Re: Losing CAN messages with socket-CAN

[Wolfgang Grandegger <wg@grandegger.com>]

On 04/04/2013 04:44 PM, j.p.lammertink wrote:
> Hi Marc,
> 
> Thanks for the fast response. I've collected answers to your questions and 
> pasted them below.
> 
> Best Regards,
> 
> Jeroen Lammertink
> 
> --------------------------------------------------------------------------
> 
> <> We are sending batches of 80 CAN messages using socket CAN to transfer an
> <> image for download. After the 80 messages we wait for a response
> 
> <Are you evaluating the return value of the write() or sendmsg() systemcall?
> 
> We do evaluate the write call to the socket with an assert. I’ve included
> the code below. (See CCanPort::Write()). I’ve also included the code in
> which the socket is created, just to be complete. When the problem (of
> losing CAN messages) happens, the Assert is not triggered. So I deduce
> that the socket remains unaware of the losing of messages.
> 
> <> consisting of a single CAN message. Within the messages we have coded a
> <> sequence number, so that we can detect lost messages. Besides the CAN
> <> device that is subject of download, we have also a (hardware) CAN sniffer
> <> on the CAN bus. We typically log CAN messages when we put them in the
> <> socket. With this set-up we typically detect losing a short series of
> <> (mostly 5) CAN messages, somewhere halfway the batch of 80. No other
> <> devices are present on the CAN bus.
> 
> <> I've dumped below some version and set-up info of Ubuntu, socket-can and
> <> ifconfig.
> 
> <> Is this problem known?
> <> May this problem be related to using an old version of socket can?
> 
> < This is either a userspace problem or a buggy driver. The driver problem
> < may have been fixed in the meantime.
> 
> <> How can I upgrade Ubuntu to the latest versions of socket-can?
> 
> < First we have to figure out what's broken.
> 
> <> --------------------------------------------------------------------------
> <> bmterra@q7buntu:~/user/jla$ uname -a
> <> Linux q7buntu 3.5.0-26-generic #42~precise1-Ubuntu SMP Mon Mar 11 22:19:42
> <> UTC 2013 i686 i686 i386 GNU/Linux
> <> bmterra@q7buntu:~/user/jla$ cat /proc/net/can/version
> <> rev 20090105 abi 8
> <> bmterra@q7buntu:~/user/jla$ ifconfig can0
> <> can0      Link encap:UNSPEC  HWaddr
> <> 00-00-00-00-00-00-00-00-00-00-00-00-00-00-00-00
> <>           UP RUNNING NOARP  MTU:16  Metric:1
> <>           RX packets:1128 errors:0 dropped:0 overruns:0 frame:0
> <>           TX packets:4400 errors:0 dropped:0 overruns:0 carrier:0
> <>           collisions:0 txqueuelen:1024
> <>           RX bytes:9024 (9.0 KB)  TX bytes:34176 (34.1 KB)
> <>           Interrupt:18
> <> --------------------------------------------------------------------------
> 
> < Which CAN hardware are you using?
> 
> Qseven module PQ7-M105IT
> with Intel Platform Controller HUB EG20T IOH
> with a “BOSCH CAN Protocol Version 2.0B Active 
>     (ISO 11519, ISO 11898, and SAEJ2411)” CAN controller

Which driver do you use? (check "ip -d -s link show can0").

> This CAN controller identifies itself (according to data sheet) as follows:
>     Vendor Identification Register    8086h
>     Device Identification Register    8818h
>     Revision Identification Register  00h
>     Class Code Register               0C0900h

Ah, the "pch_can" driver does have known issues. End of last year I sent
some patches to improve the "pch_can" or, even better, make the "c_can"
driver work on the EG20T as well. See:

  http://marc.info/?l=linux-can&m=135531975013055&w=2

Are you able to build and use a recent mainline kernel? I need to figure
out how to help you best.

Wolfgang.