ndp(7P)ndp(7P)NAMEndp - Neighbor Discovery Protocol, NDP
DESCRIPTION
Neighbor Discovery Protocol (NDP) is a protocol used by hosts and
routers to:
1. Find the link-layer address of the neighbors known to be
attached to the same link.
2. Find the neighboring routers that are willing to forward
packets on their behalf.
3. Actively keep track of which neighbors are reachable and
which are not.
4. Search for alternate routers when the path to a router
fails.
To accomplish the above mentioned tasks, NDP defines the following pro‐
cesses:
1. Router and Prefix Discovery
Router discovery is a process through which hosts locate the
neighboring routers and learn prefix plus other parameters nec‐
essary for address autoconfiguration.
Prefix discovery is used by the hosts to learn the range of IPv6
addresses that reside on-link and can be reached without going
through a router.
Routers send Router Advertisements which will make the hosts
treat them as the default routers. The Router Advertisements
will also contain prefix information options that will identify
the range of IPv6 addresses that are on-link (Subnet prefix).
2. Router and Host Requirements
Router requirements in NDP specify a set of rules for host to
act as a router. These rules include:
· Router configuration variables.
These configuration variables include intervals between suc‐
cessive unsolicited router advertisements, etc.
· How to make an interface an advertising interface.
When an interface is made an advertising interface, it means
that the node is going to send periodic router advertisements
and is willing to forward packets on behalf of hosts on that
link.
· Message content for router advertisements.
A router will send periodic as well as solicited Router
Advertisements on an advertising interface. NDP specifies
the format of these messages.
· Sending unsolicited router advertisements.
Apart from sending solicited router advertisements in
response to router solicitations, routers can send unso‐
licited router advertisements. For example, unsolicited
router advertisements can be sent to expire a prefix or to
advertise a new prefix, etc.
· Stopping router advertisements on an interface.
A router can stop advertising prefixes on an interface. This
can happen due to system management decisions when a router
may be stopped from being one. NDP specifies what the router
should be doing under these circumstances.
· Processing router solicitation messages.
Hosts as part of the stateless autoconfiguration process will
send Router Solicitations. Routers should respond to such
solicitations with a router advertisement.
· Steps to be taken when the link-local address for the router
changes.
Normally the link-local address of a Router should not
change. However, NDP still defines the steps should be taken
by the router when its link-local address changes for any of
its interfaces.
Host requirements are a set of rules that apply for a IPv6 host.
They are:
· IPv6 variables that have to be maintained.
These variables include the time between retransmissions of
neighbor solicitations, link MTU for each interface, etc.
· Processing router advertisements.
This rule discusses what actions should be taken on receipt
of router advertisements.
· Timing out prefixes and default routers.
Whenever routers send router advertisements, they include the
lifetime of the router as well as the prefixes that they
advertise. NDP specifies what actions the host should take
when these lifetimes expire.
· Selecting a default router.
When there is more than one router in the link, the default
router selection algorithm comes into effect. This algorithm
helps select the default router based on factors like reacha‐
bility, etc.
· Sending a router solicitation.
When an interface is enabled, a host need not wait for the
unsolicited router advertisement. Instead, it can send a
router solicitation and get a router advertisement as a
response. This will help in receiving the default router and
prefix information as soon as the interface is enabled.
3. Algorithm for Sending a Packet
Any IPv6 host is required to maintain some data structures that
will be used by the algorithm for sending a packet. These data
structures are:
Neighbor Cache
A set of entries that will maintain IPv6 Address to link-
layer address mappings for neighbors to which a packet
has been sent recently. In addition to that it maintains
information needed for neighbor unreachability detection
like the reachability state, etc.
Destination Cache
A set of entries for hosts to whom packets have been sent
recently. This includes hosts which are both on-link and
off-link. It contains a level of indirection to the
neighbor cache.
Prefix List
This is a list of prefixes which define the set of IPv6
address that are on-link. This information is maintained
on a per interface basis. Typically this list is built
from Router Advertisements received from the router.
Default Router List
A list of routers which will forward packets on behalf of
this host. This list will again have a pointer to a
neighbor cache entry for the respective router.
A host will use the above data structures while sending a packet
to a host. Following is the conceptual algorithm for sending a
packet to a unicast destination.
a. Before a packet is sent out, the next hop should be
determined. Normally, next hop determination is not done
on all packets. The results of a next hop determination
are stored in the destination cache. The host should
first check the destination cache for any entry that
matches with the current destination address. If it
finds a match, then it proceeds to step c, below.
b. If there is no entry for the destination in the destina‐
tion cache, a longest prefix match is made with all pre‐
fixes in the prefix list. If there is a match, the des‐
tination is determined to be on-link and the destination
address will be considered as the next hop. Otherwise,
the next hop is determined from the routing table.
c. Once the next hop is determined, the address resolution
process and neighbor unreachability detection are done
for the next hop. This process is explained in the next
section.
d. Once the neighbor is known to be reachable, the packet is
sent to that destination.
4. Address Resolution and Neighbor Unreachability Detection
Address resolution is a process used to determine the link-layer
address of a neighbor. The IPv6 Address to link-layer address
mapping found through this process is cached in the Neighbor
Cache. Following are the steps involved in Address Resolution.
a. First, the neighbor cache is checked for an entry which
matches the current destination address. If the entry is
not present, the host sends a Neighbor Solicitation Mes‐
sage to the solicited-node multicast group. This multi‐
cast address is derived based on the destination IPv6
address and all nodes with the particular IPv6 address
are required to join that group.
b. If a host with the specified IPv6 address is present in
the network, it will reply this solicitation with a
Neighbor Advertisement Message.
c. On receiving the Neighbor Advertisement, the node will
search for an entry in the neighbor cache for the
sender's IPv6 address. A new entry is created in the
neighbor cache and the reachability flag is set to REACH‐
ABLE.
Once the Address resolution is completed, neighbor unreachabil‐
ity detection will be performed. This process depends on the
reachability field of the neighbor cache. An entry in the
neighbor cache can have any of the following states:
INCOMPLETE The address resolution is in progress and the
link-layer address of the destination is yet to
be determined.
REACHABLE The destination is reachable until recently.
STALE The destination is no longer known to be reach‐
able, but reachability detection need not be made
until a packet has to be sent to that destina‐
tion.
DELAY This state is an optimization that gives addi‐
tional time for the upper layer protocols to pro‐
vide the reachability confirmation.
PROBE A reachability confirmation is actively requested
by repeatedly sending Neighbor Solicitations.
During neighbor unreachability detection, the node checks for
the state in the neighbor cache. If the state for the destina‐
tion is REACHABLE, the packet is sent. Otherwise, the following
steps are taken:
a. When an address resolution is made on a destination, an
entry is created in the neighbor cache for that destina‐
tion and the reachability state will be set to INCOM‐
PLETE. If the address resolution fails, the entry is
deleted.
b. When the address resolution passes, the entry will be
filled with the destination's link-layer address and the
state will be set to REACHABLE.
c. There is a timer maintained called the Reachability timer
which will expire the state of an entry in the neighbor
cache. Once this timer expires, the reachability state
changes from REACHABLE to STALE.
d. When a packet is being sent to a destination whose state
is STALE in the neighbor cache, the node sets the state
to DELAY and starts a timer associated with that state.
By the time the timer expires if the node received reach‐
ability confirmation, the state is set to REACHABLE.
Otherwise, it is set to PROBE.
e. Once the entry's state is in PROBE, the node sends uni‐
cast neighbor solicitations to the link-layer address
specified in the entry. If it receives a neighbor adver‐
tisement in response the state is set to REACHABLE. This
solicitation will be sent repeatedly; the maximum number
of times is configurable. If the reachability confirma‐
tion is not received after maximum solicitations, the
entry is deleted from the neighbor cache and the address
resolution is done again.
Note: Entries in the neighbor cache can also be created as a
result of node receiving unsolicited Neighbor Advertisements,
Router Advertisements and Router Solicitations, etc. However,
for the entry created under these circumstances the reachability
state will always be set to STALE.
5. Redirect Function
A router will send a host a redirect message when it finds that
there is a better next-hop router on the same link. This is a
requirement for a router.
On receiving a router redirect message, a host should update its
destination cache with the new next hop address.
AUTHOR
NDP was developed by the IPng Working Group of the Internet Engineering
Task Force.
SEE ALSOifconfig(1M), ndp(1M), ip6(7P), lan(7).
Neighbor Discovery for IPv6, RFC2461, T. Narten et al. NDP Neighbor
Discovery Protocol
ndp(7P)
