bird/proto/rip/rip.c
2000-06-07 14:06:37 +00:00

1013 lines
27 KiB
C

/*
* Rest in pieces - RIP protocol
*
* Copyright (c) 1998, 1999 Pavel Machek <pavel@ucw.cz>
*
* Can be freely distributed and used under the terms of the GNU GPL.
*
FIXME: IpV6 support: packet size
FIXME: (nonurgent) IpV6 support: receive "route using" blocks
FIXME: (nonurgent) IpV6 support: generate "nexthop" blocks
next hops are only advisory, and they are pretty ugly in IpV6.
I suggest just forgetting about them.
FIXME (nonurgent): fold rip_connection into rip_interface?
FIXME: (nonurgent) allow bigger frequencies than 1 regular update in 6 seconds (?)
FIXME: propagation of metric=infinity into main routing table may or may not be good idea.
*/
/**
* DOC: Routing Information Protocol
*
* RIP is a pretty simple protocol, so about a half of its code is interface
* with the core.
*
* We maintain our own linked list of &rip_entry structures -- it serves
* as our small routing table. RIP never adds to this linked list upon
* packet reception; instead, it lets the core know about data from the packet
* and waits for the core to call rip_rte_notify().
*
* Within rip_tx(), the list is
* walked and a packet is generated using rip_tx_prepare(). This gets
* tricky because we may need to send more than one packet to one
* destination. Struct &rip_connection is used to hold context information such as how
* many of &rip_entry's we have already sent and it's also used to protect
* against two concurrent sends to one destination. Each &rip_interface has
* at most one &rip_connection.
*
* We are not going to honor requests for sending part of
* routing table. That would need to turn split horizon off etc.
*
* About triggered updates, RFC says: when a triggered update was sent,
* don't send a new one for something between 1 and 5 seconds (and send one
* after that). We do something else: each 5 seconds,
* we look for any changed routes and broadcast them.
*/
#undef LOCAL_DEBUG
#include "nest/bird.h"
#include "nest/iface.h"
#include "nest/protocol.h"
#include "nest/route.h"
#include "lib/socket.h"
#include "lib/resource.h"
#include "lib/lists.h"
#include "lib/timer.h"
#include "lib/string.h"
#include "rip.h"
#include <assert.h>
#define P ((struct rip_proto *) p)
#define P_CF ((struct rip_proto_config *)p->cf)
#define E ((struct rip_entry *) e)
#define TRACE(level, msg, args...) do { if (p->debug & level) { log(L_TRACE "%s: " msg, p->name , ## args); } } while(0)
static struct rip_interface *new_iface(struct proto *p, struct iface *new, unsigned long flags, struct iface_patt *patt);
#define P_NAME p->name
/*
* Output processing
*
* This part is responsible for getting packets out to the network.
*/
static void
rip_tx_err( sock *s, int err )
{
struct rip_connection *c = s->data;
struct proto *p = c->proto;
log( L_ERR "%s: Unexpected error at rip transmit: %M", P_NAME, err );
}
/*
* rip_tx_prepare:
* @e: rip entry that needs to be translated to form suitable for network
* @b: block to be filled
*
* Fill one rip block with info that needs to go to the network. Handle
* nexthop and split horizont correctly. (Next hop is ignored for IPv6,
* that could be fixed but it is not real problem).
*/
static int
rip_tx_prepare(struct proto *p, ip_addr daddr, struct rip_block *b, struct rip_entry *e, struct rip_interface *rif, int pos )
{
int metric;
DBG( "." );
b->tag = htons( e->tag );
b->network = e->n.prefix;
metric = e->metric;
if (neigh_connected_to(p, &e->whotoldme, rif->iface)) {
DBG( "(split horizon)" );
metric = P_CF->infinity;
}
#ifndef IPV6
b->family = htons( 2 ); /* AF_INET */
b->netmask = ipa_mkmask( e->n.pxlen );
ipa_hton( b->netmask );
if (neigh_connected_to(p, &e->nexthop, rif->iface))
b->nexthop = e->nexthop;
else
b->nexthop = IPA_NONE;
ipa_hton( b->nexthop );
b->metric = htonl( metric );
#else
b->pxlen = e->n.pxlen;
b->metric = metric; /* it is u8 */
#endif
ipa_hton( b->network );
return pos+1;
}
/*
* rip_tx - send one rip packet to the network
*/
static void
rip_tx( sock *s )
{
struct rip_interface *rif = s->data;
struct rip_connection *c = rif->busy;
struct proto *p = c->proto;
struct rip_packet *packet = (void *) s->tbuf;
int i, packetlen;
int maxi, nullupdate = 1;
DBG( "Sending to %I\n", s->daddr );
do {
if (c->done)
goto done;
DBG( "Preparing packet to send: " );
packet->heading.command = RIPCMD_RESPONSE;
packet->heading.version = RIP_V2;
packet->heading.unused = 0;
i = !!P_CF->authtype;
#ifndef IPV6
maxi = ((P_CF->authtype == AT_MD5) ? PACKET_MD5_MAX : PACKET_MAX);
#else
maxi = 5; /* We need to have at least reserve of one at end of packet */
#endif
FIB_ITERATE_START(&P->rtable, &c->iter, z) {
struct rip_entry *e = (struct rip_entry *) z;
if (!rif->triggered || (!(e->updated < now-5))) {
nullupdate = 0;
i = rip_tx_prepare( p, s->daddr, packet->block + i, e, rif, i );
if (i >= maxi) {
FIB_ITERATE_PUT(&c->iter, z);
goto break_loop;
}
}
} FIB_ITERATE_END(z);
c->done = 1;
break_loop:
packetlen = rip_outgoing_authentication(p, (void *) &packet->block[0], packet, i);
DBG( ", sending %d blocks, ", i );
if (nullupdate) {
DBG( "not sending NULL update\n" );
c->done = 1;
goto done;
}
if (ipa_nonzero(c->daddr))
i = sk_send_to( s, packetlen, c->daddr, c->dport );
else
i = sk_send( s, packetlen );
DBG( "it wants more\n" );
} while (i>0);
if (i<0) rip_tx_err( s, i );
DBG( "blocked\n" );
return;
done:
DBG( "Looks like I'm" );
c->rif->busy = NULL;
rem_node(NODE c);
mb_free(c);
DBG( " done\n" );
return;
}
/*
* rip_sendto - send whole routing table to selected destination
* @rif: interface to use. Notice that we lock interface so that at
* most one send to one interface is done.
*/
static void
rip_sendto( struct proto *p, ip_addr daddr, int dport, struct rip_interface *rif )
{
struct iface *iface = rif->iface;
struct rip_connection *c;
static int num = 0;
if (rif->busy) {
log (L_WARN "%s: Interface %s is much too slow, dropping request", P_NAME, iface->name);
return;
}
c = mb_alloc( p->pool, sizeof( struct rip_connection ));
rif->busy = c;
c->addr = daddr;
c->proto = p;
c->num = num++;
c->rif = rif;
c->dport = dport;
c->daddr = daddr;
if (c->rif->sock->data != rif)
bug("not enough send magic");
c->done = 0;
FIB_ITERATE_INIT( &c->iter, &P->rtable );
add_head( &P->connections, NODE c );
if (ipa_nonzero(daddr))
TRACE(D_PACKETS, "Sending my routing table to %I:%d on %s", daddr, dport, rif->iface->name );
else
TRACE(D_PACKETS, "Broadcasting routing table to %s", rif->iface->name );
rip_tx(c->rif->sock);
}
static struct rip_interface*
find_interface(struct proto *p, struct iface *what)
{
struct rip_interface *i;
WALK_LIST (i, P->interfaces)
if (i->iface == what)
return i;
return NULL;
}
/*
* Input processing
*
* This part is responsible for any updates that come from network
*/
static void
rip_rte_update_if_better(rtable *tab, net *net, struct proto *p, rte *new)
{
rte *old;
old = rte_find(net, p);
if (!old || p->rte_better(new, old))
rte_update(tab, net, p, new);
}
/*
* advertise_entry - let main routing table know about our new entry
* @b: entry in network format
*
* This basically translates @b to format used by bird core and feeds
* bird core with this route.
*/
static void
advertise_entry( struct proto *p, struct rip_block *b, ip_addr whotoldme )
{
rta *a, A;
rte *r;
net *n;
neighbor *neighbor;
struct rip_interface *rif;
int pxlen;
bzero(&A, sizeof(A));
A.proto = p;
A.source = RTS_RIP;
A.scope = SCOPE_UNIVERSE;
A.cast = RTC_UNICAST;
A.dest = RTD_ROUTER;
A.flags = 0;
#ifndef IPV6
A.gw = ipa_nonzero(b->nexthop) ? b->nexthop : whotoldme;
pxlen = ipa_mklen(b->netmask);
#else
/* FIXME: next hop is in other packet for v6 */
A.gw = whotoldme;
pxlen = b->pxlen;
#endif
A.from = whotoldme;
/* No need to look if destination looks valid - ie not net 0 or 127 -- core will do for us. */
neighbor = neigh_find( p, &A.gw, 0 );
if (!neighbor) {
log( L_REMOTE "%s: %I asked me to route %I/%d using not-neighbor %I.", p->name, A.from, b->network, pxlen, A.gw );
return;
}
if (neighbor->scope == SCOPE_HOST) {
DBG("Self-destined route, ignoring.\n");
return;
}
A.iface = neighbor->iface;
if (!(rif = neighbor->data)) {
rif = neighbor->data = find_interface(p, A.iface);
}
if (!rif)
bug("Route packet using unknown interface? No.");
/* set to: interface of nexthop */
a = rta_lookup(&A);
if (pxlen==-1) {
log( L_REMOTE "%s: %I gave me invalid pxlen/netmask for %I.", p->name, A.from, b->network );
return;
}
n = net_get( p->table, b->network, pxlen );
r = rte_get_temp(a);
#ifndef IPV6
r->u.rip.metric = ntohl(b->metric) + rif->metric;
#else
r->u.rip.metric = b->metric + rif->metric;
#endif
r->u.rip.entry = NULL;
if (r->u.rip.metric > P_CF->infinity) r->u.rip.metric = P_CF->infinity;
r->u.rip.tag = ntohl(b->tag);
r->net = n;
r->pflags = 0; /* Here go my flags */
rip_rte_update_if_better( p->table, n, p, r );
DBG( "done\n" );
}
/*
* process_block - do some basic check and pass block to advertise_entry
*/
static void
process_block( struct proto *p, struct rip_block *block, ip_addr whotoldme )
{
#ifndef IPV6
int metric = ntohl( block->metric );
#else
int metric = block->metric;
#endif
ip_addr network = block->network;
CHK_MAGIC;
TRACE(D_ROUTES, "block: %I tells me: %I/??? available, metric %d... ", whotoldme, network, metric );
/* FIXME: Why `???'? If prefix is unknown, just don't print it. [mj] */
if ((!metric) || (metric > P_CF->infinity)) {
#ifdef IPV6 /* Someone is sedning us nexthop and we are ignoring it */
if (metric == 0xff)
{ DBG( "IpV6 nexthop ignored" ); return; }
#endif
log( L_WARN "%s: Got metric %d from %I", P_NAME, metric, whotoldme );
return;
}
advertise_entry( p, block, whotoldme );
}
#define BAD( x ) { log( L_REMOTE "%s: " x, P_NAME ); return 1; }
/*
* rip_process_packet - this is main routine for incoming packets.
*/
static int
rip_process_packet( struct proto *p, struct rip_packet *packet, int num, ip_addr whotoldme, int port )
{
int i;
int native_class = 0, authenticated = 0;
neighbor *neighbor;
switch( packet->heading.version ) {
case RIP_V1: DBG( "Rip1: " ); break;
case RIP_V2: DBG( "Rip2: " ); break;
default: BAD( "Unknown version" );
}
switch( packet->heading.command ) {
case RIPCMD_REQUEST: DBG( "Asked to send my routing table\n" );
if (P_CF->honor == HO_NEVER)
BAD( "They asked me to send routing table, but I was told not to do it" );
if ((P_CF->honor == HO_NEIGHBOR) && (!neigh_find( p, &whotoldme, 0 )))
BAD( "They asked me to send routing table, but he is not my neighbor" );
rip_sendto( p, whotoldme, port, HEAD(P->interfaces) ); /* no broadcast */
break;
case RIPCMD_RESPONSE: DBG( "*** Rtable from %I\n", whotoldme );
if (port != P_CF->port) {
log( L_REMOTE "%s: %I send me routing info from port %d", P_NAME, whotoldme, port );
return 1;
}
if (!(neighbor = neigh_find( p, &whotoldme, 0 )) || neighbor->scope == SCOPE_HOST) {
log( L_REMOTE "%s: %I send me routing info but he is not my neighbor", P_NAME, whotoldme );
return 0;
}
for (i=0; i<num; i++) {
struct rip_block *block = &packet->block[i];
#ifndef IPV6
/* Authentication is not defined for v6 */
if (block->family == 0xffff) {
if (i)
continue; /* md5 tail has this family */
if (rip_incoming_authentication(p, (void *) block, packet, num, whotoldme))
BAD( "Authentication failed" );
authenticated = 1;
continue;
}
#endif
if ((!authenticated) && (P_CF->authtype != AT_NONE))
BAD( "Packet is not authenticated and it should be" );
ipa_ntoh( block->network );
#ifndef IPV6
ipa_ntoh( block->netmask );
ipa_ntoh( block->nexthop );
if (packet->heading.version == RIP_V1) /* FIXME (nonurgent): switch to disable this? */
block->netmask = ipa_class_mask(block->network);
#endif
process_block( p, block, whotoldme );
}
break;
case RIPCMD_TRACEON:
case RIPCMD_TRACEOFF: BAD( "I was asked for traceon/traceoff" );
case 5: BAD( "Some Sun extension around here" );
default: BAD( "Unknown command" );
}
return 0;
}
/*
* rip_rx - Receive hook: do basic checks and pass packet to rip_process_packet
*/
static int
rip_rx(sock *s, int size)
{
struct rip_interface *i = s->data;
struct proto *p = i->proto;
int num;
CHK_MAGIC;
DBG( "RIP: message came: %d bytes from %I via %s\n", size, s->faddr, i->iface ? i->iface->name : "(dummy)" );
size -= sizeof( struct rip_packet_heading );
if (size < 0) BAD( "Too small packet" );
if (size % sizeof( struct rip_block )) BAD( "Odd sized packet" );
num = size / sizeof( struct rip_block );
if (num>PACKET_MAX) BAD( "Too many blocks" );
#ifdef IPV6
/* Try to absolutize link scope addresses */
ipa_absolutize(&s->faddr, &i->iface->addr->ip);
#endif
if (ipa_equal(i->iface->addr->ip, s->faddr)) {
DBG("My own packet\n");
return 1;
}
rip_process_packet( p, (struct rip_packet *) s->rbuf, num, s->faddr, s->fport );
return 1;
}
/*
* Interface to BIRD core
*/
static void
rip_dump_entry( struct rip_entry *e )
{
debug( "%I told me %d/%d ago: to %I/%d go via %I, metric %d ",
e->whotoldme, e->updated-now, e->changed-now, e->n.prefix, e->n.pxlen, e->nexthop, e->metric );
debug( "\n" );
}
/**
* rip_timer
* @t: timer
*
* Broadcast routing tables periodically (using rip_tx) and kill
* routes that are too old. RIP keeps a list of its own entries present
* in the core table by a linked list (functions rip_rte_insert() and
* rip_rte_delete() are responsible for that), it walks this list in the timer
* and in case an entry is too old, it is discarded.
*/
static void
rip_timer(timer *t)
{
struct proto *p = t->data;
struct fib_node *e, *et;
CHK_MAGIC;
DBG( "RIP: tick tock\n" );
WALK_LIST_DELSAFE( e, et, P->garbage ) {
rte *rte;
rte = SKIP_BACK( struct rte, u.rip.garbage, e );
#ifdef LOCAL_DEBUG
{
struct proto *p = rte->attrs->proto;
CHK_MAGIC;
}
DBG( "Garbage: (%p)", rte ); rte_dump( rte );
#endif
if (now - rte->u.rip.lastmodX > P_CF->timeout_time) {
TRACE(D_EVENTS, "entry is too old: %I", rte->net->n.prefix );
if (rte->u.rip.entry) {
rte->u.rip.entry->metric = P_CF->infinity;
rte->u.rip.metric = P_CF->infinity;
}
}
if (now - rte->u.rip.lastmodX > P_CF->garbage_time) {
TRACE(D_EVENTS, "entry is much too old: %I", rte->net->n.prefix );
rte_discard(p->table, rte);
}
}
DBG( "RIP: Broadcasting routing tables\n" );
{
struct rip_interface *rif;
WALK_LIST( rif, P->interfaces ) {
struct iface *iface = rif->iface;
if (!iface) continue;
if (rif->mode & IM_QUIET) continue;
if (!(iface->flags & IF_UP)) continue;
rif->triggered = (P->tx_count % 6);
rip_sendto( p, IPA_NONE, 0, rif );
}
P->tx_count ++;
}
DBG( "RIP: tick tock done\n" );
}
/*
* rip_start - initialize instance of rip
*/
static int
rip_start(struct proto *p)
{
struct rip_interface *rif;
DBG( "RIP: starting instance...\n" );
assert( sizeof(struct rip_packet_heading) == 4);
assert( sizeof(struct rip_block) == 20);
assert( sizeof(struct rip_block_auth) == 20);
P->magic = RIP_MAGIC;
fib_init( &P->rtable, p->pool, sizeof( struct rip_entry ), 0, NULL );
init_list( &P->connections );
init_list( &P->garbage );
init_list( &P->interfaces );
P->timer = tm_new( p->pool );
P->timer->data = p;
P->timer->randomize = 5;
P->timer->recurrent = (P_CF->period / 6)+1;
P->timer->hook = rip_timer;
tm_start( P->timer, 5 );
rif = new_iface(p, NULL, 0, NULL); /* Initialize dummy interface */
add_head( &P->interfaces, NODE rif );
CHK_MAGIC;
rip_init_instance(p);
DBG( "RIP: ...done\n");
return PS_UP;
}
static struct proto *
rip_init(struct proto_config *cfg)
{
struct proto *p = proto_new(cfg, sizeof(struct rip_proto));
return p;
}
static void
rip_dump(struct proto *p)
{
int i;
node *w, *e;
struct rip_interface *rif;
i = 0;
CHK_MAGIC;
WALK_LIST( w, P->connections ) {
struct rip_connection *n = (void *) w;
debug( "RIP: connection #%d: %I\n", n->num, n->addr );
}
i = 0;
FIB_WALK( &P->rtable, e ) {
debug( "RIP: entry #%d: ", i++ );
rip_dump_entry( E );
} FIB_WALK_END;
i = 0;
WALK_LIST( rif, P->interfaces ) {
debug( "RIP: interface #%d: %s, %I, busy = %x\n", i++, rif->iface?rif->iface->name:"(dummy)", rif->sock->daddr, rif->busy );
}
}
static void
rip_get_route_info(rte *rte, byte *buf, ea_list *attrs)
{
eattr *metric = ea_find(attrs, EA_RIP_METRIC);
eattr *tag = ea_find(attrs, EA_RIP_TAG);
buf += bsprintf(buf, " (%d/%d)", rte->pref, metric ? metric->u.data : 0);
if (tag && tag->u.data)
bsprintf(buf, " t%04x", tag->u.data);
}
static int
rip_want_this_if(struct rip_interface *iface)
{
return 1;
}
static void
kill_iface(struct proto *p, struct rip_interface *i)
{
DBG( "RIP: Interface %s disappeared\n", i->iface->name);
rfree(i->sock);
mb_free(i);
}
/**
* new_iface
* @p: myself
* @new: interface to be created or %NULL if we are creating a magic
* socket. The magic socket is used for listening and also for
* sending requested responses.
* @flags: interface flags
* @patt: pattern this interface matched, used for access to config options
*
* Create an interface structure and start listening on the interface.
*/
static struct rip_interface *
new_iface(struct proto *p, struct iface *new, unsigned long flags, struct iface_patt *patt )
{
struct rip_interface *rif;
struct rip_patt *PATT = (struct rip_patt *) patt;
rif = mb_allocz(p->pool, sizeof( struct rip_interface ));
rif->iface = new;
rif->proto = p;
rif->busy = NULL;
if (PATT) {
rif->mode = PATT->mode;
rif->metric = PATT->metric;
rif->multicast = (!(PATT->mode & IM_BROADCAST)) && (flags & IF_MULTICAST);
}
/* lookup multicasts over unnumbered links - no: rip is not defined over unnumbered links */
if (rif->multicast)
DBG( "Doing multicasts!\n" );
rif->sock = sk_new( p->pool );
rif->sock->type = rif->multicast?SK_UDP_MC:SK_UDP;
rif->sock->sport = P_CF->port;
rif->sock->rx_hook = rip_rx;
rif->sock->data = rif;
rif->sock->rbsize = 10240;
rif->sock->iface = new; /* Automagically works for dummy interface */
rif->sock->tbuf = mb_alloc( p->pool, sizeof( struct rip_packet ));
rif->sock->tx_hook = rip_tx;
rif->sock->err_hook = rip_tx_err;
rif->sock->daddr = IPA_NONE;
rif->sock->dport = P_CF->port;
if (new)
{
rif->sock->ttl = 1;
rif->sock->tos = IP_PREC_INTERNET_CONTROL;
}
if (new) {
if (new->addr->flags & IA_UNNUMBERED)
log( L_WARN "%s: rip is not defined over unnumbered links", P_NAME );
if (rif->multicast) {
#ifndef IPV6
rif->sock->daddr = ipa_from_u32(0xe0000009);
rif->sock->saddr = ipa_from_u32(0xe0000009);
#else
rif->sock->daddr = ipa_build(0xff020000, 0, 0, 9);
rif->sock->saddr = new->addr->ip;
#endif
} else {
rif->sock->daddr = new->addr->brd;
rif->sock->saddr = new->addr->brd;
}
}
if (!ipa_nonzero(rif->sock->daddr)) {
if (rif->iface)
log( L_WARN "%s: interface %s is too strange for me", P_NAME, rif->iface->name );
} else
if (!(rif->mode & IM_NOLISTEN))
if (sk_open(rif->sock)<0) {
log( L_ERR "%s: could not listen on %s", P_NAME, rif->iface ? rif->iface->name : "(dummy)" );
if (rif->iface) {
rfree(rif->sock);
mb_free(rif);
return NULL;
}
/* On dummy, we just return non-working socket, so that user gets error every time anyone requests table */
}
TRACE(D_EVENTS, "Listening on %s, port %d, mode %s (%I)", rif->iface ? rif->iface->name : "(dummy)", P_CF->port, rif->multicast ? "multicast" : "broadcast", rif->sock->daddr );
return rif;
}
static void
rip_real_if_add(struct object_lock *lock)
{
struct iface *iface = lock->iface;
struct proto *p = lock->data;
struct rip_interface *rif;
struct iface_patt *k = iface_patt_match(&P_CF->iface_list, iface);
if (!k)
bug("This can not happen! It existed few seconds ago!" );
DBG("adding interface %s\n", iface->name );
rif = new_iface(p, iface, iface->flags, k);
if (rif) {
add_head( &P->interfaces, NODE rif );
DBG("Adding object lock of %p for %p\n", lock, rif);
rif->lock = lock;
} else { rfree(lock); }
}
static void
rip_if_notify(struct proto *p, unsigned c, struct iface *iface)
{
DBG( "RIP: if notify\n" );
if (iface->flags & IF_IGNORE)
return;
if (c & IF_CHANGE_DOWN) {
struct rip_interface *i;
i = find_interface(p, iface);
if (i) {
rem_node(NODE i);
rfree(i->lock);
kill_iface(p, i);
}
}
if (c & IF_CHANGE_UP) {
struct iface_patt *k = iface_patt_match(&P_CF->iface_list, iface);
struct object_lock *lock;
if (!k) return; /* We are not interested in this interface */
lock = olock_new( p->pool );
#ifndef IPV6
lock->addr = ipa_from_u32(0xe0000009); /* This is okay: we
may actually use
other address, but
we do not want two
rips at one time,
anyway. */
#else
ip_pton("FF02::9", &lock->addr);
#endif
lock->port = P_CF->port;
lock->iface = iface;
lock->hook = rip_real_if_add;
lock->data = p;
lock->type = OBJLOCK_UDP;
olock_acquire(lock);
}
}
static struct ea_list *
rip_gen_attrs(struct proto *p, struct linpool *pool, int metric, u16 tag)
{
struct ea_list *l = lp_alloc(pool, sizeof(struct ea_list) + 2*sizeof(eattr));
l->next = NULL;
l->flags = EALF_SORTED;
l->count = 2;
l->attrs[0].id = EA_RIP_TAG;
l->attrs[0].flags = 0;
l->attrs[0].type = EAF_TYPE_INT | EAF_TEMP;
l->attrs[0].u.data = tag;
l->attrs[1].id = EA_RIP_METRIC;
l->attrs[1].flags = 0;
l->attrs[1].type = EAF_TYPE_INT | EAF_TEMP;
l->attrs[1].u.data = metric;
return l;
}
static int
rip_import_control(struct proto *p, struct rte **rt, struct ea_list **attrs, struct linpool *pool)
{
if ((*rt)->attrs->proto == p) /* My own must not be touched */
return 1;
if ((*rt)->attrs->source != RTS_RIP) {
struct ea_list *new = rip_gen_attrs(p, pool, 1, 0);
new->next = *attrs;
*attrs = new;
}
return 0;
}
static struct ea_list *
rip_make_tmp_attrs(struct rte *rt, struct linpool *pool)
{
struct proto *p = rt->attrs->proto;
return rip_gen_attrs(p, pool, rt->u.rip.metric, rt->u.rip.tag);
}
static void
rip_store_tmp_attrs(struct rte *rt, struct ea_list *attrs)
{
rt->u.rip.tag = ea_get_int(attrs, EA_RIP_TAG, 0);
rt->u.rip.metric = ea_get_int(attrs, EA_RIP_METRIC, 1);
}
/*
* rip_rt_notify - core tells us about new route (possibly our
* own), so store it into our data structures.
*/
static void
rip_rt_notify(struct proto *p, struct network *net, struct rte *new, struct rte *old, struct ea_list *attrs)
{
CHK_MAGIC;
if (old) {
struct rip_entry *e = fib_find( &P->rtable, &net->n.prefix, net->n.pxlen );
if (!e)
log( L_BUG "%s: Deleting nonexistent entry?!", P_NAME );
fib_delete( &P->rtable, e );
}
if (new) {
struct rip_entry *e;
#if 0
/* This can happen since feeding of protocols is asynchronous */
if (fib_find( &P->rtable, &net->n.prefix, net->n.pxlen ))
log( L_BUG "%s: Inserting entry which is already there?", P_NAME );
#endif
e = fib_get( &P->rtable, &net->n.prefix, net->n.pxlen );
e->nexthop = new->attrs->gw;
e->metric = 0;
e->whotoldme = IPA_NONE;
new->u.rip.entry = e;
e->tag = ea_get_int(attrs, EA_RIP_TAG, 0);
e->metric = ea_get_int(attrs, EA_RIP_METRIC, 1);
if (e->metric > P_CF->infinity)
e->metric = P_CF->infinity;
if (new->attrs->proto == p)
e->whotoldme = new->attrs->from;
if (!e->metric) /* That's okay: this way user can set his own value for external
routes in rip. */
e->metric = 5;
e->updated = e->changed = now;
e->flags = 0;
}
}
static int
rip_rte_better(struct rte *new, struct rte *old)
{
struct proto *p = new->attrs->proto;
if (ipa_equal(old->attrs->from, new->attrs->from))
return 1;
if (old->u.rip.metric < new->u.rip.metric)
return 0;
if (old->u.rip.metric > new->u.rip.metric)
return 1;
if ((old->u.rip.metric < 16) && (new->u.rip.metric == P_CF->infinity)) {
new->u.rip.lastmodX = now - P_CF->timeout_time; /* Check this: if new metric is 16, act as it was timed out */
}
if (old->attrs->proto == new->attrs->proto) /* This does not make much sense for different protocols */
if ((old->u.rip.metric == new->u.rip.metric) &&
((now - old->u.rip.lastmodX) > (P_CF->timeout_time / 2)))
return 1;
return 0;
}
/*
* rip_rte_insert - we maintain linked list of "our" entries in main
* routing table, so that we can timeout them correctly. rip_timer()
* walks the list.
*/
static void
rip_rte_insert(net *net, rte *rte)
{
struct proto *p = rte->attrs->proto;
CHK_MAGIC;
DBG( "rip_rte_insert: %p\n", rte );
rte->u.rip.lastmodX = now;
add_head( &P->garbage, &rte->u.rip.garbage );
}
/*
* rip_rte_remove - link list maintenance
*/
static void
rip_rte_remove(net *net, rte *rte)
{
struct proto *p = rte->attrs->proto;
CHK_MAGIC;
DBG( "rip_rte_remove: %p\n", rte );
rem_node( &rte->u.rip.garbage );
}
void
rip_init_instance(struct proto *p)
{
p->if_notify = rip_if_notify;
p->rt_notify = rip_rt_notify;
p->import_control = rip_import_control;
p->make_tmp_attrs = rip_make_tmp_attrs;
p->store_tmp_attrs = rip_store_tmp_attrs;
p->rte_better = rip_rte_better;
p->rte_insert = rip_rte_insert;
p->rte_remove = rip_rte_remove;
}
void
rip_init_config(struct rip_proto_config *c)
{
init_list(&c->iface_list);
c->c.preference = DEF_PREF_RIP;
c->infinity = 16;
c->port = 520;
c->period = 30;
c->garbage_time = 120+180;
c->timeout_time = 120;
c->passwords = NULL;
c->authtype = AT_NONE;
}
static int
rip_get_attr(eattr *a, byte *buf)
{
switch (a->id) {
case EA_RIP_METRIC: buf += bsprintf( buf, "metric: %d", a->u.data ); return GA_FULL;
case EA_RIP_TAG: buf += bsprintf( buf, "tag: %d", a->u.data ); return GA_FULL;
default: return GA_UNKNOWN;
}
}
static int
rip_pat_compare(struct rip_patt *a, struct rip_patt *b)
{
return ((a->metric == b->metric) &&
(a->mode == b->mode));
}
static int
rip_reconfigure(struct proto *p, struct proto_config *c)
{
struct rip_proto_config *new = (struct rip_proto_config *) c;
int generic = sizeof(struct proto_config) + sizeof(list) /* + sizeof(struct password_item *) */;
if (!iface_patts_equal(&P_CF->iface_list, &new->iface_list, (void *) rip_pat_compare))
return 0;
if (!password_same(P_CF->passwords,
new->passwords))
return 0;
return !memcmp(((byte *) P_CF) + generic,
((byte *) new) + generic,
sizeof(struct rip_proto_config) - generic);
}
struct protocol proto_rip = {
name: "RIP",
template: "rip%d",
attr_class: EAP_RIP,
get_route_info: rip_get_route_info,
get_attr: rip_get_attr,
init: rip_init,
dump: rip_dump,
start: rip_start,
reconfigure: rip_reconfigure,
};