Bug #2611

Change in IP address results in network not working

Added by phma 10 months ago. Updated 10 months ago.

Status:NewStart date:12/03/2013
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

I'm on a satellite connection and am assigned a public IPv6 address and an IPv4 address in shared address space behind NAT. Occasionally the network goes down, and I get new addresses. This does not bother the Linux boxes (except that I have to edit /etc/hosts so that they know where each other are) but the DragonFly box is unable to reach any hosts except on the LAN, and that only on IPv6.

In IPv4, after the address change, a ping from 100.116.42.28 (Linux laptop) to 100.116.42.26 (DragonFly box) is answered by a pong from 100.116.27.42 (DFly's old address, which no longer appears in ifconfig) to 100.116.42.28.

In IPv6, after the address change, the default gateway begins with fe80 instead of 2001. The attached files are:
routebad: the routing table (netstat -rn) after a previous address change
routegood: the routing table after rebooting
routebadagain: the routing table after the most recent address change.
After rebooting, I had to find the gateway by asking the Linux box and set it manually; this may not be a bug, as normally zyxomma runs a tunnel, and the IPv6 gateway was pointing to an address in the tunnel.

routebad (3.37 KB) phma, 12/03/2013 01:14 PM

routegood (3.29 KB) phma, 12/03/2013 01:14 PM

routebadagain (3.21 KB) phma, 12/03/2013 01:14 PM

History

#1 Updated by sepherosa 10 months ago

On Wed, Dec 4, 2013 at 5:14 AM, wrote:

>
> Issue #2611 has been reported by phma.
>
> ----------------------------------------
> Bug #2611: Change in IP address results in network not working
> http://bugs.dragonflybsd.org/issues/2611
>
> * Author: phma
> * Status: New
> * Priority: Normal
> * Assignee:
> * Category:
> * Target version:
> ----------------------------------------
> I'm on a satellite connection and am assigned a public IPv6 address and an
> IPv4 address in shared address space behind NAT. Occasionally the network
> goes down, and I get new addresses. This does not bother the Linux boxes
> (except that I have to edit /etc/hosts so that they know where each other
> are) but the DragonFly box is unable to reach any hosts except on the LAN,
> and that only on IPv6.
>
> In IPv4, after the address change, a ping from 100.116.42.28 (Linux
> laptop) to 100.116.42.26 (DragonFly box) is answered by a pong from
> 100.116.27.42 (DFly's old address, which no longer appears in ifconfig) to
> 100.116.42.28.
>
>
For IPv4, if you manually change the interface address, the routes
"associated" w/ that address's prefix route will be gone, e.g.

ifconfig iface0 inet 10.0.0.2
route add -net 192.168.1.0/24 10.0.0.1
ifconfig iface0 inet 10.0.0.3

Then the route to 192.168.1.0/24 is gone, you will have to manually add it
back.

The easiest way probably is to run /etc/rc.d/routing, which adds routes
listed in /etc/rc.conf, after you change your interface address manually.

Best Regards,
sephe

> In IPv6, after the address change, the default gateway begins with fe80
> instead of 2001. The attached files are:
> routebad: the routing table (netstat -rn) after a previous address change
> routegood: the routing table after rebooting
> routebadagain: the routing table after the most recent address change.
> After rebooting, I had to find the gateway by asking the Linux box and set
> it manually; this may not be a bug, as normally zyxomma runs a tunnel, and
> the IPv6 gateway was pointing to an address in the tunnel.
>
>
> --
> You have received this notification because you have either subscribed to
> it, or are involved in it.
> To change your notification preferences, please click here:
> http://bugs.dragonflybsd.org/my/account
>

--
Tomorrow Will Never Die

#2 Updated by phma 10 months ago

I ran "/etc/rc.d/routing restart" and tried pinging again. Again the packets are sent from 100.116.27.42 to 100.116.42.25.

I didn't manually change the address. It was changed by dhcp and rtadvd.

This is on zyxomma, which is running 3.7. I've booted darner, which is running 3.5; I'll let you know what darner does if this happens again. Sometime tomorrow I should be taking zyxomma and leopard to a DSL connection, which shouldn't have this problem. When I was on cable in Charlotte, the address usually changed only if the power went out, in which case I had to reboot them, or the cable was cut or jiggled.

#3 Updated by phma 10 months ago

Darner does not have the bug of sending an IPv4 packet with the old source address. It does have a default gateway beginning with fe80 after its address was changed on it, but that apparently does not cause a problem with only one network card.

Also available in: Atom PDF