0
0
mirror of https://gitlab.nic.cz/labs/bird.git synced 2024-12-22 09:41:54 +00:00
bird/proto/ospf/neighbor.c
Maria Matejka 22f54eaee6 Resource pools are now bound with domains.
Memory allocation is a fragile part of BIRD and we need checking that
everybody is using the resource pools in an appropriate way. To assure
this, all the resource pools are associated with locking domains and
every resource manipulation is thoroughly checked whether the
appropriate locking domain is locked.

With transitive resource manipulation like resource dumping or mass free
operations, domains are locked and unlocked on the go, thus we require
pool domains to have higher order than their parent to allow for this
transitive operations.

Adding pool locking revealed some cases of insecure memory manipulation
and this commit fixes that as well.
2023-04-24 10:33:28 +02:00

865 lines
20 KiB
C

/*
* BIRD -- OSPF
*
* (c) 1999--2004 Ondrej Filip <feela@network.cz>
* (c) 2009--2014 Ondrej Zajicek <santiago@crfreenet.org>
* (c) 2009--2014 CZ.NIC z.s.p.o.
*
* Can be freely distributed and used under the terms of the GNU GPL.
*/
#include "ospf.h"
const char *ospf_ns_names[] = {
"Down", "Attempt", "Init", "2-Way", "ExStart", "Exchange", "Loading", "Full"
};
const char *ospf_inm_names[] = {
"HelloReceived", "Start", "2-WayReceived", "NegotiationDone", "ExchangeDone",
"BadLSReq", "LoadingDone", "AdjOK?", "SeqNumberMismatch", "1-WayReceived",
"KillNbr", "InactivityTimer", "LLDown"
};
static int can_do_adj(struct ospf_neighbor *n);
static void inactivity_timer_hook(timer * timer);
static void dbdes_timer_hook(timer *t);
static void lsrq_timer_hook(timer *t);
static void lsrt_timer_hook(timer *t);
static void ackd_timer_hook(timer *t);
static void ospf_neigh_stop_graceful_restart_(struct ospf_neighbor *n);
static void graceful_restart_timeout(timer *t);
static void
init_lists(struct ospf_proto *p, struct ospf_neighbor *n)
{
s_init_list(&(n->lsrql));
n->lsrqi = SHEAD(n->lsrql);
n->lsrqh = ospf_top_new(p, n->pool);
s_init_list(&(n->lsrtl));
n->lsrth = ospf_top_new(p, n->pool);
}
static void
release_lsrtl(struct ospf_proto *p, struct ospf_neighbor *n)
{
struct top_hash_entry *ret, *en;
WALK_SLIST(ret, n->lsrtl)
{
en = ospf_hash_find_entry(p->gr, ret);
if (en)
en->ret_count--;
}
}
/* Resets LSA request and retransmit lists.
* We do not reset DB summary list iterator here,
* it is reset during entering EXCHANGE state.
*/
static void
reset_lists(struct ospf_proto *p, struct ospf_neighbor *n)
{
release_lsrtl(p, n);
ospf_top_free(n->lsrqh);
ospf_top_free(n->lsrth);
ospf_reset_lsack_queue(n);
tm_stop(n->dbdes_timer);
tm_stop(n->lsrq_timer);
tm_stop(n->lsrt_timer);
tm_stop(n->ackd_timer);
init_lists(p, n);
}
struct ospf_neighbor *
ospf_neighbor_new(struct ospf_iface *ifa)
{
struct ospf_proto *p = ifa->oa->po;
struct pool *pool = rp_new(p->p.pool, proto_domain(&p->p), "OSPF Neighbor");
struct ospf_neighbor *n = mb_allocz(pool, sizeof(struct ospf_neighbor));
n->pool = pool;
n->ifa = ifa;
add_tail(&ifa->neigh_list, NODE n);
n->state = NEIGHBOR_DOWN;
init_lists(p, n);
s_init(&(n->dbsi), &(p->lsal));
init_list(&n->ackl[ACKL_DIRECT]);
init_list(&n->ackl[ACKL_DELAY]);
n->inactim = tm_new_init(pool, inactivity_timer_hook, n, 0, 0);
n->dbdes_timer = tm_new_init(pool, dbdes_timer_hook, n, ifa->rxmtint S, 0);
n->lsrq_timer = tm_new_init(pool, lsrq_timer_hook, n, ifa->rxmtint S, 0);
n->lsrt_timer = tm_new_init(pool, lsrt_timer_hook, n, ifa->rxmtint S, 0);
n->ackd_timer = tm_new_init(pool, ackd_timer_hook, n, ifa->rxmtint S / 2, 0);
return (n);
}
static void
ospf_neigh_down(struct ospf_neighbor *n)
{
struct ospf_iface *ifa = n->ifa;
struct ospf_proto *p = ifa->oa->po;
u32 rid = n->rid;
if ((ifa->type == OSPF_IT_NBMA) || (ifa->type == OSPF_IT_PTMP))
{
struct nbma_node *nn = find_nbma_node(ifa, n->ip);
if (nn)
nn->found = 0;
}
s_get(&(n->dbsi));
release_lsrtl(p, n);
rem_node(NODE n);
rp_free(n->pool);
OSPF_TRACE(D_EVENTS, "Neighbor %R on %s removed", rid, ifa->ifname);
}
/**
* ospf_neigh_chstate - handles changes related to new or lod state of neighbor
* @n: OSPF neighbor
* @state: new state
*
* Many actions have to be taken acording to a change of state of a neighbor. It
* starts rxmt timers, call interface state machine etc.
*/
static void
ospf_neigh_chstate(struct ospf_neighbor *n, u8 state)
{
struct ospf_iface *ifa = n->ifa;
struct ospf_proto *p = ifa->oa->po;
u8 old_state = n->state;
int old_fadj = ifa->fadj;
if (state == old_state)
return;
OSPF_TRACE(D_EVENTS, "Neighbor %R on %s changed state from %s to %s",
n->rid, ifa->ifname, ospf_ns_names[old_state], ospf_ns_names[state]);
n->state = state;
/* Increase number of partial adjacencies */
if ((state == NEIGHBOR_EXCHANGE) || (state == NEIGHBOR_LOADING))
p->padj++;
/* Decrease number of partial adjacencies */
if ((old_state == NEIGHBOR_EXCHANGE) || (old_state == NEIGHBOR_LOADING))
p->padj--;
/* Increase number of full adjacencies */
if (state == NEIGHBOR_FULL)
ifa->fadj++;
/* Decrease number of full adjacencies */
if (old_state == NEIGHBOR_FULL)
ifa->fadj--;
if ((ifa->fadj != old_fadj) && !n->gr_active)
{
/* RFC 2328 12.4 Event 4 - neighbor enters/leaves Full state */
ospf_notify_rt_lsa(ifa->oa);
ospf_notify_net_lsa(ifa);
/* RFC 2328 12.4 Event 8 - vlink state change */
if (ifa->type == OSPF_IT_VLINK)
ospf_notify_rt_lsa(ifa->voa);
}
if (state == NEIGHBOR_EXSTART)
{
/* First time adjacency attempt */
if (old_state < NEIGHBOR_EXSTART)
n->dds = random_u32();
n->dds++;
n->myimms = DBDES_IMMS;
n->got_my_rt_lsa = 0;
tm_start(n->dbdes_timer, 0);
tm_start(n->ackd_timer, ifa->rxmtint S / 2);
}
if (state > NEIGHBOR_EXSTART)
n->myimms &= ~DBDES_I;
/* Generate NeighborChange event if needed, see RFC 2328 9.2 */
if ((state == NEIGHBOR_2WAY) && (old_state < NEIGHBOR_2WAY) && !n->gr_active)
ospf_iface_sm(ifa, ISM_NEICH);
if ((state < NEIGHBOR_2WAY) && (old_state >= NEIGHBOR_2WAY) && !n->gr_active)
ospf_iface_sm(ifa, ISM_NEICH);
}
/**
* ospf_neigh_sm - ospf neighbor state machine
* @n: neighor
* @event: actual event
*
* This part implements the neighbor state machine as described in 10.3 of
* RFC 2328. The only difference is that state %NEIGHBOR_ATTEMPT is not
* used. We discover neighbors on nonbroadcast networks in the
* same way as on broadcast networks. The only difference is in
* sending hello packets. These are sent to IPs listed in
* @ospf_iface->nbma_list .
*/
void
ospf_neigh_sm(struct ospf_neighbor *n, int event)
{
struct ospf_proto *p = n->ifa->oa->po;
DBG("Neighbor state machine for %R on %s, event %s\n",
n->rid, n->ifa->ifname, ospf_inm_names[event]);
switch (event)
{
case INM_START:
ospf_neigh_chstate(n, NEIGHBOR_ATTEMPT);
/* NBMA are used different way */
break;
case INM_HELLOREC:
if (n->state < NEIGHBOR_INIT)
ospf_neigh_chstate(n, NEIGHBOR_INIT);
/* Restart inactivity timer */
tm_start(n->inactim, n->ifa->deadint S);
break;
case INM_2WAYREC:
if (n->state < NEIGHBOR_2WAY)
ospf_neigh_chstate(n, NEIGHBOR_2WAY);
if ((n->state == NEIGHBOR_2WAY) && can_do_adj(n))
ospf_neigh_chstate(n, NEIGHBOR_EXSTART);
break;
case INM_NEGDONE:
if (n->state == NEIGHBOR_EXSTART)
{
ospf_neigh_chstate(n, NEIGHBOR_EXCHANGE);
/* Reset DB summary list iterator */
s_get(&(n->dbsi));
s_init(&(n->dbsi), &p->lsal);
/* Add MaxAge LSA entries to retransmission list */
ospf_add_flushed_to_lsrt(p, n);
}
else
bug("NEGDONE and I'm not in EXSTART?");
break;
case INM_EXDONE:
if (!EMPTY_SLIST(n->lsrql))
ospf_neigh_chstate(n, NEIGHBOR_LOADING);
else
ospf_neigh_chstate(n, NEIGHBOR_FULL);
break;
case INM_LOADDONE:
ospf_neigh_chstate(n, NEIGHBOR_FULL);
break;
case INM_ADJOK:
/* Can In build adjacency? */
if ((n->state == NEIGHBOR_2WAY) && can_do_adj(n))
{
ospf_neigh_chstate(n, NEIGHBOR_EXSTART);
}
else if ((n->state >= NEIGHBOR_EXSTART) && !can_do_adj(n))
{
reset_lists(p, n);
ospf_neigh_chstate(n, NEIGHBOR_2WAY);
}
break;
case INM_SEQMIS:
case INM_BADLSREQ:
if (n->state >= NEIGHBOR_EXCHANGE)
{
reset_lists(p, n);
ospf_neigh_chstate(n, NEIGHBOR_EXSTART);
}
break;
case INM_KILLNBR:
case INM_LLDOWN:
case INM_INACTTIM:
if (n->gr_active && (event == INM_INACTTIM))
{
/* Just down the neighbor, but do not remove it */
reset_lists(p, n);
ospf_neigh_chstate(n, NEIGHBOR_DOWN);
break;
}
if (n->gr_active)
ospf_neigh_stop_graceful_restart_(n);
/* No need for reset_lists() */
ospf_neigh_chstate(n, NEIGHBOR_DOWN);
ospf_neigh_down(n);
break;
case INM_1WAYREC:
reset_lists(p, n);
ospf_neigh_chstate(n, NEIGHBOR_INIT);
break;
default:
bug("%s: INM - Unknown event?", p->p.name);
break;
}
}
static int
can_do_adj(struct ospf_neighbor *n)
{
struct ospf_iface *ifa = n->ifa;
struct ospf_proto *p = ifa->oa->po;
int i = 0;
switch (ifa->type)
{
case OSPF_IT_PTP:
case OSPF_IT_PTMP:
case OSPF_IT_VLINK:
i = 1;
break;
case OSPF_IT_BCAST:
case OSPF_IT_NBMA:
switch (ifa->state)
{
case OSPF_IS_DOWN:
case OSPF_IS_LOOP:
bug("%s: Iface %s in down state?", p->p.name, ifa->ifname);
break;
case OSPF_IS_WAITING:
DBG("%s: Neighbor? on iface %s\n", p->p.name, ifa->ifname);
break;
case OSPF_IS_DROTHER:
if (((n->rid == ifa->drid) || (n->rid == ifa->bdrid))
&& (n->state >= NEIGHBOR_2WAY))
i = 1;
break;
case OSPF_IS_PTP:
case OSPF_IS_BACKUP:
case OSPF_IS_DR:
if (n->state >= NEIGHBOR_2WAY)
i = 1;
break;
default:
bug("%s: Iface %s in unknown state?", p->p.name, ifa->ifname);
break;
}
break;
default:
bug("%s: Iface %s is unknown type?", p->p.name, ifa->ifname);
break;
}
DBG("%s: Iface %s can_do_adj=%d\n", p->p.name, ifa->ifname, i);
return i;
}
static void
ospf_neigh_start_graceful_restart(struct ospf_neighbor *n, uint gr_time)
{
struct ospf_proto *p = n->ifa->oa->po;
OSPF_TRACE(D_EVENTS, "Neighbor %R on %s started graceful restart",
n->rid, n->ifa->ifname);
n->gr_active = 1;
p->gr_count++;
n->gr_timer = tm_new_init(n->pool, graceful_restart_timeout, n, 0, 0);
tm_start(n->gr_timer, gr_time S);
}
static void
ospf_neigh_stop_graceful_restart_(struct ospf_neighbor *n)
{
struct ospf_proto *p = n->ifa->oa->po;
struct ospf_iface *ifa = n->ifa;
n->gr_active = 0;
p->gr_count--;
rfree(n->gr_timer);
n->gr_timer = NULL;
ospf_notify_rt_lsa(ifa->oa);
ospf_notify_net_lsa(ifa);
if (ifa->type == OSPF_IT_VLINK)
ospf_notify_rt_lsa(ifa->voa);
ospf_iface_sm(ifa, ISM_NEICH);
}
static void
ospf_neigh_stop_graceful_restart(struct ospf_neighbor *n)
{
struct ospf_proto *p = n->ifa->oa->po;
OSPF_TRACE(D_EVENTS, "Neighbor %R on %s finished graceful restart",
n->rid, n->ifa->ifname);
ospf_neigh_stop_graceful_restart_(n);
}
void
ospf_neigh_cancel_graceful_restart(struct ospf_neighbor *n)
{
struct ospf_proto *p = n->ifa->oa->po;
OSPF_TRACE(D_EVENTS, "Graceful restart canceled for nbr %R on %s",
n->rid, n->ifa->ifname);
ospf_neigh_stop_graceful_restart_(n);
if (n->state == NEIGHBOR_DOWN)
ospf_neigh_down(n);
}
static void
graceful_restart_timeout(timer *t)
{
struct ospf_neighbor *n = t->data;
struct ospf_proto *p = n->ifa->oa->po;
OSPF_TRACE(D_EVENTS, "Graceful restart timer expired for nbr %R on %s",
n->rid, n->ifa->ifname);
ospf_neigh_stop_graceful_restart_(n);
if (n->state == NEIGHBOR_DOWN)
ospf_neigh_down(n);
}
static inline int
changes_in_lsrtl(struct ospf_neighbor *n)
{
/* This could be improved, see RFC 3623 3.1 (2) */
struct top_hash_entry *en;
WALK_SLIST(en, n->lsrtl)
if (LSA_FUNCTION(en->lsa_type) <= LSA_FUNCTION(LSA_T_NSSA))
return 1;
return 0;
}
void
ospf_neigh_notify_grace_lsa(struct ospf_neighbor *n, struct top_hash_entry *en)
{
struct ospf_iface *ifa = n->ifa;
struct ospf_proto *p = ifa->oa->po;
/* In OSPFv2, neighbors are identified by either IP or Router ID, based on network type */
uint t = ifa->type;
if (ospf_is_v2(p) && ((t == OSPF_IT_BCAST) || (t == OSPF_IT_NBMA) || (t == OSPF_IT_PTMP)))
{
struct ospf_tlv *tlv = lsa_get_tlv(en, LSA_GR_ADDRESS);
if (!tlv || tlv->length != 4)
return;
ip_addr addr = ipa_from_u32(tlv->data[0]);
if (!ipa_equal(n->ip, addr))
n = find_neigh_by_ip(ifa, addr);
}
else
{
if (n->rid != en->lsa.rt)
n = find_neigh(ifa, en->lsa.rt);
}
if (!n)
return;
if (en->lsa.age < LSA_MAXAGE)
{
u32 period = lsa_get_tlv_u32(en, LSA_GR_PERIOD);
/* Exception for updating grace period */
if (n->gr_active)
{
tm_start(n->gr_timer, (period S) - (en->lsa.age S));
return;
}
/* RFC 3623 3.1 (1) - full adjacency */
if (n->state != NEIGHBOR_FULL)
return;
/* RFC 3623 3.1 (2) - no changes in LSADB */
if (changes_in_lsrtl(n))
return;
/* RFC 3623 3.1 (3) - grace period not expired */
if (en->lsa.age >= period)
return;
/* RFC 3623 3.1 (4) - helper mode allowed */
if (!p->gr_mode)
return;
/* RFC 3623 3.1 (5) - no local graceful restart */
if (p->p.gr_recovery)
return;
ospf_neigh_start_graceful_restart(n, period - en->lsa.age);
}
else /* Grace-LSA is flushed */
{
if (n->gr_active)
ospf_neigh_stop_graceful_restart(n);
}
}
void
ospf_neigh_lsadb_changed_(struct ospf_proto *p, struct top_hash_entry *en)
{
struct ospf_iface *ifa;
struct ospf_neighbor *n, *nx;
if (LSA_FUNCTION(en->lsa_type) > LSA_FUNCTION(LSA_T_NSSA))
return;
/* RFC 3623 3.2 (3) - cancel graceful restart when LSdb changed */
WALK_LIST(ifa, p->iface_list)
if (lsa_flooding_allowed(en->lsa_type, en->domain, ifa))
WALK_LIST_DELSAFE(n, nx, ifa->neigh_list)
if (n->gr_active)
ospf_neigh_cancel_graceful_restart(n);
}
static inline u32 neigh_get_id(struct ospf_proto *p, struct ospf_neighbor *n)
{ return ospf_is_v2(p) ? ipa_to_u32(n->ip) : n->rid; }
static struct ospf_neighbor *
elect_bdr(struct ospf_proto *p, list nl)
{
struct ospf_neighbor *neigh, *n1, *n2;
u32 nid;
n1 = NULL;
n2 = NULL;
WALK_LIST(neigh, nl) /* First try those decl. themselves */
{
nid = neigh_get_id(p, neigh);
if (neigh->state >= NEIGHBOR_2WAY) /* Higher than 2WAY */
if (neigh->priority > 0) /* Eligible */
if (neigh->dr != nid) /* And not decl. itself DR */
{
if (neigh->bdr == nid) /* Declaring BDR */
{
if (n1 != NULL)
{
if (neigh->priority > n1->priority)
n1 = neigh;
else if (neigh->priority == n1->priority)
if (neigh->rid > n1->rid)
n1 = neigh;
}
else
{
n1 = neigh;
}
}
else /* And NOT declaring BDR */
{
if (n2 != NULL)
{
if (neigh->priority > n2->priority)
n2 = neigh;
else if (neigh->priority == n2->priority)
if (neigh->rid > n2->rid)
n2 = neigh;
}
else
{
n2 = neigh;
}
}
}
}
if (n1 == NULL)
n1 = n2;
return (n1);
}
static struct ospf_neighbor *
elect_dr(struct ospf_proto *p, list nl)
{
struct ospf_neighbor *neigh, *n;
u32 nid;
n = NULL;
WALK_LIST(neigh, nl) /* And now DR */
{
nid = neigh_get_id(p, neigh);
if (neigh->state >= NEIGHBOR_2WAY) /* Higher than 2WAY */
if (neigh->priority > 0) /* Eligible */
if (neigh->dr == nid) /* And declaring itself DR */
{
if (n != NULL)
{
if (neigh->priority > n->priority)
n = neigh;
else if (neigh->priority == n->priority)
if (neigh->rid > n->rid)
n = neigh;
}
else
{
n = neigh;
}
}
}
return (n);
}
/**
* ospf_dr_election - (Backup) Designed Router election
* @ifa: actual interface
*
* When the wait timer fires, it is time to elect (Backup) Designated Router.
* Structure describing me is added to this list so every electing router has
* the same list. Backup Designated Router is elected before Designated
* Router. This process is described in 9.4 of RFC 2328. The function is
* supposed to be called only from ospf_iface_sm() as a part of the interface
* state machine.
*/
void
ospf_dr_election(struct ospf_iface *ifa)
{
struct ospf_proto *p = ifa->oa->po;
struct ospf_neighbor *neigh, *ndr, *nbdr;
u32 myid = p->router_id;
DBG("(B)DR election.\n");
struct ospf_neighbor me = {
.state = NEIGHBOR_2WAY,
.rid = myid,
.priority = ifa->priority,
.ip = ifa->addr->ip,
.dr = ospf_is_v2(p) ? ipa_to_u32(ifa->drip) : ifa->drid,
.bdr = ospf_is_v2(p) ? ipa_to_u32(ifa->bdrip) : ifa->bdrid,
.iface_id = ifa->iface_id,
};
add_tail(&ifa->neigh_list, NODE & me);
nbdr = elect_bdr(p, ifa->neigh_list);
ndr = elect_dr(p, ifa->neigh_list);
if (ndr == NULL)
ndr = nbdr;
/* 9.4. (4) */
if (((ifa->drid == myid) && (ndr != &me))
|| ((ifa->drid != myid) && (ndr == &me))
|| ((ifa->bdrid == myid) && (nbdr != &me))
|| ((ifa->bdrid != myid) && (nbdr == &me)))
{
me.dr = ndr ? neigh_get_id(p, ndr) : 0;
me.bdr = nbdr ? neigh_get_id(p, nbdr) : 0;
nbdr = elect_bdr(p, ifa->neigh_list);
ndr = elect_dr(p, ifa->neigh_list);
if (ndr == NULL)
ndr = nbdr;
}
rem_node(NODE & me);
u32 old_drid = ifa->drid;
u32 old_bdrid = ifa->bdrid;
ip_addr none = ospf_is_v2(p) ? IPA_NONE4 : IPA_NONE6;
ifa->drid = ndr ? ndr->rid : 0;
ifa->drip = ndr ? ndr->ip : none;
ifa->dr_iface_id = ndr ? ndr->iface_id : 0;
ifa->bdrid = nbdr ? nbdr->rid : 0;
ifa->bdrip = nbdr ? nbdr->ip : none;
DBG("DR=%R, BDR=%R\n", ifa->drid, ifa->bdrid);
/* We are part of the interface state machine */
if (ifa->drid == myid)
ospf_iface_chstate(ifa, OSPF_IS_DR);
else if (ifa->bdrid == myid)
ospf_iface_chstate(ifa, OSPF_IS_BACKUP);
else
ospf_iface_chstate(ifa, OSPF_IS_DROTHER);
/* Review neighbor adjacencies if DR or BDR changed */
if ((ifa->drid != old_drid) || (ifa->bdrid != old_bdrid))
WALK_LIST(neigh, ifa->neigh_list)
if (neigh->state >= NEIGHBOR_2WAY)
ospf_neigh_sm(neigh, INM_ADJOK);
/* RFC 2328 12.4 Event 3 - DR change */
if (ifa->drid != old_drid)
ospf_notify_rt_lsa(ifa->oa);
}
struct ospf_neighbor *
find_neigh(struct ospf_iface *ifa, u32 rid)
{
struct ospf_neighbor *n;
WALK_LIST(n, ifa->neigh_list)
if (n->rid == rid)
return n;
return NULL;
}
struct ospf_neighbor *
find_neigh_by_ip(struct ospf_iface *ifa, ip_addr ip)
{
struct ospf_neighbor *n;
WALK_LIST(n, ifa->neigh_list)
if (ipa_equal(n->ip, ip))
return n;
return NULL;
}
static void
inactivity_timer_hook(timer * timer)
{
struct ospf_neighbor *n = (struct ospf_neighbor *) timer->data;
struct ospf_proto *p = n->ifa->oa->po;
OSPF_TRACE(D_EVENTS, "Inactivity timer expired for nbr %R on %s",
n->rid, n->ifa->ifname);
ospf_neigh_sm(n, INM_INACTTIM);
}
static void
ospf_neigh_bfd_hook(struct bfd_request *req)
{
struct ospf_neighbor *n = req->data;
struct ospf_proto *p = n->ifa->oa->po;
if (req->down)
{
OSPF_TRACE(D_EVENTS, "BFD session down for nbr %R on %s",
n->rid, n->ifa->ifname);
ospf_neigh_sm(n, INM_INACTTIM);
}
}
void
ospf_neigh_update_bfd(struct ospf_neighbor *n, int use_bfd)
{
struct ospf_proto *p = n->ifa->oa->po;
if (use_bfd && !n->bfd_req)
n->bfd_req = bfd_request_session(n->pool, n->ip, n->ifa->addr->ip,
n->ifa->iface, p->p.vrf,
ospf_neigh_bfd_hook, n, p->p.loop, NULL);
if (!use_bfd && n->bfd_req)
{
rfree(n->bfd_req);
n->bfd_req = NULL;
}
}
static void
dbdes_timer_hook(timer *t)
{
struct ospf_neighbor *n = t->data;
struct ospf_proto *p = n->ifa->oa->po;
// OSPF_TRACE(D_EVENTS, "DBDES timer expired for nbr %R on %s", n->rid, n->ifa->ifname);
if (n->state == NEIGHBOR_EXSTART)
ospf_send_dbdes(p, n);
if ((n->state == NEIGHBOR_EXCHANGE) && (n->myimms & DBDES_MS))
ospf_rxmt_dbdes(p, n);
if ((n->state > NEIGHBOR_LOADING) && !(n->myimms & DBDES_MS))
{
ospf_reset_ldd(p, n);
tm_stop(n->dbdes_timer);
}
}
static void
lsrq_timer_hook(timer *t)
{
struct ospf_neighbor *n = t->data;
struct ospf_proto *p = n->ifa->oa->po;
// OSPF_TRACE(D_EVENTS, "LSRQ timer expired for nbr %R on %s", n->rid, n->ifa->ifname);
if ((n->state >= NEIGHBOR_EXCHANGE) && !EMPTY_SLIST(n->lsrql))
ospf_send_lsreq(p, n);
}
static void
lsrt_timer_hook(timer *t)
{
struct ospf_neighbor *n = t->data;
struct ospf_proto *p = n->ifa->oa->po;
// OSPF_TRACE(D_EVENTS, "LSRT timer expired for nbr %R on %s", n->rid, n->ifa->ifname);
if ((n->state >= NEIGHBOR_EXCHANGE) && !EMPTY_SLIST(n->lsrtl))
ospf_rxmt_lsupd(p, n);
}
static void
ackd_timer_hook(timer *t)
{
struct ospf_neighbor *n = t->data;
struct ospf_proto *p = n->ifa->oa->po;
ospf_send_lsack(p, n, ACKL_DELAY);
}
void
ospf_sh_neigh_info(struct ospf_neighbor *n)
{
struct ospf_iface *ifa = n->ifa;
char *pos = "PtP ";
if ((ifa->type == OSPF_IT_BCAST) || (ifa->type == OSPF_IT_NBMA))
{
if (n->rid == ifa->drid)
pos = "DR ";
else if (n->rid == ifa->bdrid)
pos = "BDR ";
else
pos = "Other";
}
cli_msg(-1013, "%-12R\t%3u\t%s/%s\t%6t\t%-10s %I",
n->rid, n->priority, ospf_ns_names[n->state], pos,
tm_remains(n->inactim), ifa->ifname, n->ip);
}