Re: [Linuxptp-users] Wrong source IP Address in Multicast

[gntech h. <gt...@gm...>]

Any ideas as to why the patch previously posted was necessary to resolve
the incorrect source IP address in multicast packets?  Is this more
appropriate for the development list?  I posted here based on
https://lists.nwtime.org/sympa/info/linuxptp-users.

Thank you,
gthweng

On Mon, Dec 4, 2023 at 3:08 PM geontech hweng <gt...@gm...> wrote:

> Hi everyone,
>
> I am standing up PTP between an Ubuntu 20.04 system (server) and the 1GbE
> PS interface of an AMD Ultrascale+ MPSoC board (client).  The client has
> PTP enabled as part of the Petalinux build v3.1.1.  On the Ubuntu host,
> v3.1.1 was built from source and installed because v1.9.2 that is packaged
> for 20.04 has published vulnerabilities.  Tests were performed with the
> systems connected using a small desktop switch and connected directly.  In
> both cases, the interfaces are configured for static IP addresses and the
> interfaces of both systems have multicast enabled.  The client never prints
> anything beyond the following.
>
> ```bash
> sudo ptp4l -i eth0 --slaveOnly=1 -m --tx_timestamp_timeout=10
> Password:
> ptp4l[6063.829]: selected /dev/ptp0 as PTP clock
> ptp4l[6063.830]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
> ptp4l[6063.830]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
> ptp4l[6066.310]: port 1: new foreign master f4939f.fffe.f45b6d-1
> ptp4l[6070.310]: selected best master clock f4939f.fffe.f45b6d
> ```
> The command used on the Ubuntu server is 'sudo ptp4l -i eno1
> --masterOnly=1 -m --logSyncInterval=-3 --tx_timestamp_timeout=10"
>
> Using wireshark and tcpdump, we have ascertained that the source IP
> address for the multicast packets from the server contains the IP address
> for another interface of the Ubuntu system.  The source IP address is
> incorrect.  During the investigation, we discovered a post on this list is
> 2017 that described the same issue and included a patch that fixed the
> issue.  The post is titled "Wrong Interface" and was posted on 017-06-22
> 20:59:12.  The links to the original message and patch are below.
>
> https://sourceforge.net/p/linuxptp/mailman/message/35907900/
>
> https://sourceforge.net/p/linuxptp/mailman/attachment/CAP6mCQSva38taSYXvXZaW2JoQ2YzZxfk3scM8QfYMxDFjqg68g%40mail.gmail.com/1/
>
> Application of the changes in the patch corrected the issue for v3.1.1 on
> the Ubuntu system.  The client now converges to sub-100ns rms offsets with
> HWTS enabled on both systems.  A comparison of udp.c for v3.1.1 and v4.1
> indicates that nothing was changed to address this behavior in the most
> recent release.  Do you consider this an issue?  If yes, will it be
> corrected?  If not, how do you recommend that users mitigate this issue?
>
> Thank you!
>
> gthweng
>
>