Page 1 of 1

Ethernet port slow to be up

Posted: 2020/06/24 05:17:08
by filovitch
Hi,

I have proceeded to a new installation of Centos 8 on a ThinkPad M720q. I was thinking the boot was quite slow for this hardware. When I checked the dmesg I seen that the ethernet card took 12 seconds to be up. I comparaison, this same card, on the same hardware, take 5 seconds to be up on a debian 10.4 (see dmesg log in attachment).

Anyone have a idea where is the problem?

Centos:

Code: Select all

[    2.486567] e1000e 0000:00:1f.6 eno1: renamed from eth0
[    6.178365] IPv6: ADDRCONF(NETDEV_UP): eno1: link is not ready
[    6.360210] IPv6: ADDRCONF(NETDEV_UP): eno1: link is not ready
[    6.364497] IPv6: ADDRCONF(NETDEV_UP): eno1: link is not ready
[   12.736632] e1000e: eno1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None
[   12.736711] IPv6: ADDRCONF(NETDEV_CHANGE): eno1: link becomes ready
Debian:

Code: Select all

[    1.773458] e1000e 0000:00:1f.6 eno1: renamed from eth0
[    2.878061] IPv6: ADDRCONF(NETDEV_UP): eno1: link is not ready
[    5.684948] e1000e: eno1 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: None
[    5.685057] IPv6: ADDRCONF(NETDEV_CHANGE): eno1: link becomes ready
Centos complete dmesg: https://pastebin.com/XJjBUSSa
debian complete dmesg: https://pastebin.com/7iW3qQ1w

Re: Ethernet port slow to be up

Posted: 2020/06/26 22:43:04
by aks
I'd start here:
" If the kernel is appropriately configured, lockdown may be lifted by typing the appropriate sequence on a directly attached physical keyboard. For x86 machines, this is SysRq+x."
(it's from the man page mentioned).
https://en.wikipedia.org/wiki/Magic_SysRq_key

Re: Ethernet port slow to be up

Posted: 2020/07/07 19:35:30
by filovitch
Unfortunately, I have the same problem even if I disable secure boot
https://paste.centos.org/view/a5b06991
I have done a lot of research and really do not know how to debug the problem.
The difference between debian and centos is the kernel version.

Re: Ethernet port slow to be up

Posted: 2020/07/07 19:42:54
by TrevorH
The difference in times is not as great as you say. You're measuring them from different points. The CentOS system doesn't start to look at bringing the network interface up until 6.178365 and takes until 12.736632, so roughly 6.5s. The Debian system starts at 2.878061 and is done by 5.684948 so about 3s.

If it's slow enough to notice then I doubt if 3s difference is enough to explain it. Try running systemd-analyze blame and see what that reports.