can-j1939: semantics question

can-j1939: semantics question

[Kurt Van Dijck]

Hey,

I eventually got into investigating the necessary changes to drop
iproute2 from can-j1939.
We somewhat agreed that opening a can-j1939 socket and binding
to a CAN interface should activate can-j1939 processing for that
interface.
But how about deactivation?
Should the last can-j1939 socket on a CAN iface deactivate can-j1939 for
that iface when that socket closes?

Kind regards,
Kurt

Re: can-j1939: semantics question

[Oliver Hartkopp]

On 11.11.2015 20:52, Kurt Van Dijck wrote:

> I eventually got into investigating the necessary changes to drop
> iproute2 from can-j1939.
> We somewhat agreed that opening a can-j1939 socket and binding
> to a CAN interface should activate can-j1939 processing for that
> interface.
> But how about deactivation?
> Should the last can-j1939 socket on a CAN iface deactivate can-j1939 for
> that iface when that socket closes?

At least this sounds like some kind of natural behaviour.

What would be the other option?

Would you still monitor address claiming inside the kernel without doing 
anything else?

Or is the address claiming in user space the last application that closes the 
last socket anyway??

Regards,
Oliver

Re: can-j1939: semantics question

[Kurt Van Dijck]

> On 11.11.2015 20:52, Kurt Van Dijck wrote:
> 
> >I eventually got into investigating the necessary changes to drop
> >iproute2 from can-j1939.
> >We somewhat agreed that opening a can-j1939 socket and binding
> >to a CAN interface should activate can-j1939 processing for that
> >interface.
> >But how about deactivation?
> >Should the last can-j1939 socket on a CAN iface deactivate can-j1939 for
> >that iface when that socket closes?
> 
> At least this sounds like some kind of natural behaviour.
> 
> What would be the other option?

A special tool, intentionally not called 'ip' :-), that turns off
j1939 processing. Looks bad.

I'll stick to the 'natural behaviour'.

> 
> Would you still monitor address claiming inside the kernel without doing
> anything else?

no. And noone needs it anyway.

> 
> Or is the address claiming in user space the last application that closes
> the last socket anyway??

indeed, address claiming requires an open socket.

Thanks,
Kurt