[Ovmsdev] OVMS going comatose

[Michael Balzer]

Steve,

in the two weeks since disabling TLS on the server V2 connection I 
haven't had a single crash. While that's not a proof yet the watchdog 
issue is TLS related, it's at least a strong indicator.

The watchdog triggers if the idle task on a core doesn't get a CPU share 
for 120 seconds. If the TLS functions block a CPU for more than a few 
seconds, that's already pretty bad, as that means TLS will cause delays 
in CAN processing (disrupting protocol transfers) and can possibly cause 
frame drops and queue overflows. Blocking the whole system for more than 
120 seconds is totally unacceptable.

This doesn't feel like a calculation / math performance issue, it rather 
feels like a bug – and that may imply a security issue as well.

But I don't think this is caused by WolfSSL, as the issue has been 
present with mbedTLS as well, just didn't occur that frequently. Maybe 
some race condition with the LwIP task?

Regards,
Michael


Am 11.04.21 um 09:44 schrieb Michael Balzer:
> Steve,
>
> I can confirm an increase of these events since we changed to WolfSSL, 
> about once every three days currently for me. The frequency was much 
> lower before, more like once or twice per month.
>
> I've disabled TLS on my module now and will report if that helps.
>
> Regards,
> Michael
>
>
> Am 10.04.21 um 21:20 schrieb Stephen Casner:
>> Michael,
>>
>> As you saw from my earlier emails, I was getting these crashes
>> typically after less than 24 hours of operation.  I changed my config
>> to disable TLS on server v2 and rebooted 2021-04-05 23:36:04.648 PDT
>> and there has not been a crash since.  So it definitely appears to be
>> correlated with the additional processing to support TLS.
>>
>>                                                          -- Steve
>>
>> On Sun, 4 Apr 2021, Michael Balzer wrote:
>>
>>> Steve,
>>>
>>> that's the problem with this issue, it's totally unclear what causes 
>>> this.
>>>
>>> The signal dropping begins when the queue is full, which happens 
>>> after the
>>> task has been blocked for ~as many seconds as the queue is big. So 
>>> there is no
>>> logged activity that could cause this, your module basically went 
>>> into this
>>> from idling.
>>>
>>> Regards,
>>> Michael
>>
>
>
> _______________________________________________
> 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/20210424/a135db47/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 203 bytes
Desc: OpenPGP digital signature
URL: <http://lists.openvehicles.com/pipermail/ovmsdev/attachments/20210424/a135db47/attachment.sig>