Bug 9811

Latest working kernel version: none
Earliest failing kernel version: 2.6.18 (verified, but most likely "any")
Distribution: Debian
Hardware Environment: 
Software Environment:
Problem Description: 

From gary.manchon@gmail.com :

After a bad network interface configuration (ifconfig eth0 127.0.0.1),
I cannot recover the network without rebooting the kernel.

This is my working configuration :

# ifconfig
eth0      Link encap:Ethernet  HWaddr 00:90:3E:1F:1C:17
          inet addr:192.168.240.195  Bcast:192.168.247.255  Mask:255.255.248.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:26 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:2633 (2.5 KiB)  TX bytes:0 (0.0 B)

lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          UP LOOPBACK RUNNING  MTU:16436  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:0 (0.0 B)  TX bytes:0 (0.0 B)

# route
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
192.168.240.0   *               255.255.248.0   U     0      0        0 eth0
default         192.168.240.1   0.0.0.0         UG    0      0        0 eth0

# ping -c 1 66.102.11.99
PING 66.102.11.99 (66.102.11.99): 56 data bytes
64 bytes from 66.102.11.99: icmp_seq=0 ttl=63 time=4.5 ms

--- 66.102.11.99 ping statistics ---
1 packets transmitted, 1 packets received, 0% packet loss
round-trip min/avg/max = 4.5/4.5/4.5 ms

Steps to reproduce:

# ifconfig eth0 127.0.0.1
# ifconfig eth0 192.168.240.195 netmask 255.255.248.0
# route add default gw 192.168.240.1


From Bjørn Mork <bjorn@mork.no> :

I suspect that the problem might be this code in net/ipv4/devinet.c ,
which sets ifa_scope to RT_SCOPE_HOST if you configure a loopback
address (127/8) on any interface.  I guess it's there to protect us from
sending packets with a loopback source address, which woulnd't look too
good:

static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
{
	struct in_device *in_dev = __in_dev_get_rtnl(dev);

	ASSERT_RTNL();

	if (!in_dev) {
		inet_free_ifa(ifa);
		return -ENOBUFS;
	}
	ipv4_devconf_setall(in_dev);
	if (ifa->ifa_dev != in_dev) {
		BUG_TRAP(!ifa->ifa_dev);
		in_dev_hold(in_dev);
		ifa->ifa_dev = in_dev;
	}
	if (LOOPBACK(ifa->ifa_local))
		ifa->ifa_scope = RT_SCOPE_HOST;
	return inet_insert_ifa(ifa);
}



The real problem is that there's never anything resetting this scope if
you change the address later.  The attached patch fixes this.


dhcp232:~# ip addr show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,10000> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:aa:00:ff:00:ff brd ff:ff:ff:ff:ff:ff
    inet 192.168.3.232/24 brd 192.168.3.255 scope global eth0
    inet6 2001:16d8:ffb4:0:2aa:ff:feff:ff/64 scope global dynamic 
       valid_lft 2591971sec preferred_lft 604771sec
    inet6 fe80::2aa:ff:feff:ff/64 scope link 
       valid_lft forever preferred_lft forever
dhcp232:~# ifconfig eth0 127.0.0.1
dhcp232:~# ip addr show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,10000> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:aa:00:ff:00:ff brd ff:ff:ff:ff:ff:ff
    inet 127.0.0.1/8 brd 127.255.255.255 scope host eth0
    inet6 2001:16d8:ffb4:0:2aa:ff:feff:ff/64 scope global dynamic 
       valid_lft 2591951sec preferred_lft 604751sec
    inet6 fe80::2aa:ff:feff:ff/64 scope link 
       valid_lft forever preferred_lft forever
dhcp232:~# ifconfig eth0 192.168.3.232
dhcp232:~# ip addr show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,10000> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:aa:00:ff:00:ff brd ff:ff:ff:ff:ff:ff
    inet 192.168.3.232/24 brd 192.168.3.255 scope host eth0
    inet6 2001:16d8:ffb4:0:2aa:ff:feff:ff/64 scope global dynamic 
       valid_lft 2591933sec preferred_lft 604733sec
    inet6 fe80::2aa:ff:feff:ff/64 scope link 
       valid_lft forever preferred_lft forever



Notice how the scope changes from "global" to "host" when configuring
127.0.0.1, and just never changes back.  It will stay that way forever,
or until something changes the scope or deletes the address.

Deleting the addresss and re-adding it will work around the problem:

dhcp232:~# ip addr del 192.168.3.232/24 dev eth0
dhcp232:~# ifconfig eth0 192.168.3.232
dhcp232:~# ip addr show dev eth0
2: eth0: <BROADCAST,MULTICAST,UP,10000> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:aa:00:ff:00:ff brd ff:ff:ff:ff:ff:ff
    inet 192.168.3.232/24 brd 192.168.3.255 scope global eth0
    inet6 2001:16d8:ffb4:0:2aa:ff:feff:ff/64 scope global dynamic 
       valid_lft 2591963sec preferred_lft 604763sec
    inet6 fe80::2aa:ff:feff:ff/64 scope link 
       valid_lft forever preferred_lft forever


Bjørn