Hi,
This fix introduced a regression with ThunderX nodes (CRB-1S, CRB-2S) and our 10G switch (Extreme Networks x670 10GE L3).
We have opened a downstream bug report [1], where we temporarily bypassed this by pinning the kernel to 4.4.0-45.
I also tested 4.8 (multiple builds), 4.10 and 4.11-rc1 (vanilla); all are still affected by link training issues with our switch, with 4.11-rc1 not working at all and reporting more issues (logs attached in a different LP comment [2]).
Hi,
This fix introduced a regression with ThunderX nodes (CRB-1S, CRB-2S) and our 10G switch (Extreme Networks x670 10GE L3).
We have opened a downstream bug report [1], where we temporarily bypassed this by pinning the kernel to 4.4.0-45.
I also tested 4.8 (multiple builds), 4.10 and 4.11-rc1 (vanilla); all are still affected by link training issues with our switch, with 4.11-rc1 not working at all and reporting more issues (logs attached in a different LP comment [2]).
BR,
Alex
[1] https:/ /jira.opnfv. org/browse/ ARMBAND- 168 /bugs.launchpad .net/ubuntu/ +source/ linux/+ bug/1672521/ comments/ 17
[2] https:/