[Ovmsdev] OVMS v2 Memory/Settings Corruption

Julien Banchet jaxx at jaxx.org
Sun Dec 11 21:07:52 HKT 2016


Sorry Michael,

Haven't really had the time yet (medical emergency last weekend, on call
with work this weekend... the Telco industry never sleeps :-) )

(The hiccup happened again on thursday, phone number became "+3365" instead
of "+33651886877")

I'll be jobless after december 13th (no problem, got a year and a half of
paid vacation to get paid ;-) ) so I'll have a little bit of time to flash
the code in before visiting family.

JaXX./.


On Sun, Dec 11, 2016 at 8:57 AM, Michael Balzer <dexter at expeedo.de> wrote:

> Got the buffer overrun effect yesterday with the new release.
>
> As I now...
>
> - Keep volatile features over reset
>
>
> ...it now also shows up with random data in features 0-7. Annoying.
>
> Julien, Pierre, any crash debug data collected yet?
>
> Regards,
> Michael
>
>
>
> Am 29.11.2016 um 22:13 schrieb Julien Banchet:
>
> Thank you Michael,
> Will flash it this weekend!
>
> MfG,
> JaXX./.
>
> On Tue, Nov 29, 2016 at 10:10 PM, Michael Balzer <dexter at expeedo.de>
> wrote:
>
>> Julien, Pierre,
>>
>> here's a firmware build with crash debug enabled:
>> https://dexters-web.de/f/tw-beta/OVMS-Twizy-3.8.2-crashdebug.zip
>>
>> It excludes the diag mode (serial interface command shell) but is
>> otherwise complete.
>>
>> The crash debug logs to the server (when it is/becomes available),
>> historical record type "*-OVM-DebugCrash". These records expire after 30
>> days, so you can collect some days.
>>
>> The crash count and data from the most recent crash will also be shown by
>> the SMS/text command "DIAG".
>>
>> The crash records contain a crash count, crash reasons (flags from RCON)
>> and the last checkpoint seen before the crash.
>>
>> Regards,
>> Michael
>>
>>
>>
>> Am 28.11.2016 um 22:37 schrieb Julien Banchet:
>>
>> Sure Michael,
>>
>> I've never got around compiling it right, and I'm out of a Windows since,
>> wow, I don't even know if grub would boot the partition anymore... If you
>> have it under hand (i hope), I'd be happy to give it a shot, it would also
>> be the occasion for me next weekend to pull a ribbon cable out once and for
>> all (or find a clever way to keep it hidden though programmable)
>>
>> How does crash recording work ? over Diag ? (PS: 50M/month data if needed)
>>
>> Julien
>> ./.
>>
>> On Mon, Nov 28, 2016 at 10:16 PM, Michael Balzer <dexter at expeedo.de>
>> wrote:
>>
>>> There's no memory management in the module, there's too few RAM
>>> available to do dynamic allocations.
>>>
>>> All buffers are static, the error can be some pointer corruption or
>>> missing length check / end of string check.
>>>
>>> A stack overflow could also be possible, the software stack is 256 bytes
>>> (1 page). As there's not enough ROM left to support the crash debug
>>> framework anymore in the Twizy build, I cannot tell if that happens again.
>>>
>>> Julien, you could try a reduced build with crash debug recording
>>> enabled...
>>>
>>> Regards,
>>> Michael
>>>
>>>
>>>
>>> Am 28.11.2016 um 22:00 schrieb Greg D.:
>>>
>>> Ah!  That sounds much more likely.
>>>
>>> Is there an error (timeout) path that might be freeing the buffer when
>>> it's still in use?
>>>
>>> Greg.
>>>
>>>
>>> Michael Balzer wrote:
>>>
>>> Hi Greg, welcome :)
>>>
>>> Juliens parameter #0 has been overwritten with "Topping o", which is
>>> part of the charge status message ("Topping off").
>>>
>>> So it's much more likely we've got a buffer corruption bug that's
>>> triggered by (maybe) bad connectivity situations.
>>>
>>> Regards,
>>> Michael
>>>
>>>
>>> Am 28.11.2016 um 06:38 schrieb Greg D.:
>>>
>>> Hi all,
>>>
>>> New here, so this may be totally off the wall.
>>>
>>> I notice that the first example, with a "5" going to a "k", it may be
>>> more easily explained as a bit *shift* instead of a set of flips.  Is
>>> there anywhere in the system where the bytes are clocked serially between
>>> devices, where there may be some marginal timing?
>>>
>>> Just a thought.
>>>
>>> Greg.
>>>
>>>
>>> Mark Webb-Johnson wrote:
>>>
>>> $ perl -e 'printf "%08b\n",ord("5");'
>>> 00110101
>>> $ perl -e 'printf "%08b\n",ord("k");'
>>> 01101011
>>> $ perl -e 'printf "%08b\n",ord("8");'
>>> 00111000
>>> $ perl -e 'printf "%08b\n",ord("Z");'
>>> 01011010
>>>
>>> Strange. Seems more than a simple bit flip. I haven’t seen this myself, or heard of it much from other users.
>>>
>>> Is the Twizy code using EEPROM for anything other than settings storage? I know that there are a very limit number of write cycles into that EEPROM space.
>>>
>>> I did some googling, but all I can find is corruption during writing (either power down, or interrupt during write). But, for our case of very very rare EEPROM writes, that shouldn’t be an issue.
>>>
>>> Regards, Mark.
>>>
>>>
>>> On 25 Nov 2016, at 5:47 AM, Michael Balzer <dexter at expeedo.de> <dexter at expeedo.de> wrote:
>>>
>>> The EEPROM data loss is a known issue, and I'm afraid there's nothing we can do about that. It seems to be a bug in the PIC18.
>>>
>>> Regards
>>> Michael
>>>
>>> Am 24.11.2016 um 22:10 schrieb Julien Banchet:
>>>
>>> Hi all,
>>>
>>> Thanks to Michaels last message to remind me to ask a question here:
>>>
>>> Since a few months, maybe a bit more because it happens only once in a while, I have noticed I wasn't receiving the SMS alerts (feature on SMSIP mode) , and sending a "Stats?" only gave me the nasty access denied answer.
>>>
>>> It turns out, using the application, I noticed that the phone number was getting corrupted (afaik, it's the only setting affected)
>>>
>>> Normally: "+33651886877" it became "+336k1886877" or "+33651886Z77" etc... like a solar flare bit flip !_______________________________________________
>>>
>>> _______________________________________________
>>> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/mailman/listinfo/ovmsdev
>>>
>>> _______________________________________________
>>> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/mailman/listinfo/ovmsdev
>>>
>>> --
>>> Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
>>> Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
>>>
>>> _______________________________________________
>>> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/mailman/listinfo/ovmsdev
>>>
>>> _______________________________________________
>>> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/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.teslaclub.hk http://lists.teslaclub.hk/mail
>>> man/listinfo/ovmsdev
>>
>> _______________________________________________
>> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/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.teslaclub.hk http://lists.teslaclub.hk/mail
>> man/listinfo/ovmsdev
>
> _______________________________________________
> OvmsDev mailing listOvmsDev at lists.teslaclub.hkhttp://lists.teslaclub.hk/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.teslaclub.hk
> http://lists.teslaclub.hk/mailman/listinfo/ovmsdev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openvehicles.com/pipermail/ovmsdev/attachments/20161211/35047229/attachment.htm>


More information about the OvmsDev mailing list