Linux – How to handle DHCP on two network interfaces, one clobbers the other’s /etc/resolv.conf

dhcplinuxmodemnetworking

I have an embedded Debian Linux system. It has two network interfaces, eth0 and ppp0 via a cell modem. The problem I have is that when the system boots up it gets an address via DHCP on eth0. If the modem connects that then also gets it's own IP address and clobbers the contents of /etc/resolv.conf. In my application I need eth0 to be the default route if eth0 is alive. If I make eth0 the default route, DNS no longer works as the nameservers from the modem DHCP request are not reachable.

Curiously if I then run "dhclient eth0" after all this is running, it does the right thing and adds it's nameservers to the existing file, rather than clobbering it.

Is there any good and standard way of handling this dual DHCP network situation?

Best Answer

man interface-order first.

There is a way as well to ignore some settings that DHCP provides. Check for prepend and request options at man dhclient.conf

Related Solutions

Linux – Ubuntu Linux Router: Set outbound NAT on 2 interfaces

The problem lies in getting a response from the client LAN I think, and so yes you need to use NAT to the client-LAN as well.

Let me explain:

Nodes on your LAN have your gateway as their default gateway.

Thus any traffic from them, be it to the client network (10.0.0.0/8) or the wider world (0.0.0.0/8) goes to it.

Traffic to the wider world is then NAT'd onto the Internet.

Traffic to the client's LAN is forwarded as is (I'm assuming here forwarding is working which the NAT working indicates it is).

So a packet from a node on your internal LAN, let us say it's IP is 192.168.1.200 goes to your gateway and is forward to the client LAN. However it's source address is still 192.168.1.200.

The client machine receives this and tried to reply, to 192.168.1.200.

Unless the client LAN machines have routes set for 192.168.1.0/24, or their default gateway is able to forward the packets to your gateway, they will not be able to route.

With NAT enabled then at the gateway the source address 192.168.1.200 is NAT'd to your client LAN gateway address, which client LAN nodes can respond to, where it will be re-addressed by your gateway and returned.

Linux – How to determine eth0 gateway address when it is not the default gateway

Assume eth0 is DHCP client interface.
One option is to check the DHCP client lease files dhcpd.leases
Place and name depends on the system; on some Fedora systems, the files under /var/lib/dhclient/ are lease files, where the interesting string is like that :

  option routers 192.168.1.1;

Another option, which worked for me on a funtoo box: dhcpcd -U eth0
prints a nice table, ready to source in scripts

broadcast_address=192.168.1.255
dhcp_lease_time=86400
dhcp_message_type=5
dhcp_server_identifier=192.168.1.1
domain_name_servers='192.168.1.1 192.168.1.101'
ip_address=192.168.1.101
network_number=192.168.1.0
routers=192.168.1.1
subnet_cidr=24
subnet_mask=255.255.255.0

There other options like dhcping, dhclient -n, according to google and manpages, but they fail on my boxes, but may work for you.

Best Answer

Related Solutions

Linux – Ubuntu Linux Router: Set outbound NAT on 2 interfaces

Linux – How to determine eth0 gateway address when it is not the default gateway

Related Question