On 03/03/18 13:59, Claudio Jeker wrote:
On Fri, Mar 02, 2018 at 07:22:05AM +0100, Claudio Jeker wrote:
On Thu, Mar 01, 2018 at 07:18:29PM +0100, Andreas Bartelt wrote:
On 03/01/18 09:07, Jiri B wrote:
On Thu, Mar 01, 2018 at 08:37:38AM +0100, Andreas Bartelt wrote:
Please let me know if you need further infos on my specific config.
I use this (OpenBSD 6.2-current (GENERIC.MP) #6: Tue Feb 13 20:16:11 MST 2018):
/etc/hostname.vether20
rdomain 20
inet 172.16.20.2 255.255.255.0 172.16.20.255 description "archive rdomain"
group onion
!ifconfig lo20 inet 127.0.0.1/8
!route -T20 -qn add -net 127 127.0.0.1 -reject
!route -T20 -qn add default 172.16.20.1
thanks -- your variant for setting an IP address for lo(4) in the underlying
interface's rdomain is less evil than delaying the ifconfig statement until
/etc/rc.local is run. Your variant also still works on current.
Nevertheless, I suppose there was nothing inherently wrong with my previous
approach to configuring an IP address for loX via /etc/hostname.loX which is
now broken on current. Or is there a good reason I'm not aware of that this
variant doesn't work anymore?
I will revert the changes to automatically configure the lo interfaces in
-current. That should get us back to where it was before. Still I think
rdomains should automatically have a 127.0.0.1 and ::1 IP configured just
need to figure out how to do this properly.
Atfer some more thinking I think this has nothing todo with my reverted
diff. The problem is actually the new netstart and the way interfaces are
now configured. Before hostname.lo1 was run (and created) after em0 was
configured which was adding rdomain 1. Now lo(4) is considered a cloning
interface and ifconfig lo1 create is run very early on. Because of
this creation of rdomain 1 fails and so the config fails.
I can confirm that the /etc/hostname.lo1 config variant from my original
mail also doesn't work with yesterday's build which included your
reverted patch with regard to 127.0.0.1 autocreate. However, behavior
seems to have slightly changed since rdomain 1 is actually created now
and lo1 also is inside rdomain 1. However, the underlying interface is
not inside rdomain 1 (although it's hostname.if(5) contains a rdomain 1
statement).
