Issue with finding net port for a MBIM modem

[Aleksander Morgado]

>> >> > > > But now I wonder, maybe we could just remove the whole block? Is
>> >> > > > there
>> >> > > > any case nowadays were we don't get port event removals? If this
>> >> > > > was a
>> >> > > > hack for a bug in kernel 2.6.31, maybe we could consider it
>> >> > > > already
>> >> > > > obsoleted?
>> >> > >
>> >> > >
>> >> > >
>> >> > > Yeah, this comes from an old commit
>> >> > > 53af144f49b0d81bd4dc1f5ee9eea6d61ccae992
>> >> > > (udev: handle removal of parent usb devices) dated back to 2009,
>> >> > > which seems
>> >> > > to imply that the kernel doesn't remove the tty when the usb device
>> >> > > is
>> >> > > removed. I guess it's probably not hard to reproduce the steps, but
>> >> > > I'm not
>> >> > > sure if the original fix was related to a particular kernel version
>> >> > > or a
>> >> > > particular devices or a combination of both.
>> >> > >
>> >> >
>> >> > Dan, what do you think?
>> >> >
>> >> > From my POV, I would nuke that device removal logic right away now...
>> >>
>> >> I'll have to think about that one, but I'm on vacation all this coming
>> >> week so I might not have a reply soon.  IIRC it may have been related
>> >> to before the tty subsystem was fully sysfs/kevent ported.  I think we
>> >> could be reasonably sure about it if we test a bunch of old modems with
>> >> the block removed.
>> >
>> >
>> > Aleksander, in the meantime, should I submit a patch to guard this block
>> > of
>> > code with a check DEVTYPE==usb_device? That would at least address the
>> > issue
>> > I observed on my Huawei modem. WDYT?
>> >
>>
>> Yes, this is something we can do, and would also be a safe update for
>> stable branches.
>
>
> sgtm, let me know if the previously attached patch looks reasonable to you.
> I'm not sure if we need to do anything if udev is disabled.
>

When udev is disabled, we rely on the user reporting port
addition/removals via the ReportKernelEvent() API. This API only
allows reporting port addition/removals (i.e. no interface or device
removals). So in theory, even without the #ifdef WITH_UDEV the code
would work correctly, but I guess it doesn't harm for now having it.

-- 
Aleksander
https://aleksander.es