On 09/07/2009 11:16 AM, Sebastian Haas wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi,
>
> Wolfgang Grandegger schrieb:
>> Hi Vladislav,
>>
>> On 09/07/2009 10:31 AM, [email protected] wrote:
>>> Hi,
>>>
>>> my interpretation of ISO 11898-1 is that it is indeed allowed
>>> to recover from bus off to error-active after 128*11 recessive
>>> bits AND "user request", where "user request" is SW or HW
>>> request.
>>> So the HW (CAN controller) is allowed to implement auto recovery
>>> feature (may even make auto recovery to the permanent and the
>>> only behavior).
>>>
>>>>  From the CAN point of view the bus off is a protection mechanism,
>>> the CAN controller goes to bus off because something is seriously
>>> wrong with itself, wires or other nodes. So it is very often not
>>> really good idea to make automatic recovery which is with other
>>> words simply ignoring that serious problems (well, the reception
>>> of 128*11 recessive bits is a good sign that the problem has gone,
>>> but still..). The application may want the CAN controller to
>>> signal the bus off and stay off the bus until the application
>>> decides to recover the communication.
>>>
>>> So if the CAN controller only supports auto recovery and cannot
>>> keep hands off from the bus, the option for the manual recovery
>>> in the socket CAN driver is to reset CAN controller in the bus off
>>> interrupt (I remember this was proposes by Sebastian, and I
>>> agree on this).
>>
>> Yes, the question is if we should support auto-recovery at all, e.g. as
>> proposed in
>> https://lists.berlios.de/pipermail/socketcan-core/2009-September/002971.html
>>
>> I tend to disallow it, also to have a common behavior of auto and manual
>> recovery.
> Why not supporting it and let the user choose to enable or disable it.
> But by default it should be disabled, IMHO.

I would like to avoid another configuration option if possible. Because 
firstly to keep the code simply and secondly to not confuse the user 
even more.

Wolfgang.
_______________________________________________
Socketcan-core mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-core

Reply via email to