The first thing you'll want to do is list the IPs on your machine. The ip address tool will display IP (and terse encapsulation information) when invoked with the show verb. To specify that you wish to see the IP information for only one interface, you can add dev <device-name>
Example C.6. Displaying IP information with ip address
You should notice some similarity between the output of ip address and ifconfig. Each device is given an sequential number as an identifying number. This is merely a convenience, and should not be used to refer to devices. The second field in an entry is the interface name (which usually corresponds to the device name). Next, we see the familiar device flags and maximum transmission unit size.
The final fields in the first line of output for each device entry refer to the traffic control queueing discipline (qdisc) and the Ethernet buffer transmit queue length (qlen). For more on understanding and using traffic control under linux, see the LARTC documentation.
The second line of output describes the link layer characteristics of the device. For Ethernet devices, this will always say "link/ether" followed by the hardware address of the device and the media broadcast address. For more detail on the link layer characteristics of a device see Section B.3, “ip link”.
Subsequent lines of output describe the IP addresses available on each interface. In a typical installation only one address is used on each interface, although an arbitrary number of addresses can also be used on each interface.
Each line contains the IP address and netmask in CIDR notation, an optional broadcast address, scope information and a label. Let's examine the scope and label first and then discuss IP addressing and broadcast calculation. The possible values for scope are outlined in the following table.
Table C.2. IP Scope under ip address
|valid only within this site (IPv6)
|valid only on this device
|valid only inside this host (machine)
Scope is normally determined by the ip utility without explicit use on the command line. For example, an IP address in the 127.0.0.0/8 range falls in the range of localhost IPs, so should not be routed out any device. This explains the presence of the host scope for addresses bound to interface lo. Usually, addresses on other interfaces are public interfaces, which means that their scope will be global. We will revisit scope again when we discuss routing with ip route, and there we will also encounter the link scope.
Now, let's examine IP addressing with the ip address utility by adding and removing IP addresses from active interfaces.
If you need to host an additional IP address on
tristan, here's how you
would accomplish this task.
Example C.7. Adding IP addresses to an interface with ip address
There are a few items of note. You can use ip address add even if the link layer on the device is down. This means that you can readdress an interface without bringing it up. When you add an address within the same CIDR network as another address on the same interface, the second address becomes a secondary address, meaning that if the first address is removed, the second address will also be purged from the interface.
In order to support compatibility with ifconfig the ip address command allows the user to specify a label on every hosted address on a given device. After adding an address to an interface as we did in Example C.7, “Adding IP addresses to an interface with ip address”, ifconfig will not report that the new IP 192.168.99.37 is hosted on the same device as the primary IP 192.168.99.35. In order to prevent this sort of confusion or apparently contradictory output, you should get in the habit of using the label option to identify each IP hosted on a device. Let's take a look at how to remove the 192.168.99.37 IP from eth0 and add it back so that ifconfig will report the presence of another IP on the eth0 device.
There is a difference between IPs considered as primary addresses on an interface and secondary addresses. If in the output, an address is listed as a secondary address, removing the primary address will also remove the secondary address.
A workaround is to set the netmask on the second address added to the interface to /32. Unfortunately, this subterfuge will prevent the kernel from entering the correct corresponding network and broadcast routes.
Example C.8. Removing IP addresses from interfaces with ip address
Taking the minor precaution of using labels on IP addresses added to an interface will prevent confusion if there are multiple administrators of a machine, some of whom use ifconfig.
Finally, let's look at the use of ip address flush. If an interface has already had IP addresses assigned to it, and all of the addresses need to be removed (along with their routes), there is one handy command to accomplish all of these tasks. ip address flush takes an interface name as an argument. Let's look at the output of ip address show just before and just after removing all IPs.
Example C.9. Removing all IPs on an interface with ip address flush
As you can see, the ip address utility provides a wealth of information and a great deal of control over the IPs associated with each device. For more detailed information about the iproute2 package and included tools, see Section I.1.6, “iproute2 documentation”.