WDS Bind Mux Port?

[Daniele Palmas]

Hi Bjørn,

2017-02-15 10:34 GMT+01:00 Daniele Palmas <dnlplm at gmail.com>:
> Hi Bjørn and all,
>
> 2017-01-17 16:12 GMT+01:00 Bjørn Mork <bjorn at mork.no>:
>> Daniele Palmas <dnlplm at gmail.com> writes:
>>> 2017-01-12 14:53 GMT+01:00 Bjørn Mork <bjorn at mork.no>:
>>>> Daniele Palmas <dnlplm at gmail.com> writes:
>>>>
>>>>> 2017-01-11 21:21 GMT+01:00 Bjørn Mork <bjorn at mork.no>:
>>>>>
>>>>> Thanks!! I'll try to take a look at this and, if possible, do some tests!
>>>>
>>>> I know I wasn't going to spend more time on this, but the attached
>>>> "Doh!" fix showed up in the back of my head while I was sleeping.  It
>>>> saves you from testing how many netdes your system can handle before it
>>>> slowly stops ;)
>>>>
>>>> Note that it still doesn't work. I'm doing something wrong when looking
>>>> up the interfaces. But the lookup code should probably be made much more
>>>> efficient than looping over a list anyway, since we'll do it for each
>>>> packet received.
>>>
>>> I'm slowly taking a look and I'm not an expert, so please forgive me
>>> if I say something silly.
>>>
>>> In qmimux_find_dev should netdev_for_each_lower_dev replaced by
>>> something like netdev_for_each_upper_dev_rcu ?
>>>
>>> Similar fix also in the other parts (e.g. add_mux_show).
>>
>> Yes, that is probably correct.  I have a vertigo issue with network
>> device stacking ;)
>>
>
> attached my fixes to Bjørn's implementation to be applied on top of
> Bjørn's patches, not sure if this is how they should be done, but
> things seem to work properly since I'm now able to have two concurrent
> data connections.
>
> qmi_disconnect implementation is due to a problem when disconnecting
> the device (or removing the kernel module) if a qmimux network
> interface was still present.
>

I'm not sure which is the proper way we can go forward with this on
the kernel side.

Are you going to submit a patch for kernel inclusion or would you like
me to do so?

In the latter case, which way should I follow for giving you proper
credit, since it is 99% code written by you?

Thanks,
Daniele

> Carlo is patching libqmi to add support at the userspace level and
> will send the patches once available.
>
> Daniele
>
>>
>> Bjørn