NETDEV WATCHDOG: eth0（mlx5_core）: transmit queue timed out

Infrastructure & Networking Software And Drivers Mellanox OFED
1

hi all, we use openeuler 20.03 sp3 and centos 7, mlx5_core version is 5.0-0,some time show mlx5_core transmit queue timed out in /var/log/message ,and we upgrade mlx5_core drvier version to 5.8-2.0.3 ,still has the timeout problem.
any solution for how to resolve this? thanks you

2

What HCA model? Are you use IPoIB?

You can update to latest OFED and Firmware.

3

We do not use IPoIB,When the server use ip-in-ip, we discvoer anther problem, client can not connect to the server,and client’s TcpInCsumErrors increace in “nstat -az” command.when the server run “ethtool -k eth0 tx off”,then client can connect to server.
latest EN and OFED has this problem.
Is it a bug ?thank you

4

That is TX CHECKSUM offload, it should be off in your case.

5

In the kernel default driver 5.0-0,Do not need to turn it off, Update to new version, need to ,why is that? Will it fix in then next version?