Description of problem: When you have two interfaces on the same network (i.e. eth0=192.1.1.1 and eth1=192.1.1.2), 'route' shows two entries for 192.1.1.0, but both with interface eth1! Therefore when you do an ifdown eth1, it takes BOTH 192.1.1.0 routes out, leaving a dead network. Version-Release number of selected component (if applicable): 1.60-3 How reproducible: Always Steps to Reproduce: 1. Configure eth0 as 192.1.1.1, bring it up 2. Configure eth1 as 192.1.1.2, bring it up 3. Do a 'route': it shows two 192.1.1.0 entries but both bound to eth1 (incorrect) 4. ifdown eth0, one route should go away (doesn't) 5. ifdown eth1, both 192.1.1.0 routes go away (only one should) Actual Results: Both routes got bound to the second interface, eth1. Expected Results: One route should be bound to eth0, one route should be bound to eth1.
Could you verify if this still happens with the latest initscripts from here: http://www.redhat.com/support/errata/RHBA-2002-022.html and see if it works? Thanks, Read ya, Phil
Still happens. It may just be that it's sufficiently confused by two interfaces on the same 192.1.1.x network. I could understand if this is not a valid config...
I see this problem on my machine too. It has a motherboard ethernet interface and a PCI card 10/100 NIC. eth1 gets assigned twice to the base network address at boot, which disables the eth0 connection. I think the situation can be fixed temporarily by removing the route for eth1 (twice): # route del -net 192.1.1.1 netmask 255.255.255.0 dev eth1 # route del -net 192.1.1.1 netmask 255.255.255.0 dev eth1 ...then adding back routes for each interface. This doesn't require brigging down the interfaces.
Works fine with a RH 7.2 with the following packages installed: glibc-2.2.4-24 kernel-2.4.18-3 initscripts-6.43-1 net-tools-1.60-3 Read ya, Phil