Re: [PATCH RESEND] net: can: Introduce MEN 16Z192-00 CAN controller driver

[Oliver Hartkopp]

On 08/11/2016 10:58 AM, Andreas Werner wrote:
> On Thu, Aug 11, 2016 at 10:45:00AM +0200, Oliver Hartkopp wrote:


> > When you still have the possibility to change the IP core I would suggest to
> > create some kind of 16/32 bit value which you can pass to the CAN controller
> > along with the CAN frame to be sent.
> >
> > And when this frame comes back due to the loopback you can use this non-zero
> > 16/32 bit value to match into a list of tx skb pointers for IFF_ECHO.
> >
> > E.g. when this 16/32 bit value is zero this CAN frame obviously was received
> > from another CAN node.
> >
> > Just an idea.
>
> I am not sure if we have a way to change the IP but i will try to talk with
> my IC designer. He will be available next week.
>
> Your idea sounds good. I will check a few more driver to get more information
> how they did the implementation.

I just looked into your patch at 
http://marc.info/?l=linux-can&m=146952497113100&w=2

The

	struct men_z192_cf_buf {
		u32 can_id;
		u32 data[2];
		u32 length;
	};

has a u32 for the length which is masked by

	#define MEN_Z192_CFBUF_LEN	GENMASK(3, 0)

in men_z192_read_frame() and is just copied in men_z192_xmit()

	writel(cf->can_dlc, &cf_buf->length);

as only 4 bits are used in the u32 length you probably already can use 
the upper 16 bits for the discussed IFF_ECHO purpose.

Don't know how your IP core handles this u32 length when you enable the 
loopback - maybe the upper 16 bits are still there in the receive path 
and you can implement this idea directly :-)

Regards,
Oliver