loopback/recv own messages option not working as expected with plx_pci (Adlink Technology PCI-7841)

loopback/recv own messages option not working as expected with plx_pci (Adlink Technology PCI-7841)

[Felix Obenhuber]

Hi,

I just encountered a problem with the loopback and receive own
messages option in combination with the plx_pci driver serving a
Adlink PCI-7841. It driver behaves a the CAN_RAW_RECV_OWN_MSGS and
LOOPBACK option would have been set (and they aren't).

One box with the Adlink card connected to another host with a approved
and well initialized CAN card. Sending single frames from each hosts
works fine. When I start the canecho test from trunk/test and send a
frame on  the bus (from the other node) its duplicated each time
canecho reads/writes the frame. You would get the same result if you
set the CAN_RAW_RECV_OWN_MSGS and LOOPBACK option on the socket in the
canecho test.
I have the same behavior with a tool written by me and tested with the
latest trunk revision - same result.
Running canecho on a vcan or peak usb device works as expected.

Has someone seen something similar with that driver? I'd just like to
ask before digging into it.

cheers,

Felix

Re: loopback/recv own messages option not working as expected with plx_pci (Adlink Technology PCI-7841)

[Oliver Hartkopp]

Hi Felix,

On 12.01.2012 14:53, Felix Obenhuber wrote:


> I just encountered a problem with the loopback and receive own
> messages option in combination with the plx_pci driver serving a
> Adlink PCI-7841. It driver behaves a the CAN_RAW_RECV_OWN_MSGS and
> LOOPBACK option would have been set (and they aren't).
> 
> One box with the Adlink card connected to another host with a approved
> and well initialized CAN card. Sending single frames from each hosts
> works fine. When I start the canecho test from trunk/test and send a
> frame on  the bus (from the other node) its duplicated each time
> canecho reads/writes the frame. You would get the same result if you
> set the CAN_RAW_RECV_OWN_MSGS and LOOPBACK option on the socket in the
> canecho test.
> I have the same behavior with a tool written by me and tested with the
> latest trunk revision - same result.
> Running canecho on a vcan or peak usb device works as expected.


Both vcan & the PEAK driver use a fallback solution for that - so it is not a
real statement when they work.

> 
> Has someone seen something similar with that driver? I'd just like to
> ask before digging into it.


No - i assume you are using a 2.6.30.x (0 <= x < 5) kernel, right?

See

http://old.nabble.com/Local-loopback-broken--td24108048.html

Regards,
Oliver

Re: loopback/recv own messages option not working as expected with plx_pci (Adlink Technology PCI-7841)

[Felix Obenhuber]

Hi Oliver,

On Thu, Jan 12, 2012 at 5:57 PM, Oliver Hartkopp <socketcan@hartkopp.net> wrote:
>> Has someone seen something similar with that driver? I'd just like to
>> ask before digging into it.
>
> No - i assume you are using a 2.6.30.x (0 <= x < 5) kernel, right?

No, it's a 2.6.38 packaged by Ubuntu - nothing special:

2.6.38-11-generic-pae #49-Ubuntu SMP

> See
>
> http://old.nabble.com/Local-loopback-broken--td24108048.html

Hm. Corresponding to your post this issues was fixed in 2.6.30.5, so
it should be included in my kernel for sure.

I have another machine here, equipped with the same card:

2.6.32-122-rtai #rtai SMP from
http://www.linuxcnc.org/emc2 lucid base emc2.4

Here everything works fine.

I'll try a different kernel on the box that is not working.

Thanks,

Felix

Re: loopback/recv own messages option not working as expected with plx_pci (Adlink Technology PCI-7841)

[Felix Obenhuber]

> I'll try a different kernel on the box that is not working.

I updated to the stock ubuntu kernel

3.0.0-15-generic #25-Ubuntu SMP

The loopback option works as expected.

Felix