18.6. Configuring a Network Connection Manually

Manual configuration of the network software should always be the last alternative. Using YaST is recommended. However, this background information about the network configuration can also assist your work with YaST.

All built-in network cards and hotplug network cards (PCMCIA, USB, some PCI cards) are detected and configured via hotplug. The system sees a network card in two different ways: first as a physical device and second as an interface. The insertion or detection of a device triggers a hotplug event. This hotplug event triggers the initialization of the device with the script hwup. When the network card is initialized as a new network interface, the kernel generates another hotplug event that triggers the setup of the interface with ifup.

The kernel numbers interface names according to the temporal order of their registration. The initialization sequence is decisive for the assignment of names. If one of several network card fails, the numbering of all subsequently initialized cards is shifted. For real hotpluggable cards, the order in which the devices are connected is what matters.

To achieve a flexible configuration, the configuration of the device (hardware) and the interface has been separated and the mapping of configurations to devices and interfaces is no longer managed on the basis of the interface names. The device configurations are located in /etc/sysconfig/hardware/hwcfg-*. The interface configurations are located in /etc/sysconfig/network/ifcfg-*. The names of the configurations are assigned in such a way that they describe the devices and interfaces with which they are associated. Because the former mapping of drivers to interface name required static interface names, this mapping can no longer take place in /etc/modprobe.conf. In the new concept, alias entries in this file would cause undesirable side effects.

The configuration names—everything after hwcfg- or ifcfg-—can describe the devices by means of the slot, a device-specific ID, or the interface name. For example, the configuration name for a PCI card could be bus-pci-0000:02:01.0 (PCI slot) or vpid-0x8086-0x1014-0x0549 (vendor and product ID). The name of the associated interface could be bus-pci-0000:02:01.0 or wlan-id-00:05:4e:42:31:7a (MAC address).

To assign a certain network configuration to any card of a certain type (of which only one is inserted at a time) instead of a certain card, select less specific configuration names. For example, bus-pcmcia would be used for all PCMCIA cards. On the other hand, the names can be limited by a preceding interface type. For example, wlan-bus-usb would be assigned to WLAN cards connected to a USB port.

The system always uses the configuration that best describes an interface or the device providing the interface. The search for the most suitable configuration is handled by getcfg. The output of getcfg delivers all information that can be used for describing a device. Details regarding the specification of configuration names are available in the manual page of getcfg.

With the described method, a network interface is configured with the correct configuration even if the network devices are not always initialized in the same order. However, the name of the interface still depends on the initialization sequence. There are two ways to ensure reliable access to the interface of a certain network card:

ifup requires an existing interface, because it does not initialize the hardware. The initialization of the hardware is handled by the command hwup (executed by hotplug or coldplug). When a device is initialized, ifup is automatically executed for the new interface via hotplug and the interface is set up if the start mode is onboot, hotplug, or auto and the network service was started. Formerly, the command ifup interfacename triggered the hardware initialization. Now the procedure has been reversed. First, a hardware component is initialized then all other actions follow. In this way, a varying number of devices can always be configured in the best way possible with an existing set of configurations.

Table 18.5, “Manual Network Configuration Scripts” summarizes the most important scripts involved in the network configuration. Where possible, the scripts are distinguished by hardware and interface.

Table 18.5. Manual Network Configuration Scripts

Configuration Stage

Command

Function

Hardware

hw{up,down,status}

The hw* scripts are executed by the hotplug subsystem to initialize a device, undo the initialization, or query the status of a device. More information is available in the manual page of hwup.

Interface

getcfg

getcfg can be used to query the interface name associated with a configuration name or a hardware description. More information is available in the manual page of getcfg.

Interface

if{up,down,status}

The if* scripts start existing network interfaces or return the status of the specified interface. More information is available in the manual page of ifup.

More information about hotplug and persistent device names is available in Chapter 12, Dynamic Kernel Device Management with udev.

18.6.1. Configuration Files

This section provides an overview of the network configuration files and explains their purpose and the format used.

18.6.1.1. /etc/syconfig/hardware/hwcfg-*

These files contain the hardware configurations of network cards and other devices. They contain the needed parameters, such as the kernel module, start mode, and script associations. Refer to the manual page of hwup for details. Regardless of the existing hardware, the hwcfg-static-* configurations are applied when coldplug is started.

18.6.1.2. /etc/sysconfig/network/ifcfg-*

These files contain the configurations for network interface. They include information such as the start mode and the IP address. Possible parameters are described in the manual page of ifup. Additionally, all variables from the files dhcp, wireless, and config can be used in the ifcfg-* files if a general setting should be used for only one interface.

18.6.1.3. /etc/sysconfig/network/config, dhcp, wireless

The file config contains general settings for the behavior of ifup, ifdown, and ifstatus. dhcp contains settings for DHCP and wireless for wireless LAN cards. The variables in all three configuration files are commented and can also be used in ifcfg-* files, where they are treated with higher priority.

18.6.1.4. /etc/sysconfig/network/routes,ifroute-*

The static routing of TCP/IP packets is determined here. All the static routes required by the various system tasks can be entered in the /etc/sysconfig/network/routes file: routes to a host, routes to a host via a gateway, and routes to a network. For each interface that needs individual routing, define an additional configuration file: /etc/sysconfig/network/ifroute-*. Replace * with the name of the interface. The entries in the routing configuration files look like this:

# Destination     Dummy/Gateway     Netmask            Device
#
127.0.0.0         0.0.0.0           255.255.255.0      lo
204.127.235.0     0.0.0.0           255.255.255.0      eth0
default           204.127.235.41    0.0.0.0            eth0
207.68.156.51     207.68.145.45     255.255.255.255    eth1
192.168.0.0       207.68.156.51     255.255.0.0        eth1

The route's destination is in the first column. This column may contain the IP address of a network or host or, in the case of reachable name servers, the fully qualified network or hostname.

The second column contains the default gateway or a gateway through which a host or network can be accessed. The third column contains the netmask for networks or hosts behind a gateway. For example, the mask is 255.255.255.255 for a host behind a gateway.

The fourth column is only relevant for networks connected to the local host such as loopback, Ethernet, ISDN, PPP, and dummy device. The device name must be entered here.

An (optional) fifth column can be used to specify the type of a route. Columns that are not needed should contain a minus sign - to ensure that the parser correctly interprets the command. For details, refer to the routes(5) man page.

18.6.1.5. /etc/resolv.conf

The domain to which the host belongs is specified in this file (keyword search). Also listed is the status of the name server address to access (keyword nameserver). Multiple domain names can be specified. When resolving a name that is not fully qualified, an attempt is made to generate one by attaching the individual search entries. Use multiple name servers by entering several lines, each beginning with nameserver. Precede comments with # signs. YaST enters the specified name server in this file. Example 18.5, “/etc/resolv.conf shows what /etc/resolv.conf could look like.

Example 18.5. /etc/resolv.conf


# Our domain
search example.com
#
# We use sun (192.168.0.20) as nameserver
nameserver 192.168.0.20

Some services, like pppd (wvdial), ipppd (isdn), dhcp (dhcpcd and dhclient), pcmcia, and hotplug, modify the file /etc/resolv.conf by means of the script modify_resolvconf. If the file /etc/resolv.conf has been temporarily modified by this script, it contains a predefined comment giving information about the service that modified it, the location where the original file has been backed up, and how to turn off the automatic modification mechanism. If /etc/resolv.conf is modified several times, the file includes modifications in a nested form. These can be reverted in a clean way even if this reversal takes place in an order different from the order in which modifications were introduced. Services that may need this flexibility include isdn, pcmcia, and hotplug.

If a service was not terminated in a normal, clean way, modify_resolvconf can be used to restore the original file. Also, on system boot, a check is performed to see whether there is an uncleaned, modified resolv.conf, for example, after a system crash, in which case the original (unmodified) resolv.conf is restored.

YaST uses the command modify_resolvconf check to find out whether resolv.conf has been modified and subsequently warns the user that changes will be lost after restoring the file. Apart from this, YaST does not rely on modify_resolvconf, which means that the impact of changing resolv.conf through YaST is the same as that of any manual change. In both cases, changes have a permanent effect. Modifications requested by the mentioned services are only temporary.

18.6.1.6.  /etc/hosts

In this file, shown in Example 18.6, “/etc/hosts, IP addresses are assigned to hostnames. If no name server is implemented, all hosts to which an IP connection will be set up must be listed here. For each host, enter a line consisting of the IP address, the fully qualified hostname, and the hostname into the file. The IP address must be at the beginning of the line and the entries separated by blanks and tabs. Comments are always preceded by the # sign.

Example 18.6. /etc/hosts


127.0.0.1 localhost
192.168.0.20 sun.example.com sun
192.168.0.0 earth.example.com earth

18.6.1.7. /etc/networks

Here, network names are converted to network addresses. The format is similar to that of the hosts file, except the network names precede the addresses. See Example 18.7, “/etc/networks.

Example 18.7. /etc/networks


loopback     127.0.0.0
localnet     192.168.0.0

18.6.1.8. /etc/host.conf

Name resolution—the translation of host and network names via the resolver library—is controlled by this file. This file is only used for programs linked to libc4 or libc5. For current glibc programs, refer to the settings in /etc/nsswitch.conf. A parameter must always stand alone in its own line. Comments are preceded by a # sign. Table 18.6, “Parameters for /etc/host.conf” shows the parameters available. A sample /etc/host.conf is shown in Example 18.8, “ /etc/host.conf.

Table 18.6. Parameters for /etc/host.conf

order hosts, bind

Specifies in which order the services are accessed for the name resolution. Available arguments are (separated by blank spaces or commas):

hosts: Searches the /etc/hosts file

bind: Accesses a name server

nis: Uses NIS

multi on/off

Defines if a host entered in /etc/hosts can have multiple IP addresses.

nospoof on spoofalert on/off

These parameters influence the name server spoofing, but, apart from that, do not exert any influence on the network configuration.

trim domainname

The specified domain name is separated from the hostname after hostname resolution (as long as the hostname includes the domain name). This option is useful if only names from the local domain are in the /etc/hosts file, but should still be recognized with the attached domain names.

Example 18.8.  /etc/host.conf


# We have named running
order hosts bind
# Allow multiple addrs
multi on

18.6.1.9. /etc/nsswitch.conf

The introduction of the GNU C Library 2.0 was accompanied by the introduction of the Name Service Switch (NSS). Refer to the nsswitch.conf(5) man page and The GNU C Library Reference Manual for details.

The order for queries is defined in the file /etc/nsswitch.conf. A sample nsswitch.conf is shown in Example 18.9, “/etc/nsswitch.conf. Comments are introduced by # signs. In this example, the entry under the hosts database means that a request is sent to /etc/hosts (files) via DNS (see Chapter 20, The Domain Name System).

Example 18.9. /etc/nsswitch.conf


passwd:     compat
group:      compat

hosts:      files dns
networks:   files dns

services:   db files
protocols:  db files

netgroup:   files
automount:  files nis

The “databases” available over NSS are listed in Table 18.7, “Databases Available via /etc/nsswitch.conf”. In addition, automount, bootparams, netmasks, and publickey are expected in the near future. The configuration options for NSS databases are listed in Table 18.8, “Configuration Options for NSS “Databases””.

Table 18.7. Databases Available via /etc/nsswitch.conf

aliases

Mail aliases implemented by sendmail; see man 5 aliases.

ethers

Ethernet addresses.

group

For user groups, used by getgrent. See also the man page for group.

hosts

For hostnames and IP addresses, used by gethostbyname and similar functions.

netgroup

Valid host and user lists in the network for the purpose of controlling access permissions; see the netgroup(5) man page.

networks

Network names and addresses, used by getnetent.

passwd

User passwords, used by getpwent; see the passwd(5) man page.

protocols

Network protocols, used by getprotoent; see the protocols(5) man page.

rpc

Remote procedure call names and addresses, used by getrpcbyname and similar functions.

services

Network services, used by getservent.

shadow

Shadow passwords of users, used by getspnam; see the shadow(5) man page.

Table 18.8. Configuration Options for NSS “Databases

files

directly access files, for example, /etc/aliases

db

access via a database

nis, nisplus

NIS, see also Chapter 21, Using NIS

dns

can only be used as an extension for hosts and networks

compat

can only be used as an extension for passwd, shadow, and group

18.6.1.10. /etc/nscd.conf

This file is used to configure nscd (name service cache daemon). See the nscd(8) and nscd.conf(5) man pages. By default, the system entries of passwd and groups are cached by nscd. This is important for the performance of directory services, like NIS and LDAP, because otherwise the network connection needs to be used for every access to names or groups. hosts is not cached by default, because the mechanism in nscd to cache hosts makes the local system unable to trust forward and reverse lookup checks. Instead of asking nscd to cache names, set up a caching DNS server.

If the caching for passwd is activated, it usually takes about fifteen seconds until a newly added local user is recognized. Reduce this waiting time by restarting nscd with the command rcnscd restart.

18.6.1.11. /etc/HOSTNAME

This contains the hostname without the domain name attached. This file is read by several scripts while the machine is booting. It may only contain one line in which the hostname is set.

18.6.2. Start-Up Scripts

Apart from the configuration files described above, there are also various scripts that load the network programs while the machine is booting. These are started as soon as the system is switched to one of the multiuser runlevels. Some of these scripts are described in Table 18.9, “Some Start-Up Scripts for Network Programs”.

Table 18.9. Some Start-Up Scripts for Network Programs

/etc/init.d/network

This script handles the configuration of the network interfaces. The hardware must already have been initialized by /etc/init.d/coldplug (via hotplug). If the network service was not started, no network interfaces are implemented when they are inserted via hotplug.

/etc/init.d/inetd

Starts xinetd. xinetd can be used to make server services available on the system. For example, it can start vsftpd whenever an FTP connection is initiated.

/etc/init.d/portmap

Starts the portmapper needed for the RPC server, such as an NFS server.

/etc/init.d/nfsserver

Starts the NFS server.

/etc/init.d/sendmail

Controls the sendmail process.

/etc/init.d/ypserv

Starts the NIS server.

/etc/init.d/ypbind

Starts the NIS client.