If we get a HW FIFO overflow in MUX mode, there is little point in continuing. Probably simplest to reset the modem at that point.
At the moment, we pickup on various failures (like no MUX traffic, unable to make cellular connection, etc) and reset the modem. But I too have seen cases where the cellular connection can’t be made, that only a module reset seems to fix.

Perhaps the problem is in the async driver? Better to reset that as well?
Hardware flow control would be nice, but sadly we simply don’t have the GPIOs for it. We only have two free unused, and they are desperately needed for expansion modules. The MUX protocol does have a software flow control mechanism, but I am not sure if that would help.
I wonder why you are getting HW FIFO overflows? I am not seeing these.

On 12 Dec 2021, at 6:10 AM, Craig Leres <leres@xse.com> wrote:

I've made a tiny bit of progress with my sim7600a ppp issue.

First I realized that I could now run the master branch. The first test I did with my dev/7600 actually stayed connected to ppp for more than 14 hours... I went looking for differences in my for-v3.3 and master builds and discovered my for-v3.3 tree was still using wolfssl. So I thought I had found the problem until 12 hours later when the dev unit lost ppp again.

I also see that it's possible to get ppp back up without rebooting by powering cell down (power cell off), waiting for that to complete, and then powering back up.

I went looking for additional things to log. I have cellular, cellular-modem-auto, and cellular-modem-factory set to verbose and gsm-mux and gsm-ppp set to debug. What else would be helpful?

I was reviewing this thread and found this:

On 9/6/21 23:28, Michael Balzer wrote:
>
> I also added counters for these, you can query them with the "simcom
> status debug" command:
>
>> OVMS# sim stat deb
>> Network Registration: RegisteredHome
>> Provider: congstar
>> Signal: -97 dBm
>>
>> State: NetMode
>>   Ticker: 73099
>>   User Data: 0
>> *** HW FIFO overflows: 21*
>>   Buffer overflows: 0
>>
>>   Mux
>>     Status: up
>>     Open Channels: 4
>> ***    Framing Errors: 24*
>>     Last RX frame: 1 sec(s) ago
>>     RX frames: 151452
>>     TX frames: 5472
>>
>> PPP: Connected on channel: #2
>>
>> GPS: Connected on channel: #1
>>      Status: enabled

Indeed I see one fifo overflow on my dev (which has lots ppp one time since boot):

   OVMS# cell status debug
   MODEM Status
     Model: SIM7600
     Network Registration: RegisteredRoaming
       Provider: AT&T Hologram
       Signal: -83 dBm
       Mode: LTE,Online
     State: NetMode
       Ticker: 1434
       User Data: 0
     UART:
       FIFO overflows: 1
       Buffer overflows: 0
       Parity errors: 0
       Frame errors: 0
       Driver Buffer overflows: 0
     Mux: Status up
       Open Channels: 4
       Framing Errors: 0
       RX frames: 1864
       TX frames: 78
       Last RX frame: 0 sec(s) ago
     PPP: Connected on channel: #2
     GPS: Connected on channel: #1
        Status: enabled
        Time: enabled

I'm not sure if modem::modem() gets called when you power down/up the modem so I don't know if m_err_uart_fifo_ovf has been zero'd since boot.

On 9/6/21 18:35, Mark Webb-Johnson wrote:
>> Is it too soon to add the v3.3 and v1.3 pdfs to the source tree?)
>
> Yes, we haven’t finalised those, or gone through certification yet.
> We’ll release the v3.3 main board and modem layouts and schematics when
> that is complete (probably November/December timeframe).

I was thinking if I had the v3.3 modem schematic I could upgrade my boards to have uart hardware flow control. (I'm guessing if I had this I would not expect to see fifo overflows?)

Would it be useful for me to setup a way for you (Mark) to connect to my dev unit and interact with it while it is in the failed state?

Craig