linuxptp-users

[Linuxptp-users] ptp4l over ethernet failed

[Raymond G. <ray...@ya...>]

Hi,

We are trying to run ptp4l (linuxptp-1.3) over ethernet (the -2 option).

On the slave side, we use the command:

"./ptp4l -H -2 -H -i eth0 -s -l 7 -m -q -p /dev/ptp0".

However because of some "bad message" errors (see below),
the slave side is not working.

ptp4l[63896.503]: PI servo: sync interval 1.000 kp 0.100 ki 0.001000
ptp4l[63896.503]: port 1: get_ts_info not supported
ptp4l[63896.600]: port 1: INITIALIZING to LISTENING on INITIALIZE
ptp4l[63896.600]: port 0: INITIALIZING to LISTENING on INITIALIZE
ptp4l[63897.324]: port 1: setting asCapable
ptp4l[63897.324]: port 1: new foreign master 080028.fffe.329976-1
ptp4l[63897.526]: port 1: bad message

We notice that the "bad message" is caused by the slave side receiving
frames that includes the FCS.

This FCS is treated as TLV, for example in a follow_up message from
host side, in the suffix processing function suffix_post_recv.

We experimented by letting ptp4l ignore the last 4 bytes of FCS in
the messages received, then it works fine.

I cannot find any reference that says the linux kernel or the hardware or
the ethernet driver is not supposed to pass up the FCS to apps when
communicating over raw sockets.

Is this an issue lying in the kernel, ethernet driver or hardware which is
not supposed to deliver the FCS?

Otherwise is ptp4l supposed to properly handle the FCS (such as ignoring
the last 4 bytes), or setting a socket option (if one exists) to kernel to
ask it to discard the FCS before delivering to user space?

Any suggestion will be appreciated.

Thanks
Raymond

Re: [Linuxptp-users] ptp4l over ethernet failed

[Richard C. <ric...@gm...>]

On Sun, Oct 27, 2013 at 06:24:53AM -0700, Raymond Guo wrote:

> I cannot find any reference that says the linux kernel or the hardware or
> the ethernet driver is not supposed to pass up the FCS to apps when
> communicating over raw sockets.

I found this:

http://wiki.wireshark.org/Ethernet

   As the Ethernet hardware filters the preamble, it is not given to
   Wireshark or any other application. Most Ethernet interfaces also
   either don't supply the FCS to Wireshark or other applications, or
   aren't configured by their driver to do so; therefore, Wireshark
   will typically only be given the green fields, although on some
   platforms, with some interfaces, the FCS will be supplied on
   incoming packets.
 
> Is this an issue lying in the kernel, ethernet driver or hardware which is
> not supposed to deliver the FCS?

I would bet this is a driver issue. What kernel and driver are you using?
 
> Otherwise is ptp4l supposed to properly handle the FCS (such as ignoring
> the last 4 bytes), or setting a socket option (if one exists) to kernel to
> ask it to discard the FCS before delivering to user space?

AFAICT, there is no such socket option.

Sorry,
Richard