NetBSD/dist/dhcp/doc/rfc1542.txt

1291 lines
52 KiB
Plaintext
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

Network Working Group W. Wimer
Request for Comments: 1542 Carnegie Mellon University
Updates: 951 October 1993
Obsoletes: 1532
Category: Standards Track
Clarifications and Extensions for the Bootstrap Protocol
Status of this Memo
This RFC specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" for the standardization state and status
of this protocol. Distribution of this memo is unlimited.
Abstract
Some aspects of the BOOTP protocol were rather loosely defined in its
original specification. In particular, only a general description
was provided for the behavior of "BOOTP relay agents" (originally
called BOOTP forwarding agents"). The client behavior description
also suffered in certain ways. This memo attempts to clarify and
strengthen the specification in these areas. Due to some errors
introduced into RFC 1532 in the editorial process, this memo is
reissued as RFC 1542.
In addition, new issues have arisen since the original specification
was written. This memo also attempts to address some of these.
Table of Contents
1. Introduction................................................. 2
1.1 Requirements................................................ 3
1.2 Terminology................................................. 3
1.3 Data Transmission Order..................................... 4
2. General Issues............................................... 5
2.1 General BOOTP Processing.................................... 5
2.2 Definition of the 'flags' Field............................. 5
2.3 Bit Ordering of Hardware Addresses.......................... 7
2.4 BOOTP Over IEEE 802.5 Token Ring Networks................... 8
3. BOOTP Client Behavior........................................ 9
3.1 Client use of the 'flags' field............................. 9
3.1.1 The BROADCAST flag........................................ 9
3.1.2 The remainder of the 'flags' field........................ 9
3.2 Definition of the 'secs' field.............................. 10
3.3 Use of the 'ciaddr' and 'yiaddr' fields..................... 10
Wimer [Page 1]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
3.4 Interpretation of the 'giaddr' field........................ 11
3.5 Vendor information "magic cookie"........................... 12
4. BOOTP Relay Agents........................................... 13
4.1 General BOOTP Processing for Relay Agents................... 14
4.1.1 BOOTREQUEST Messages...................................... 14
4.1.2 BOOTREPLY Messages........................................ 17
5. BOOTP Server Behavior........................................ 18
5.1 Reception of BOOTREQUEST Messages........................... 18
5.2 Use of the 'secs' field..................................... 19
5.3 Use of the 'ciaddr' field................................... 19
5.4 Strategy for Delivery of BOOTREPLY Messages................. 20
Acknowledgements................................................ 21
References...................................................... 22
Security Considerations......................................... 23
Author's Address................................................ 23
1. Introduction
The Bootstrap Protocol (BOOTP) is a UDP/IP-based protocol which
allows a booting host to configure itself dynamically and without
user supervision. BOOTP provides a means to notify a host of its
assigned IP address, the IP address of a boot server host, and the
name of a file to be loaded into memory and executed [1]. Other
configuration information such as the local subnet mask, the local
time offset, the addresses of default routers, and the addresses of
various Internet servers can also be communicated to a host using
BOOTP [2].
Unfortunately, the original BOOTP specification [1] left some issues
of the protocol open to question. The exact behavior of BOOTP relay
agents formerly called "BOOTP forwarding agents") was not clearly
specified. Some parts of the overall protocol specification actually
conflict, while other parts have been subject to misinterpretation,
indicating that clarification is needed. This memo addresses these
problems.
Since the introduction of BOOTP, the IEEE 802.5 Token Ring Network
has been developed which presents a unique problem for BOOTP's
particular message-transfer paradigm. This memo also suggests a
solution for this problem.
NOTE: Unless otherwise specified in this document or a later
document, the information and requirements specified througout this
document also apply to extensions to BOOTP such as the Dynamic Host
Configuration Protocol (DHCP) [3].
Wimer [Page 2]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
1.1 Requirements
In this memo, the words that are used to define the significance of
particular requirements are capitalized. These words are:
o "MUST"
This word or the adjective "REQUIRED" means that the item
is an absolute requirement of the specification.
o "MUST NOT"
This phrase means that the item is an absolute prohibition
of the specification.
o "SHOULD"
This word or the adjective "RECOMMENDED" means that there
may exist valid reasons in particular circumstances to
ignore this item, but the full implications should be
understood and the case carefully weighed before choosing a
different course.
o "SHOULD NOT"
This phrase means that there may exist valid reasons in
particular circumstances when the listed behavior is
acceptable or even useful, but the full implications should
be understood and the case carefully weighed before
implementing any behavior described with this label.
o "MAY"
This word or the adjective "OPTIONAL" means that this item
is truly optional. One vendor may choose to include the
item because a particular marketplace requires it or
because it enhances the product, for example; another
vendor may omit the same item.
1.2 Terminology
This memo uses the following terms:
BOOTREQUEST
A BOOTREQUEST message is a BOOTP message sent from a BOOTP
client to a BOOTP server, requesting configuration information.
Wimer [Page 3]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
BOOTREPLY
A BOOTREPLY message is a BOOTP message sent from a BOOTP server
to a BOOTP client, providing configuration information.
Silently discard
This memo specifies several cases where a BOOTP entity is to
"silently discard" a received BOOTP message. This means that
the entity is to discard the message without further
processing, and that the entity will not send any ICMP error
message as a result. However, for diagnosis of problems, the
entity SHOULD provide the capability of logging the error,
including the contents of the silently-discarded message, and
SHOULD record the event in a statistics counter.
1.3 Data Transmission Order
The order of transmission of the header and data described in this
document is resolved to the octet level. Whenever a diagram shows a
group of octets, the order of transmission of those octets is the
normal order in which they are read in English. For example, in the
following diagram, the octets are transmitted in the order they are
numbered.
0 1
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 1 | 2 |
+-------------------------------+
| 3 | 4 |
+-------------------------------+
| 5 | 6 |
+-------------------------------+
Whenever an octet represents a numeric quantity, the leftmost bit in
the diagram is the high order or most significant bit. That is, the
bit labeled 0 is the most significant bit. For example, the
following diagram represents the value 170 (decimal).
0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+
|1 0 1 0 1 0 1 0|
+---------------+
Similarly, whenever a multi-octet field represents a numeric quantity
the leftmost bit of the whole field is the most significant bit.
When a multi-octet quantity is transmitted the most significant octet
Wimer [Page 4]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
is transmitted first.
2. General Issues
This section covers issues of general relevance to all BOOTP entities
(clients, servers, and relay agents).
2.1 General BOOTP Processing
The following consistency checks SHOULD be performed on BOOTP
messages:
o The IP Total Length and UDP Length must be large enough to
contain the minimal BOOTP header of 300 octets (in the UDP
data field) specified in [1].
NOTE: Future extensions to the BOOTP protocol may increase the size
of BOOTP messages. Therefore, BOOTP messages which, according to the
IP Total Length and UDP Length fields, are larger than the minimum
size specified by [1] MUST also be accepted.
o The 'op' (opcode) field of the message must contain either the
code for a BOOTREQUEST (1) or the code for a BOOTREPLY (2).
BOOTP messages not meeting these consistency checks MUST be silently
discarded.
2.2 Definition of the 'flags' Field
The standard BOOTP message format defined in [1] includes a two-octet
field located between the 'secs' field and the 'ciaddr' field. This
field is merely designated as "unused" and its contents left
unspecified, although Section 7.1 of [1] does offer the following
suggestion:
"Before setting up the packet for the first time, it is a good
idea to clear the entire packet buffer to all zeros; this will
place all fields in their default state."
This memo hereby designates this two-octet field as the 'flags'
field.
This memo hereby defines the most significant bit of the 'flags'
field as the BROADCAST (B) flag. The semantics of this flag are
discussed in Sections 3.1.1 and 4.1.2 of this memo.
The remaining bits of the 'flags' field are reserved for future
use. They MUST be set to zero by clients and ignored by servers
Wimer [Page 5]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
and relay agents.
The 'flags' field, then, appears as follows:
0 1
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|B| MBZ |
+-+-----------------------------+
where:
B BROADCAST flag (discussed in Sections 3.1.1 and 4.1.2)
MBZ MUST BE ZERO (reserved for future use)
The format of a BOOTP message is shown below. The numbers in
parentheses indicate the size of each field in octets.
Wimer [Page 6]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| op (1) | htype (1) | hlen (1) | hops (1) |
+---------------+---------------+---------------+---------------+
| xid (4) |
+-------------------------------+-------------------------------+
| secs (2) | flags (2) |
+-------------------------------+-------------------------------+
| ciaddr (4) |
+---------------------------------------------------------------+
| yiaddr (4) |
+---------------------------------------------------------------+
| siaddr (4) |
+---------------------------------------------------------------+
| giaddr (4) |
+---------------------------------------------------------------+
| |
| chaddr (16) |
| |
| |
+---------------------------------------------------------------+
| |
| sname (64) |
+---------------------------------------------------------------+
| |
| file (128) |
+---------------------------------------------------------------+
| |
| vend (64) |
+---------------------------------------------------------------+
2.3 Bit Ordering of Hardware Addresses
The bit ordering used for link-level hardware addresses in the
'chaddr' field SHOULD be the same as the ordering used for the ARP
protocol [4] on the client's link-level network (assuming ARP is
defined for that network).
The 'chaddr' field MUST be preserved as it was specified by the BOOTP
client. A relay agent MUST NOT reverse the bit ordering of the
'chaddr' field even if it happens to be relaying a BOOTREQUEST
between two networks which use different bit orderings.
DISCUSSION:
One of the primary reasons the 'chaddr' field exists is to
enable BOOTP servers and relay agents to communicate directly
Wimer [Page 7]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
with clients without the use of broadcasts. In practice, the
contents of the 'chaddr' field is often used to create an ARP-
cache entry in exactly the same way the normal ARP protocol
would have. Clearly, interoperability can only be achieved if
a consistent interpretation of the 'chaddr' field is used.
As a practical example, this means that the bit ordering used
for the 'chaddr' field by a BOOTP client on an IEEE 802.5 Token
Ring network is the opposite of the bit ordering used by a
BOOTP client on a DIX ethernet network.
2.4 BOOTP Over IEEE 802.5 Token Ring Networks
Special consideration of the client/server and client/relay agent
interactions must be given to IEEE 802.5 networks because of non-
transparent bridging.
The client SHOULD send its broadcast BOOTREQUEST with an All Routes
Explorer RIF. This will enable servers/relay agents to cache the
return route if they choose to do so. For those server/relay agents
which cannot cache the return route (because they are stateless, for
example), the BOOTREPLY message SHOULD be sent to the client's
hardware address, as taken from the BOOTP message, with a Spanning
Tree Rooted RIF. The actual bridge route will be recorded by the
client and server/relay agent by normal ARP processing code.
DISCUSSION:
In the simplest case, an unbridged, single ring network, the
broadcast behavior of the BOOTP protocol is identical to that
of Ethernet networks. However, a BOOTP client cannot know, a
priori, that an 802.5 network is not bridged. In fact, the
likelihood is that the server, or relay agent, will not know
either.
Of the four possible scenerios, only two are interesting: where
the assumption is that the 802.5 network is not bridged and it
is, and the assumption that the network is bridged and it is
not. In the former case, the Routing Information Field (RIF)
will not be used; therefore, if the server/relay agent are on
another segment of the ring, the client cannot reach it. In
the latter case, the RIF field will be used, resulting in a few
extraneous bytes on the ring. It is obvious that an almost
immeasurable inefficiency is to be preferred over a complete
failure to communicate.
Wimer [Page 8]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
Given that the assumption is that RIF fields will be needed, it
is necesary to determine the optimum method for the client to
reach the server/relay agent, and the optimum method for the
response to be returned.
3. BOOTP Client Behavior
This section clarifies various issues regarding BOOTP client
behavior.
3.1 Client use of the 'flags' field
3.1.1 The BROADCAST flag
Normally, BOOTP servers and relay agents attempt to deliver BOOTREPLY
messages directly to a client using unicast delivery. The IP
destination address (in the IP header) is set to the BOOTP 'yiaddr'
address and the link-layer destination address is set to the BOOTP
'chaddr' address. Unfortunately, some client implementations are
unable to receive such unicast IP datagrams until they know their own
IP address (thus we have a "chicken and egg" issue). Often, however,
they can receive broadcast IP datagrams (those with a valid IP
broadcast address as the IP destination and the link-layer broadcast
address as the link-layer destination).
If a client falls into this category, it SHOULD set (to 1) the
newly-defined BROADCAST flag in the 'flags' field of BOOTREPLY
messages it generates. This will provide a hint to BOOTP servers and
relay agents that they should attempt to broadcast their BOOTREPLY
messages to the client.
If a client does not have this limitation (i.e., it is perfectly able
to receive unicast BOOTREPLY messages), it SHOULD NOT set the
BROADCAST flag (i.e., it SHOULD clear the BROADCAST flag to 0).
DISCUSSION:
This addition to the protocol is a workaround for old host
implementations. Such implementations SHOULD be modified so
that they may receive unicast BOOTREPLY messages, thus making
use of this workaround unnecessary. In general, the use of
this mechanism is discouraged.
3.1.2 The remainder of the 'flags' field
The remaining bits of the 'flags' field are reserved for future use.
A client MUST set these bits to zero in all BOOTREQUEST messages it
generates. A client MUST ignore these bits in all BOOTREPLY messages
Wimer [Page 9]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
it receives.
3.2 Definition of the 'secs' field
The 'secs' field of a BOOTREQUEST message SHOULD represent the
elapsed time, in seconds, since the client sent its first BOOTREQUEST
message. Note that this implies that the 'secs' field of the first
BOOTREQUEST message SHOULD be set to zero.
Clients SHOULD NOT set the 'secs' field to a value which is constant
for all BOOTREQUEST messages.
DISCUSSION:
The original definition of the 'secs' field was vague. It was
not clear whether it represented the time since the first
BOOTREQUEST message was sent or some other time period such as
the time since the client machine was powered-up. This has
limited its usefulness as a policy control mechanism for BOOTP
servers and relay agents. Furthermore, certain client
implementations have been known to simply set this field to a
constant value or use incorrect byte-ordering. Incorrect
byte-ordering usually makes it appear as if a client has been
waiting much longer than it really has, so a relay agent will
relay the BOOTREQUEST sooner than desired (usually
immediately). These implementation errors have further
undermined the usefulness of the 'secs' field. These incorrect
implementations SHOULD be corrected.
3.3 Use of the 'ciaddr' and 'yiaddr' fields
If a BOOTP client does not know what IP address it should be using,
the client SHOULD set the 'ciaddr' field to 0.0.0.0. If the client
has the ability to remember the last IP address it was assigned, or
it has been preconfigured with an IP address via some alternate
mechanism, the client MAY fill the 'ciaddr' field with that IP
address. If the client does place a non-zero IP address in the
'ciaddr' field, the client MUST be prepared to accept incoming
unicast datagrams addressed to that IP address and also answer ARP
requests for that IP address (if ARP is used on that network).
The BOOTP server is free to assign a different IP address (in the
'yiaddr' field) than the client expressed in 'ciaddr'. The client
SHOULD adopt the IP address specified in 'yiaddr' and begin using it
as soon as possible.
Wimer [Page 10]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
DISCUSSION:
There are various interpretations about the purpose of the
'ciaddr' field and, unfortunately, no agreement on a single
correct interpretation. One interpretation is that if a client
is willing to accept whatever IP address the BOOTP server
assigns to it, the client should always place 0.0.0.0 in the
'ciaddr' field, regardless of whether it knows its previously-
assigned address. Conversely, if the client wishes to assert
that it must have a particular IP address (e.g., the IP address
was hand-configured by the host administrator and BOOTP is only
being used to obtain a boot file and/or information from the
'vend' field), the client will then fill the 'ciaddr' field
with the desired IP address and ignore the IP address assigned
by the BOOTP server as indicated in the 'yiaddr' field. An
alternate interpretation holds that the client always fills the
'ciaddr' field with its most recently-assigned IP address (if
known) even if that address may be incorrect. Such a client
will still accept and use the address assigned by the BOOTP
server as indicated in the 'yiaddr' field. The motivation for
this interpretation is to aid the server in identifying the
client and/or in delivering the BOOTREPLY to the client. Yet a
third (mis)interpretation allows the client to use 'ciaddr' to
express the client's desired IP address, even if the client has
never used that address before or is not currently using that
address.
The last interpretation is incorrect as it may prevent the
BOOTREPLY from reaching the client. The server will usually
unicast the reply to the address given in 'ciaddr' but the
client may not be listening on that address yet, or the client
may be connected to an incorrect subnet such that normal IP
routing (correctly) routes the reply to a different subnet.
The second interpretation also suffers from the "incorrect
subnet" problem.
The first interpretation seems to be the safest and most likely
to promote interoperability.
3.4 Interpretation of the 'giaddr' field
The 'giaddr' field is rather poorly named. It exists to facilitate
the transfer of BOOTREQUEST messages from a client, through BOOTP
relay agents, to servers on different networks than the client.
Similarly, it facilitates the delivery of BOOTREPLY messages from the
servers, through BOOTP relay agents, back to the client. In no case
does it represent a general IP router to be used by the client. A
Wimer [Page 11]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
BOOTP client MUST set the 'giaddr' field to zero (0.0.0.0) in all
BOOTREQUEST messages it generates.
A BOOTP client MUST NOT interpret the 'giaddr' field of a BOOTREPLY
message to be the IP address of an IP router. A BOOTP client SHOULD
completely ignore the contents of the 'giaddr' field in BOOTREPLY
messages.
DISCUSSION:
The semantics of the 'giaddr' field were poorly defined.
Section 7.5 of [1] states:
"If 'giaddr' (gateway address) is nonzero, then the packets
should be forwarded there first, in order to get to the
server."
In that sentence, "get to" refers to communication from the client to
the server subsequent to the BOOTP exchange, such as a TFTP session.
Unfortunately, the 'giaddr' field may contain the address of a BOOTP
relay agent that is not itself an IP router (according to [1],
Section 8, fifth paragraph), in which case, it will be useless as a
first-hop for TFTP packets sent to the server (since, by definition,
non-routers don't forward datagrams at the IP layer).
Although now prohibited by Section 4.1.1 of this memo, the 'giaddr'
field might contain a broadcast address according to Section 8, sixth
paragraph of [1]. Not only would such an address be useless as a
router address, it might also cause the client to ARP for the
broadcast address (since, if the client didn't receive a subnet mask
in the BOOTREPLY message, it would be unable to recognize a subnet
broadcast address). This is clearly undesirable.
To reach a non-local server, clients can obtain a first-hop router
address from the "Gateway" subfield of the "Vendor Information
Extensions" [2] (if present), or via the ICMP router discovery
protocol [5] or other similar mechanism.
3.5 Vendor information "magic cookie"
It is RECOMMENDED that a BOOTP client always fill the first four
octets of the 'vend' (vendor information) field of a BOOTREQUEST with
a four-octet identifier called a "magic cookie." A BOOTP client
SHOULD do this even if it has no special information to communicate
to the BOOTP server using the 'vend' field. This aids the BOOTP
server in determining what vendor information format it should use in
its BOOTREPLY messages.
Wimer [Page 12]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
If a special vendor-specific magic cookie is not being used, a BOOTP
client SHOULD use the dotted decimal value 99.130.83.99 as specified
in [2]. In this case, if the client has no information to
communicate to the server, the octet immediately following the magic
cookie SHOULD be set to the "End" tag (255) and the remaining octets
of the 'vend' field SHOULD be set to zero.
DISCUSSION:
Sometimes different operating systems or networking packages
are run on the same machine at different times (or even at the
same time!). Since the hardware address placed in the 'chaddr'
field will likely be the same, BOOTREQUESTs from completely
different BOOTP clients on the same machine will likely be
difficult for a BOOTP server to differentiate. If the client
includes a magic cookie in its BOOTREQUESTs, the server will at
least know what format the client expects and can understand in
corresponding BOOTREPLY messages.
4. BOOTP Relay Agents
In many cases, BOOTP clients and their associated BOOTP
server(s) do not reside on the same IP network or subnet. In
such cases, some kind of third-party agent is required to
transfer BOOTP messages between clients and servers. Such an
agent was originally referred to as a "BOOTP forwarding agent."
However, in order to avoid confusion with the IP forwarding
function of an IP router, the name "BOOTP relay agent" is
hereby adopted instead.
DISCUSSION:
A BOOTP relay agent performs a task which is distinct from an
IP router's normal IP forwarding function. While a router
normally switches IP datagrams between networks more-or-less
transparently, a BOOTP relay agent may more properly be thought
to receive BOOTP messages as a final destination and then
generate new BOOTP messages as a result. It is incorrect for a
relay agent implementation to simply forward a BOOTP message
"straight through like a regular packet."
This relay-agent functionality is most conveniently located in
the routers which interconnect the clients and servers, but may
alternatively be located in a host which is directly connected
to the client subnet.
Any Internet host or router which provides BOOTP relay-agent
capability MUST conform to the specifications in this memo.
Wimer [Page 13]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
4.1 General BOOTP Processing for Relay Agents
All locally delivered UDP messages whose UDP destination port number
is BOOTPS (67) are considered for special processing by the host or
router's logical BOOTP relay agent.
In the case of a host, locally delivered datagrams are simply all
datagrams normally received by that host, i.e., broadcast and
multicast datagrams as well as unicast datagrams addressed to IP
addresses of that host.
In the case of a router, locally delivered datagrams are broadcast
and multicast datagrams as well as unicast datagrams addressed to IP
addresses of that router. These are datagrams for which the router
should be considered an end destination as opposed to an intermediate
switching node. Thus a unicast datagram with an IP destination not
matching any of the router's IP addresses is not considered for
processing by the router's logical BOOTP relay agent.
Hosts and routers are usually required to silently discard incoming
datagrams containing illegal IP source addresses. This is generally
known as "Martian address filtering." One of these illegal addresses
is 0.0.0.0 (or actually anything on network 0). However, hosts or
routers which support a BOOTP relay agent MUST accept for local
delivery to the relay agent BOOTREQUEST messages whose IP source
address is 0.0.0.0. BOOTREQUEST messages from legal IP source
addresses MUST also be accepted.
A relay agent MUST silently discard any received UDP messages whose
UDP destination port number is BOOTPC (68).
DISCUSSION:
There should be no need for a relay agent to process messages
addressed to the BOOTPC port. Careful reading of the original
BOOTP specification [1] will show this. Nevertheless, some
relay agent implementations incorrectly relay such messages.
The consistency checks specified in Section 2.1 SHOULD be performed
by the relay agent. BOOTP messages not meeting these consistency
checks MUST be silently discarded.
4.1.1 BOOTREQUEST Messages
Some configuration mechanism MUST exist to enable or disable the
relaying of BOOTREQUEST messages. Relaying MUST be disabled by
default.
Wimer [Page 14]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
When the BOOTP relay agent receives a BOOTREQUEST message, it MAY use
the value of the 'secs' (seconds since client began booting) field of
the request as a factor in deciding whether to relay the request. If
such a policy mechanism is implemented, its threshold SHOULD be
configurable.
DISCUSSION:
To date, this feature of the BOOTP protocol has not necessarily
been shown to be useful. See Section 3.2 for a discussion.
The relay agent MUST silently discard BOOTREQUEST messages whose
'hops' field exceeds the value 16. A configuration option SHOULD be
provided to set this threshold to a smaller value if desired by the
network manager. The default setting for a configurable threshold
SHOULD be 4.
If the relay agent does decide to relay the request, it MUST examine
the 'giaddr' ("gateway" IP address) field. If this field is zero,
the relay agent MUST fill this field with the IP address of the
interface on which the request was received. If the interface has
more than one IP address logically associated with it, the relay
agent SHOULD choose one IP address associated with that interface and
use it consistently for all BOOTP messages it relays. If the
'giaddr' field contains some non-zero value, the 'giaddr' field MUST
NOT be modified. The relay agent MUST NOT, under any circumstances,
fill the 'giaddr' field with a broadcast address as is suggested in
[1] (Section 8, sixth paragraph).
The value of the 'hops' field MUST be incremented.
All other BOOTP fields MUST be preserved intact.
At this point, the request is relayed to its new destination (or
destinations). This destination MUST be configurable. Further, this
destination configuration SHOULD be independent of the destination
configuration for any other so-called "broadcast forwarders" (e.g.,
for the UDP-based TFTP, DNS, Time, etc. protocols).
DISCUSSION:
The network manager may wish the relaying destination to be an
IP unicast, multicast, broadcast, or some combination. A
configurable list of destination IP addresses provides good
flexibility. More flexible configuration schemes are
encouraged. For example, it may be desirable to send to the
limited broadcast address (255.255.255.255) on specific
physical interfaces. However, if the BOOTREQUEST message was
Wimer [Page 15]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
received as a broadcast, the relay agent MUST NOT rebroadcast
the BOOTREQUEST on the physical interface from whence it came.
A relay agent MUST use the same destination (or set of
destinations) for all BOOTREQUEST messages it relays from a
given client.
DISCUSSION:
At least one known relay agent implementation uses a round-
robin scheme to provide load balancing across multiple BOOTP
servers. Each time it receives a new BOOTREQUEST message, it
relays the message to the next BOOTP server in a list of
servers. Thus, with this relay agent, multiple consecutive
BOOTREQUEST messages from a given client will be delivered to
different servers.
Unfortunately, this well-intentioned scheme reacts badly with
DHCP [3] and perhaps other variations of the BOOTP protocol
which depend on multiple exchanges of BOOTREQUEST and BOOTREPLY
messages between clients and servers. Therefore, all
BOOTREQUEST messages from a given client MUST be relayed to the
same destination (or set of destinations).
One way to meet this requirement while providing some load-
balancing benefit is to hash the client's link-layer address
(or some other reliable client-identifying information) and use
the resulting hash value to select the appropriate relay
destination (or set of destinations). The simplest solution,
of course, is to not use a load-balancing scheme and just relay
ALL received BOOTREQUEST messages to the same destination (or
set of destinations).
When transmitting the request to its next destination, the
relay agent may set the IP Time-To-Live field to either the
default value for new datagrams originated by the relay agent,
or to the TTL of the original BOOTREQUEST decremented by (at
least) one.
DISCUSSION:
As an extra precaution against BOOTREQUEST loops, it is
preferable to use the decremented TTL from the original
BOOTREQUEST. Unfortunately, this may be difficult to do in
some implementations.
If the BOOTREQUEST has a UDP checksum (i.e., the UDP checksum
is non-zero), the checksum must be recalculated before
Wimer [Page 16]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
transmitting the request.
4.1.2 BOOTREPLY Messages
BOOTP relay agents relay BOOTREPLY messages only to BOOTP clients.
It is the responsibility of BOOTP servers to send BOOTREPLY messages
directly to the relay agent identified in the 'giaddr' field.
Therefore, a relay agent may assume that all BOOTREPLY messages it
receives are intended for BOOTP clients on its directly-connected
networks.
When a relay agent receives a BOOTREPLY message, it should examine
the BOOTP 'giaddr', 'yiaddr', 'chaddr', 'htype', and 'hlen' fields.
These fields should provide adequate information for the relay agent
to deliver the BOOTREPLY message to the client.
The 'giaddr' field can be used to identify the logical interface from
which the reply must be sent (i.e., the host or router interface
connected to the same network as the BOOTP client). If the content
of the 'giaddr' field does not match one of the relay agent's
directly-connected logical interfaces, the BOOTREPLY messsage MUST be
silently discarded.
The 'htype', 'hlen', and 'chaddr' fields supply the link-layer
hardware type, hardware address length, and hardware address of the
client as defined in the ARP protocol [4] and the Assigned Numbers
document [6]. The 'yiaddr' field is the IP address of the client, as
assigned by the BOOTP server.
The relay agent SHOULD examine the newly-defined BROADCAST flag (see
Sections 2.2 and 3.1.1 for more information). If this flag is set to
1, the reply SHOULD be sent as an IP broadcast using the IP limited
broadcast address 255.255.255.255 as the IP destination address and
the link-layer broadcast address as the link-layer destination
address. If the BROADCAST flag is cleared (0), the reply SHOULD be
sent as an IP unicast to the IP address specified by the 'yiaddr'
field and the link-layer address specified in the 'chaddr' field. If
unicasting is not possible, the reply MAY be sent as a broadcast, in
which case it SHOULD be sent to the link-layer broadcast address
using the IP limited broadcast address 255.255.255.255 as the IP
destination address.
DISCUSSION:
The addition of the BROADCAST flag to the protocol is a
workaround to help promote interoperability with certain client
implementations.
Wimer [Page 17]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
Note that since the 'flags' field was previously defined in [1]
simply as an "unused" field, it is possible that old client or
server implementations may accidentally and unknowingly set the
new BROADCAST flag. It is actually expected that such
implementations will be rare (most implementations seem to
zero-out this field), but interactions with such
implementations must nevertheless be considered. If an old
client or server does set the BROADCAST flag to 1 incorrectly,
conforming relay agents will generate broadcast BOOTREPLY
messages to the corresponding client. The BOOTREPLY messages
should still properly reach the client, at the cost of one
(otherwise unnecessary) additional broadcast. This, however,
is no worse than a server or relay agent which always
broadcasts its BOOTREPLY messages.
Older client or server implementations which accidentally set
the BROADCAST flag SHOULD be corrected to properly comply with
this newer specification.
All BOOTP fields MUST be preserved intact. The relay agent
MUST NOT modify any BOOTP field of the BOOTREPLY message when
relaying it to the client.
The reply MUST have its UDP destination port set to BOOTPC
(68).
If the BOOTREPLY has a UDP checksum (i.e., the UDP checksum is
non-zero), the checksum must be recalculated before
transmitting the reply.
5. BOOTP Server Behavior
This section provides clarifications on the behavior of BOOTP
servers.
5.1 Reception of BOOTREQUEST Messages
All received UDP messages whose UDP destination port number is BOOTPS
(67) are considered for processing by the BOOTP server.
Hosts and routers are usually required to silently discard incoming
datagrams containing illegal IP source addresses. This is generally
known as "Martian address filtering." One of these illegal addresses
is 0.0.0.0 (or actually anything on network 0). However, hosts or
routers which support a BOOTP server MUST accept for local delivery
to the server BOOTREQUEST messages whose IP source address is
0.0.0.0. BOOTREQUEST messages from legal IP source addresses MUST
also be accepted.
Wimer [Page 18]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
A BOOTP server MUST silently discard any received UDP messages whose
UDP destination port number is BOOTPC (68).
DISCUSSION:
There should be no need for a BOOTP server to process messages
addressed to the BOOTPC port. Careful reading of the original
BOOTP specification [1] will show this.
The consistency checks specified in Section 2.1 SHOULD be
performed by the BOOTP server. BOOTP messages not meeting
these consistency checks MUST be silently discarded.
5.2 Use of the 'secs' field
When the BOOTP server receives a BOOTREQUEST message, it MAY use the
value of the 'secs' (seconds since client began booting) field of the
request as a factor in deciding whether and/or how to reply to the
request.
DISCUSSION:
To date, this feature of the BOOTP protocol has not necessarily
been shown to be useful. See Section 3.2 for a discussion.
5.3 Use of the 'ciaddr' field
There have been various client interpretations of the 'ciaddr' field
for which Section 3.3 should be consulted. A BOOTP server SHOULD be
prepared to deal with these varying interpretations. In general, the
'ciaddr' field SHOULD NOT be trusted as a sole key in identifying a
client; the contents of the 'ciaddr', 'chaddr', 'htype', and 'hlen'
fields, and probably other information (perhaps in the 'file' and
'vend' fields) SHOULD all be considered together in deciding how to
respond to a given client.
BOOTP servers SHOULD preserve the contents of the 'ciaddr' field in
BOOTREPLY messages; the contents of 'ciaddr' in a BOOTREPLY message
SHOULD exactly match the contents of 'ciaddr' in the corresponding
BOOTREQUEST message.
DISCUSSION:
It has been suggested that a client may wish to use the
contents of 'ciaddr' to further verify that a particular
BOOTREPLY message was indeed intended for it.
Wimer [Page 19]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
5.4 Strategy for Delivery of BOOTREPLY Messages
Once the BOOTP server has created an appropriate BOOTREPLY message,
that BOOTREPLY message must be properly delivered to the client.
The server SHOULD first check the 'ciaddr' field. If the 'ciaddr'
field is non-zero, the BOOTREPLY message SHOULD be sent as an IP
unicast to the IP address identified in the 'ciaddr' field. The UDP
destination port MUST be set to BOOTPC (68). However, the server
MUST be aware of the problems identified in Section 3.3. The server
MAY choose to ignore the 'ciaddr' field and act as if the 'ciaddr'
field contains 0.0.0.0 (and thus continue with the rest of the
delivery algorithm below).
The server SHOULD next check the 'giaddr' field. If this field is
non-zero, the server SHOULD send the BOOTREPLY as an IP unicast to
the IP address identified in the 'giaddr' field. The UDP destination
port MUST be set to BOOTPS (67). This action will deliver the
BOOTREPLY message directly to the BOOTP relay agent closest to the
client; the relay agent will then perform the final delivery to the
client. If the BOOTP server has prior knowledge that a particular
client cannot receive unicast BOOTREPLY messages (e.g., the network
manager has explicitly configured the server with such knowledge),
the server MAY set the newly-defined BROADCAST flag to indicate that
relay agents SHOULD broadcast the BOOTREPLY message to the client.
Otherwise, the server MUST preserve the state of the BROADCAST flag
so that the relay agent can correctly act upon it.
If the 'giaddr' field is set to 0.0.0.0, then the client resides on
one of the same networks as the BOOTP server. The server SHOULD
examine the newly-defined BROADCAST flag (see Sections 2.2, 3.1.1 and
4.1.2 for more information). If this flag is set to 1 or the server
has prior knowledge that the client is unable to receive unicast
BOOTREPLY messages, the reply SHOULD be sent as an IP broadcast using
the IP limited broadcast address 255.255.255.255 as the IP
destination address and the link-layer broadcast address as the
link-layer destination address. If the BROADCAST flag is cleared
(0), the reply SHOULD be sent as an IP unicast to the IP address
specified by the 'yiaddr' field and the link-layer address specified
in the 'chaddr' field. If unicasting is not possible, the reply MAY
be sent as a broadcast in which case it SHOULD be sent to the link-
layer broadcast address using the IP limited broadcast address
255.255.255.255 as the IP destination address. In any case, the UDP
destination port MUST be set to BOOTPC (68).
Wimer [Page 20]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
DISCUSSION:
The addition of the BROADCAST flag to the protocol is a
workaround to help promote interoperability with certain client
implementations.
The following table summarizes server delivery decisions for
BOOTREPLY messages based upon information in BOOTREQUEST
messages:
BOOTREQUEST fields BOOTREPLY values for UDP, IP, link-layer
+-----------------------+-----------------------------------------+
| 'ciaddr' 'giaddr' B | UDP dest IP destination link dest |
+-----------------------+-----------------------------------------+
| non-zero X X | BOOTPC (68) 'ciaddr' normal |
| 0.0.0.0 non-zero X | BOOTPS (67) 'giaddr' normal |
| 0.0.0.0 0.0.0.0 0 | BOOTPC (68) 'yiaddr' 'chaddr' |
| 0.0.0.0 0.0.0.0 1 | BOOTPC (68) 255.255.255.255 broadcast |
+-----------------------+-----------------------------------------+
B = BROADCAST flag
X = Don't care
normal = determine from the given IP destination using normal
IP routing mechanisms and/or ARP as for any other
normal datagram
Acknowledgements
The author would like to thank Gary Malkin for his contribution of
the "BOOTP over IEEE 802.5 Token Ring Networks" section, and Steve
Deering for his observations on the problems associated with the
'giaddr' field.
Ralph Droms and the many members of the IETF Dynamic Host
Configuration and Router Requirements working groups provided ideas
for this memo as well as encouragement to write it.
Philip Almquist and David Piscitello offered many helpful suggestions
for improving the clarity, accuracy, and organization of this memo.
These contributions are graciously acknowledged.
Wimer [Page 21]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
References
[1] Croft, B., and J. Gilmore, "Bootstrap Protocol (BOOTP)", RFC 951,
Stanford University and Sun Microsystems, September 1985.
[2] Reynolds, J., "BOOTP Vendor Information Extensions", RFC 1497,
USC/Information Sciences Institute, August 1993. This RFC is
occasionally reissued with a new number. Please be sure to
consult the latest version.
[3] Droms, R., "Dynamic Host Configuration Protocol", RFC 1541,
Bucknell University, October 1993.
[4] Plummer, D., "An Ethernet Address Resolution Protocol", STD 37,
RFC 826, MIT, November 1982.
[5] Deering, S., "ICMP Router Discovery Messages", RFC 1256, Xerox
PARC, September 1991.
[6] Reynolds, J., and J. Postel, "Assigned Numbers", STD 2, RFC 1340,
USC/Information Sciences Institute, July, 1992. This RFC is
periodically reissued with a new number. Please be sure to
consult the latest version.
Wimer [Page 22]
RFC 1542 Clarifications and Extensions for BOOTP October 1993
Security Considerations
There are many factors which make BOOTP in its current form quite
insecure. BOOTP is built directly upon UDP and IP which are as yet
inherently insecure themselves. Furthermore, BOOTP is generally
intended to make maintenance of remote and/or diskless hosts easier.
While perhaps not impossible, configuring such hosts with passwords or
keys may be difficult and inconvenient. This makes it difficult to
provide any form of reasonable authentication between servers and
clients.
Unauthorized BOOTP servers may easily be set up. Such servers can
then send false and potentially disruptive information to clients such
as incorrect or duplicate IP addresses, incorrect routing information
(including spoof routers, etc.), incorrect domain nameserver addresses
(such as spoof nameservers), and so on. Clearly, once this "seed"
mis-information is planted, an attacker can further compromise the
affected systems.
Unauthorized BOOTP relay agents may present some of the same problems
as unauthorized BOOTP servers.
Malicious BOOTP clients could masquerade as legitimate clients and
retrieve information intended for those legitimate clients. Where
dynamic allocation of resources is used, a malicious client could
claim all resources for itself, thereby denying resources to
legitimate clients.
Author's Address
Walt Wimer
Network Development
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213-3890
Phone: (412) 268-6252
EMail: Walter.Wimer@CMU.EDU
Wimer [Page 23]