Dear List,ÿ
I'm using some post-up commands for additinal ip routes within /etc/network/interfaces. For many years and several debian versions this was fine. Like for example with (sanetized version):
iface eth0 inet6 staticÿ
ÿÿaddress aaaa:bbbb:10:60:11::4ÿ
ÿÿnetmask 80ÿ
ÿÿgateway aaaa:bbbb:10:60:11::2
ÿ ÿpost-up ip -6 route add aaaa:bbbb:10:60:105::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3
With trixie this does no longer work. It seems the ip address gets valid several seconds AFTER the post-up script gets run. The logs show some error: ?Error: Invalid source address?
Does anybody know the reason for this?
regards
hede
ÿ
Some more information. The following commands get run by ifup:
ip addr flush dev eth0 mngtmpaddrÿ
ip link set dev eth0 ÿupÿ
ip -6 addr add aaaa:bbbb:10:60:11::4/80 ÿdev eth0 ÿ
ip -6 route replace default via aaaa:bbbb:10:60:11::2 ÿdev eth0 onlink ÿ
ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3
The following shows that it needs about 2 seconds for the address to get valid / the route gets applied only after two one-seconds sleeps:
root@system:~# ip addr flush dev eth0 mngtmpaddr && ip link set dev eth0 ÿup && ip -6 addr add aaaa:bbbb:10:60:11::4/80 ÿdev eth0 ÿ&& ip -6 route replace default via aaaa:bbbb:10:60:11::2 ÿdev eth0 onlink ÿ&& echo iface is up; echo first: ; ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3 ; sleep 1; echo second: ; ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3 ; sleep 1; echo third: ; ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3 ; sleep 1; echo forth: ; ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3 ; sleep 1; echo fifth: ; ip -6 route add aaaa:bbbb:10:60:13::/80 dev eth0 src aaaa:bbbb:10:60:11::4 via aaaa:bbbb:10:60:11::3 ;ÿ
iface is up
first:
Error: Invalid source address.
second:
Error: Invalid source address.
third:
forth:
RTNETLINK answers: File exists
fifth:
RTNETLINK answers: File exists
and btw: I do have my solution now, simply by running a postup-script some seconds after ifup finishes. But I'm still interested why this happens.ÿ


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Hi,

On Fri, Jan 09, 2026 at 08:50:31PM +0100, hede wrote:

I do have my solution now, simply by running a postup-script some
seconds after ifup finishes. But I'm still interested why this
happens.ÿ

Could it be waiting for duplicate address detection (DAD)? You could try
with:

iface eth0 inet6 staticÿ
ÿÿ address aaaa:bbbb:10:60:11::4/80
ÿÿ gateway aaaa:bbbb:10:60:11::2
dad-attempts 0
?

to see if that helps. Obviously you will then have to be careful not to
use this address elsewhere as well.

Failing that, you could improve the post-up sleep a bit by calling a
script that:

1. reads the list of IPs for that interface

2. If source address found, add the route

3. sleep for a second

4. go to (1), probably check if already waited for too long and abort if
so.

Thanks,
Andy

--
https://bitfolk.com/ -- No-nonsense VPS hosting

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Hi Andy
Am 09.01.26 um 21:38 schrieb Andy Smith:

...
Could it be waiting for duplicate address detection (DAD)? You could try with:

iface eth0 inet6 static
?? address aaaa:bbbb:10:60:11::4/80
?? gateway aaaa:bbbb:10:60:11::2
dad-attempts 0
\u2026

That's it. Thank you. Quite easy to understand.
Yet I am even more interested in why that was no problem with previous
Debian versions. And indeed as it's a system within a network for me
/dad/ will stay enabled - obviously ;-).
regards
hede


--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)

Could it be waiting for duplicate address detection (DAD)? You could try
with:

iface eth0 inet6 static
ÿÿ address aaaa:bbbb:10:60:11::4/80
ÿÿ gateway aaaa:bbbb:10:60:11::2
dad-attempts 0
?

That's it. Thank you. Quite easy to understand.

Yet I am even more interested in why that was no problem with previous
Debian versions.

I suggest `reportbug`: It seems like the `post-up` script is run
a bit too early.


Stefan

--- PyGate Linux v1.5.2
* Origin: Dragon's Lair, PyGate NNTP<>Fido Gate (3:633/10)