When should gpio-restart be used/avoided?

[Lech Perczak]

W dniu 2022-08-14 o 21:15, Bjørn Mork pisze:
> Lech Perczak <lech.perczak at gmail.com> writes:
>> W dniu 2022-08-14 o 18:40, Bjørn Mork pisze:
>>> Lech Perczak <lech.perczak at gmail.com> writes:
>>>
>>>> I'll try your settings, on MF286D and on older MF286 as well.
>>>> My setup is a bit more convoluted, since my build has patches to use
>>>> split APN for IPv4v6 dual stack and I use DHCP.
>>> Are patches still required for dual-stack with uqmi?
>> For generic ipv4v6 they are not, however my usecase is even more special.
>> I use "internet" APN for v4 and "internetipv6" APN for v6, hence
>> "split APN".
> Ah, right.  That's unusual.
>
>> My configuration is like this:
>> config interface 'wan'
>>          option device '/dev/cdc-wdm0'
>>          option proto 'qmi'
>>          option apn 'internet'
>>          option v6apn 'internetipv6'
>>          option profile '1'
>>          option v6profile '5'
>>          option autoconnect '1'
>>          option dhcpv6 '1'
>>          option auth 'none'
>>          option pdptype 'ipv4-and-ipv6'
> Very strange.  I would not have guessed that this was possible.  I've
> always assumed that the QMI dual stack model depended on two sessions to
> the same APN, and a single IPV4V6 context to the network.  Learning
> something every day.  And forgetting two things.
>
>> Do you think it's worth upstreaming? Users of Orange Poland would
>> benefit from that, not sure how in other parts of the world.
> If it's useful to you, then I'm sure it will be useful to someone else
> as well.  Please do upstream it.  Assuming that it's not too ugly :-)
Ok, I'll open a RFC PR soon.
>
>> It seems doable with ModemManager as well, by creating two separate
>> bearers (and that's done under the hood for single APN as well),
>> but I haven't tried so far, uqmi works fine for me, albeit it lacks
>> connection drop monitoring.
> I'm not sure if MM can do that with one network interface on the host
> side.  Two bearers need two interfaces.  But if we have QMAP support
> then there shouldn't be any problem having one IPv4 and one IPv6
> interface connected to separate APNs.
>
>>> That warning isn't there when the "power button blocker" is disabled.
>> This part was expected. Just checked on my unit, and this still did
>> reboot the modem.
> How do you check for modem reboot?
I checked for modem LEDs to turn off after executing reboot (with 
GPIO421 set high)
>
>> So it seems, that the modem reset is still possible by toggling this
>> GPIO from userspace, while holding power button blocker,
>> even without rebooting whole device.
> That sounds more useful than having this wired up as a router reboot
> device.
Yes, but fiddling with two GPIOs to do just that is finicky. Let's see 
what Paweł thinks on that,
I'll try digging into MF286 a bit more to get the modem to restart 
without rebooting device,
for newer ones we can redefine this to be two separate GPIO switches.
>
>> This won't be possible on MF286 though, it lacks the power button
>> blocker line in hardware - it's physically present only from MF286A
>> onwards,
>> It would be certainly nice, if the behaviour is consistent across the
>> whole line of devices.
> Yes, dream on.  The only consistent part of a ZTE device is the model
> number :-)

But in OpenWrt we can still at least try to achieve that.

Based on MF286 series I tend to disagree, they are very similar to each 
other.
Router part of MF286A and MF286R is the very same device.
Too bad, that all of them have just "MF286" written on the boilerplate.

>
>
> Bjørn


-- 
Pozdrawiam,
Lech Perczak