-1

As the title says, after deleting a rule from the main routing able I am not able to add any route till I do 

service network restart

or

ifup eth0

To reproduce the problem:

ip route list table main
192.168.2.0/24 dev eth0  proto kernel  scope link  src 192.168.2.47 
default via 192.168.2.1 dev eth0

I can delete and add the default route

#ip route del default
#ip route list
192.168.2.0/24 dev eth0  proto kernel  scope link  src 192.168.2.47 
#ip route add default via 192.168.2.1

#ip route list
192.168.2.0/24 dev eth0  proto kernel  scope link  src 192.168.2.47 
default via 192.168.2.1 dev eth0

However, if I delete the 192.168.2.0/24 route I can't add any route anymore.

# ip route del 192.168.2.0/24

# ip route list
default via 192.168.2.1 dev eth0 
# ip route add 192.168.2.0/24
RTNETLINK answers: No such device

# ip route add 192.168.2.0/24 via 0.0.0.0
RTNETLINK answers: No such device
# ip route add 192.168.2.0/24 src 192.168.2.47
RTNETLINK answers: No such device
# ip route list
default via 192.168.2.1 dev eth0
MadHatter
  • 79,770
  • 20
  • 184
  • 232
chandank
  • 847
  • 3
  • 14
  • 31
  • Why on **earth** would you want to do that? Is there a business need for this, or are you just asking a question that boils down to "*when I break the routing table, routing stops working*"? – MadHatter Aug 04 '14 at 14:14
  • basically we ship centos on a small pc based device to our customers. Occasionally they do break the routing table by doing something like this. I was wondering if there any way to fix it without rebooting the box. – chandank Aug 04 '14 at 15:31

2 Answers2

2

The reason for this is that you actually have two routing table entries. First there is the default, which points at the gateway 192.168.2.1 in 192.168.2.0/24 network, and then there is the network route for network for interface eth0 (ie. 192.168.2.0/24) as well.

If you delete the network route of an active, it effectively shuts it down. In this particular case deleting the route 192.168.2.0/24 means you have no usable interfaces for routes and thus adding routes will fail.

Summa summarum: You are simply doing it wrong or at lest in wrong order. Don't delete the 192.168.2.0/24 network, if you have an active interface in that network.

  • 1
    I question your use of "*happens to be*"; next-hop-route generally *has* to be in the local broadcast network, or you can't use it. – MadHatter Aug 04 '14 at 14:20
  • After reading what I wrote, I agree. –  Aug 04 '14 at 14:24
  • Fair enough; +1 from me. – MadHatter Aug 04 '14 at 14:28
  • Thanks for answering the question, however, I can't accept the fact that in Linux you can't fix something without reboot. – chandank Aug 04 '14 at 21:14
  • That is not at all what is written above. There is no reboot required. Only that whatever you're trying to do is simply not going to work, reboot or not. –  Aug 05 '14 at 11:50
0

Finally I got myself the problem.

Basically while adding a route we also need to define the scope and proto as well.

In my case the reason I was not able to add it back because I was not specifying the scope. Scope defined visibility of the IP in the network. It could be host, link [limited to LAN], gloabal etc.

So the correct syntax to add it would be

ip route add 192.168.2.0/24 via 0.0.0.0 dev eth0 scope link
chandank
  • 847
  • 3
  • 14
  • 31