On Mon, Feb 08, 2010 at 10:12:12AM +0100, Wolfgang Grandegger wrote: > Kurt Van Dijck wrote: > > On Sun, Feb 07, 2010 at 06:25:39PM +0100, Wolfgang Grandegger wrote: > >> christian pellegrin wrote: > >>> On Sun, Feb 7, 2010 at 6:05 PM, Wolfgang Grandegger <[email protected]> > >>> wrote: > >>>> christian pellegrin wrote: > >>>> increase). Can you confirm that? Normal state changes are interrupt > >>>> driver. So, if the hardware does not trigger an interrupt, we have a > >>>> problem. > >>> On the mcp251x we get an interrupt when we get back from error-warning > >>> to error-active but I don't know if we have to send some kind of error > >>> frame in this case. Now nothing is sent, I was worried if this is > >>> right. > >> Good question. I think we should send an error message for any state > >> change also for passive->warning->active, which we currently do not > >> handle by software. We speak about controller *problems* and there is > >> currently no CAN_ERR_CRTL_ACTIVE. Well, that's another weak point :-(. > >> This needs some more thoughts/discussion. > > We had some discussion lately, but having an interrupt (and some message > > to userspace) seems like a non-optional requirement to me. > > I don't understand. What is "non-optional"? Currently we just report > controller *problems*, meaning state changes > active->warning->passive->bus-off. sorry for the confusion. I meant: having an 'bus state change' interrupt seems a requirement to me (therefore : non-optional, but it depends on CAN chip capabilities). This (bus-state-change interrupt) is, IMO, more important than the bus error interrupts.
Therefore, I think that it may not be very good not sending anything on the 'error-warning to error-active interrupt', as was mentioned above. Implementing bus-error interrupts, but poor bus state change interrupts requires the user to enable bus-error interrupts, and tracking the bus state itself. I do not like that (But I did no mcp251x work either). I think we agree, but I didn't express my opinion very clear :-) > > Wolfgang. Kurt _______________________________________________ Socketcan-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/socketcan-users
