lfs-book/chapter07/network.xml

<?xml version="1.0" encoding="ISO-8859-1"?>
<!DOCTYPE sect1 PUBLIC "-//OASIS//DTD DocBook XML V4.4//EN"
  "http://www.oasis-open.org/docbook/xml/4.4/docbookx.dtd" [
  <!ENTITY % general-entities SYSTEM "../general.ent">
  %general-entities;
]>

<sect1 id="ch-scripts-network">
  <?dbhtml filename="network.html"?>

  <title>Configuring the network Script</title>

  <indexterm zone="ch-scripts-network">
    <primary sortas="d-network">network</primary>
  <secondary>configuring</secondary></indexterm>

  <para>This section only applies if a network card is to be
  configured.</para>

  <para>If a network card will not be used, there is likely no need to
  create any configuration files relating to network cards. If that is
  the case, remove the <filename class="symlink">network</filename>
  symlinks from all run-level directories (<filename
  class="directory">/etc/rc.d/rc*.d</filename>).</para>

  <sect2>
    <title>Creating stable names for network interfaces</title>

    <para>Instructions in this section are optional if you have only one
    network card.</para>

    <para>With Udev and modular network drivers, the network interface numbering
    is not persistent across reboots by default, because the drivers are loaded
    in parallel and, thus, in random order. For example, on a computer having
    two network cards made by Intel and Realtek, the network card manufactured
    by Intel may become <filename class="devicefile">eth0</filename> and the
    Realtek card becomes  <filename class="devicefile">eth1</filename>. In some
    cases, after a reboot the cards get renumbered the other way around. To
    avoid this, create Udev rules that assign stable names to network cards
    based on their MAC addresses or bus positions.</para>

    <para>If you are going to use MAC addresses to identify your network
    cards, find the addresses with the following command:</para>

<screen role="nodump"><userinput>grep -H . /sys/class/net/*/address</userinput></screen>

    <para>For each network card (but not for the loopback interface),
    invent a descriptive name, such as <quote>realtek</quote>, and create
    Udev rules similar to the following:</para>

<screen role="nodump"><userinput>cat &gt; /etc/udev/rules.d/70-persistent-net.rules &lt;&lt; EOF
<literal>ACTION=="add", SUBSYSTEM=="net", SYSFS{address}=="<replaceable>00:e0:4c:12:34:56</replaceable>", \
    NAME="<replaceable>realtek</replaceable>"
ACTION=="add", SUBSYSTEM=="net", SYSFS{address}=="<replaceable>00:a0:c9:78:9a:bc</replaceable>", \
    NAME="<replaceable>intel</replaceable>"</literal>
EOF</userinput></screen>

<!-- Yes, I know that VLANs are beyond BLFS. This is not the reason to get them
     incorrect by default when every distro does this right. -->

    <note>
      <para>Be aware that Udev does not recognize the backslash for line
      continuation.  The examples in this book work properly because both
      the backslash and newline are ignored by the shell.  This makes the
      shell send each rule to cat on only one line.  (The shell ignores
      this sequence because the EOF string used in the here-document
      redirection is not enclosed in either double or single quotes.  For
      more details, see the bash(1) manpage, and search it for "Here
      Documents".)</para>
      <para>If modifying Udev rules with an editor, be sure to leave each
      rule on one physical line.</para>
    </note>

    <para>If you are going to use the bus position as the key, find the
    position of each card with the following commands:</para>

<screen role="nodump"><userinput>for dir in /sys/class/net/* ; do
    [ -e $dir/device ] &amp;&amp; {
        basename $dir ; readlink -f $dir/device
    }
done</userinput></screen>

    <para>This will yield output similar to:</para>

<screen role="nodump"><userinput><replaceable>eth0</replaceable>
/sys/devices/pci0000:00/<replaceable>0000:00:0c.0</replaceable>
<replaceable>eth1</replaceable>
/sys/devices/pci0000:00/<replaceable>0000:00:0d.0</replaceable></userinput></screen>

    <para>In this example, <replaceable>eth0</replaceable> has PCI bus position
    <replaceable>0000:00:0c.0</replaceable> (domain 0000, bus 00, device 0c,
    function 0), and <replaceable>eth1</replaceable> has PCI bus position
    <replaceable>0000:00:0d.0</replaceable> (domain 0000, bus 00, device 0d,
    function 0).</para>
  
    <para>Now create Udev rules similar to the following:</para>

<screen role="nodump"><userinput>cat &gt; /etc/udev/rules.d/70-persistent-net.rules &lt;&lt; EOF
<literal>ACTION=="add", SUBSYSTEM=="net", BUS=="<replaceable>pci</replaceable>", KERNELS=="<replaceable>0000:00:0c.0</replaceable>", \
    NAME="<replaceable>realtek</replaceable>"
ACTION=="add", SUBSYSTEM=="net", BUS=="<replaceable>pci</replaceable>", KERNELS=="<replaceable>0000:00:0d.0</replaceable>", \
    NAME="<replaceable>intel</replaceable>"</literal>
EOF</userinput></screen>

    <para>Udev has installed a rule_generator rules file that uses MAC
    addresses, not bus positions.  Rules generated by this file will conflict
    with the rules you just created, so delete the file:</para>

<screen role="nodump"><userinput>rm /etc/udev/rules.d/75-persistent-net-generator.rules</userinput></screen>

    <note>
      <para>You will also have to remember to create a new bus-position-based
      rule each time you plug in an additional network card.  In a MAC address
      based persistence scheme, the rule_generator rules file would do this
      automatically.</para>
    </note>

    <para>Regardless of which method you use, these rules will always rename
    the network cards to <quote>realtek</quote> and <quote>intel</quote>,
    independently of the original numbering provided by the kernel (i.e.: the
    original <quote>eth0</quote> and <quote>eth1</quote> interfaces will no
    longer exist, unless you put such <quote>descriptive</quote> names in the
    NAME key). Use the descriptive names from the Udev rules instead of
    <quote>eth0</quote> in the network interface configuration files
    below.</para>

    <para>Note that the rules above don't work for every setup. For example,
    MAC-based rules break when bridges or VLANs are used, because bridges and
    VLANs have the same MAC address as the network card. One wants to rename
    only the network card interface, not the bridge or VLAN interface, but the
    example rule matches both. If you use such virtual interfaces, you have two
    potential solutions. One is to add the DRIVER=="?*" key after
    SUBSYSTEM=="net" in MAC-based rules which will stop matching the virtual
    interfaces.  This is known to fail with some older Ethernet cards because
    they don't have the DRIVER variable in the uevent and thus the rule does
    not match with such cards. Another solution is to switch to rules that use
    the bus position as a key.</para>

    <para>The second known non-working case is with wireless cards using the
    MadWifi or HostAP drivers, because they create at least two interfaces with
    the same MAC address and bus position. For example, the Madwifi driver
    creates both an athX and a wifiX interface where X is a digit.  To
    differentiate these interfaces, add an appropriate KERNEL parameter such as
    KERNEL=="ath*" after SUBSYSTEM=="net".</para>

    <para>There may be other cases where the rules above don't work. Currently,
    bugs on this topic are still being reported to Linux distributions, and no
    solution that covers every case is available.</para>

  </sect2>

  <sect2>
    <title>Creating Network Interface Configuration Files</title>

    <para>Which interfaces are brought up and down by the network script
    depends on the files and directories in the <filename
    class="directory">/etc/sysconfig/network-devices</filename> hierarchy.
    This directory should contain a sub-directory for each interface to be
    configured, such as <filename>ifconfig.xyz</filename>, where
    <quote>xyz</quote> is a network interface name. Inside this directory
    would be files defining the attributes to this interface, such as its IP
    address(es), subnet masks, and so forth.</para>

    <para>The following command creates a sample <filename>ipv4</filename>
    file for the <emphasis>eth0</emphasis> device:</para>

<screen><userinput>cd /etc/sysconfig/network-devices &amp;&amp;
mkdir -v ifconfig.eth0 &amp;&amp;
cat &gt; ifconfig.eth0/ipv4 &lt;&lt; "EOF"
<literal>ONBOOT=yes
SERVICE=ipv4-static
IP=192.168.1.1
GATEWAY=192.168.1.2
PREFIX=24
BROADCAST=192.168.1.255</literal>
EOF</userinput></screen>

    <para>The values of these variables must be changed in every file to match
    the proper setup. If the <envar>ONBOOT</envar> variable is set to
    <quote>yes</quote> the network script will bring up the Network Interface
    Card (NIC) during booting of the system. If set to anything but
    <quote>yes</quote> the NIC will be ignored by the network script and not
    be brought up.</para>

    <para>The <envar>SERVICE</envar> variable defines the method used for
    obtaining the IP address. The LFS-Bootscripts package has a modular IP
    assignment format, and creating additional files in the <filename
    class="directory">/etc/sysconfig/network-devices/services</filename>
    directory allows other IP assignment methods. This is commonly used for
    Dynamic Host Configuration Protocol (DHCP), which is addressed in the
    BLFS book.</para>

    <para>The <envar>GATEWAY</envar> variable should contain the default
    gateway IP address, if one is present. If not, then comment out the
    variable entirely.</para>

    <para>The <envar>PREFIX</envar> variable needs to contain the number of
    bits used in the subnet. Each octet in an IP address is 8 bits. If the
    subnet's netmask is 255.255.255.0, then it is using the first three octets
    (24 bits) to specify the network number. If the netmask is 255.255.255.240,
    it would be using the first 28 bits.  Prefixes longer than 24 bits are
    commonly used by DSL and cable-based Internet Service Providers (ISPs).
    In this example (PREFIX=24), the netmask is 255.255.255.0. Adjust the
    <envar>PREFIX</envar> variable according to your specific subnet.</para>

  </sect2>

  <sect2 id="resolv.conf">
    <title>Creating the /etc/resolv.conf File</title>

    <indexterm zone="resolv.conf">
      <primary sortas="e-/etc/resolv.conf">/etc/resolv.conf</primary>
    </indexterm>

    <para>If the system is going to be connected to the Internet, it will
    need some means of Domain Name Service (DNS) name resolution to
    resolve Internet domain names to IP addresses, and vice versa. This is
    best achieved by placing the IP address of the DNS server, available
    from the ISP or network administrator, into
    <filename>/etc/resolv.conf</filename>. Create the file by running the
    following:</para>

<screen><userinput>cat &gt; /etc/resolv.conf &lt;&lt; "EOF"
<literal># Begin /etc/resolv.conf

domain <replaceable>&lt;Your Domain Name&gt;</replaceable>
nameserver <replaceable>&lt;IP address of your primary nameserver&gt;</replaceable>
nameserver <replaceable>&lt;IP address of your secondary nameserver&gt;</replaceable>

# End /etc/resolv.conf</literal>
EOF</userinput></screen>

    <para>Replace <replaceable>&lt;IP address of the nameserver&gt;</replaceable>
    with the IP address of the DNS most appropriate for the setup. There will
    often be more than one entry (requirements demand secondary servers for
    fallback capability). If you only need or want one DNS server, remove the
    second <emphasis>nameserver</emphasis> line from the file. The IP address
    may also be a router on the local network.</para>

  </sect2>

</sect1>