[systemd-devel] systemd-networkd, IPv6PrivacyExtensions=kernel, sysctl and devicenames

[Ede Wolf]

> 
> Given lack of errors after interface rename, settings were most probably
> applied correctly.

No. According to the log, the lack of errors after the rename result 
simlpy in there are no more settings left that could be applied. Because 
they all have been tried before. And failed. sysctl.conf is most likely 
only read once during boot. And in this case that happens before the 
rename.


>>
> 
> tw:~ # systemctl stop NetworkManager.service
> tw:~ # ip l set dev enp0s5 down
> tw:~ # cat /proc/sys/net/ipv6/conf/enp0s5/use_tempaddr
> 1
> tw:~ # cat /proc/sys/net/ipv6/conf/enp0s5/addr_gen_mode
> 1
> tw:~ # echo 3 > /proc/sys/net/ipv6/conf/enp0s5/addr_gen_mode
> tw:~ # echo 2 > /proc/sys/net/ipv6/conf/enp0s5/use_tempaddr
> tw:~ # cat /proc/sys/net/ipv6/conf/enp0s5/use_tempaddr
> 2
> tw:~ # cat /proc/sys/net/ipv6/conf/enp0s5/addr_gen_mode
> 3
> tw:~ # ip l set dev enp0s5 name ififif
> tw:~ # ip l set dev ififif up
> tw:~ # cat /proc/sys/net/ipv6/conf/ififif/use_tempaddr
> 2
> tw:~ # cat /proc/sys/net/ipv6/conf/ififif/addr_gen_mode
> 3
> 

Thanks for taking the time to demonstrate this. But it does not comapre, 
as you do not remane the interface. It is alway enp0s5, and therefore 
the kernel is keeping the /proc/sys/net/ipv6/conf/enp0s5/ settings.

You you would need to apply the sysctl to enp0s5, then rename it to 
lan01, then the /proc/sys/net/ipv6/conf/enp0s5/... hierachy would be 
gone and instead you would have a
/proc/sys/net/ipv6/conf/lan01/ hierachie, that would NOT inherit the 
settings you've applied to enp0s5.
 From a kernel perspective point of view these two are not related at 
all, they are independent interfaces, that just appear and go. At least, 
that's what it seems like.

And that is what is happening here on boot.

systemd-sysctl applies the eth0 setting from sysctl.conf to the 
interface. It failes to to do the same for ens3, as this interface does 
not exist yet. As it is still named eth0

Then systemd renames eth0 to ens3. The eth0 settings are gone, the ones 
for ens3 are not applied any more, as systemd-sysctl has already 
processed sysctl.conf. Therefore the default values are used.

As written in a follow up, the ordering seems to be different on deepin 
linux using systemd 241 instead of arch and systemd 245.

However, even with deepin, if you rename an interface with an udev rule 
or .link file, you run into the same problems again. Just, that there 
are now three renames and sysctl is read after the second, but before 
the third rename. Same result however.

> 
> What likely happens in your case is something else changes settings
> after they had been applied by udev. Notice I had to stop NetworkManager
> to avoid interfering as soon as link comes up.

I do not use any other network manager but systemd-networkd. In fact, I 
do not even have any installed. netctl, the default of arch, has been 
removed straight after install.

And, in addition, journal is pretty clear about systemd-networkd doing 
the name change:

Mai 23 10:41:27 systemd-networkd[464]: ens3: Interface name change 
detected, ens3 has been renamed to eth0.
Mai 23 10:41:27 systemd-networkd[464]: eth0: Interface name change 
detected, eth0 has been renamed to ens3.


But, in case this is still true, what would be the best way to figure out?

How can we hunt down this issue?