[Ovmsdev] SIMCOM UART FIFO overflows (was Kona electric and http routing)

Michael Balzer dexter at expeedo.de
Thu Oct 24 02:53:43 HKT 2019 

For comparison, this is my bench module with the old CP2102:

    OVMS# simcom status debug
    Network Registration: RegisteredRoaming
    Provider: Vodafone.de Hologram
    Signal: -91 dBm

    State: NetMode
      Ticker: 593
      User Data: 0
      HW FIFO overflows: 63
      Buffer overflows: 0

      Mux
        Status: up
        Open Channels: 4
        Framing Errors: 61
        Last RX frame: 0 sec(s) ago
        RX frames: 1069
        TX frames: 47

    PPP: Connected on channel: #2

    GPS: Connected on channel: #1
         Status: enabled
         Time: enabled


The bench module has both my main Wifi AP and the repeater in reach, so that may also be due to the Wifi situation.

Regards,
Michael


Am 23.10.19 um 19:31 schrieb Michael Balzer:
> I've pushed my SIMCOM changes. With these optimizations, fifo overflows and frame errors have dropped significantly:
>
>     OVMS# sim stat de
>     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
>          Time: enabled
>
> …but the overflow frequency differs between my modules (both v3.1, the one with the first gen CP2102 has more overflows), and I can still see a direct
> relation to Wifi.
>
> With Wifi switched completely off, no overflows occurred over a period of ~5 hours. As soon as Wifi was reactivated (regardless of the mode), the errors were
> back.
>
>
> Am 23.10.19 um 09:31 schrieb Mark Webb-Johnson:
>> While there is flow control on the USB side, I don’t think there is any between the ESP32 and the CP2102. See Espressif’s example DEVKIT-C schematic:
>>
>>     https://dl.espressif.com/dl/schematics/ESP32-Core-Board-V2_sch.pdf
>>
>>
>> or the OVMS one (which we based on DEVKIT-C). Just RX and TX data lines to the ESP32.
>
> What about the RTS/CTS lines connected to IO13 & IO15 in the core board?
>
>> However, during flashing there is pretty much nothing else running on the system and no high level operating system.
>
> Indeed.
>
> Regards,
> Michael
>
>> The good news is that these are very very short data lines. Just a few inches, I think. I did look at the signals in the early days of the project, and they
>> seem quite clean.
>>
>> The GMS MUX flow control is at the frame level. I would guess that several frames would still need to be fit into the buffer for it to be effective. It is
>> implemented on a per-channel basis, and a short description (from the blox manual) is:
>>
>>     http://read.pudn.com/downloads406/ebook/1729787/MuxImplementation_ApplicationNote_(WLS-CS-11002).pdf
>>
>>     6.3 FlowControlonvirtualchannels
>>
>>     The Flow control of the virtual channel is implemented in terms of MSC packets with the FC bit. If the application processor sets the FC bit to 1 for a
>>     particular DLC, the TE does not send data to the application processor for that DLC until FC returns to 0 for the same DLC. The TE has limited resources
>>     for buffering data, so if the DLC is involved in large data transfers (for example downloading data through a GPRS connection) a buffer overflow may
>>     occur if the time between FC=1 and FC=0 is too long; in this case data may be lost and there is no error indication.
>>
>>     The application processor should avoid (if possible) the use of this feature or keep the time interval with the FC=1 as small as possible. 
>>
>>
>> I never really did any optimisation of the SIMCOM (or MUX) driver at all. The MUX in particular was written by looking at other implementations, as the
>> protocol specification has gone the way of most standards body specifications and is mostly undecipherable. I think there is some opportunity for improvement
>> (but given our low bandwidth requirements at the time we never had the incentive). Flow control is not implemented at all.
>>
>> Regards, Mark
>>
>>> On 23 Oct 2019, at 3:01 PM, Michael Balzer <dexter at expeedo.de <mailto:dexter at expeedo.de>> wrote:
>>>
>>> Mark,
>>>
>>> also not trying to sound too negative… but please read: https://github.com/openvehicles/Open-Vehicle-Monitoring-System-3/issues/274
>>>
>>> The flashing process is done with hardware flow control. With Wifi enabled, we cannot even handle 115 kbit without fifo overflows.
>>>
>>> We can try the MUX flow control you mentioned, but will it be able to throttle transmission within frames longer than the HW FIFO?
>>>
>>> Btw, I've got an improvement on fifo overflow recovery in testing, if you want to work on this, I can push my changes this evening.
>>>
>>> Regards,
>>> Michael
>>>
>>>
>>> Am 23.10.19 um 08:50 schrieb Mark Webb-Johnson:
>>>> But perhaps I am unintentionally sounding too negative...
>>>>
>>>> The biggest hurdle to this working technically is undoubtedly the LWIP support for SNAT and routing. If the library you caption (jonask1337/esp-lwip)
>>>> solves that, it makes this technically feasible.
>>>>
>>>> My comments on baud rate on the UART link between ESP32 and SIMCOM are more about the practicality of it. That could be tested with a few simple
>>>> modifications to our SIMCOM driver, to see how fast it could actually be driven. I know we get up to about 1Mbps for firmware flashing without an issue,
>>>> and the ESP32 hardware UART is up to 5Mbps.
>>>>
>>>> It would be a fantastic feature to have, and incredibly useful.
>>>>
>>>> Regards, Mark.
>>>>
>>>>> On 23 Oct 2019, at 2:33 PM, Mark Webb-Johnson <mark at webb-johnson.net <mailto:mark at webb-johnson.net>> wrote:
>>>>>
>>>>> The connection between the SIMCOM and the ESP32 is 115,200 baud. That could be increased (in software), and there is software flow control on the GSM MUX
>>>>> we use (although I have no idea if SIMCOM implements it); but without hardware flow control lines I don’t think it could/would approach 3G speeds.
>>>>>
>>>>> The alternative is to swap it around. Put the modem and the SIM in some other device designed for that purpose, and have OVMS connect to that as a WiFi
>>>>> client.
>>>>>
>>>>> Regards, Mark.
>>>>>
>>>>>> On 23 Oct 2019, at 2:23 PM, Peter Lord <plord12 at gmail.com <mailto:plord12 at gmail.com>> wrote:
>>>>>>
>>>>>> Hi All,
>>>>>>
>>>>>> I've been lurking here for a while still debating wether to ditch my autopi in favour of OVMS.
>>>>>>
>>>>>> One thing thats held me back is to find a way to use the wifi hotspot as a NAT router - this is
>>>>>> useful to allow my sat nav to get traffic and charging point updates.  As far as I can see on
>>>>>> the web page this isn't currently supported.
>>>>>>
>>>>>> However I did see a couple of projects that adds NAT support to lwip :
>>>>>>
>>>>>> https://github.com/martin-ger/lwip_nat_arduino
>>>>>> https://github.com/jonask1337/esp-lwip
>>>>>>
>>>>>> Does anyone know if adding NAT has a fighting chance ?
>>>>>>
>>>>>> Cheers,
>>>>>>
>>>>>> Pete
>>>>>> _______________________________________________
>>>>>> OvmsDev mailing list
>>>>>> OvmsDev at lists.openvehicles.com
>>>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>> _______________________________________________
>>>> OvmsDev mailing list
>>>> OvmsDev at lists.openvehicles.com <mailto:OvmsDev at lists.openvehicles.com>
>>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>>
>>>
>>> -- 
>>> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
>>> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>>>
>>>
>>> _______________________________________________
>>> OvmsDev mailing list
>>> OvmsDev at lists.openvehicles.com <mailto:OvmsDev at lists.openvehicles.com>
>>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>>
>>
>> _______________________________________________
>> OvmsDev mailing list
>> OvmsDev at lists.openvehicles.com
>> http://lists.openvehicles.com/mailman/listinfo/ovmsdev
>
> -- 
> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>
> _______________________________________________
> OvmsDev mailing list
> OvmsDev at lists.openvehicles.com
> http://lists.openvehicles.com/mailman/listinfo/ovmsdev

-- 
Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
Fon 02333 / 833 5735 * Handy 0176 / 206 989 26

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvehicles.com/pipermail/ovmsdev/attachments/20191023/87343dc0/attachment.htm>

More information about the OvmsDev mailing list