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
_______________________________________________
Linuxptp-users mailing list
Linuxptp-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-users

Reply via email to