Huawei me906s-158

[Bjørn Mork]

Andreas Fett <andreas.fett at secunet.com> writes:
> On 07/03/16 09:38, Bjørn Mork wrote:
>> Use "lsof /dev/cdc-wdm0" to make sure there are no unwanted listeners.
> I checked that.
>
>> Make sure nothing probes the device for any other protocol than MBIM
>> (some Sierra Wireless Qualcomm based devices are for example known to
>> switch control protocol to whatever probes it first, regardless of USB
>> descriptors saying MBIM).
> The attached dump is started after the xhci and usbmon module are loaded
> (the device is not visible when only ehci/uhci are loaded).
>
> I then switched the device manually to configuration 3 via sysfs and
> loaded the cdc_mbim module and it's dependencies via modprobe.
>
> Afterwards I ran the mbimcli --query-device-caps command as before.
>
> Maybe you can see anything from that.

Thanks.  I don't know if it helps much, but at least this eliminates all
theories about any unwanted process interfering.  I see only the parts I
expect from the driver and initial MBIM OPEN message from libmbim:

 set configuration = 3
 set interface(1) = 0
 get NTB parameters returning pretty standard stuff
 set NTB format = 16bit
 set interface(1) = 1
 set NTB input size = 16384

The device responds to all these as expected.  No problems so far.  At
this point we expect the MBIM function to be set up properly and ready
for MBIM commands. But the device is silent.

I guess the only things left are all the long shots:

Make sure the modem is switched to MBIM mode as soon as possible (use an
udev rule).  The USB configuration switch appears to be fine, but there
could still be some problem if it initialized "something" before the
switch.

We could also try to make the driver setup sequence a little bit closer
to what Windows does and cross fingers.  Requires repeated attempts with
all sort of meaningless patches, unfortunately.  Difficult to do, and
probably futile.

But one simple change to test is this patch:
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/net/usb/cdc_ncm.c?id=48906f62c96cc2cd35753e59310cb70eb08cc6a5

Which will be in v4.5, and hopefully asserted stable releases too soon, if
you don't want to do your own patching.  FWIW, this should not make any
difference at all.  And I don't think it will. But I have been surprised
before...

The next step would be to add MBIM function reset request(s) etc.  Don't
know if we want to go down that route.  It didn't help the last device I
tried it on, and it turned out to require a number of pointless patches.


Bjørn