When it starts, dhcpd reads its configuration from the /etc/dhcpd.conf file. dhcpd.conf defines the network being served by the DHCP server and the configuration information the server provides to its clients.
dhcpd.conf is an ASCII text file. Comments in the file begin with a sharp sign (#). Keywords are case-insensitive. Whitespace can be used to format the file. Related statements are enclosed in curly braces. IP address can be entered as numeric addresses or as hostnames that resolve to addresses.
Statements in the configuration file define the topology of the network being served. In the documentation these statements are called "declarations" because they declare something about the network topology. The statements that define the topology are: server-identifier, shared-network, subnet, group, and host. When used, there is only one server-identifier. All the other statements can appear multiple times in the configuration file. The statements define a hierarchical structure. The shared-network contains subnets, and subnets can contain hosts.
Parameters and options can be associated with each of these statements. Parameters define things about the server and the protocol, such as the length of time for an address lease or where the boot file is located. The options provide the clients with values for the standard DHCP configuration options defined by the RFCs: for example, whether the client should enable IP forwarding. Parameters and options specified outside of a specific topology statement apply to all networks served by this server. Those specified in the group statement apply to all of the shared networks, subnets or hosts grouped together by the statement. The shared-network statement options and parameters apply to all subnets on the shared network. Subnet options and parameters apply to everything on the subnet. Host options and parameters only apply to the individual host. Options applied at a general level can be overridden by the same option applied at a more specific level. Subnet options override global options and host options override subnet options. This structure allows the network administrator to define configuration information for the entire network and all of its parts.
In the following sections, we examine the syntax of all of the topology statements and of all the parameters and options that can be associated with them. We include many more parameters and options than you will ever use, and there is no need to study them all. Use this reference to look up the details of individual parameters and options when you need them. See Chapter 9 for examples of how these statements, parameters, and options are actually used in a real-world configuration.
The server-identifier statement documents the IP address of the server. It is sometimes used at the start of the file as the first statement of a group of parameter statements and option statements that apply to every network served by this server. The documentation calls these "global parameters."
The group statement groups together shared-network, subnet, host, or other group statements to apply a set of parameters or options to all members of the group.
The shared-network statement is used only if more than one IP subnet
shares the same physical network. In most cases, different subnets are
on different physical networks. The
name, which must be provided,
can be any descriptive name. It is used only in debugging messages.
Parameters and options associated with the shared network are declared
within the curly braces and apply to all subnets in the shared network.
The subnets in a shared network must be defined within the curly braces
of the shared-network statement. It is assumed that each
statement contains at least two subnet statements; otherwise there is no
need to use the shared-subnet statement. dhcpd cannot tell on which
subnet of a shared network a client should boot. Therefore, dynamically
allocated addresses are taken from the available range of all subnets
on the shared network and assigned as needed.
The subnet statement defines the IP address and address mask of every subnet the daemon will serve. The address and mask are used to identify the clients that belong to the subnet. The parameters and options defined within the curly braces apply to every client on the subnet. Every subnet physically connected to the server must have a subnet statement even if the subnet does not have any clients.
The host statement defines parameters and options for individual clients. Every BOOTP client must have a host statement in the dhcpd.conf file. For DHCP clients, the host statement is optional. It is matched to an actual DHCP or BOOTP clients by matching the dhcp-client-identifier provided by the client or by matching the hardware parameter to the hardware address of the client. BOOTP clients do not provide a dhcp-client-identifier, so use the hardware address for BOOTP clients. DHCP clients can be identified by either the dhcp-client-identifier or the hardware address.
The parameter statements defined in this section control the operation of the DHCP server and the DHCP protocol. The standard DHCP configuration values that are passed to clients are defined in option statements, which are covered in the next section. Some parameter statements can be associated with any of the topology statements discussed above. Others can only be used with specific statements. These are noted in the description of the parameter.
The range parameter defines the scope of addresses that are available for dynamic assignment by defining the lowest and highest IP addresses available for assignment. The range parameter must be associated with a subnet statement. All addresses in the scope of the range parameter must be in the subnet in which the range parameter is declared. The dynamic-bootp flag is specified if addresses may be automatically assigned to BOOTP clients as well as DHCP clients. The range parameter must be defined if you intend to use dynamic address assignment. If the subnet statement does not include a range parameter, dynamic address assignments are not made to clients on the subnet.
The life of an address lease in seconds that is used if the client does not request a specific lease length.
The maximum life of an address lease in seconds regardless of the lease length the client requests.
Defines a client's hardware address. At present,
type must be
address must be
an appropriate physical address for the type of hardware. The hardware
parameter must be associated with a host statement. It is required for
a BOOTP client to be recognized. It is optional for DHCP clients for
which it is an alternative to the dhcp-client-identifier option.
Identifies the boot file for diskless clients.
file is an ASCII
string enclosed in quotation marks.
The hostname of the DHCP server that is provided to the client.
is an ASCII string enclosed in quotation marks.
The hostname or address of the server from which the boot file is to be loaded.
Assigns one or more fixed IP addresses to a host. The fixed-address parameter is valid only when associated with a host statement. If more than one address is supplied, the client is assigned the address that is valid for the subnet on which it is booting. If none of the addresses is valid for the subnet, no configuration data is sent to the client.
Sets a termination date for addresses assigned to BOOTP clients. BOOTP clients do not have a way of renewing leases and don't know that address leases expire. By default, dhcpd assigns permanent address to BOOTP clients. This parameter changes that behavior. It is used only in special circumstances where the life of all systems is known in advance - for example, on a college campus where it is known that all student systems will be removed by June.
Defines the life of an address lease in seconds for an address automatically assigned to a BOOTP client. As noted above, BOOTP clients do not understand address leases. This parameter is used only in special circumstances where clients use a BOOTP boot PROM and run an operating system that supports DHCP. During the boot the client acts as a BOOTP client, but once it boots the client runs DHCP and knows how to renew a lease. Use this parameter, and the previous one, with caution.
Tells dhcpd whether or not to dynamically assign addresses to
unknown clients. If
flag is "false," addresses are provided
only to clients that have a host statement in the configuration file.
By default, the flag is "true" and addresses are dynamically assigned
to any client on a valid subnet.
Tells dhcpd if it should send a DNS hostname to the client when
it dynamically assigns it an IP address. If
flag is "true,"
dhcpd uses DNS to look up the hostnames for all dynamically
assigned addresses, which dramatically slows DHCP performance. By
flag is "false" and no lookups are done.
The option statements available with dhcpd cover all of the standard DHCP configuration options currently defined in the RFCs. Furthermore the syntax of the dhcpd.conf option statement is extensible. A new option can be identified by its decimal option code. All options are assigned a decimal option code, either in the RFC that describes the option or in the vendor documentation if it is vendor-specific. The value assigned to the new option can be expressed as a string enclosed in quotes or as a colon-separated list of hexadecimal numbers. Imagine that a new DHCP option is created and assigned an option code of 133. Further, imagine that the value carried by this option is a 16-bit binary mask and that you want your clients to "turn on" the high-order 4-bit and "turn off" all other bits in the mask. You could add the following option to your configuration:
option option-133 F0:00
All option statements begin with the keyword
option. The keyword
is then followed by the name of the option and the value assigned to
the option, in that order. In the example above, the option name is in
the form option-
nnn is the decimal option code
assigned to the option. In this manner any new option that appears can
be added to dhcpd.conf file. The value assigned to this imaginary
option is F000.
Looking at the huge list of standard options, you may well wonder if they will ever need to be extended. The standard options are listed in the following section. The types of values that are assigned to options are:
An IP address written in dotted decimal notation or a host name that resolves to an address
A series of characters enclosed in quotation marks
A numeric value
A switch containing either 1 or 0
In this book, the list of options is divided into "Commonly used options" and "Other options."
Specifies the subnet mask in dotted decimal notation. If the subnet mask option is not provided, dhcpd uses the network mask from the subnet statement.
Specifies the number of seconds this time zone is offset from Coordinated Universal Time (ETC).
Lists the routers the client should use, in order of preference.
Lists the Domain Name System (DNS) name servers the client should use, in order of preference.
Lists line printer (LPR) servers the client should use, in order of preference.
Defines the hostname the client should use.
Defines the domain name.
Defines the MTU the client should use. The minimum legal value for the MTU is 68.
Defines the broadcast address for the client's subnet.
destination gateway... ];
Lists the static routes the client should use. The default route cannot be specified in this manner. Use the routers option for the default route.
Specifies if the client should use trailer encapsulation. See the discussion of trailer encapsulation in Chapter 6, Configuring the Interface . 0 means "no" the client shouldn't and 1 means "yes" the client should use trailer encapsulation.
A character string that defines the name of the Network Information Services (NIS) domain.
Lists IP addresses of the NIS servers the client should use, in order of preference.
Lists the time servers the client should use, in order of preference.
Lists the IEN 116 name servers the client should use, in order of preference. IEN 116 is an obsolete name service. Avoid this and use DNS.
Lists the MIT-LCS UDP log servers the client should use, in order of preference.
Lists the cookie servers available to the client, in order of preference.
Lists the Image Impress servers available to the client, in order of preference.
Lists the Resource Location servers the client should use, in order of preference.
The number of 512-octet blocks in boot file.
path is a character string that identifies the location of the
file the client should dump core to in the event of a crash.
Specifies the IP address of the client's swap server.
path is a character string that identifies the location of the
client's root disk.
Specifies if the client should do IP forwarding. 0 disables IP forwarding, and 1 enables it.
Specifies if the client should allow non-local source routes. Source routes are a potential security problem as they can be used by intruders to route data off the local network in ways not intended by the local network administrator. 0 disables forwarding of non-local source routed datagrams, and 1 enables forwarding. 0 is the more secure setting.
Lists the IP addresses and masks that specify the only valid destination/mask pairs for incoming source routes. Any source-routed datagram whose next-hop address does not match one of the filters is discarded by the client.
Defines, in bytes, the largest datagram the client should be
prepared to reassemble. The value of
bytes cannot be less than 576.
Defines the default time-to-live (ttl) for outgoing datagrams. See the discussion of traceroute in Chapter 11, Troubleshooting TCP/IP , for information about ttl.
Set the number of seconds for timing out Path MTU values discovered by the mechanism defined in RFC 1191.
Defines a table of MTU sizes to use when performing Path MTU Discovery as defined in RFC 1191. The minimum MTU value cannot be smaller than 68.
Tells the client if all subnets of the local network use the same MTU. 1 means that all subnets share the same MTU. 0 means that some subnets have smaller MTUs.
Specifies if the client should use ICMP to discover the subnet mask. 0 enables ICMP mask discovery, and 1 disables it. Because the DHCP server can provide the correct subnet mask, ICMP mask discovery is rarely used on networks that have a DHCP server.
Specifies if the client should respond to ICMP subnet mask requests. 0 means "no" and 1 means "yes" it should respond.
Specifies if the client should use the Router Discovery mechanism defined in RFC 1256 to locate routers. 0 means "no" it shouldn't, and 1 means "yes" the client should perform router discovery. Because the DHCP server provides the correct list of routers, router discovery is rarely used on networks that have a DHCP server.
Defines the address to which the client should transmit a router solicitation request if router discovery is enabled.
Defines the number of seconds entries are maintained in the ARP cache.
Specifies if the client should use Ethernet II (DIX) or IEEE 802.3 Ethernet encapsulation on the network. 0 tells the client to use Ethernet II and 1 tells the client to use IEEE 802.3 encapsulation.
Defines the default TTL for TCP segments. Possible values are 1 to 255.
The number of seconds TCP should wait before sending a keepalive message. 0 means that TCP should not generate keepalive messages. Keepalive messages are generally discouraged.
Specifies if the client should send TCP keepalive messages with an octet of garbage for compatibility with older implementations. 0 means don't send a garbage octet and 1 means send it. Keepalives are generally discouraged.
Lists the IP addresses of the Network Time Protocol (NTP) servers the client should use, in order of preference.
Lists the NetBIOS name servers (NBNS) the client should use, in order of preference.
Lists the NetBIOS datagram distribution servers (NBDD) the client should use, in order of preference.
Defines the NetBIOS node type of the client. A
type of 1 is a
NetBIOS B-node; 2 is a P-node; 4 is an M-node; 8 is an H-node.
A character string that defines the NetBIOS over TCP/IP scope parameter as specified in RFC 1001/1002.
Lists the X Window System Font servers the client should use, in order of preference.