xref: /src/usr.sbin/mrouted/prune.c

/*	$NetBSD: prune.c,v 1.17.28.1 2009/05/13 19:20:30 jym Exp $	*/

/*
 * The mrouted program is covered by the license in the accompanying file
 * named "LICENSE".  Use of the mrouted program represents acceptance of
 * the terms and conditions listed in that file.
 *
 * The mrouted program is COPYRIGHT 1989 by The Board of Trustees of
 * Leland Stanford Junior University.
 */


#include "defs.h"

extern int cache_lifetime;
extern int max_prune_lifetime;
extern struct rtentry *routing_table;

extern int phys_vif;

/*
 * dither cache lifetime to obtain a value between x and 2*x
 */
#define CACHE_LIFETIME(x) ((x) + (arc4random() % (x)))

#define CHK_GS(x, y) {	\
		switch(x) { \
			case 2:	\
			case 4:	\
			case 8:	\
			case 16: \
			case 32: \
			case 64: \
			case 128: \
			/*case 256:*/ y = 1; \
				  break; \
			default:  y = 0; \
		} \
	}

struct gtable *kernel_table;		/* ptr to list of kernel grp entries*/
static struct gtable *kernel_no_route;	/* list of grp entries w/o routes   */
struct gtable *gtp;			/* pointer for kernel rt entries    */
unsigned int kroutes;			/* current number of cache entries  */

/****************************************************************************
                       Functions that are local to prune.c
****************************************************************************/
static void		prun_add_ttls(struct gtable *gt);
static int		pruning_neighbor(vifi_t vifi, u_int32_t addr);
static int		can_mtrace(vifi_t vifi, u_int32_t addr);
static struct ptable *	find_prune_entry(u_int32_t vr, struct ptable *pt);
static void		expire_prune(vifi_t vifi, struct gtable *gt);
static void		send_prune(struct gtable *gt);
static void		send_graft(struct gtable *gt);
static void		send_graft_ack(u_int32_t src, u_int32_t dst,
				       u_int32_t origin, u_int32_t grp);
static void		update_kernel(struct gtable *g);
static const char *	scaletime(u_long t);

/*
 * Updates the ttl values for each vif.
 */
static void
prun_add_ttls(struct gtable *gt)
{
    struct uvif *v;
    vifi_t vifi;

    for (vifi = 0, v = uvifs; vifi < numvifs; ++vifi, ++v) {
	if (VIFM_ISSET(vifi, gt->gt_grpmems))
	    gt->gt_ttls[vifi] = v->uv_threshold;
	else
	    gt->gt_ttls[vifi] = 0;
    }
}

/*
 * checks for scoped multicast addresses
 */
#define GET_SCOPE(gt) { \
	vifi_t _i; \
	if ((ntohl((gt)->gt_mcastgrp) & 0xff000000) == 0xef000000) \
	    for (_i = 0; _i < numvifs; _i++) \
		if (scoped_addr(_i, (gt)->gt_mcastgrp)) \
		    VIFM_SET(_i, (gt)->gt_scope); \
	}

int
scoped_addr(vifi_t vifi, u_int32_t addr)
{
    struct vif_acl *acl;

    for (acl = uvifs[vifi].uv_acl; acl; acl = acl->acl_next)
	if ((addr & acl->acl_mask) == acl->acl_addr)
	    return 1;

    return 0;
}

/*
 * Determine if mcastgrp has a listener on vifi
 */
int
grplst_mem(vifi_t vifi, u_int32_t mcastgrp)
{
    struct listaddr *g;
    struct uvif *v;

    v = &uvifs[vifi];

    for (g = v->uv_groups; g != NULL; g = g->al_next)
	if (mcastgrp == g->al_addr)
	    return 1;

    return 0;
}

/*
 * Finds the group entry with the specified source and netmask.
 * If netmask is 0, it uses the route's netmask.
 *
 * Returns TRUE if found a match, and the global variable gtp is left
 * pointing to entry before the found entry.
 * Returns FALSE if no exact match found, gtp is left pointing to before
 * the entry in question belongs, or is NULL if the it belongs at the
 * head of the list.
 */
int
find_src_grp(u_int32_t src, u_int32_t mask, u_int32_t grp)
{
    struct gtable *gt;

    gtp = NULL;
    gt = kernel_table;
    while (gt != NULL) {
	if (grp == gt->gt_mcastgrp &&
	    (mask ? (gt->gt_route->rt_origin == src &&
		     gt->gt_route->rt_originmask == mask) :
		    ((src & gt->gt_route->rt_originmask) ==
		     gt->gt_route->rt_origin)))
	    return TRUE;
	if (ntohl(grp) > ntohl(gt->gt_mcastgrp) ||
	    (grp == gt->gt_mcastgrp &&
	     (ntohl(mask) < ntohl(gt->gt_route->rt_originmask) ||
	      (mask == gt->gt_route->rt_originmask &&
	       (ntohl(src) > ntohl(gt->gt_route->rt_origin)))))) {
	    gtp = gt;
	    gt = gt->gt_gnext;
	}
	else break;
    }
    return FALSE;
}

/*
 * Check if the neighbor supports pruning
 */
static int
pruning_neighbor(vifi_t vifi, u_int32_t addr)
{
    struct listaddr *n = neighbor_info(vifi, addr);
    int vers;

    if (n == NULL)
	return 0;

    if (n->al_flags & NF_PRUNE)
	return 1;

    /*
     * Versions from 3.0 to 3.4 relied on the version number to identify
     * that they could handle pruning.
     */
    vers = NBR_VERS(n);
    return (vers >= 0x0300 && vers <= 0x0304);
}

/*
 * Can the neighbor in question handle multicast traceroute?
 */
static int
can_mtrace(vifi_t vifi, u_int32_t addr)
{
    struct listaddr *n = neighbor_info(vifi, addr);
    int vers;

    if (n == NULL)
	return 0;

    if (n->al_flags & NF_MTRACE)
	return 1;

    /*
     * Versions 3.3 and 3.4 relied on the version number to identify
     * that they could handle traceroute.
     */
    vers = NBR_VERS(n);
    return (vers >= 0x0303 && vers <= 0x0304);
}

/*
 * Returns the prune entry of the router, or NULL if none exists
 */
static struct ptable *
find_prune_entry(u_int32_t vr, struct ptable *pt)
{
    while (pt) {
	if (pt->pt_router == vr)
	    return pt;
	pt = pt->pt_next;
    }

    return NULL;
}

/*
 * Send a prune message to the dominant router for
 * this source.
 *
 * Record an entry that a prune was sent for this group
 */
static void
send_prune(struct gtable *gt)
{
    struct ptable *pt;
    char *p;
    int i;
    int datalen;
    u_int32_t src;
    u_int32_t dst;
    u_int32_t tmp;

    /* Don't process any prunes if router is not pruning */
    if (pruning == 0)
	return;

    /* Can't process a prune if we don't have an associated route */
    if (gt->gt_route == NULL)
	return;

    /* Don't send a prune to a non-pruning router */
    if (!pruning_neighbor(gt->gt_route->rt_parent, gt->gt_route->rt_gateway))
	return;

    /*
     * sends a prune message to the router upstream.
     */
    src = uvifs[gt->gt_route->rt_parent].uv_lcl_addr;
    dst = gt->gt_route->rt_gateway;

    p = send_buf + MIN_IP_HEADER_LEN + IGMP_MINLEN;
    datalen = 0;

    /*
     * determine prune lifetime
     */
    gt->gt_prsent_timer = gt->gt_timer;
    for (pt = gt->gt_pruntbl; pt; pt = pt->pt_next)
	if (pt->pt_timer < gt->gt_prsent_timer)
	    gt->gt_prsent_timer = pt->pt_timer;

    /*
     * If we have a graft pending, cancel graft retransmission
     */
    gt->gt_grftsnt = 0;

    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(gt->gt_route->rt_origin))[i];
    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(gt->gt_mcastgrp))[i];
    tmp = htonl(gt->gt_prsent_timer);
    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(tmp))[i];
    datalen += 12;

    send_igmp(src, dst, IGMP_DVMRP, DVMRP_PRUNE,
	      htonl(MROUTED_LEVEL), datalen);

    logit(LOG_DEBUG, 0, "sent prune for (%s %s)/%d on vif %d to %s",
      inet_fmts(gt->gt_route->rt_origin, gt->gt_route->rt_originmask),
      inet_fmt(gt->gt_mcastgrp),
      gt->gt_prsent_timer, gt->gt_route->rt_parent,
      inet_fmt(gt->gt_route->rt_gateway));
}

/*
 * a prune was sent upstream
 * so, a graft has to be sent to annul the prune
 * set up a graft timer so that if an ack is not
 * heard within that time, another graft request
 * is sent out.
 */
static void
send_graft(struct gtable *gt)
{
    char *p;
    int i;
    int datalen;
    u_int32_t src;
    u_int32_t dst;

    /* Can't send a graft without an associated route */
    if (gt->gt_route == NULL)
	return;

    src = uvifs[gt->gt_route->rt_parent].uv_lcl_addr;
    dst = gt->gt_route->rt_gateway;

    p = send_buf + MIN_IP_HEADER_LEN + IGMP_MINLEN;
    datalen = 0;

    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(gt->gt_route->rt_origin))[i];
    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(gt->gt_mcastgrp))[i];
    datalen += 8;

    if (datalen != 0) {
	send_igmp(src, dst, IGMP_DVMRP, DVMRP_GRAFT,
		  htonl(MROUTED_LEVEL), datalen);
    }
    logit(LOG_DEBUG, 0, "sent graft for (%s %s) to %s on vif %d",
	inet_fmts(gt->gt_route->rt_origin, gt->gt_route->rt_originmask),
	inet_fmt(gt->gt_mcastgrp),
	inet_fmt(gt->gt_route->rt_gateway),
	gt->gt_route->rt_parent);
}

/*
 * Send an ack that a graft was received
 */
static void
send_graft_ack(u_int32_t src, u_int32_t dst, u_int32_t origin, u_int32_t grp)
{
    char *p;
    int i;
    int datalen;

    p = send_buf + MIN_IP_HEADER_LEN + IGMP_MINLEN;
    datalen = 0;

    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(origin))[i];
    for (i = 0; i < 4; i++)
	*p++ = ((char *)&(grp))[i];
    datalen += 8;

    send_igmp(src, dst, IGMP_DVMRP, DVMRP_GRAFT_ACK,
	      htonl(MROUTED_LEVEL), datalen);

    logit(LOG_DEBUG, 0, "sent graft ack for (%s, %s) to %s",
	inet_fmt(origin), inet_fmt(grp),
	inet_fmt(dst));
}

/*
 * Update the kernel cache with all the routes hanging off the group entry
 */
static void
update_kernel(struct gtable *g)
{
    struct stable *st;

    for (st = g->gt_srctbl; st; st = st->st_next)
	k_add_rg(st->st_origin, g);
}

/****************************************************************************
                          Functions that are used externally
****************************************************************************/

#ifdef SNMP
#include <sys/types.h>
#include "snmp.h"

/*
 * Find a specific group entry in the group table
 */
struct gtable *
find_grp(u_long grp)
{
   struct gtable *gt;

   for (gt = kernel_table; gt; gt = gt->gt_gnext) {
      if (ntohl(grp) < ntohl(gt->gt_mcastgrp))
      	 break;
      if (gt->gt_mcastgrp == grp)
         return gt;
   }
   return NULL;
}

/*
 * Given a group entry and source, find the corresponding source table
 * entry
 */
struct stable *
find_grp_src(struct gtable *gt, u_long src)
{
   struct stable *st;
   u_long grp = gt->gt_mcastgrp;
   struct gtable *gtcurr;

   for (gtcurr = gt; gtcurr->gt_mcastgrp == grp; gtcurr = gtcurr->gt_gnext) {
      for (st = gtcurr->gt_srctbl; st; st = st->st_next)
	 if (st->st_origin == src)
	    return st;
   }
   return NULL;
}

/*
 * Find next entry > specification
 *
 * gtpp: ordered by group
 * stpp: ordered by source
 */
int
next_grp_src_mask(struct gtable **gtpp, struct stable **stpp, u_long grp,
		  u_long src, u_long mask)
{
   struct gtable *gt, *gbest = NULL;
   struct stable *st, *sbest = NULL;

   /* Find first group entry >= grp spec */
   (*gtpp) = kernel_table;
   while ((*gtpp) && ntohl((*gtpp)->gt_mcastgrp) < ntohl(grp))
      (*gtpp)=(*gtpp)->gt_gnext;
   if (!(*gtpp))
      return 0; /* no more groups */

   for (gt = kernel_table; gt; gt=gt->gt_gnext) {
      /* Since grps are ordered, we can stop when group changes from gbest */
      if (gbest && gbest->gt_mcastgrp != gt->gt_mcastgrp)
         break;
      for (st = gt->gt_srctbl; st; st=st->st_next) {

         /* Among those entries > spec, find "lowest" one */
         if (((ntohl(gt->gt_mcastgrp)> ntohl(grp))
           || (ntohl(gt->gt_mcastgrp)==ntohl(grp)
              && ntohl(st->st_origin)> ntohl(src))
           || (ntohl(gt->gt_mcastgrp)==ntohl(grp)
              && ntohl(st->st_origin)==src && 0xFFFFFFFF>ntohl(mask)))
          && (!gbest
           || (ntohl(gt->gt_mcastgrp)< ntohl(gbest->gt_mcastgrp))
           || (ntohl(gt->gt_mcastgrp)==ntohl(gbest->gt_mcastgrp)
              && ntohl(st->st_origin)< ntohl(sbest->st_origin)))) {
               gbest = gt;
               sbest = st;
         }
      }
   }
   (*gtpp) = gbest;
   (*stpp) = sbest;
   return (*gtpp)!=0;
}

/*
 * Ensure that sg contains current information for the given group,source.
 * This is fetched from the kernel as a unit so that counts for the entry
 * are consistent, i.e. packet and byte counts for the same entry are
 * read at the same time.
 */
void
refresh_sg(struct sioc_sg_req *sg, struct gtable *gt, struct stable *st)
{
   static   int lastq = -1;

   if (quantum != lastq || sg->src.s_addr!=st->st_origin
    || sg->grp.s_addr!=gt->gt_mcastgrp) {
       lastq = quantum;
       sg->src.s_addr = st->st_origin;
       sg->grp.s_addr = gt->gt_mcastgrp;
       ioctl(igmp_socket, SIOCGETSGCNT, (char *)sg);
   }
}

/*
 * Return pointer to a specific route entry.  This must be a separate
 * function from find_route() which modifies rtp.
 */
struct rtentry *
snmp_find_route(u_long src, u_long mask)
{
    struct rtentry *rt;

   for (rt = routing_table; rt; rt = rt->rt_next) {
      if (src == rt->rt_origin && mask == rt->rt_originmask)
         return rt;
   }
   return NULL;
}

/*
 * Find next route entry > specification
 */
int
next_route(struct rtentry **rtpp, u_long src, u_long mask)
{
   struct rtentry *rt, *rbest = NULL;

   /* Among all entries > spec, find "lowest" one in order */
   for (rt = routing_table; rt; rt=rt->rt_next) {
      if ((ntohl(rt->rt_origin) > ntohl(src)
          || (ntohl(rt->rt_origin) == ntohl(src)
             && ntohl(rt->rt_originmask) > ntohl(mask)))
       && (!rbest || (ntohl(rt->rt_origin) < ntohl(rbest->rt_origin))
          || (ntohl(rt->rt_origin) == ntohl(rbest->rt_origin)
             && ntohl(rt->rt_originmask) < ntohl(rbest->rt_originmask))))
               rbest = rt;
   }
   (*rtpp) = rbest;
   return (*rtpp)!=0;
}

/*
 * Given a routing table entry, and a vifi, find the next vifi/entry
 *
 * vifi: vifi at which to start looking
 */
int
next_route_child(struct rtentry **rtpp, u_long src, u_long mask, vifi_t *vifi)
{
   struct rtentry *rt;

   /* Get (S,M) entry */
   if (!((*rtpp) = snmp_find_route(src,mask)))
      if (!next_route(rtpp, src, mask))
         return 0;

   /* Continue until we get one with a valid next vif */
   do {
      for (; (*rtpp)->rt_children && *vifi<numvifs; (*vifi)++)
         if (VIFM_ISSET(*vifi, (*rtpp)->rt_children))
            return 1;
      *vifi = 0;
   } while( next_route(rtpp, (*rtpp)->rt_origin, (*rtpp)->rt_originmask) );

   return 0;
}

/*
 * Given a routing table entry, and a vifi, find the next entry
 * equal to or greater than those
 *
 * vifi: vifi at which to start looking
 */
int
next_child(struct gtable **gtpp, struct stable **stpp, u_long grp, u_long src,
	   u_long mask, vifi_t *vifi)
{
   struct stable *st;

   /* Get (G,S,M) entry */
   if (mask!=0xFFFFFFFF
    || !((*gtpp) = find_grp(grp))
    || !((*stpp) = find_grp_src((*gtpp),src)))
      if (!next_grp_src_mask(gtpp, stpp, grp, src, mask))
         return 0;

   /* Continue until we get one with a valid next vif */
   do {
      for (; (*gtpp)->gt_route->rt_children && *vifi<numvifs; (*vifi)++)
         if (VIFM_ISSET(*vifi, (*gtpp)->gt_route->rt_children))
            return 1;
      *vifi = 0;
   } while (next_grp_src_mask(gtpp, stpp, (*gtpp)->gt_mcastgrp,
		(*stpp)->st_origin, 0xFFFFFFFF) );

   return 0;
}
#endif /* SNMP */

/*
 * Initialize the kernel table structure
 */
void
init_ktable(void)
{
    kernel_table 	= NULL;
    kernel_no_route	= NULL;
    kroutes		= 0;
}

/*
 * Add a new table entry for (origin, mcastgrp)
 */
void
add_table_entry(u_int32_t origin, u_int32_t mcastgrp)
{
    struct rtentry *r;
    struct gtable *gt,**gtnp,*prev_gt;
    struct stable *st,**stnp;
    vifi_t i;

#ifdef DEBUG_MFC
    md_log(MD_MISS, origin, mcastgrp);
#endif

    r = determine_route(origin);
    prev_gt = NULL;
    if (r == NULL) {
	/*
	 * Look for it on the no_route table; if it is found then
	 * it will be detected as a duplicate below.
	 */
	for (gt = kernel_no_route; gt; gt = gt->gt_next)
	    if (mcastgrp == gt->gt_mcastgrp &&
		gt->gt_srctbl && gt->gt_srctbl->st_origin == origin)
			break;
	gtnp = &kernel_no_route;
    } else {
	gtnp = &r->rt_groups;
	while ((gt = *gtnp) != NULL) {
	    if (gt->gt_mcastgrp >= mcastgrp)
		break;
	    gtnp = &gt->gt_next;
	    prev_gt = gt;
	}
    }

    if (gt == NULL || gt->gt_mcastgrp != mcastgrp) {
	gt = (struct gtable *)malloc(sizeof(struct gtable));
	if (gt == NULL) {
	    logit(LOG_ERR, 0, "ran out of memory");
	    return;
	}

	gt->gt_mcastgrp	    = mcastgrp;
	gt->gt_timer   	    = CACHE_LIFETIME(cache_lifetime);
	time(&gt->gt_ctime);
	gt->gt_grpmems	    = 0;
	gt->gt_scope	    = 0;
	gt->gt_prsent_timer = 0;
	gt->gt_grftsnt	    = 0;
	gt->gt_srctbl	    = NULL;
	gt->gt_pruntbl	    = NULL;
	gt->gt_route	    = r;
#ifdef RSRR
	gt->gt_rsrr_cache   = NULL;
#endif

	if (r != NULL) {
	    /* obtain the multicast group membership list */
	    for (i = 0; i < numvifs; i++) {
		if (VIFM_ISSET(i, r->rt_children) &&
		    !(VIFM_ISSET(i, r->rt_leaves)))
		    VIFM_SET(i, gt->gt_grpmems);

		if (VIFM_ISSET(i, r->rt_leaves) && grplst_mem(i, mcastgrp))
		    VIFM_SET(i, gt->gt_grpmems);
	    }
	    GET_SCOPE(gt);
	    if (VIFM_ISSET(r->rt_parent, gt->gt_scope))
		gt->gt_scope = -1;
	    gt->gt_grpmems &= ~gt->gt_scope;
	} else {
	    gt->gt_scope = -1;
	    gt->gt_grpmems = 0;
	}

	/* update ttls */
	prun_add_ttls(gt);

	gt->gt_next = *gtnp;
	*gtnp = gt;
	if (gt->gt_next)
	    gt->gt_next->gt_prev = gt;
	gt->gt_prev = prev_gt;

	if (r) {
	    if (find_src_grp(r->rt_origin, r->rt_originmask, gt->gt_mcastgrp)) {
		struct gtable *g;

		g = gtp ? gtp->gt_gnext : kernel_table;
		logit(LOG_WARNING, 0, "Entry for (%s %s) (rt:%p) exists (rt:%p)",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp),
		    r, g->gt_route);
	    } else {
		if (gtp) {
		    gt->gt_gnext = gtp->gt_gnext;
		    gt->gt_gprev = gtp;
		    gtp->gt_gnext = gt;
		} else {
		    gt->gt_gnext = kernel_table;
		    gt->gt_gprev = NULL;
		    kernel_table = gt;
		}
		if (gt->gt_gnext)
		    gt->gt_gnext->gt_gprev = gt;
	    }
	} else {
	    gt->gt_gnext = gt->gt_gprev = NULL;
	}
    }

    stnp = &gt->gt_srctbl;
    while ((st = *stnp) != NULL) {
	if (ntohl(st->st_origin) >= ntohl(origin))
	    break;
	stnp = &st->st_next;
    }

    if (st == NULL || st->st_origin != origin) {
	st = (struct stable *)malloc(sizeof(struct stable));
	if (st == NULL)
	    logit(LOG_ERR, 0, "ran out of memory");

	st->st_origin = origin;
	st->st_pktcnt = 0;
	st->st_next = *stnp;
	*stnp = st;
    } else {
#ifdef DEBUG_MFC
	md_log(MD_DUPE, origin, mcastgrp);
#endif
	logit(LOG_WARNING, 0, "kernel entry already exists for (%s %s)",
		inet_fmt(origin),
		inet_fmt(mcastgrp));
	/* XXX Doing this should cause no harm, and may ensure
	 * kernel<>mrouted synchronization */
	k_add_rg(origin, gt);
	return;
    }

    kroutes++;
    k_add_rg(origin, gt);

    logit(LOG_DEBUG, 0, "add cache entry (%s %s) gm:%x, parent-vif:%d",
	inet_fmt(origin),
	inet_fmt(mcastgrp),
	gt->gt_grpmems, r ? r->rt_parent : -1);

    /* If there are no leaf vifs
     * which have this group, then
     * mark this src-grp as a prune candidate.
     */
    if (!gt->gt_prsent_timer && !gt->gt_grpmems && r && r->rt_gateway)
	send_prune(gt);
}

/*
 * An mrouter has gone down and come up on an interface
 * Forward on that interface immediately
 */
void
reset_neighbor_state(vifi_t vifi, u_int32_t addr)
{
    struct rtentry *r;
    struct gtable *g;
    struct ptable *pt, **ptnp;
    struct stable *st;

    for (g = kernel_table; g; g = g->gt_gnext) {
	r = g->gt_route;

	/*
	 * If neighbor was the parent, remove the prune sent state
	 * and all of the source cache info so that prunes get
	 * regenerated.
	 */
	if (vifi == r->rt_parent) {
	    if (addr == r->rt_gateway) {
		logit(LOG_DEBUG, 0, "reset_neighbor_state parent reset (%s %s)",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp));

		g->gt_prsent_timer = 0;
		g->gt_grftsnt = 0;
		while ((st = g->gt_srctbl) != NULL) {
		    g->gt_srctbl = st->st_next;
		    k_del_rg(st->st_origin, g);
		    kroutes--;
		    free(st);
		}
	    }
	} else {
	    /*
	     * Neighbor was not the parent, send grafts to join the groups
	     */
	    if (g->gt_prsent_timer) {
		g->gt_grftsnt = 1;
		send_graft(g);
		g->gt_prsent_timer = 0;
	    }

	    /*
	     * Remove any prunes that this router has sent us.
	     */
	    ptnp = &g->gt_pruntbl;
	    while ((pt = *ptnp) != NULL) {
		if (pt->pt_vifi == vifi && pt->pt_router == addr) {
		    *ptnp = pt->pt_next;
		    free(pt);
		} else
		    ptnp = &pt->pt_next;
	    }

	    /*
	     * And see if we want to forward again.
	     */
	    if (!VIFM_ISSET(vifi, g->gt_grpmems)) {
		if (VIFM_ISSET(vifi, r->rt_children) &&
		    !(VIFM_ISSET(vifi, r->rt_leaves)))
		    VIFM_SET(vifi, g->gt_grpmems);

		if (VIFM_ISSET(vifi, r->rt_leaves) &&
		    grplst_mem(vifi, g->gt_mcastgrp))
		    VIFM_SET(vifi, g->gt_grpmems);

		g->gt_grpmems &= ~g->gt_scope;
		prun_add_ttls(g);

		/* Update kernel state */
		update_kernel(g);
#ifdef RSRR
		/* Send route change notification to reservation protocol. */
		rsrr_cache_send(g,1);
#endif /* RSRR */

		logit(LOG_DEBUG, 0, "reset member state (%s %s) gm:%x",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp), g->gt_grpmems);
	    }
	}
    }
}

/*
 * Delete table entry from the kernel
 * del_flag determines how many entries to delete
 */
void
del_table_entry(struct rtentry *r, u_int32_t mcastgrp, u_int del_flag)
{
    struct gtable *g, *prev_g;
    struct stable *st, *prev_st;
    struct ptable *pt, *prev_pt;

    if (del_flag == DEL_ALL_ROUTES) {
	g = r->rt_groups;
	while (g) {
	    logit(LOG_DEBUG, 0, "del_table_entry deleting (%s %s)",
		inet_fmts(r->rt_origin, r->rt_originmask),
		inet_fmt(g->gt_mcastgrp));
	    st = g->gt_srctbl;
	    while (st) {
		if (k_del_rg(st->st_origin, g) < 0) {
		    logit(LOG_WARNING, errno,
			"del_table_entry trying to delete (%s, %s)",
			inet_fmt(st->st_origin),
			inet_fmt(g->gt_mcastgrp));
		}
		kroutes--;
		prev_st = st;
		st = st->st_next;
		free(prev_st);
	    }
	    g->gt_srctbl = NULL;

	    pt = g->gt_pruntbl;
	    while (pt) {
		prev_pt = pt;
		pt = pt->pt_next;
		free(prev_pt);
	    }
	    g->gt_pruntbl = NULL;

	    if (g->gt_gnext)
		g->gt_gnext->gt_gprev = g->gt_gprev;
	    if (g->gt_gprev)
		g->gt_gprev->gt_gnext = g->gt_gnext;
	    else
		kernel_table = g->gt_gnext;

#ifdef RSRR
	    /* Send route change notification to reservation protocol. */
	    rsrr_cache_send(g,0);
	    rsrr_cache_clean(g);
#endif /* RSRR */
	    prev_g = g;
	    g = g->gt_next;
	    free(prev_g);
	}
	r->rt_groups = NULL;
    }

    /*
     * Dummy routine - someday this may be needed, so it is just there
     */
    if (del_flag == DEL_RTE_GROUP) {
	prev_g = (struct gtable *)&r->rt_groups;
	for (g = r->rt_groups; g; g = g->gt_next) {
	    if (g->gt_mcastgrp == mcastgrp) {
		logit(LOG_DEBUG, 0, "del_table_entry deleting (%s %s)",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp));
		st = g->gt_srctbl;
		while (st) {
		    if (k_del_rg(st->st_origin, g) < 0) {
			logit(LOG_WARNING, errno,
			    "del_table_entry trying to delete (%s, %s)",
			    inet_fmt(st->st_origin),
			    inet_fmt(g->gt_mcastgrp));
		    }
		    kroutes--;
		    prev_st = st;
		    st = st->st_next;
		    free(prev_st);
		}
		g->gt_srctbl = NULL;

		pt = g->gt_pruntbl;
		while (pt) {
		    prev_pt = pt;
		    pt = pt->pt_next;
		    free(prev_pt);
		}
		g->gt_pruntbl = NULL;

		if (g->gt_gnext)
		    g->gt_gnext->gt_gprev = g->gt_gprev;
		if (g->gt_gprev)
		    g->gt_gprev->gt_gnext = g->gt_gnext;
		else
		    kernel_table = g->gt_gnext;

		if (prev_g != (struct gtable *)&r->rt_groups)
		    g->gt_next->gt_prev = prev_g;
		else
		    g->gt_next->gt_prev = NULL;
		prev_g->gt_next = g->gt_next;

#ifdef RSRR
		/* Send route change notification to reservation protocol. */
		rsrr_cache_send(g,0);
		rsrr_cache_clean(g);
#endif /* RSRR */
		free(g);
		g = prev_g;
	    } else {
		prev_g = g;
	    }
	}
    }
}

/*
 * update kernel table entry when a route entry changes
 */
void
update_table_entry(struct rtentry *r)
{
    struct gtable *g;
    struct ptable *pt, *prev_pt;
    vifi_t i;

    for (g = r->rt_groups; g; g = g->gt_next) {
	pt = g->gt_pruntbl;
	while (pt) {
	    prev_pt = pt->pt_next;
	    free(pt);
	    pt = prev_pt;
	}
	g->gt_pruntbl = NULL;

	g->gt_grpmems = 0;

	/* obtain the multicast group membership list */
	for (i = 0; i < numvifs; i++) {
	    if (VIFM_ISSET(i, r->rt_children) &&
		!(VIFM_ISSET(i, r->rt_leaves)))
		VIFM_SET(i, g->gt_grpmems);

	    if (VIFM_ISSET(i, r->rt_leaves) && grplst_mem(i, g->gt_mcastgrp))
		VIFM_SET(i, g->gt_grpmems);
	}
	if (VIFM_ISSET(r->rt_parent, g->gt_scope))
	    g->gt_scope = -1;
	g->gt_grpmems &= ~g->gt_scope;

	logit(LOG_DEBUG, 0, "updating cache entries (%s %s) gm:%x",
	    inet_fmts(r->rt_origin, r->rt_originmask),
	    inet_fmt(g->gt_mcastgrp),
	    g->gt_grpmems);

	if (g->gt_grpmems && g->gt_prsent_timer) {
	    g->gt_grftsnt = 1;
	    send_graft(g);
	    g->gt_prsent_timer = 0;
	}

	/* update ttls and add entry into kernel */
	prun_add_ttls(g);
	update_kernel(g);
#ifdef RSRR
	/* Send route change notification to reservation protocol. */
	rsrr_cache_send(g,1);
#endif /* RSRR */

	/* Check if we want to prune this group */
	if (!g->gt_prsent_timer && g->gt_grpmems == 0 && r->rt_gateway) {
	    g->gt_timer = CACHE_LIFETIME(cache_lifetime);
	    send_prune(g);
	}
    }
}

/*
 * set the forwarding flag for all mcastgrps on this vifi
 */
void
update_lclgrp(vifi_t vifi, u_int32_t mcastgrp)
{
    struct rtentry *r;
    struct gtable *g;

    logit(LOG_DEBUG, 0, "group %s joined on vif %d",
	inet_fmt(mcastgrp), vifi);

    for (g = kernel_table; g; g = g->gt_gnext) {
	if (ntohl(mcastgrp) < ntohl(g->gt_mcastgrp))
	    break;

	r = g->gt_route;
	if (g->gt_mcastgrp == mcastgrp &&
	    VIFM_ISSET(vifi, r->rt_children)) {

	    VIFM_SET(vifi, g->gt_grpmems);
	    g->gt_grpmems &= ~g->gt_scope;
	    if (g->gt_grpmems == 0)
		continue;

	    prun_add_ttls(g);
	    logit(LOG_DEBUG, 0, "update lclgrp (%s %s) gm:%x",
		inet_fmts(r->rt_origin, r->rt_originmask),
		inet_fmt(g->gt_mcastgrp), g->gt_grpmems);

	    update_kernel(g);
#ifdef RSRR
	    /* Send route change notification to reservation protocol. */
	    rsrr_cache_send(g,1);
#endif /* RSRR */
	}
    }
}

/*
 * reset forwarding flag for all mcastgrps on this vifi
 */
void
delete_lclgrp(vifi_t vifi, u_int32_t mcastgrp)
{
    struct rtentry *r;
    struct gtable *g;

    logit(LOG_DEBUG, 0, "group %s left on vif %d",
	inet_fmt(mcastgrp), vifi);

    for (g = kernel_table; g; g = g->gt_gnext) {
	if (ntohl(mcastgrp) < ntohl(g->gt_mcastgrp))
	    break;

	if (g->gt_mcastgrp == mcastgrp) {
	    int stop_sending = 1;

	    r = g->gt_route;
	    /*
	     * If this is not a leaf, then we have router neighbors on this
	     * vif.  Only turn off forwarding if they have all pruned.
	     */
	    if (!VIFM_ISSET(vifi, r->rt_leaves)) {
		struct listaddr *vr;

		for (vr = uvifs[vifi].uv_neighbors; vr; vr = vr->al_next)
		  if (find_prune_entry(vr->al_addr, g->gt_pruntbl) == NULL) {
		      stop_sending = 0;
		      break;
		  }
	    }

	    if (stop_sending) {
		VIFM_CLR(vifi, g->gt_grpmems);
		logit(LOG_DEBUG, 0, "delete lclgrp (%s %s) gm:%x",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp), g->gt_grpmems);

		prun_add_ttls(g);
		update_kernel(g);
#ifdef RSRR
		/* Send route change notification to reservation protocol. */
		rsrr_cache_send(g,1);
#endif /* RSRR */

		/*
		 * If there are no more members of this particular group,
		 *  send prune upstream
		 */
		if (!g->gt_prsent_timer && g->gt_grpmems == 0 && r->rt_gateway)
		    send_prune(g);
	    }
	}
    }
}

/*
 * Takes the prune message received and then strips it to
 * determine the (src, grp) pair to be pruned.
 *
 * Adds the router to the (src, grp) entry then.
 *
 * Determines if further packets have to be sent down that vif
 *
 * Determines if a corresponding prune message has to be generated
 */
void
accept_prune(u_int32_t src, u_int32_t dst, char *p, int datalen)
{
    u_int32_t prun_src;
    u_int32_t prun_grp;
    u_int32_t prun_tmr;
    vifi_t vifi;
    int i;
    int stop_sending;
    struct rtentry *r;
    struct gtable *g;
    struct ptable *pt;
    struct listaddr *vr;

    /* Don't process any prunes if router is not pruning */
    if (pruning == 0)
	return;

    if ((vifi = find_vif(src, dst)) == NO_VIF) {
	logit(LOG_INFO, 0,
    	    "ignoring prune report from non-neighbor %s",
	    inet_fmt(src));
	return;
    }

    /* Check if enough data is present */
    if (datalen < 12)
	{
	    logit(LOG_WARNING, 0,
		"non-decipherable prune from %s",
		inet_fmt(src));
	    return;
	}

    for (i = 0; i< 4; i++)
	((char *)&prun_src)[i] = *p++;
    for (i = 0; i< 4; i++)
	((char *)&prun_grp)[i] = *p++;
    for (i = 0; i< 4; i++)
	((char *)&prun_tmr)[i] = *p++;
    prun_tmr = ntohl(prun_tmr);

    logit(LOG_DEBUG, 0, "%s on vif %d prunes (%s %s)/%d",
	inet_fmt(src), vifi,
	inet_fmt(prun_src), inet_fmt(prun_grp), prun_tmr);

    /*
     * Find the subnet for the prune
     */
    if (find_src_grp(prun_src, 0, prun_grp)) {
	g = gtp ? gtp->gt_gnext : kernel_table;
    	r = g->gt_route;

	if (!VIFM_ISSET(vifi, r->rt_children)) {
	    logit(LOG_WARNING, 0, "prune received from non-child %s for (%s %s)",
		inet_fmt(src), inet_fmt(prun_src),
		inet_fmt(prun_grp));
	    return;
	}
	if (VIFM_ISSET(vifi, g->gt_scope)) {
	    logit(LOG_WARNING, 0, "prune received from %s on scoped grp (%s %s)",
		inet_fmt(src), inet_fmt(prun_src),
		inet_fmt(prun_grp));
	    return;
	}
	if ((pt = find_prune_entry(src, g->gt_pruntbl)) != NULL) {
	    /*
	     * If it's about to expire, then it's only still around because
	     * of timer granularity, so don't warn about it.
	     */
	    if (pt->pt_timer > 10) {
		logit(LOG_WARNING, 0, "%s %d from %s for (%s %s)/%d %s %d %s %x",
		    "duplicate prune received on vif",
		    vifi, inet_fmt(src), inet_fmt(prun_src),
		    inet_fmt(prun_grp), prun_tmr,
		    "old timer:", pt->pt_timer, "cur gm:", g->gt_grpmems);
	    }
	    pt->pt_timer = prun_tmr;
	} else {
	    /* allocate space for the prune structure */
	    pt = (struct ptable *)(malloc(sizeof(struct ptable)));
	    if (pt == NULL) {
	      logit(LOG_ERR, 0, "pt: ran out of memory");
	      return;
	    }

	    pt->pt_vifi = vifi;
	    pt->pt_router = src;
	    pt->pt_timer = prun_tmr;

	    pt->pt_next = g->gt_pruntbl;
	    g->gt_pruntbl = pt;
	}

	/* Refresh the group's lifetime */
	g->gt_timer = CACHE_LIFETIME(cache_lifetime);
	if ((u_int32_t)g->gt_timer < prun_tmr)
	    g->gt_timer = prun_tmr;

	/*
	 * check if any more packets need to be sent on the
	 * vif which sent this message
	 */
	stop_sending = 1;
	for (vr = uvifs[vifi].uv_neighbors; vr; vr = vr->al_next)
	  if (find_prune_entry(vr->al_addr, g->gt_pruntbl) == NULL)  {
	      stop_sending = 0;
	      break;
	  }

	if (stop_sending && !grplst_mem(vifi, prun_grp)) {
	    VIFM_CLR(vifi, g->gt_grpmems);
	    logit(LOG_DEBUG, 0, "prune (%s %s), stop sending on vif %d, gm:%x",
		inet_fmts(r->rt_origin, r->rt_originmask),
		inet_fmt(g->gt_mcastgrp), vifi, g->gt_grpmems);

	    prun_add_ttls(g);
	    update_kernel(g);
#ifdef RSRR
	    /* Send route change notification to reservation protocol. */
	    rsrr_cache_send(g,1);
#endif /* RSRR */
	}

	/*
	 * check if all the child routers have expressed no interest
	 * in this group and if this group does not exist in the
	 * interface
	 * Send a prune message then upstream
	 */
	if (!g->gt_prsent_timer && g->gt_grpmems == 0 && r->rt_gateway) {
	    send_prune(g);
	}
    } else {
	/*
	 * There is no kernel entry for this group.  Therefore, we can
	 * simply ignore the prune, as we are not forwarding this traffic
	 * downstream.
	 */
	logit(LOG_DEBUG, 0, "%s (%s %s)/%d from %s",
	    "prune message received with no kernel entry for",
	    inet_fmt(prun_src), inet_fmt(prun_grp),
	    prun_tmr, inet_fmt(src));
	return;
    }
}

/*
 * Checks if this mcastgrp is present in the kernel table
 * If so and if a prune was sent, it sends a graft upwards
 */
void
chkgrp_graft(vifi_t vifi, u_int32_t mcastgrp)
{
    struct rtentry *r;
    struct gtable *g;

    for (g = kernel_table; g; g = g->gt_gnext) {
	if (ntohl(mcastgrp) < ntohl(g->gt_mcastgrp))
	    break;

	r = g->gt_route;
	if (g->gt_mcastgrp == mcastgrp && VIFM_ISSET(vifi, r->rt_children))
	    if (g->gt_prsent_timer) {
		VIFM_SET(vifi, g->gt_grpmems);

		/*
		 * If the vif that was joined was a scoped vif,
		 * ignore it ; don't graft back
		 */
		g->gt_grpmems &= ~g->gt_scope;
		if (g->gt_grpmems == 0)
		    continue;

		/* set the flag for graft retransmission */
		g->gt_grftsnt = 1;

		/* send graft upwards */
		send_graft(g);

		/* reset the prune timer and update cache timer*/
		g->gt_prsent_timer = 0;
		g->gt_timer = max_prune_lifetime;

		logit(LOG_DEBUG, 0, "chkgrp graft (%s %s) gm:%x",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp), g->gt_grpmems);

		prun_add_ttls(g);
		update_kernel(g);
#ifdef RSRR
		/* Send route change notification to reservation protocol. */
		rsrr_cache_send(g,1);
#endif /* RSRR */
	    }
    }
}

/* determine the multicast group and src
 *
 * if it does, then determine if a prune was sent
 * upstream.
 * if prune sent upstream, send graft upstream and send
 * ack downstream.
 *
 * if no prune sent upstream, change the forwarding bit
 * for this interface and send ack downstream.
 *
 * if no entry exists for this group send ack downstream.
 */
void
accept_graft(u_int32_t src, u_int32_t dst, char *p, int datalen)
{
    vifi_t 	vifi;
    u_int32_t 	graft_src;
    u_int32_t	graft_grp;
    int 	i;
    struct rtentry *r;
    struct gtable *g;
    struct ptable *pt, **ptnp;

    if ((vifi = find_vif(src, dst)) == NO_VIF) {
	logit(LOG_INFO, 0,
    	    "ignoring graft from non-neighbor %s",
	    inet_fmt(src));
	return;
    }

    if (datalen < 8) {
	logit(LOG_WARNING, 0,
	    "received non-decipherable graft from %s",
	    inet_fmt(src));
	return;
    }

    for (i = 0; i< 4; i++)
	((char *)&graft_src)[i] = *p++;
    for (i = 0; i< 4; i++)
	((char *)&graft_grp)[i] = *p++;

    logit(LOG_DEBUG, 0, "%s on vif %d grafts (%s %s)",
	inet_fmt(src), vifi,
	inet_fmt(graft_src), inet_fmt(graft_grp));

    /*
     * Find the subnet for the graft
     */
    if (find_src_grp(graft_src, 0, graft_grp)) {
	g = gtp ? gtp->gt_gnext : kernel_table;
	r = g->gt_route;

	if (VIFM_ISSET(vifi, g->gt_scope)) {
	    logit(LOG_WARNING, 0, "graft received from %s on scoped grp (%s %s)",
		inet_fmt(src), inet_fmt(graft_src),
		inet_fmt(graft_grp));
	    return;
	}

	ptnp = &g->gt_pruntbl;
	while ((pt = *ptnp) != NULL) {
	    if ((pt->pt_vifi == vifi) && (pt->pt_router == src)) {
		*ptnp = pt->pt_next;
		free(pt);

		VIFM_SET(vifi, g->gt_grpmems);
		logit(LOG_DEBUG, 0, "accept graft (%s %s) gm:%x",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(g->gt_mcastgrp), g->gt_grpmems);

		prun_add_ttls(g);
		update_kernel(g);
#ifdef RSRR
		/* Send route change notification to reservation protocol. */
		rsrr_cache_send(g,1);
#endif /* RSRR */
		break;
	    } else {
		ptnp = &pt->pt_next;
	    }
	}

	/* send ack downstream */
	send_graft_ack(dst, src, graft_src, graft_grp);
	g->gt_timer = max_prune_lifetime;

	if (g->gt_prsent_timer) {
	    /* set the flag for graft retransmission */
	    g->gt_grftsnt = 1;

	    /* send graft upwards */
	    send_graft(g);

	    /* reset the prune sent timer */
	    g->gt_prsent_timer = 0;
	}
    } else {
	/*
	 * We have no state for the source and group in question.
	 * We can simply acknowledge the graft, since we know
	 * that we have no prune state, and grafts are requests
	 * to remove prune state.
	 */
	send_graft_ack(dst, src, graft_src, graft_grp);
	logit(LOG_DEBUG, 0, "%s (%s %s) from %s",
	    "graft received with no kernel entry for",
	    inet_fmt(graft_src), inet_fmt(graft_grp),
	    inet_fmt(src));
	return;
    }
}

/*
 * find out which group is involved first of all
 * then determine if a graft was sent.
 * if no graft sent, ignore the message
 * if graft was sent and the ack is from the right
 * source, remove the graft timer so that we don't
 * have send a graft again
 */
void
accept_g_ack(u_int32_t src, u_int32_t dst, char *p, int datalen)
{
    struct gtable *g;
    vifi_t 	vifi;
    u_int32_t 	grft_src;
    u_int32_t	grft_grp;
    int 	i;

    if ((vifi = find_vif(src, dst)) == NO_VIF) {
	logit(LOG_INFO, 0,
    	    "ignoring graft ack from non-neighbor %s",
	    inet_fmt(src));
	return;
    }

    if (datalen < 0  || datalen > 8) {
	logit(LOG_WARNING, 0,
	    "received non-decipherable graft ack from %s",
	    inet_fmt(src));
	return;
    }

    for (i = 0; i< 4; i++)
	((char *)&grft_src)[i] = *p++;
    for (i = 0; i< 4; i++)
	((char *)&grft_grp)[i] = *p++;

    logit(LOG_DEBUG, 0, "%s on vif %d acks graft (%s, %s)",
	inet_fmt(src), vifi,
	inet_fmt(grft_src), inet_fmt(grft_grp));

    /*
     * Find the subnet for the graft ack
     */
    if (find_src_grp(grft_src, 0, grft_grp)) {
	g = gtp ? gtp->gt_gnext : kernel_table;
	g->gt_grftsnt = 0;
    } else {
	logit(LOG_WARNING, 0, "%s (%s, %s) from %s",
	    "rcvd graft ack with no kernel entry for",
	    inet_fmt(grft_src), inet_fmt(grft_grp),
	    inet_fmt(src));
	return;
    }
}


/*
 * free all prune entries and kernel routes
 * normally, this should inform the kernel that all of its routes
 * are going away, but this is only called by restart(), which is
 * about to call MRT_DONE which does that anyway.
 */
void
free_all_prunes(void)
{
    struct rtentry *r;
    struct gtable *g, *prev_g;
    struct stable *s, *prev_s;
    struct ptable *p, *prev_p;

    for (r = routing_table; r; r = r->rt_next) {
	g = r->rt_groups;
	while (g) {
	    s = g->gt_srctbl;
	    while (s) {
		prev_s = s;
		s = s->st_next;
		free(prev_s);
	    }

	    p = g->gt_pruntbl;
	    while (p) {
		prev_p = p;
		p = p->pt_next;
		free(prev_p);
	    }

	    prev_g = g;
	    g = g->gt_next;
	    free(prev_g);
	}
	r->rt_groups = NULL;
    }
    kernel_table = NULL;

    g = kernel_no_route;
    while (g) {
	if (g->gt_srctbl)
	    free(g->gt_srctbl);

	prev_g = g;
	g = g->gt_next;
	free(prev_g);
    }
    kernel_no_route = NULL;
}

/*
 * When a new route is created, search
 * a) The less-specific part of the routing table
 * b) The route-less kernel table
 * for sources that the new route might want to handle.
 *
 * "Inheriting" these sources might be cleanest, but simply deleting
 * them is easier, and letting the kernel re-request them.
 */
void
steal_sources(struct rtentry *rt)
{
    struct rtentry *rp;
    struct gtable *gt, **gtnp;
    struct stable *st, **stnp;

    for (rp = rt->rt_next; rp; rp = rp->rt_next) {
	if ((rt->rt_origin & rp->rt_originmask) == rp->rt_origin) {
	    logit(LOG_DEBUG, 0, "Route for %s stealing sources from %s",
		inet_fmts(rt->rt_origin, rt->rt_originmask),
		inet_fmts(rp->rt_origin, rp->rt_originmask));
	    for (gt = rp->rt_groups; gt; gt = gt->gt_next) {
		stnp = &gt->gt_srctbl;
		while ((st = *stnp) != NULL) {
		    if ((st->st_origin & rt->rt_originmask) == rt->rt_origin) {
			logit(LOG_DEBUG, 0, "%s stealing (%s %s) from %s",
			    inet_fmts(rt->rt_origin, rt->rt_originmask),
			    inet_fmt(st->st_origin),
			    inet_fmt(gt->gt_mcastgrp),
			    inet_fmts(rp->rt_origin, rp->rt_originmask));
			if (k_del_rg(st->st_origin, gt) < 0) {
			    logit(LOG_WARNING, errno, "%s (%s, %s)",
				"steal_sources trying to delete",
				inet_fmt(st->st_origin),
				inet_fmt(gt->gt_mcastgrp));
			}
			*stnp = st->st_next;
			kroutes--;
			free(st);
		    } else {
			stnp = &st->st_next;
		    }
		}
	    }
	}
    }

    gtnp = &kernel_no_route;
    while ((gt = *gtnp) != NULL) {
	if (gt->gt_srctbl && ((gt->gt_srctbl->st_origin & rt->rt_originmask)
				    == rt->rt_origin)) {
	    logit(LOG_DEBUG, 0, "%s stealing (%s %s) from %s",
		inet_fmts(rt->rt_origin, rt->rt_originmask),
		inet_fmt(gt->gt_srctbl->st_origin),
		inet_fmt(gt->gt_mcastgrp),
		"no_route table");
	    if (k_del_rg(gt->gt_srctbl->st_origin, gt) < 0) {
		logit(LOG_WARNING, errno, "%s (%s %s)",
		    "steal_sources trying to delete",
		    inet_fmt(gt->gt_srctbl->st_origin),
		    inet_fmt(gt->gt_mcastgrp));
	    }
	    kroutes--;
	    free(gt->gt_srctbl);
	    *gtnp = gt->gt_next;
	    if (gt->gt_next)
		gt->gt_next->gt_prev = gt->gt_prev;
	    free(gt);
	} else {
	    gtnp = &gt->gt_next;
	}
    }
}

/*
 * Advance the timers on all the cache entries.
 * If there are any entries whose timers have expired,
 * remove these entries from the kernel cache.
 */
void
age_table_entry(void)
{
    struct rtentry *r;
    struct gtable *gt, **gtnptr;
    struct stable *st, **stnp;
    struct ptable *pt, **ptnp;
    struct sioc_sg_req sg_req;

    logit(LOG_DEBUG, 0, "ageing entries");

    gtnptr = &kernel_table;
    while ((gt = *gtnptr) != NULL) {
	r = gt->gt_route;

	/* advance the timer for the kernel entry */
	gt->gt_timer -= ROUTE_MAX_REPORT_DELAY;

	/* decrement prune timer if need be */
	if (gt->gt_prsent_timer > 0) {
	    gt->gt_prsent_timer -= ROUTE_MAX_REPORT_DELAY;
	    if (gt->gt_prsent_timer <= 0) {
		logit(LOG_DEBUG, 0, "upstream prune tmo (%s %s)",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(gt->gt_mcastgrp));
		gt->gt_prsent_timer = -1;
	    }
	}

	/* retransmit graft if graft sent flag is still set */
	if (gt->gt_grftsnt) {
	    int y;
	    CHK_GS(gt->gt_grftsnt++, y);
	    if (y)
		send_graft(gt);
	}

	/*
	 * Age prunes
	 *
	 * If a prune expires, forward again on that vif.
	 */
	ptnp = &gt->gt_pruntbl;
	while ((pt = *ptnp) != NULL) {
	    if ((pt->pt_timer -= ROUTE_MAX_REPORT_DELAY) <= 0) {
		logit(LOG_DEBUG, 0, "expire prune (%s %s) from %s on vif %d",
		    inet_fmts(r->rt_origin, r->rt_originmask),
		    inet_fmt(gt->gt_mcastgrp),
		    inet_fmt(pt->pt_router),
		    pt->pt_vifi);

		expire_prune(pt->pt_vifi, gt);

		/* remove the router's prune entry and await new one */
		*ptnp = pt->pt_next;
		free(pt);
	    } else {
		ptnp = &pt->pt_next;
	    }
	}

	/*
	 * If the cache entry has expired, delete source table entries for
	 * silent sources.  If there are no source entries left, and there
	 * are no downstream prunes, then the entry is deleted.
	 * Otherwise, the cache entry's timer is refreshed.
	 */
	if (gt->gt_timer <= 0) {
	    /* Check for traffic before deleting source entries */
	    sg_req.grp.s_addr = gt->gt_mcastgrp;
	    stnp = &gt->gt_srctbl;
	    while ((st = *stnp) != NULL) {
		sg_req.src.s_addr = st->st_origin;
		if (ioctl(igmp_socket, SIOCGETSGCNT, (char *)&sg_req) < 0) {
		    logit(LOG_WARNING, errno, "%s (%s %s)",
			"age_table_entry: SIOCGETSGCNT failing for",
			inet_fmt(st->st_origin),
			inet_fmt(gt->gt_mcastgrp));
		    /* Make sure it gets deleted below */
		    sg_req.pktcnt = st->st_pktcnt;
		}
		if (sg_req.pktcnt == st->st_pktcnt) {
		    *stnp = st->st_next;
		    logit(LOG_DEBUG, 0, "age_table_entry deleting (%s %s)",
			inet_fmt(st->st_origin),
			inet_fmt(gt->gt_mcastgrp));
		    if (k_del_rg(st->st_origin, gt) < 0) {
			logit(LOG_WARNING, errno,
			    "age_table_entry trying to delete (%s %s)",
			    inet_fmt(st->st_origin),
			    inet_fmt(gt->gt_mcastgrp));
		    }
		    kroutes--;
		    free(st);
		} else {
		    st->st_pktcnt = sg_req.pktcnt;
		    stnp = &st->st_next;
		}
	    }

	    /*
	     * Retain the group entry if we have downstream prunes or if
	     * there is at least one source in the list that still has
	     * traffic, or if our upstream prune timer is running.
	     */
	    if (gt->gt_pruntbl != NULL || gt->gt_srctbl != NULL ||
		gt->gt_prsent_timer > 0) {
		gt->gt_timer = CACHE_LIFETIME(cache_lifetime);
		if (gt->gt_prsent_timer == -1) {
		    if (gt->gt_grpmems == 0)
			send_prune(gt);
		    else
			gt->gt_prsent_timer = 0;
		}
		gtnptr = &gt->gt_gnext;
		continue;
	    }

	    logit(LOG_DEBUG, 0, "timeout cache entry (%s, %s)",
		inet_fmts(r->rt_origin, r->rt_originmask),
		inet_fmt(gt->gt_mcastgrp));

	    if (gt->gt_prev)
		gt->gt_prev->gt_next = gt->gt_next;
	    else
		gt->gt_route->rt_groups = gt->gt_next;
	    if (gt->gt_next)
		gt->gt_next->gt_prev = gt->gt_prev;

	    if (gt->gt_gprev) {
		gt->gt_gprev->gt_gnext = gt->gt_gnext;
		gtnptr = &gt->gt_gprev->gt_gnext;
	    } else {
		kernel_table = gt->gt_gnext;
		gtnptr = &kernel_table;
	    }
	    if (gt->gt_gnext)
		gt->gt_gnext->gt_gprev = gt->gt_gprev;

#ifdef RSRR
	    /* Send route change notification to reservation protocol. */
	    rsrr_cache_send(gt,0);
	    rsrr_cache_clean(gt);
#endif /* RSRR */
	    free((char *)gt);
	} else {
	    if (gt->gt_prsent_timer == -1) {
		if (gt->gt_grpmems == 0)
		    send_prune(gt);
		else
		    gt->gt_prsent_timer = 0;
	    }
	    gtnptr = &gt->gt_gnext;
	}
    }

    /*
     * When traversing the no_route table, the decision is much easier.
     * Just delete it if it has timed out.
     */
    gtnptr = &kernel_no_route;
    while ((gt = *gtnptr) != NULL) {
	/* advance the timer for the kernel entry */
	gt->gt_timer -= ROUTE_MAX_REPORT_DELAY;

	if (gt->gt_timer < 0) {
	    if (gt->gt_srctbl) {
		if (k_del_rg(gt->gt_srctbl->st_origin, gt) < 0) {
		    logit(LOG_WARNING, errno, "%s (%s %s)",
			"age_table_entry trying to delete no-route",
			inet_fmt(gt->gt_srctbl->st_origin),
			inet_fmt(gt->gt_mcastgrp));
		}
		free(gt->gt_srctbl);
	    }
	    *gtnptr = gt->gt_next;
	    if (gt->gt_next)
		gt->gt_next->gt_prev = gt->gt_prev;

	    free((char *)gt);
	} else {
	    gtnptr = &gt->gt_next;
	}
    }
}

/*
 * Modify the kernel to forward packets when one or multiple prunes that
 * were received on the vif given by vifi, for the group given by gt,
 * have expired.
 */
static void
expire_prune(vifi_t vifi, struct gtable *gt)
{
    /*
     * No need to send a graft, any prunes that we sent
     * will expire before any prunes that we have received.
     */
    if (gt->gt_prsent_timer > 0) {
        logit(LOG_DEBUG, 0, "prune expired with %d left on %s",
		gt->gt_prsent_timer, "prsent_timer");
        gt->gt_prsent_timer = 0;
    }

    /* modify the kernel entry to forward packets */
    if (!VIFM_ISSET(vifi, gt->gt_grpmems)) {
        struct rtentry *rt = gt->gt_route;
        VIFM_SET(vifi, gt->gt_grpmems);
        logit(LOG_DEBUG, 0, "forw again (%s %s) gm:%x vif:%d",
	inet_fmts(rt->rt_origin, rt->rt_originmask),
	inet_fmt(gt->gt_mcastgrp), gt->gt_grpmems, vifi);

        prun_add_ttls(gt);
        update_kernel(gt);
#ifdef RSRR
        /* Send route change notification to reservation protocol. */
        rsrr_cache_send(gt,1);
#endif /* RSRR */
    }
}


static const char *
scaletime(u_long t)
{
    static char buf1[5];
    static char buf2[5];
    static char *buf=buf1;
    char s;
    char *p;

    p = buf;
    if (buf == buf1)
	buf = buf2;
    else
	buf = buf1;

    if (t < 120) {
	s = 's';
    } else if (t < 3600) {
	t /= 60;
	s = 'm';
    } else if (t < 86400) {
	t /= 3600;
	s = 'h';
    } else if (t < 864000) {
	t /= 86400;
	s = 'd';
    } else {
	t /= 604800;
	s = 'w';
    }
    if (t > 999)
	return "*** ";

    snprintf(p, 5, "%3d%c", (int)t, s);

    return p;
}

/*
 * Print the contents of the cache table on file 'fp2'.
 */
void
dump_cache(FILE *fp2)
{
    struct rtentry *r;
    struct gtable *gt;
    struct stable *st;
    vifi_t i;
    time_t thyme = time(0);

    fprintf(fp2,
	    "Multicast Routing Cache Table (%d entries)\n%s", kroutes,
    " Origin             Mcast-group     CTmr  Age Ptmr IVif Forwvifs\n");

    for (gt = kernel_no_route; gt; gt = gt->gt_next) {
	if (gt->gt_srctbl) {
	    fprintf(fp2, " %-18s %-15s %-4s %-4s    - -1\n",
		inet_fmts(gt->gt_srctbl->st_origin, 0xffffffff),
		inet_fmt(gt->gt_mcastgrp), scaletime(gt->gt_timer),
		scaletime(thyme - gt->gt_ctime));
	    fprintf(fp2, ">%s\n", inet_fmt(gt->gt_srctbl->st_origin));
	}
    }

    for (gt = kernel_table; gt; gt = gt->gt_gnext) {
	r = gt->gt_route;
	fprintf(fp2, " %-18s %-15s",
	    inet_fmts(r->rt_origin, r->rt_originmask),
	    inet_fmt(gt->gt_mcastgrp));

	fprintf(fp2, " %-4s", scaletime(gt->gt_timer));

	fprintf(fp2, " %-4s %-4s ", scaletime(thyme - gt->gt_ctime),
			gt->gt_prsent_timer ? scaletime(gt->gt_prsent_timer) :
					      "   -");

	fprintf(fp2, "%2u%c%c ", r->rt_parent,
	    gt->gt_prsent_timer ? 'P' : ' ',
	    VIFM_ISSET(r->rt_parent, gt->gt_scope) ? 'B' : ' ');

	for (i = 0; i < numvifs; ++i) {
	    if (VIFM_ISSET(i, gt->gt_grpmems))
		fprintf(fp2, " %u ", i);
	    else if (VIFM_ISSET(i, r->rt_children) &&
		     !VIFM_ISSET(i, r->rt_leaves))
		fprintf(fp2, " %u%c", i,
			VIFM_ISSET(i, gt->gt_scope) ? 'b' : 'p');
	}
	fprintf(fp2, "\n");
	for (st = gt->gt_srctbl; st; st = st->st_next) {
	    fprintf(fp2, ">%s\n", inet_fmt(st->st_origin));
	}
#ifdef DEBUG_PRUNES
	for (pt = gt->gt_pruntbl; pt; pt = pt->pt_next) {
	    fprintf(fp2, "<r:%s v:%d t:%d\n", inet_fmt(pt->pt_router),
		pt->pt_vifi, pt->pt_timer);
	}
#endif
    }
}

/*
 * Traceroute function which returns traceroute replies to the requesting
 * router. Also forwards the request to downstream routers.
 *
 * no: promoted u_char
 */
void
accept_mtrace(u_int32_t src, u_int32_t dst, u_int32_t group, char *data,
	      u_int no, int datalen)
{
    u_char type;
    struct rtentry *rt;
    struct gtable *gt;
    struct tr_query *qry;
    struct tr_resp  *resp;
    int vifi;
    char *p;
    int rcount;
    int errcode = TR_NO_ERR;
    int resptype;
    struct timeval tp;
    struct sioc_vif_req v_req;
    struct sioc_sg_req sg_req;

    /* Remember qid across invocations */
    static u_int32_t oqid = 0;

    /* timestamp the request/response */
    gettimeofday(&tp, 0);

    /*
     * Check if it is a query or a response
     */
    if (datalen == QLEN) {
	type = QUERY;
	logit(LOG_DEBUG, 0, "Initial traceroute query rcvd from %s to %s",
	    inet_fmt(src), inet_fmt(dst));
    }
    else if ((datalen - QLEN) % RLEN == 0) {
	type = RESP;
	logit(LOG_DEBUG, 0, "In-transit traceroute query rcvd from %s to %s",
	    inet_fmt(src), inet_fmt(dst));
	if (IN_MULTICAST(ntohl(dst))) {
	    logit(LOG_DEBUG, 0, "Dropping multicast response");
	    return;
	}
    }
    else {
	logit(LOG_WARNING, 0, "%s from %s to %s",
	    "Non decipherable traceroute request received",
	    inet_fmt(src), inet_fmt(dst));
	return;
    }

    qry = (struct tr_query *)data;

    /*
     * if it is a packet with all reports filled, drop it
     */
    if ((u_int)(rcount = (datalen - QLEN)/RLEN) == no) {
	logit(LOG_DEBUG, 0, "packet with all reports filled in");
	return;
    }

    logit(LOG_DEBUG, 0, "s: %s g: %s d: %s ",
	    inet_fmt(qry->tr_src),
	    inet_fmt(group),
	    inet_fmt(qry->tr_dst));
    logit(LOG_DEBUG, 0, "rttl: %d rd: %s", qry->tr_rttl,
	    inet_fmt(qry->tr_raddr));
    logit(LOG_DEBUG, 0, "rcount:%d, qid:%06x", rcount, qry->tr_qid);

    /* determine the routing table entry for this traceroute */
    rt = determine_route(qry->tr_src);
    if (rt) {
	logit(LOG_DEBUG, 0, "rt parent vif: %d rtr: %s metric: %d",
		rt->rt_parent, inet_fmt(rt->rt_gateway),
		rt->rt_metric);
	logit(LOG_DEBUG, 0, "rt origin %s",
		inet_fmts(rt->rt_origin, rt->rt_originmask));
    } else
	logit(LOG_DEBUG, 0, "...no route");

    /*
     * Query type packet - check if rte exists
     * Check if the query destination is a vif connected to me.
     * and if so, whether I should start response back
     */
    if (type == QUERY) {
	if (oqid == qry->tr_qid) {
	    /*
	     * If the multicast router is a member of the group being
	     * queried, and the query is multicasted, then the router can
	     * receive multiple copies of the same query.  If we have already
	     * replied to this traceroute, just ignore it this time.
	     *
	     * This is not a total solution, but since if this fails you
	     * only get N copies, N <= the number of interfaces on the router,
	     * it is not fatal.
	     */
	    logit(LOG_DEBUG, 0, "ignoring duplicate traceroute packet");
	    return;
	}

	if (rt == NULL) {
	    logit(LOG_DEBUG, 0, "Mcast traceroute: no route entry %s",
		   inet_fmt(qry->tr_src));
	    if (IN_MULTICAST(ntohl(dst)))
		return;
	}
	vifi = find_vif(qry->tr_dst, 0);

	if (vifi == NO_VIF) {
	    /* The traceroute destination is not on one of my subnet vifs. */
	    logit(LOG_DEBUG, 0, "Destination %s not an interface",
		   inet_fmt(qry->tr_dst));
	    if (IN_MULTICAST(ntohl(dst)))
		return;
	    errcode = TR_WRONG_IF;
	} else if (rt != NULL && !VIFM_ISSET(vifi, rt->rt_children)) {
	    logit(LOG_DEBUG, 0,
		    "Destination %s not on forwarding tree for src %s",
		   inet_fmt(qry->tr_dst),
		   inet_fmt(qry->tr_src));
	    if (IN_MULTICAST(ntohl(dst)))
		return;
	    errcode = TR_WRONG_IF;
	}
    }
    else {
	/*
	 * determine which interface the packet came in on
	 * RESP packets travel hop-by-hop so this either traversed
	 * a tunnel or came from a directly attached mrouter.
	 */
	if ((vifi = find_vif(src, dst)) == NO_VIF) {
	    logit(LOG_DEBUG, 0, "Wrong interface for packet");
	    errcode = TR_WRONG_IF;
	}
    }

    /* Now that we've decided to send a response, save the qid */
    oqid = qry->tr_qid;

    logit(LOG_DEBUG, 0, "Sending traceroute response");

    /* copy the packet to the sending buffer */
    p = send_buf + MIN_IP_HEADER_LEN + IGMP_MINLEN;

    bcopy(data, p, datalen);

    p += datalen;

    /*
     * If there is no room to insert our reply, coopt the previous hop
     * error indication to relay this fact.
     */
    if (p + sizeof(struct tr_resp) > send_buf + RECV_BUF_SIZE) {
	resp = (struct tr_resp *)p - 1;
	resp->tr_rflags = TR_NO_SPACE;
	rt = NULL;
	goto sendit;
    }

    /*
     * fill in initial response fields
     */
    resp = (struct tr_resp *)p;
    bzero(resp, sizeof(struct tr_resp));
    datalen += RLEN;

    resp->tr_qarr    = htonl((tp.tv_sec + JAN_1970) << 16) +
				((tp.tv_usec >> 4) & 0xffff);

    resp->tr_rproto  = PROTO_DVMRP;
    if (errcode != TR_NO_ERR) {
	resp->tr_rflags	 = errcode;
	rt = NULL;	/* hack to enforce send straight to requestor */
	goto sendit;
    }
    resp->tr_outaddr = uvifs[vifi].uv_lcl_addr;
    resp->tr_fttl    = uvifs[vifi].uv_threshold;
    resp->tr_rflags  = TR_NO_ERR;

    /*
     * obtain # of packets out on interface
     */
    v_req.vifi = vifi;
    if (ioctl(igmp_socket, SIOCGETVIFCNT, (char *)&v_req) >= 0)
	resp->tr_vifout  =  htonl(v_req.ocount);

    /*
     * fill in scoping & pruning information
     */
    if (rt)
	for (gt = rt->rt_groups; gt; gt = gt->gt_next) {
	    if (gt->gt_mcastgrp >= group)
		break;
	}
    else
	gt = NULL;

    if (gt && gt->gt_mcastgrp == group) {
	sg_req.src.s_addr = qry->tr_src;
	sg_req.grp.s_addr = group;
	if (ioctl(igmp_socket, SIOCGETSGCNT, (char *)&sg_req) >= 0)
	    resp->tr_pktcnt = htonl(sg_req.pktcnt);

	if (VIFM_ISSET(vifi, gt->gt_scope))
	    resp->tr_rflags = TR_SCOPED;
	else if (gt->gt_prsent_timer)
	    resp->tr_rflags = TR_PRUNED;
	else if (!VIFM_ISSET(vifi, gt->gt_grpmems)) {
	    if (VIFM_ISSET(vifi, rt->rt_children) &&
		!VIFM_ISSET(vifi, rt->rt_leaves))
		resp->tr_rflags = TR_OPRUNED;
	    else
		resp->tr_rflags = TR_NO_FWD;
	}
    } else {
	if (scoped_addr(vifi, group))
	    resp->tr_rflags = TR_SCOPED;
	else if (rt && !VIFM_ISSET(vifi, rt->rt_children))
	    resp->tr_rflags = TR_NO_FWD;
    }

    /*
     *  if no rte exists, set NO_RTE error
     */
    if (rt == NULL) {
	src = dst;		/* the dst address of resp. pkt */
	resp->tr_inaddr   = 0;
	resp->tr_rflags   = TR_NO_RTE;
	resp->tr_rmtaddr  = 0;
    } else {
	/* get # of packets in on interface */
	v_req.vifi = rt->rt_parent;
	if (ioctl(igmp_socket, SIOCGETVIFCNT, (char *)&v_req) >= 0)
	    resp->tr_vifin = htonl(v_req.icount);

	MASK_TO_VAL(rt->rt_originmask, resp->tr_smask);
	src = uvifs[rt->rt_parent].uv_lcl_addr;
	resp->tr_inaddr = src;
	resp->tr_rmtaddr = rt->rt_gateway;
	if (!VIFM_ISSET(vifi, rt->rt_children)) {
	    logit(LOG_DEBUG, 0,
		    "Destination %s not on forwarding tree for src %s",
		   inet_fmt(qry->tr_dst),
		   inet_fmt(qry->tr_src));
	    resp->tr_rflags = TR_WRONG_IF;
	}
	if (rt->rt_metric >= UNREACHABLE) {
	    resp->tr_rflags = TR_NO_RTE;
	    /* Hack to send reply directly */
	    rt = NULL;
	}
    }

sendit:
    /*
     * if metric is 1 or no. of reports is 1, send response to requestor
     * else send to upstream router.  If the upstream router can't handle
     * mtrace, set an error code and send to requestor anyway.
     */
    logit(LOG_DEBUG, 0, "rcount:%d, no:%d", rcount, no);

    if (((u_int)(rcount + 1) == no) || (rt == NULL) || (rt->rt_metric == 1)) {
	resptype = IGMP_MTRACE_REPLY;
	dst = qry->tr_raddr;
    } else
	if (!can_mtrace(rt->rt_parent, rt->rt_gateway)) {
	    dst = qry->tr_raddr;
	    resp->tr_rflags = TR_OLD_ROUTER;
	    resptype = IGMP_MTRACE_REPLY;
	} else {
	    dst = rt->rt_gateway;
	    resptype = IGMP_MTRACE_QUERY;
	}

    if (IN_MULTICAST(ntohl(dst))) {
	/*
	 * Send the reply on a known multicast capable vif.
	 * If we don't have one, we can't source any multicasts anyway.
	 */
	if (phys_vif != -1) {
	    logit(LOG_DEBUG, 0, "Sending reply to %s from %s",
		inet_fmt(dst), inet_fmt(uvifs[phys_vif].uv_lcl_addr));
	    k_set_ttl(qry->tr_rttl);
	    send_igmp(uvifs[phys_vif].uv_lcl_addr, dst,
		      resptype, no, group,
		      datalen);
	    k_set_ttl(1);
	} else
	    logit(LOG_INFO, 0, "No enabled phyints -- %s",
			"dropping traceroute reply");
    } else {
	logit(LOG_DEBUG, 0, "Sending %s to %s from %s",
	    resptype == IGMP_MTRACE_REPLY ?  "reply" : "request on",
	    inet_fmt(dst), inet_fmt(src));

	send_igmp(src, dst,
		  resptype, no, group,
		  datalen);
    }
    return;
}