.  Maybe it
actually works on that device?  Any volunteers for testing?

>> No idea whether adding these to qmi_wwan would be useful or not, you've
>> done more work than I with these devices, and I'm not sure where we
>> landed on whether or not to expose the non-functional QMI interfaces
>> through qmi_wwan.
>
> It was decided not to expose the non-functional ones.
>
> E.g. the MC7304 (0x1199, 0x68c0) exposed interfaces 8, 10 and 11. 8
> and 10 were both QMI (one with raw-ip by default, the other one with
> 802-3 by default), and 11 was the non-functional one, so it was
> removed from the driver.

Yes, and Sierra (unofficially) confirmed that the 3rd QMI function is
non-functional - at least on the older modules/firmwares:
https://forum.sierrawireless.com/viewtopic.php?f=117&t=6110#p32629

We do of course not know whether this will change (or has changed?) for
future devices.

I believe it should be simple to test for anyone having a device like
for example the 1199:9057 without having to patch the driver.  Dynamic
IDs take priority over static IDs since some recent kernel version (I do
of course not rememeber exactly which version :-), so if you do e.g.

 echo "1199 9057" >/sys/bus/usb/drivers/qmi_wwan/new_id

then the driver should attempt to bind to all interfaces without a
current driver.  Find the QMI device corresponding to USB interface #11
and see if you can get any QMI responses out of it.


Bjørn