ModemManager and modems discovery in OpenWRt

[Enrico Mioso]

Ok, I'll now will try to see if i can break it.

BTW - is there any plan to try backporting MM OpenWRt feeds to 19.07? A good number of feeds is being backported it seems, so it may still be possible.



On Mon, 18 Nov 2019, Aleksander Morgado wrote:

> Date: Mon, 18 Nov 2019 12:25:39
> From: Aleksander Morgado <aleksander at aleksander.es>
> To: Enrico Mioso <mrkiko.rs at gmail.com>
> Cc: "ModemManager (development)" <modemmanager-devel at lists.freedesktop.org>,
>     Bjørn Mork <bjorn at mork.no>
> Subject: Re: ModemManager and modems discovery in OpenWRt
> 
> Hey!
>
>> Aleksander and Bjørn, I am CC'ing you since this question came up reading your discussion about udev rules usage.
>>
>> So, modems discovery mechanism in ModemManager is implemented via --report-kernel-event.
>> And we have --report-kernel-event-auto-scan as well.
>>
>> the current method of reporting even tough, is not as robust: and sometimes it may happen a modem is present but not detected.
>
> That would be a bug to analyze. Do you have details on how that happened?
>
>> With all the respect, while --report-kernel-event seems to be a very good decision / design choice, I don't like as much the current hotplug--based method / scripts we use to report for modem events.
>> The cache, and all that, they seem too fragile to me. What's your opinion on that?
>
> Well, as long as hotplug events happen when they should, the logic
> should be consistent IMO. There's one thing I don't like, though, and
> I may be wrong about how it works, but if for any reason there is
> another hotplug script taking too much time to run, the reporting of
> the events to MM may happen with a lot of time in between events, and
> the port probing mechanism may get delayed too much. I have a MR
> trying to mitigate that, please see:
> https://gitlab.freedesktop.org/mobile-broadband/ModemManager/merge_requests/207
>
>> I would be set out to replace that, maybe implementing some code to listen to netlink events and directly "inject" events in MM, or something like that.
>> I would like to have your opinion on that, and directions on the best way to proceed.
>>
>
> Not sure what to say :) IMO, if the hotplug scripts is the way openwrt
> has to report device events to userspace, we should stick to that. If
> that is not working well, we should fix it.
>
> -- 
> Aleksander
> https://aleksander.es
>