ARP
NAME DESCRIPTION IOCTLS SYSCTLS BUGS VERSIONS SEE ALSO
arp - Linux ARP kernel module.
This kernel protocol module implements the Address Resolution Protocol defined in RFC 826. It is used to convert between Layer2 hardware addresses and IPv4 protocol addresses on directly connected networks. The user normally doesn't interact directly with this module except to configure it; instead it provides a service for other protocols in the kernel.
A user process can receive ARP packets by using packet(7) sockets. There is also a mechanism for managing the ARP cache in user-space by using netlink(7) sockets. The ARP table can also be controlled via ioctl (2) on any PF_INET socket.
The ARP module maintains a cache of mappings between hardware addresses and protocol addresses. The cache has a limited size so old and less frequently used entries are garbage-collected. Entries which are marked as permanent are never deleted by the garbage-collector. The cache can be directly manipulated by the use of ioctls and its behaviour can be tuned by the sysctls defined below.
When there is no positive feedback for an existing mapping after some time (see the sysctls below) a neighbour cache entry is considered stale. Positive feedback can be gotten from a higher layer; for example from a successful TCP ACK. Other protocols can signal forward progress using the MSG_CONFIRM flag to sendmsg(2). When there is no forward progress ARP tries to reprobe. It first tries to ask a local arp daemon app_solicit times for an updated MAC address. If that fails and an old MAC address is known an unicast probe is send ucast_solicit times. If that fails too it will broadcast a new ARP request to the network. Requests are only send when there is data queued for sending.
Linux will automatically add a non-permanent proxy arp entry when it receives a request for an address it forwards to and proxy arp is enabled on the receiving interface. When there is a reject route for the target no proxy arp entry is added.
Three ioctls are available on all PF_INET sockets. They take a pointer to a struct arpreq as their parameter.
struct arpreq { struct sockaddr arp_pa; /* protocol address / struct sockaddr arp_ha; / hardware address / int arp_flags; / flags / struct sockaddr arp_netmask; / netmask of protocol address */ char arp_dev[16?; }; SIOCSARP, SIOCDARP and SIOCGARP respectively set, delete and get an ARP mapping. Setting CAP_NET_ADMIN capability or an effective UID of 0.
arp_pa must be an AF_INET socket and arp_ha must have the same type as the device which is specified in arp_dev. arp_dev is a zero-terminated string which names a device.
If the ATF_NETMASK flag is set, then arp_netmask should be valid. Linux 2.2 does not support proxy network ARP entries, so this should be set to 0xffffffff, or 0 to remove an existing proxy arp entry. ATF_USETRAILERS is obsolete and should not be used.
ARP supports a sysctl interface to configure parameters on a global or per-interface basis. The sysctls can be accessed by reading or writing the /proc/sys/net/ipv4/neigh// files or with the sysctl(2) interface. Each interface in the system has its own directory in /proc/sys/net/ipv4/neigh/. The setting in the `default' directory is used for all newly created devices. Unless otherwise specified time related sysctls are specified in seconds.
anycast_delay
The maximum number of jiffies to delay before replying to a IPv6 neighbour solicitation message. Anycast support is not yet implemented. Defaults to 1 second.
app_solicit
The maximum number of probes to send to the user space ARP daemon via netlink before dropping back to multicast probes (see mcast_solicit). Defaults to 0.
base_reachable_time
Once a neighbour has been found, the entry is considered to be valid for at least a random value between base_reachable_time/2 and 3*base_reachable_time/2. An entry's validity will be extended if it receives positive feedback from higher level protocols. Defaults to 30 seconds.
delay_first_probe_time
Delay before first probe after it has been decided that a neighbour is stale. Defaults to 5 seconds.
gc_interval
How frequently the garbage collector for neighbour entries should attempt to run. Defaults to 30 seconds.
gc_stale_time
Determines how often to check for stale neighbour entries. When a neighbour entry is considered stale it is resolved again before sending data to it. Defaults to 60 seconds.
gc_thresh1
The minimum number of entries to keep in the ARP cache. The garbage collector will not run if there are fewer than this number of entries in the cache. Defaults to 128.
gc_thresh2
The soft maximum number of entries to keep in the ARP cache. The garbage collector will allow the number of entries to exceed this for 5 seconds before collection will be performed. Defaults to 512.
gc_thresh3
The hard maximum number of entries to keep in the ARP cache. The garbage collector will always run if there are more than this number of entries in the cache. Defaults to 1024.
locktime
The minimum number of jiffies to keep an ARP entry in the cache. This prevents ARP cache thrashing if there is more than one potential mapping (generally due to network misconfiguration). Defaults to 1 second.
mcast_solicit
The maximum number of attempts to resolve an address by multicast/broadcast before marking the entry as unreachable. Defaults to 3.
proxy_delay
When an ARP request for a known proxy-ARP address is received, delay up to proxy_delay jiffies before replying. This is used to prevent network flooding in some cases. Defaults to 0.8 seconds.
proxy_qlen
The maximum number of packets which may be queued to proxy-ARP addresses. Defaults to 64.
retrans_time
The number of jiffies to delay before retransmitting a request. Defaults to 1 second.
ucast_solicit
The maximum number of attempts to send unicast probes before asking the ARP daemon (see app_solicit). Defaults to 3.
unres_qlen
The maximum number of packets which may be queued for each unresolved address by other network layers. Defaults to 3.
Some timer settings are specified in jiffies, which is architecture related. On the Alpha a jiffy is 1/1024 of a second, on most other architectures it is 1/100s.
There is no way to signal positive feedback from user space. This means connection oriented protocols implemented in user space will generate excessive ARP traffic, because ndisc will regularly reprobe the MAC address. The same problem applies for some kernel protocols (e.g. NFS over UDP).
This man page mashes IPv4 specific and shared between IPv4 and IPv6 functionality together.
The struct arpreq changed in Linux 2.0 to include the arp_dev member and the ioctl numbers changed at the same time. Support for the old ioctls was dropped in Linux 2.2.
Support for proxy arp entries for networks (netmask not equal 0xffffffff) was dropped in Linux 2.2. It is replaced by automatic proxy arp setup by the kernel for all reachable hosts on other interfaces (when forwarding and proxy arp is enabled for the interface).
The neigh/* sysctls did not exist before Linux 2.2.
RFC826 for a description of ARP. RFC2461 for a description of IPv6 neighbour discovery and the base algorithms used.
Linux 2.2+ IPv4 ARP uses the IPv6 algorithms when applicable.
