Implement proper LSA ID generation.

nest/route.h

@@ -37,6 +37,7 @@ struct fib_node {
   byte pxlen;
   byte flags;				/* User-defined */
   byte x0, x1;				/* User-defined */
+  u32 uid;				/* Unique ID based on hash */
   ip_addr prefix;			/* In host order */
 };
 

nest/rt-fib.c

@@ -172,6 +172,28 @@ fib_find(struct fib *f, ip_addr *a, int len)
   return e;
 }
 
+/*
+int
+fib_histogram(struct fib *f)
+{
+  log(L_WARN "Histogram dump start %d %d", f->hash_size, f->entries);
+
+  int i, j;
+  struct fib_node *e;
+
+  for (i = 0; i < f->hash_size; i++)
+    {
+      j = 0;
+      for (e = f->hash_table[i]; e != NULL; e = e->next)
+	j++;
+      if (j > 0)
+        log(L_WARN "Histogram line %d: %d", i, j);
+    }
+
+  log(L_WARN "Histogram dump end");
+}
+*/
+
 /**
  * fib_get - find or create a FIB node
  * @f: FIB to work with
@@ -187,6 +209,7 @@ fib_get(struct fib *f, ip_addr *a, int len)
   unsigned int h = ipa_hash(*a);
   struct fib_node **ee = f->hash_table + (h >> f->hash_shift);
   struct fib_node *g, *e = *ee;
+  u32 uid = h << 16;
 
   while (e && (e->pxlen != len || !ipa_equal(*a, e->prefix)))
     e = e->next;
@@ -196,17 +219,31 @@ fib_get(struct fib *f, ip_addr *a, int len)
   if (len < 0 || len > BITS_PER_IP_ADDRESS || !ip_is_prefix(*a,len))
     bug("fib_get() called for invalid address");
 #endif
+
+  while ((g = *ee) && g->uid < uid)
+    ee = &g->next;
+  while ((g = *ee) && g->uid == uid)
+    {
+      ee = &g->next;
+      uid++;
+    }
+
+  if ((uid >> 16) != h)
+    log(L_ERR "FIB hash table chains are too long");
+
+  // log (L_WARN "FIB_GET %I %x %x", *a, h, uid);
+
   e = sl_alloc(f->fib_slab);
   e->prefix = *a;
   e->pxlen = len;
-  while ((g = *ee) && ipa_hash(g->prefix) < h)
-    ee = &g->next;
   e->next = *ee;
+  e->uid = uid;
   *ee = e;
   e->readers = NULL;
   f->init(e);
   if (f->entries++ > f->entries_max)
     fib_rehash(f, HASH_HI_STEP);
+
   return e;
 }
 

proto/ospf/ospf.c

@@ -605,7 +605,9 @@ ospf_reconfigure(struct proto *p, struct proto_config *c)
   struct area_net_config *anc;
   struct area_net *an;
 
-  po->rfc1583 = new->rfc1583;
+  if (po->rfc1583 != new->rfc1583)
+    return 0;
+
   schedule_rtcalc(po);
 
   po->tick = new->tick;

proto/ospf/rt.c

@@ -18,7 +18,7 @@ static void ospf_ext_spf(struct proto_ospf *po);
 static void rt_sync(struct proto_ospf *po);
 
 /* In ospf_area->rtr we store paths to routers, but we use RID (and not IP address)
-   as index, so we need to encapsulate RID to IP addresss */
+   as index, so we need to encapsulate RID to IP address */
 #ifdef OSPFv2
 #define ipa_from_rid(x) _MI(x)
 #else /* OSPFv3 */

proto/ospf/topology.c

@@ -30,11 +30,64 @@ void flush_prefix_net_lsa(struct ospf_iface *ifa);
 #define ipa_to_rid(x) _I3(x)
 #endif
 
-/* FIXME very ugly hack */
+
 #ifdef OSPFv2
-#define ipa_to_lsaid(x) _I(x)
+static inline u32
+fibnode_to_lsaid(struct proto_ospf *po, struct fib_node *fn)
+{
+  /* We have to map IP prefixes to u32 in such manner that resulting
+     u32 interpreted as IP address is a member of given
+     prefix. Therefore, /32 prefix have to be mapped on itself.
+     All received prefixes have to be mapped on different u32s.
+
+     We have an assumption that if there is nontrivial (non-/32)
+     network prefix, then there is not /32 prefix for the first
+     and the last IP address of the network (these are usually
+     reserved, therefore it is not an important restriction).
+     The network prefix is mapped to the first or the last
+     IP address in the manner that disallow collisions - we
+     use IP address that cannot be used by parent prefix.
+
+     For example:
+     192.168.0.0/24 maps to 192.168.0.255
+     192.168.1.0/24 maps to 192.168.1.0
+     because 192.168.0.0 and 192.168.1.255 might be used by
+     192.168.0.0/23 .
+
+     This is not compatible with older RFC 1583, so we have an option
+     to the RFC 1583 compatible assignment (that always uses the first
+     address) which disallows subnetting.
+
+     Appendig E of RFC 2328 suggests different algorithm, that tries
+     to maximize both compatibility and subnetting. But as it is not
+     possible to have both reliably and the suggested algorithm was
+     unnecessary complicated and it does crazy things like changing
+     LSA ID for a network because different network appeared, we
+     choose a different way. */
+
+  u32 id = _I(fn->prefix);
+
+  if ((po->rfc1583) || (fn->pxlen == 0) || (fn->pxlen == 32))
+    return id;
+
+  if (id & (1 << (32 - fn->pxlen)))
+    return id;
+  else
+    return id | ~u32_mkmask(fn->pxlen);
+}
+
 #else /* OSPFv3 */
-#define ipa_to_lsaid(x) _I0(x) ^ _I1(x) ^ _I2(x) ^ _I3(x)
+
+static inline u32
+fibnode_to_lsaid(struct proto_ospf *po, struct fib_node *fn)
+{
+  /* In OSPFv3, it is simpler. There is not a requirement
+     for membership of the result in input network, so we
+     just use hash-based unique ID. */
+
+  return fn->uid;
+}
+
 #endif
 
 
@@ -660,8 +713,7 @@ originate_sum_lsa(struct ospf_area *oa, struct fib_node *fn, int type, int metri
 
   if (type == ORT_NET)
     {
-      /* FIXME proper handling of LSA IDs and check for the same network */
+      lsa.id = fibnode_to_lsaid(po, fn);
-      lsa.id = ipa_to_lsaid(fn->prefix);
       lsa.type = LSA_T_SUM_NET;
     }
   else
@@ -710,8 +762,7 @@ flush_sum_lsa(struct ospf_area *oa, struct fib_node *fn, int type)
   lsa.rt = rid;
   if (type == ORT_NET)
     {
-      /* FIXME proper handling of LSA IDs and check for the same network */
+      lsa.id = fibnode_to_lsaid(po, fn);
-      lsa.id = ipa_to_lsaid(fn->prefix);
       lsa.type = LSA_T_SUM_NET;
     }
   else
@@ -721,6 +772,8 @@ flush_sum_lsa(struct ospf_area *oa, struct fib_node *fn, int type)
       lsa.type = LSA_T_SUM_RT;
     }
 
+  /* FIXME check for the same network */
+
   if ((en = ospf_hash_find_header(po->gr, oa->areaid, &lsa)) != NULL)
     {
       struct ospf_lsa_sum *sum = en->lsa_body;
@@ -892,7 +945,7 @@ originate_ext_lsa(net * n, rte * e, struct proto_ospf *po,
 #endif
 
   /* FIXME proper handling of LSA IDs and check for the same network */
-  lsa.id = ipa_to_lsaid(n->n.prefix);
+  lsa.id = fibnode_to_lsaid(po, &n->n);
 
   if ((en = ospf_hash_find_header(po->gr, 0, &lsa)) != NULL)
     {
@@ -927,7 +980,7 @@ flush_ext_lsa(net *n, struct proto_ospf *po)
 	     n->n.prefix, n->n.pxlen);
 
   /* FIXME proper handling of LSA IDs and check for the same network */
-  u32 lsaid = ipa_to_lsaid(n->n.prefix);
+  u32 lsaid = fibnode_to_lsaid(po, &n->n);
 
   if (en = ospf_hash_find(po->gr, 0, lsaid, rid, LSA_T_EXT))
     {