Hardware Devices for Networking
| As stated at the beginning of this chapter, networking is one of the strong points of the Linux operating system. This section covers the classes of devices used for basic networking. Note that this section talks about hardware devices, and not Linux networking devices, which are discussed in the section "Using Network Configuration Tools." Network Interface CardsA computer must have a network interface card (NIC) to connect to a network. Currently, there are several topologies (ways of connecting computers) for network connections. These topologies range from the old and mostly outdated 10BASE-2 to the much newer and popular wireless Wi-Fi or 802.11 networking.
Each NIC has a unique address (the hardware address, known as media access control, or MAC), which identifies that NIC. This address is six pairs of hexadecimal bits separated by colons (:). A MAC address looks similar to this: 00:60:08:8F:5A:D9. The hardware address is used by DHCP (see "DHCP" later in this chapter) to identify a specific host. It is also used by the Address Resolution Protocol (ARP) and Reverse Address Resolution Protocol (RARP) to map hosts to IP addresses. This section covers some of the different types of NIC used to connect to your network. Token RingToken ring networking was developed by IBM. As the name implies, the network is set up in a ring. A single token is passed from host to host, indicating the receiving host's permission to transmit data. Token ring has a maximum transfer rate of 16Mbps (16 million bits per second). Unlike 10BASE-2 and 10BASE-5, token ring uses what is called unshielded twisted pair (UTP) cable. This cable looks a lot like the cable that connects your phone to the wall. Almost all token ring NICs are recognized by Linux. 10BASE-T10BASE-T was the standard for a long time. A large number of networks still use it. 10BASE-T also uses UTP cable. Instead of being configured in a ring, 10BASE-T mostly uses a star architecture. In this architecture, the hosts all connect to a central location (usually a hub, which you learn about later in the section titled "Hubs and Switches"). All the data is sent to all hosts, but only the destination host takes action on individual packets. 10BASE-T has a transfer rate of 10Mbps. 10BASE-T has a maximum segment length of 100 meters. There are many manufacturers of 10BASE-T NICs, and most are recognized by Fedora. 100BASE-T100BASE-T was popular around the turn of the millennium, keeping the same ease of administration as 10BASE-T while increasing the speed by a factor of 10. For most networks, the step from 10BASE-T to 100BASE-T is as simple as replacing NICs and hubs. Most 100BASE-T NICs and hubs can also handle 10BASE-T and can automatically detect which is in use. This allows for a gradual network upgrade and usually does not require rewiring your whole network. Nearly every known 100BASE-T NIC and most generic NICs are compatible with Linux, thanks to Donald Becker of http://www.scyld.com/. 100BASE-T requires category 5 unshielded twisted pair cabling. 1000BASE-T1000BASE-Tusually referred to as gigabit ethernetis the accepted standard in enterprise networking, with most NICs being detected and configured correctly by Fedora. Like 100BASE-T NICs, gigabit NICs automatically downgrade if they are plugged in to a slower network. Also like 100BASE-T, gigabit NICs require category 5 unshielded twisted pair cabling; however, many institutions are now deploying category 6 cables because they have much longer range and so are often worth the extra cost. You will find that many newer computers tend to be fitted with gigabit NICs as standard. Fiber Optic and Gigabit EthernetFiber optic is more commonly used in newer and high-end installations because the cost of upgrading can be prohibitive for older sites. Fiber optics were originally used on fiber distributed data interface (FDDI) networks, similar to token ring in structure except that there are two ringsone is primary, whereas the other is secondary. The primary ring is used exclusively, and the secondary sits idle until there is a break in the primary ring. At this point, the secondary ring takes over, keeping the network alive. FDDI has a speed of 100Mbps and has a maximum ring length of 62 miles. FDDI uses several tokens at the same time that, along with the faster speed of fiber optics, account for the drastic increase in network speed. As stated, switching to a fiber optic network can be very costly. To make the upgrade, the whole network has to be rewired (as much as U.S. $150 per network connection), and all NICs must be replaced at the same time. Most FDDI NICs are recognized by Linux. Fiber-related gigabit is termed 1000BASE-X, whereas 1000BASE-T gigabit ethernet uses twisted-pair (see the "Unshielded Twisted Pair" section, later in this chapter). Wireless Network InterfacesWireless has really taken off in the past year or two. Wireless networking, as the name states, works without network cables. Upgrading is as easy as replacing network cards and equipment, such as routers and switches. Wireless networking equipment can also work along with the traditional wired networking using existing equipment. It might not be practical to upgrade a desktop or large server to wireless just yet if the wiring is already in place. Wireless networking is still generally slower than a traditional wired network. However, this situation is changing with wider adoption of newer protocols, such as 802.11g (supporting the common 802.11b and faster but less popular 802.11a), along with the introduction of more compliant and inexpensive wireless NICs. Some 802.11g NICs work at up to 108Mbps, which appears faster than 100BASE-T wired networking on the surface. However, in practice, it is a great deal slower: Unless your networking environment has paper-thin walls, you can usually halve the reported speed of Wi-Fi network devices. 108Mbps works about half the speed of 100BASE-T. Over the past few years, wireless networking has really taken off for home users. No longer do you have to go around on your hands and knees, running cables around your home. Instead you can install a central wireless router, and with a couple of wireless NICs and PC cards, you can have a safe and secure wireless network. Thankfully, with each new version of Linux, more and more wireless NICs are compatible. (That said, it is usually better to get brand name wireless NICs, because you have a better chance of compatibility.) Check the http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/web page for more specific hardware compatibility information. More on wireless networking is discussed later in this chapter. Network CableCurrently, three types of network cable exist: coaxial, unshielded twisted pair (UTP), and fiber. Coaxial cable (rarely used today) looks a lot like the coaxial cable used to connect your television to the cable jack or antenna. UTP looks a lot like the cable that runs from your phone to the wall jack (the jacks are a bit wider). Fiber cable looks sort of like the RCA cables used on your stereo or like the cable used on your electrical appliances in your house (two separate segments connected together). The following sections discuss UTP and fiber network cable in more detail. Unshielded Twisted PairUnshielded twisted pair (UTP) uses color-coded pairs of thin copper wire to transmit data. Six categories of UTP existeach serving a different purpose:
Fiber Optic CableFiber optic cable (fiber) is usually orange or red in color. The transmission rate is 100Mbps and has a maximum length of 62 miles. Fiber uses a two-pronged plug to connect to devices. A couple of advantages to fiber are that because it uses light instead of electricity to transmit its signal, it is free from the possibility of electromagnetic interference and is also more difficult to tap into and eavesdrop. Hubs and SwitchesHubs and switches are used to connect several hosts together on a star architecture network. They can have any number of connections; the common sizes are 4, 8, 16, 24, and 48 connections (ports)each port has a light that comes on when a network connection is made (link light). Their use enables you to expand your network easily; you can just add new hubs or switches when you need to add new connections. Each unit can connect to the other hubs or switches on the network, typically, through a port on the hub or switch called an uplink port. This enables two hubs or switches, connected by their uplink ports, to act as one hub or switch. Having a central location where all the hosts on your network can connect allows for easier troubleshooting of problems. If one host goes down, none of the other hosts are affected (depending on the purpose of the downed host). Because hubs and switches are not directly involved with the Linux operating system, compatibility is not an issue. If you are constructing a small to mid-size network, it is important to consider whether you intend to use either hubs or switches. Hubs and switches are visually the same in that they have rows of network ports. However, under the hood, the difference is quite important. Data is sent as packets of information across the network; with a hub the data is transmitted simultaneously to all the network ports, irrespective of which port the destination computer is attached to. Switches, however, are more intelligent because they can direct packets of information directly to the correct network port that leads to the destination computer. They do this by "learning" the MAC addresses of each computer that is attached to them. In short, using switches minimizes excess packets being sent across the network, thus increasing network bandwidth. In a small network with a handful of computers, the use of hubs might be perfectly acceptable and you will find that hubs are generally cheaper than switches. However, for larger networks of 15 computers or more, you might want consider implementing a switched network. Tip Troubleshooting network connections can be a challenge, especially on large networks. If a user complains that he has lost his network connection, the hub is a good place to start. If the link light for the user's port is lit, chances are the problem is with the user's network configuration. If the link light is not on, the host's NIC is bad, the cable is not inserted properly, or the cable has gone bad for some reason. Routers and BridgesRouters and bridges are used to connect different networks to your network and to connect different subnets within your network. Routers and bridges both serve the same purpose of connecting networks and subnets, but they do so with different techniques. The information in the following sections will help you choose the connection method that best suits your needs. BridgesBridges are used within a network to connect different subnets. A bridge blindly relays all information from one subnet to another without any filtering and is often referred to as a dumb gateway. This can be helpful if one subnet in your network is becoming overburdened and you need to lighten the load. A bridge is not very good for connecting to the Internet, however, because it lacks filtering. You really do not want all traffic traveling the Internet to be able to get through to your network. RoutersRouters can pass data from one network to another, and they allow for filtering of data. Routers are best suited to connect your network to an outside network, such as the Internet. If you have a web server for an internal intranet that you do not want people to access from the Internet, for example, you can use a router's filter to block port 80 from your network. These filters can be used to block specific hosts from accessing the Internet as well. For these reasons, routers are also called smart gateways. Routers range in complexity and price from a Cisco brand router that can cost thousands of dollars to other brands that can be less than two hundred dollars. You are most likely to come into contact with a router if you have a broadband Internet connection to your home or small office. Users who have DSL connections need a router with a built-in DSL modem to connect to their ISP. It is also very common to have a wireless element built into the router, so you end up with a wired and wireless router with an internal DSL modem. Initializing New Network HardwareAll the initial network configuration and hardware initialization for Fedora is normally done during installation. At times, however, you will have to reconfigure networking on your system, such as when a host needs to be moved to a different subnet or a different network, or if you replace any of your computer's networking hardware. Linux creates network interfaces in memory when the kernel recognizes that a NIC or other network device is attached to the system. These interfaces are unlike other Linux interfaces, such as serial communications ports, and do not have a corresponding device file in the /dev directory. Unless support for a particular NIC is built in to your kernel, Linux must be told to load a specific kernel module to support your NIC. More than 100 such modules are located in the /lib/modules/2.6.XX-XX/kernel/drivers/net directory (where XX-XX is your version of the kernel). You can initialize a NIC in several ways when using Linux. When you first install Fedora, automatic hardware probing detects and configures your system to use any installed NICs. If you remove the original NIC and replace it with a different make and model, your system will not automatically detect and initialize the device unless you configure Fedora to use automatic hardware detection when booting. If you have enabled the kudzu hardware daemon, (refer to Chapters 7, "Managing Software," and 15, "Automating Tasks"), it detects the absence of the old NIC and the presence of the new NIC at boot time and prompts you for permission to change your system's NIC configuration information. If you do not use automatic hardware detection and configuration, you can initialize network hardware by
The following sections explain the first and last of the preceding methods. Editing the /etc/modprobe.conf FileYou can manually edit the /etc/modprobe.conf file to add a module dependency entry (also known as a directive) to support a new NIC or other network device. This entry includes the device's name and its corresponding kernel module. After you add this entry, the Linux kernel recognizes your new networking hardware upon reboot. Fedora runs a module dependency check upon booting. For example, if your system uses a RealTek NIC, you could use an entry like this: alias eth0 8139too The example entry tells the Linux kernel to load the 8139too.o kernel module to support the eth0 network device. On the other hand, if you have an Intel Ethernet Pro NIC installed, you would use an entry like this: alias eth0 eepro100 Other parameters can be passed to a kernel module using one or more option entries, if need be, to properly configure your NIC. See the modprobe.conf man page for more information on using entries. For more specifics regarding NIC kernel modules, examine the module's source code because no man pages exist (a good opportunity for anyone willing to write the documentation). Tip Linux kernel and network tools can be used to diagnose problems or troubleshoot problematic NICs. However, if you browse to Don Becker's Linux Ethercard Status, Diagnostic and Setup Utilities page at http://www.scyld.com/ethercard_diag.html, you will find more than two dozen hardware-specific utilities for a variety of PCI and legacy ISA Ethernet network cards. These tools can be extremely helpful if you run into trouble during NIC recognition or configuration. Using modprobe to Manually Load Kernel ModulesYou do not have to use an /etc/modprobe.conf enTRy to initialize kernel support for your new network device. As root, you can manually load or unload the device's kernel module using the modprobe command, along with the module's name. For example, use the following command line to enable the example RealTek NIC: # modprobe 8139tooAfter you press Enter, you will see this device reported from the kernel's ring buffer messages, which can be displayed by the dmesg command. Here's a portion of that command's output: $ dmesg ... eth0: RealTek RTL8139 Fast Ethernet at 0xce8ee000, 00:30:1b:0b:07:0d, IRQ 11 eth0: Identified 8139 chip type 'RTL-8139C' eth0: Setting half-duplex based on auto-negotiated partner ability 0000. ...Note that at this point, an IP address or other settings have not been assigned to the device. Linux can use multiple ethernet interfaces, and the first ethernet device will be numbered eth0, the second eth1, and so on. Each different ethernet device recognized by the kernel might have additional or different information reported, depending on its kernel module. For example,
In this example, an Intel Ethernet Pro 100 NIC has been recognized. To disable support for a NIC, the kernel module can be unloaded, but usually only after the device is no longer in use. Read the next section to learn how to configure a NIC after it has been recognized by the Linux kernel and how to control its behavior. |
sg> and others PCI: Found IRQ 10 for device 00:0d.0 eth0: Intel Corporation 82557 [Ethernet Pro 100], 00:90:27:91:92:B5, IRQ 10. Board assembly 721383-007, Physical connectors present: RJ45 Primary interface chip i82555 PHY #1. General self-test: passed. Serial sub-system self-test: passed. Internal registers self-test: passed. ROM checksum self-test: passed (0x04f4518b). ...
EAN: 2147483647
Pages: 362