ROUTE(4) NetBSD Programmer's Manual ROUTE(4)
NAME
route - kernel packet forwarding database
SYNOPSIS
#include <sys/socket.h> #include <net/if.h> #include <net/route.h> int socket(PF_ROUTE, SOCK_RAW, int family);
DESCRIPTION
UNIX provides some packet routing facilities. The kernel maintains a routing information database, which is used in selecting the appropriate network interface when transmitting packets. A user process (or possibly multiple co-operating processes) maintains this database by sending messages over a special kind of socket. This supplants fixed size ioctl(2)'s used in earlier releases. Routing table changes may only be carried out by the super user. The operating system may spontaneously emit routing messages in response to external events, such as receipt of a re-direct, or failure to locate a suitable route for a request. The message types are described in greater detail below. Routing database entries come in two flavors: for a specific host, or for all hosts on a generic subnetwork (as specified by a bit mask and value under the mask. The effect of wildcard or default route may be achieved by using a mask of all zeros, and there may be hierarchical routes. When the system is booted and addresses are assigned to the network in- terfaces, each protocol family installs a routing table entry for each interface when it is ready for traffic. Normally the protocol specifies the route through each interface as a ``direct'' connection to the desti- nation host or network. If the route is direct, the transport layer of a protocol family usually requests the packet be sent to the same host specified in the packet. Otherwise, the interface is requested to ad- dress the packet to the gateway listed in the routing entry (i.e. the packet is forwarded). When routing a packet, the kernel will attempt to find the most specific route matching the destination. (If there are two different mask and value-under-the-mask pairs that match, the more specific is the one with more bits in the mask. A route to a host is regarded as being supplied with a mask of as many ones as there are bits in the destination). If no entry is found, the destination is declared to be unreachable, and a routing-miss message is generated if there are any listers on the routing control socket described below. A wildcard routing entry is specified with a zero destination address value, and a mask of all zeroes. Wildcard routes will be used when the system fails to find other routes matching the destination. The combina- tion of wildcard routes and routing redirects can provide an economical mechanism for routing traffic. One opens the channel for passing routing control messages by using the socket call shown in the synopsis above: The family parameter may be AF_UNSPEC which will provide routing informa- tion for all address families, or can be restricted to a specific address family by specifying which one is desired. There can be more than one routing socket open per system. Messages are formed by a header followed by a small number of sockaddrs (now variable length particularly in the ISO case), interpreted by posi- tion, and delimited by the new length entry in the sockaddr. An example of a message with four addresses might be an ISO redirect: Destination, Netmask, Gateway, and Author of the redirect. The interpretation of which address are present is given by a bit mask within the header, and the sequence is least significant to most significant bit within the vec- tor. Any messages sent to the kernel are returned, and copies are sent to all interested listeners. The kernel will provide the process ID for the sender, and the sender may use an additional sequence field to distin- guish between outstanding messages. However, message replies may be lost when kernel buffers are exhausted. The kernel may reject certain messages, and will indicate this by filling in the rtm_errno field. The routing code returns EEXIST if requested to duplicate an existing entry, ESRCH if requested to delete a non-existent entry, or ENOBUFS if insufficient resources were available to install a new route. In the current implementation, all routing processes run lo- cally, and the values for rtm_errno are available through the normal errno mechanism, even if the routing reply message is lost. A process may avoid the expense of reading replies to its own messages by issuing a setsockopt(2) call indicating that the SO_USELOOPBACK option at the SOL_SOCKET level is to be turned off. A process may ignore all mes- sages from the routing socket by doing a shutdown(2) system call for fur- ther input. If a route is in use when it is deleted, the routing entry will be marked down and removed from the routing table, but the resources associated with it will not be reclaimed until all references to it are released. User processes can obtain information about the routing entry to a spe- cific destination by using a RTM_GET message, or by reading the /dev/kmem device, or by calling sysctl(3). The messages are: #define RTM_ADD 0x1 /* Add Route */ #define RTM_DELETE 0x2 /* Delete Route */ #define RTM_CHANGE 0x3 /* Change Metrics, Flags, or Gateway */ #define RTM_GET 0x4 /* Report Information */ #define RTM_LOSING 0x5 /* Kernel Suspects Partitioning */ #define RTM_REDIRECT 0x6 /* Told to use different route */ #define RTM_MISS 0x7 /* Lookup failed on this address */ #define RTM_RESOLVE 0xb /* request to resolve dst to LL addr */ #define RTM_NEWADDR 0xc /* address being added to iface */ #define RTM_DELADDR 0xd /* address being removed from iface */ #define RTM_IFINFO 0xe /* iface going up/down etc. */ A message header consists of one of the following: struct rt_msghdr { u_short rtm_msglen; /* to skip over non-understood messages */ u_char rtm_version; /* future binary compatibility */ u_char rtm_type; /* message type */ u_short rtm_index; /* index for associated ifp */ int rtm_flags; /* flags, incl kern & message, e.g. DONE */ int rtm_addrs; /* bitmask identifying sockaddrs in msg */ pid_t rtm_pid; /* identify sender */ int rtm_seq; /* for sender to identify action */ int rtm_errno; /* why failed */ int rtm_use; /* from rtentry */ u_long rtm_inits; /* which metrics we are initializing */ struct rt_metrics rtm_rmx; /* metrics themselves */ }; struct if_msghdr { u_short ifm_msglen; /* to skip over non-understood messages */ u_char ifm_version; /* future binary compatability */ u_char ifm_type; /* message type */ int ifm_addrs; /* like rtm_addrs */ int ifm_flags; /* value of if_flags */ u_short ifm_index; /* index for associated ifp */ struct if_data ifm_data; /* statistics and other data about if */ }; struct ifa_msghdr { u_short ifam_msglen; /* to skip over non-understood messages */ u_char ifam_version; /* future binary compatability */ u_char ifam_type; /* message type */ int ifam_addrs; /* like rtm_addrs */ int ifam_flags; /* value of ifa_flags */ u_short ifam_index; /* index for associated ifp */ int ifam_metric; /* value of ifa_metric */ }; The RTM_IFINFO message uses a if_msghdr header, the RTM_NEWADDR and RTM_DELADDR messages use a ifa_msghdr header, and all other messages use the rt_msghdr header. The metrics structure is: struct rt_metrics { u_long rmx_locks; /* Kernel must leave these values alone */ u_long rmx_mtu; /* MTU for this path */ u_long rmx_hopcount; /* max hops expected */ u_long rmx_expire; /* lifetime for route, e.g. redirect */ u_long rmx_recvpipe; /* inbound delay-bandwith product */ u_long rmx_sendpipe; /* outbound delay-bandwith product */ u_long rmx_ssthresh; /* outbound gateway buffer limit */ u_long rmx_rtt; /* estimated round trip time */ u_long rmx_rttvar; /* estimated rtt variance */ u_long rmx_pksent; /* packets sent using this route */ }; Flags include the values: #define RTF_UP 0x1 /* route usable */ #define RTF_GATEWAY 0x2 /* destination is a gateway */ #define RTF_HOST 0x4 /* host entry (net otherwise) */ #define RTF_REJECT 0x8 /* host or net unreachable */ #define RTF_DYNAMIC 0x10 /* created dynamically (by redirect) */ #define RTF_MODIFIED 0x20 /* modified dynamically (by redirect) */ #define RTF_DONE 0x40 /* message confirmed */ #define RTF_MASK 0x80 /* subnet mask present */ #define RTF_CLONING 0x100 /* generate new routes on use */ #define RTF_XRESOLVE 0x200 /* external daemon resolves name */ #define RTF_LLINFO 0x400 /* generated by ARP or ESIS */ #define RTF_STATIC 0x800 /* manually added */ #define RTF_BLACKHOLE 0x1000 /* just discard pkts (during updates) */ #define RTF_PROTO2 0x4000 /* protocol specific routing flag */ #define RTF_PROTO1 0x8000 /* protocol specific routing flag */ Specifiers for metric values in rmx_locks and rtm_inits are: #define RTV_MTU 0x1 /* init or lock _mtu */ #define RTV_HOPCOUNT 0x2 /* init or lock _hopcount */ #define RTV_EXPIRE 0x4 /* init or lock _expire */ #define RTV_RPIPE 0x8 /* init or lock _recvpipe */ #define RTV_SPIPE 0x10 /* init or lock _sendpipe */ #define RTV_SSTHRESH 0x20 /* init or lock _ssthresh */ #define RTV_RTT 0x40 /* init or lock _rtt */ #define RTV_RTTVAR 0x80 /* init or lock _rttvar */ Specifiers for which addresses are present in the messages are: #define RTA_DST 0x1 /* destination sockaddr present */ #define RTA_GATEWAY 0x2 /* gateway sockaddr present */ #define RTA_NETMASK 0x4 /* netmask sockaddr present */ #define RTA_GENMASK 0x8 /* cloning mask sockaddr present */ #define RTA_IFP 0x10 /* interface name sockaddr present */ #define RTA_IFA 0x20 /* interface addr sockaddr present */ #define RTA_AUTHOR 0x40 /* sockaddr for author of redirect */ #define RTA_BRD 0x80 /* for NEWADDR, broadcast or p-p dest addr */
SEE ALSO
socket(2), sysctl(3) 4.4BSD April 19, 1994 4
Powered by man-cgi (2024-08-26). Maintained for NetBSD by Kimmo Suominen. Based on man-cgi by Panagiotis Christias.