freebsd-skq/usr.sbin/pim6sd/rp.c
Kris Kennaway 5bca844f39 Sync with latest KAME code.
Obtained from:	KAME
2000-07-06 01:48:08 +00:00

1208 lines
38 KiB
C

/*
* Copyright (C) 1999 LSIIT Laboratory.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the project nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* Copyright (c) 1998 by the University of Southern California.
* All rights reserved.
*
* Permission to use, copy, modify, and distribute this software and
* its documentation in source and binary forms for lawful
* purposes and without fee is hereby granted, provided
* that the above copyright notice appear in all copies and that both
* the copyright notice and this permission notice appear in supporting
* documentation, and that any documentation, advertising materials,
* and other materials related to such distribution and use acknowledge
* that the software was developed by the University of Southern
* California and/or Information Sciences Institute.
* The name of the University of Southern California may not
* be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THE UNIVERSITY OF SOUTHERN CALIFORNIA DOES NOT MAKE ANY REPRESENTATIONS
* ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. THIS SOFTWARE IS
* PROVIDED "AS IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES,
* INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, TITLE, AND
* NON-INFRINGEMENT.
*
* IN NO EVENT SHALL USC, OR ANY OTHER CONTRIBUTOR BE LIABLE FOR ANY
* SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES, WHETHER IN CONTRACT,
* TORT, OR OTHER FORM OF ACTION, ARISING OUT OF OR IN CONNECTION WITH,
* THE USE OR PERFORMANCE OF THIS SOFTWARE.
*
* Other copyrights might apply to parts of this software and are so
* noted when applicable.
*/
/*
* Questions concerning this software should be directed to
* Mickael Hoerdt (hoerdt@clarinet.u-strasbg.fr) LSIIT Strasbourg.
*
*/
/*
* This program has been derived from pim6dd.
* The pim6dd program is covered by the license in the accompanying file
* named "LICENSE.pim6dd".
*/
/*
* This program has been derived from pimd.
* The pimd program is covered by the license in the accompanying file
* named "LICENSE.pimd".
*
* $FreeBSD$
*/
#include <stdlib.h>
#include <syslog.h>
#include "rp.h"
#include "pim6_proto.h"
#include "pimd.h"
#include <netinet6/pim6.h>
#include "timer.h"
#include "inet6.h"
#include "route.h"
#include "pimd.h"
#include "debug.h"
#include "crc.h"
/*
* The hash function. Stollen from Eddy's (eddy@isi.edu) implementation (for
* compatibility ;)
*/
#define SEED1 1103515245
#define SEED2 12345
#define RP_HASH_VALUE(G, M, C) (((SEED1) * (((SEED1) * ((G) & (M)) + (SEED2)) ^ (C)) + (SEED2)) % 0x80000000)
#define RP_HASH_VALUE2(P, C) (((SEED1) * (((SEED1) * (P) + (SEED2)) ^ (C)) + (SEED2)) % 0x80000000)
cand_rp_t *cand_rp_list;
grp_mask_t *grp_mask_list;
cand_rp_t *segmented_cand_rp_list;
grp_mask_t *segmented_grp_mask_list;
u_int16 curr_bsr_fragment_tag;
u_int8 curr_bsr_priority;
struct sockaddr_in6 curr_bsr_address;
struct in6_addr curr_bsr_hash_mask;
u_int16 pim_bootstrap_timer; /* For electing the BSR and sending
* Cand-RP-set msgs */
u_int8 my_bsr_priority;
struct sockaddr_in6 my_bsr_address;
struct in6_addr my_bsr_hash_mask;
u_int8 cand_bsr_flag = FALSE; /* Set to TRUE if I am a candidate
* BSR */
struct sockaddr_in6 my_cand_rp_address;
u_int8 my_cand_rp_priority;
u_int16 my_cand_rp_holdtime;
u_int16 my_cand_rp_adv_period; /* The locally configured Cand-RP
* adv. period. */
u_int16 my_bsr_period; /* The locally configured BSR
period */
u_int16 pim_cand_rp_adv_timer;
u_int8 cand_rp_flag = FALSE; /* Candidate RP flag */
struct cand_rp_adv_message_ cand_rp_adv_message;
struct in6_addr rp_my_ipv6_hashmask;
/*
* Local functions definition.
*/
static cand_rp_t *add_cand_rp __P((cand_rp_t **used_cand_rp_list ,
struct sockaddr_in6 *address));
static grp_mask_t *add_grp_mask __P((grp_mask_t ** used_grp_mask_list,
struct sockaddr_in6 *group_addr,
struct in6_addr group_mask,
struct in6_addr hash_mask));
static void delete_grp_mask_entry __P((cand_rp_t ** used_cand_rp_list,
grp_mask_t ** used_grp_mask_list,
grp_mask_t * grp_mask_delete));
static void delete_rp_entry __P((cand_rp_t ** used_cand_rp_list,
grp_mask_t ** used_grp_mask_list,
cand_rp_t * cand_rp_ptr));
void
init_rp6_and_bsr6()
{
/* TODO: if the grplist is not NULL, remap all groups ASAP! */
delete_rp_list(&cand_rp_list, &grp_mask_list);
delete_rp_list(&segmented_cand_rp_list, &segmented_grp_mask_list);
if (cand_bsr_flag == FALSE)
{
/*
* If I am not candidat BSR, initialize the "current BSR" as having
* the lowest priority.
*/
curr_bsr_fragment_tag = 0;
curr_bsr_priority = 0; /* Lowest priority */
curr_bsr_address = sockaddr6_any; /* Lowest priority */
MASKLEN_TO_MASK6(RP_DEFAULT_IPV6_HASHMASKLEN, curr_bsr_hash_mask);
SET_TIMER(pim_bootstrap_timer, PIM_BOOTSTRAP_TIMEOUT);
}
else
{
curr_bsr_fragment_tag = RANDOM();
curr_bsr_priority = my_bsr_priority;
curr_bsr_address = my_bsr_address;
curr_bsr_hash_mask = my_bsr_hash_mask;
SET_TIMER(pim_bootstrap_timer, bootstrap_initial_delay());
}
if (cand_rp_flag != FALSE)
{
MASKLEN_TO_MASK6(RP_DEFAULT_IPV6_HASHMASKLEN, rp_my_ipv6_hashmask);
/* Setup the Cand-RP-Adv-Timer */
SET_TIMER(pim_cand_rp_adv_timer, RANDOM() % my_cand_rp_adv_period);
}
}
/*
* XXX: This implementation is based on section 6.2 of RFC 2362, which
* is highly dependent on IPv4.
* We'll have to rewrite the function...
*/
u_int16
bootstrap_initial_delay()
{
// long AddrDelay;
// long Delay;
// long log_mask;
// int log_of_2;
// u_int8 bestPriority;
/*
* The bootstrap timer initial value (if Cand-BSR). It depends of the
* bootstrap router priority: higher priority has shorter value:
*
* Delay = 5 + 2*log_2(1 + bestPriority - myPriority) + AddrDelay;
*
* bestPriority = Max(storedPriority, myPriority); if (bestPriority ==
* myPriority) AddrDelay = log_2(bestAddr - myAddr)/16; else AddrDelay =
* 2 - (myAddr/2^31);
*/
// bestPriority = max(curr_bsr_priority, my_bsr_priority);
// if (bestPriority == my_bsr_priority)
// {
// AddrDelay = ntohl(curr_bsr_address) - ntohl(my_bsr_address);
/* Calculate the integer part of log_2 of (bestAddr - myAddr) */
/*
* To do so, have to find the position number of the first bit from
* left which is `1`
*/
// log_mask = sizeof(AddrDelay) << 3;
// log_mask = (1 << (log_mask - 1)); /* Set the leftmost bit to
// * `1` */
/* for (log_of_2 = (sizeof(AddrDelay) << 3) - 1; log_of_2; log_of_2--)
{
if (AddrDelay & log_mask)
break;
else
*/
// log_mask >>= 1; /* Start shifting `1` on right */
/* }
AddrDelay = log_of_2 / 16;
}
else
AddrDelay = 2 - (ntohl(my_bsr_address) / (1 << 31));
Delay = 1 + bestPriority - my_bsr_priority;
*/
/* Calculate log_2(Delay) */
// log_mask = sizeof(Delay) << 3;
// log_mask = (1 << (log_mask - 1));
/* Set the leftmost bit to `1`
*/
/* for (log_of_2 = (sizeof(Delay) << 3) - 1; log_of_2; log_of_2--)
{
if (Delay & log_mask)
break;
else
*/
// log_mask >>= 1; /* Start shifting `1` on right */
/* }
Delay = 5 + 2 * Delay + AddrDelay;
return (u_int16) Delay;
*/
/* Temporary implementation */
return (RANDOM()%my_bsr_period);
}
static cand_rp_t *
add_cand_rp(used_cand_rp_list, address)
cand_rp_t **used_cand_rp_list;
struct sockaddr_in6 *address;
{
cand_rp_t *cand_rp_prev = (cand_rp_t *) NULL;
cand_rp_t *cand_rp;
cand_rp_t *cand_rp_new;
rpentry_t *rpentry_ptr;
/* The ordering is the bigger first */
for (cand_rp = *used_cand_rp_list; cand_rp != (cand_rp_t *) NULL;
cand_rp_prev = cand_rp, cand_rp = cand_rp->next)
{
if (inet6_greaterthan(&cand_rp->rpentry->address, address))
continue;
if (inet6_equal(&cand_rp->rpentry->address , address))
return (cand_rp);
else
break;
}
/* Create and insert the new entry between cand_rp_prev and cand_rp */
cand_rp_new = (cand_rp_t *) malloc(sizeof(cand_rp_t));
cand_rp_new->rp_grp_next = (rp_grp_entry_t *) NULL;
cand_rp_new->next = cand_rp;
cand_rp_new->prev = cand_rp_prev;
if (cand_rp != (cand_rp_t *) NULL)
cand_rp->prev = cand_rp_new;
if (cand_rp_prev == (cand_rp_t *) NULL)
{
*used_cand_rp_list = cand_rp_new;
}
else
{
cand_rp_prev->next = cand_rp_new;
}
rpentry_ptr = (rpentry_t *) malloc(sizeof(rpentry_t));
cand_rp_new->rpentry = rpentry_ptr;
rpentry_ptr->next = (srcentry_t *) NULL;
rpentry_ptr->prev = (srcentry_t *) NULL;
rpentry_ptr->address = *address;
rpentry_ptr->mrtlink = (mrtentry_t *) NULL;
rpentry_ptr->incoming = NO_VIF;
rpentry_ptr->upstream = (pim_nbr_entry_t *) NULL;
/* TODO: setup the metric and the preference as ~0 (the lowest)? */
rpentry_ptr->metric = ~0;
rpentry_ptr->preference = ~0;
RESET_TIMER(rpentry_ptr->timer);
rpentry_ptr->cand_rp = cand_rp_new;
/*
* TODO: XXX: check whether there is a route to that RP: if return value
* is FALSE, then no route.
*/
if (local_address(&rpentry_ptr->address) == NO_VIF)
{
/* TODO: check for error and delete */
set_incoming(rpentry_ptr, PIM_IIF_RP);
}
else
{
/* TODO: XXX: CHECK!!! */
rpentry_ptr->incoming = reg_vif_num;
}
return (cand_rp_new);
}
static grp_mask_t *
add_grp_mask(used_grp_mask_list, group_addr, group_mask, hash_mask)
grp_mask_t **used_grp_mask_list;
struct sockaddr_in6 *group_addr;
struct in6_addr group_mask;
struct in6_addr hash_mask;
{
grp_mask_t *grp_mask_prev = (grp_mask_t *) NULL;
grp_mask_t *grp_mask;
grp_mask_t *grp_mask_tmp;
struct sockaddr_in6 prefix_h;
struct sockaddr_in6 prefix_h2;
int i;
/* I compare on the adresses, inet6_equal use the scope, too */
prefix_h.sin6_scope_id = prefix_h2.sin6_scope_id = 0;
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h.sin6_addr.s6_addr[i] =
group_addr->sin6_addr.s6_addr[i] & group_mask.s6_addr[i];
/* The ordering is: bigger first */
for (grp_mask = *used_grp_mask_list; grp_mask != (grp_mask_t *) NULL;
grp_mask_prev = grp_mask, grp_mask = grp_mask->next)
{
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h2.sin6_addr.s6_addr[i] =
(grp_mask->group_addr.sin6_addr.s6_addr[i] &
grp_mask->group_mask.s6_addr[i]);
if (inet6_greaterthan(&prefix_h2, &prefix_h) )
continue;
if (inet6_equal(&prefix_h2, &prefix_h))
return (grp_mask);
else
break;
}
grp_mask_tmp = (grp_mask_t *) malloc(sizeof(grp_mask_t));
grp_mask_tmp->grp_rp_next = (rp_grp_entry_t *) NULL;
grp_mask_tmp->next = grp_mask;
grp_mask_tmp->prev = grp_mask_prev;
if (grp_mask != (grp_mask_t *) NULL)
grp_mask->prev = grp_mask_tmp;
if (grp_mask_prev == (grp_mask_t *) NULL)
{
*used_grp_mask_list = grp_mask_tmp;
}
else
{
grp_mask_prev->next = grp_mask_tmp;
}
grp_mask_tmp->group_addr = *group_addr;
grp_mask_tmp->group_mask = group_mask;
grp_mask_tmp->hash_mask = hash_mask;
grp_mask_tmp->group_rp_number = 0;
grp_mask_tmp->fragment_tag = 0;
return (grp_mask_tmp);
}
/*
* TODO: XXX: BUG: a remapping for some groups currently using some other
* grp_mask may be required by the addition of the new entry!!! Remapping all
* groups might be a costly process...
*/
rp_grp_entry_t *
add_rp_grp_entry(used_cand_rp_list, used_grp_mask_list,
rp_addr, rp_priority, rp_holdtime, group_addr, group_mask,
bsr_hash_mask,
fragment_tag)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
struct sockaddr_in6 *rp_addr;
u_int8 rp_priority;
u_int16 rp_holdtime;
struct sockaddr_in6 *group_addr;
struct in6_addr group_mask;
struct in6_addr bsr_hash_mask;
u_int16 fragment_tag;
{
cand_rp_t *cand_rp_ptr;
grp_mask_t *grp_mask_ptr;
rpentry_t *rpentry_ptr;
rp_grp_entry_t *grp_rp_entry_next;
rp_grp_entry_t *grp_rp_entry_new;
rp_grp_entry_t *grp_rp_entry_prev = (rp_grp_entry_t *) NULL;
grpentry_t *grpentry_ptr_prev;
grpentry_t *grpentry_ptr_next;
u_int8 old_highest_priority = ~0; /* Smaller value means
* "higher" */
/* Input data verification */
if (!inet6_valid_host(rp_addr))
return (rp_grp_entry_t *) NULL;
if (!IN6_IS_ADDR_MULTICAST(&group_addr->sin6_addr))
{
return (rp_grp_entry_t *) NULL;
}
grp_mask_ptr = add_grp_mask(used_grp_mask_list, group_addr, group_mask,
bsr_hash_mask);
if (grp_mask_ptr == (grp_mask_t *) NULL)
return (rp_grp_entry_t *) NULL;
/* TODO: delete */
#if 0
if (grp_mask_ptr->grp_rp_next != (rp_grp_entry_t *) NULL)
{
/* Check for obsolete grp_rp chain */
if ((my_bsr_address != curr_bsr_address)
&& (grp_mask_ptr->grp_rp_next->fragment_tag != fragment_tag))
{
/* This grp_rp chain is obsolete. Delete it. */
delete_grp_mask(used_cand_rp_list, used_grp_mask_list,
group_addr, group_mask);
grp_mask_ptr = add_grp_mask(used_grp_mask_list, group_addr,
group_mask, bsr_hash_mask);
if (grp_mask_ptr == (grp_mask_t *) NULL)
return (rp_grp_entry_t *) NULL;
}
}
#endif /* 0 */
cand_rp_ptr = add_cand_rp(used_cand_rp_list, rp_addr);
if (cand_rp_ptr == (cand_rp_t *) NULL)
{
if (grp_mask_ptr->grp_rp_next == (rp_grp_entry_t *) NULL)
delete_grp_mask(used_cand_rp_list, used_grp_mask_list,
group_addr, group_mask);
return (rp_grp_entry_t *) NULL;
}
rpentry_ptr = cand_rp_ptr->rpentry;
SET_TIMER(rpentry_ptr->timer, rp_holdtime);
grp_mask_ptr->fragment_tag = fragment_tag; /* For garbage collection */
grp_rp_entry_prev = (rp_grp_entry_t *) NULL;
grp_rp_entry_next = grp_mask_ptr->grp_rp_next;
/* TODO: improve it */
if (grp_rp_entry_next != (rp_grp_entry_t *) NULL)
old_highest_priority = grp_rp_entry_next->priority;
for (; grp_rp_entry_next != (rp_grp_entry_t *) NULL;
grp_rp_entry_prev = grp_rp_entry_next,
grp_rp_entry_next = grp_rp_entry_next->grp_rp_next)
{
/*
* Smaller value means higher priority. The entries are sorted with
* the highest priority first.
*/
if (grp_rp_entry_next->priority < rp_priority)
continue;
if (grp_rp_entry_next->priority > rp_priority)
break;
/*
* Here we don't care about higher/lower addresses, because higher
* address does not guarantee higher hash_value, but anyway we do
* order with the higher address first, so it will be easier to find
* an existing entry and update the holdtime.
*/
if (inet6_greaterthan(&grp_rp_entry_next->rp->rpentry->address , rp_addr))
continue;
if (inet6_lessthan(&grp_rp_entry_next->rp->rpentry->address , rp_addr))
break;
/* We already have this entry. Update the holdtime */
/*
* TODO: We shoudn't have old existing entry, because with the
* current implementation all of them will be deleted (different
* fragment_tag). Debug and check and eventually delete.
*/
grp_rp_entry_next->holdtime = rp_holdtime;
grp_rp_entry_next->advholdtime = rp_holdtime;
grp_rp_entry_next->fragment_tag = fragment_tag;
return (grp_rp_entry_next);
}
/* Create and link the new entry */
grp_rp_entry_new = (rp_grp_entry_t *) malloc(sizeof(rp_grp_entry_t));
grp_rp_entry_new->grp_rp_next = grp_rp_entry_next;
grp_rp_entry_new->grp_rp_prev = grp_rp_entry_prev;
if (grp_rp_entry_next != (rp_grp_entry_t *) NULL)
grp_rp_entry_next->grp_rp_prev = grp_rp_entry_new;
if (grp_rp_entry_prev == (rp_grp_entry_t *) NULL)
grp_mask_ptr->grp_rp_next = grp_rp_entry_new;
else
grp_rp_entry_prev->grp_rp_next = grp_rp_entry_new;
/*
* The rp_grp_entry chain is not ordered, so just plug the new entry at
* the head.
*/
grp_rp_entry_new->rp_grp_next = cand_rp_ptr->rp_grp_next;
if (cand_rp_ptr->rp_grp_next != (rp_grp_entry_t *) NULL)
cand_rp_ptr->rp_grp_next->rp_grp_prev = grp_rp_entry_new;
grp_rp_entry_new->rp_grp_prev = (rp_grp_entry_t *) NULL;
cand_rp_ptr->rp_grp_next = grp_rp_entry_new;
grp_rp_entry_new->holdtime = rp_holdtime;
grp_rp_entry_new->advholdtime = rp_holdtime;
grp_rp_entry_new->fragment_tag = fragment_tag;
grp_rp_entry_new->priority = rp_priority;
grp_rp_entry_new->group = grp_mask_ptr;
grp_rp_entry_new->rp = cand_rp_ptr;
grp_rp_entry_new->grplink = (grpentry_t *) NULL;
grp_mask_ptr->group_rp_number++;
if (grp_mask_ptr->grp_rp_next->priority == rp_priority)
{
/* The first entries are with the best priority. */
/* Adding this rp_grp_entry may result in group_to_rp remapping */
for (grp_rp_entry_next = grp_mask_ptr->grp_rp_next;
grp_rp_entry_next != (rp_grp_entry_t *) NULL;
grp_rp_entry_next = grp_rp_entry_next->grp_rp_next)
{
if (grp_rp_entry_next->priority > old_highest_priority)
break;
for (grpentry_ptr_prev = grp_rp_entry_next->grplink;
grpentry_ptr_prev != (grpentry_t *) NULL;)
{
grpentry_ptr_next = grpentry_ptr_prev->rpnext;
remap_grpentry(grpentry_ptr_prev);
grpentry_ptr_prev = grpentry_ptr_next;
}
}
}
return (grp_rp_entry_new);
}
void
delete_rp_grp_entry(used_cand_rp_list, used_grp_mask_list,
rp_grp_entry_delete)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
rp_grp_entry_t *rp_grp_entry_delete;
{
grpentry_t *grpentry_ptr;
grpentry_t *grpentry_ptr_next;
if (rp_grp_entry_delete == (rp_grp_entry_t *) NULL)
return;
rp_grp_entry_delete->group->group_rp_number--;
/* Free the rp_grp* and grp_rp* links */
if (rp_grp_entry_delete->rp_grp_prev != (rp_grp_entry_t *) NULL)
rp_grp_entry_delete->rp_grp_prev->rp_grp_next =
rp_grp_entry_delete->rp_grp_next;
else
rp_grp_entry_delete->rp->rp_grp_next =
rp_grp_entry_delete->rp_grp_next;
if (rp_grp_entry_delete->rp_grp_next != (rp_grp_entry_t *) NULL)
rp_grp_entry_delete->rp_grp_next->rp_grp_prev =
rp_grp_entry_delete->rp_grp_prev;
if (rp_grp_entry_delete->grp_rp_prev != (rp_grp_entry_t *) NULL)
rp_grp_entry_delete->grp_rp_prev->grp_rp_next =
rp_grp_entry_delete->grp_rp_next;
else
rp_grp_entry_delete->group->grp_rp_next =
rp_grp_entry_delete->grp_rp_next;
if (rp_grp_entry_delete->grp_rp_next != (rp_grp_entry_t *) NULL)
rp_grp_entry_delete->grp_rp_next->grp_rp_prev =
rp_grp_entry_delete->grp_rp_prev;
/* Delete Cand-RP or Group-prefix if useless */
if (rp_grp_entry_delete->group->grp_rp_next ==
(rp_grp_entry_t *) NULL)
delete_grp_mask_entry(used_cand_rp_list, used_grp_mask_list,
rp_grp_entry_delete->group);
if (rp_grp_entry_delete->rp->rp_grp_next ==
(rp_grp_entry_t *) NULL)
delete_rp_entry(used_cand_rp_list, used_grp_mask_list,
rp_grp_entry_delete->rp);
/* Remap all affected groups */
for (grpentry_ptr = rp_grp_entry_delete->grplink;
grpentry_ptr != (grpentry_t *) NULL;
grpentry_ptr = grpentry_ptr_next)
{
grpentry_ptr_next = grpentry_ptr->rpnext;
remap_grpentry(grpentry_ptr);
}
free((char *) rp_grp_entry_delete);
}
/*
* TODO: XXX: the affected group entries will be partially setup, because may
* have group routing entry, but NULL pointers to RP. After the call to this
* function, must remap all group entries ASAP.
*/
void
delete_rp_list(used_cand_rp_list, used_grp_mask_list)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
{
cand_rp_t *cand_rp_ptr,
*cand_rp_next;
grp_mask_t *grp_mask_ptr,
*grp_mask_next;
rp_grp_entry_t *rp_grp_entry_ptr,
*rp_grp_entry_next;
grpentry_t *grpentry_ptr,
*grpentry_ptr_next;
for (cand_rp_ptr = *used_cand_rp_list;
cand_rp_ptr != (cand_rp_t *) NULL;)
{
cand_rp_next = cand_rp_ptr->next;
/* Free the mrtentry (if any) for this RP */
if (cand_rp_ptr->rpentry->mrtlink != (mrtentry_t *) NULL)
{
if (cand_rp_ptr->rpentry->mrtlink->flags & MRTF_KERNEL_CACHE)
delete_mrtentry_all_kernel_cache(cand_rp_ptr->rpentry->mrtlink);
FREE_MRTENTRY(cand_rp_ptr->rpentry->mrtlink);
}
free(cand_rp_ptr->rpentry);
/* Free the whole chain of rp_grp_entry for this RP */
for (rp_grp_entry_ptr = cand_rp_ptr->rp_grp_next;
rp_grp_entry_ptr != (rp_grp_entry_t *) NULL;
rp_grp_entry_ptr = rp_grp_entry_next)
{
rp_grp_entry_next = rp_grp_entry_ptr->rp_grp_next;
/* Clear the RP related invalid pointers for all group entries */
for (grpentry_ptr = rp_grp_entry_ptr->grplink;
grpentry_ptr != (grpentry_t *) NULL;
grpentry_ptr = grpentry_ptr_next)
{
grpentry_ptr_next = grpentry_ptr->rpnext;
grpentry_ptr->rpnext = (grpentry_t *) NULL;
grpentry_ptr->rpprev = (grpentry_t *) NULL;
grpentry_ptr->active_rp_grp = (rp_grp_entry_t *) NULL;
grpentry_ptr->rpaddr = sockaddr6_any;
}
free(rp_grp_entry_ptr);
}
cand_rp_ptr = cand_rp_next;
}
*used_cand_rp_list = (cand_rp_t *) NULL;
for (grp_mask_ptr = *used_grp_mask_list;
grp_mask_ptr != (grp_mask_t *) NULL;
grp_mask_ptr = grp_mask_next)
{
grp_mask_next = grp_mask_ptr->next;
free(grp_mask_ptr);
}
*used_grp_mask_list = (grp_mask_t *) NULL;
}
void
delete_grp_mask(used_cand_rp_list, used_grp_mask_list, group_addr, group_mask)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
struct sockaddr_in6 *group_addr;
struct in6_addr group_mask;
{
grp_mask_t *grp_mask_ptr;
struct sockaddr_in6 prefix_h;
struct sockaddr_in6 prefix_h2;
int i;
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h.sin6_addr.s6_addr[i] = group_addr->sin6_addr.s6_addr[i]&group_mask.s6_addr[i];
for (grp_mask_ptr = *used_grp_mask_list;
grp_mask_ptr != (grp_mask_t *) NULL;
grp_mask_ptr = grp_mask_ptr->next)
{
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h2.sin6_addr.s6_addr[i] =
grp_mask_ptr->group_addr.sin6_addr.s6_addr[i]&grp_mask_ptr->group_mask.s6_addr[i];
if (inet6_greaterthan(&prefix_h2, &prefix_h))
continue;
if (IN6_ARE_ADDR_EQUAL(&grp_mask_ptr->group_addr.sin6_addr,
&group_addr->sin6_addr) &&
IN6_ARE_ADDR_EQUAL(&grp_mask_ptr->group_mask, &group_mask))
break;
else
return; /* Not found */
}
if (grp_mask_ptr == (grp_mask_t *) NULL)
return; /* Not found */
delete_grp_mask_entry(used_cand_rp_list, used_grp_mask_list,
grp_mask_ptr);
}
static void
delete_grp_mask_entry(used_cand_rp_list, used_grp_mask_list, grp_mask_delete)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
grp_mask_t *grp_mask_delete;
{
grpentry_t *grpentry_ptr,
*grpentry_ptr_next;
rp_grp_entry_t *grp_rp_entry_ptr;
rp_grp_entry_t *grp_rp_entry_next;
if (grp_mask_delete == (grp_mask_t *) NULL)
return;
/* Remove from the grp_mask_list first */
if (grp_mask_delete->prev != (grp_mask_t *) NULL)
grp_mask_delete->prev->next = grp_mask_delete->next;
else
*used_grp_mask_list = grp_mask_delete->next;
if (grp_mask_delete->next != (grp_mask_t *) NULL)
grp_mask_delete->next->prev = grp_mask_delete->prev;
/* Remove all grp_rp entries for this grp_mask */
for (grp_rp_entry_ptr = grp_mask_delete->grp_rp_next;
grp_rp_entry_ptr != (rp_grp_entry_t *) NULL;
grp_rp_entry_ptr = grp_rp_entry_next)
{
grp_rp_entry_next = grp_rp_entry_ptr->grp_rp_next;
/* Remap all related grpentry */
for (grpentry_ptr = grp_rp_entry_ptr->grplink;
grpentry_ptr != (grpentry_t *) NULL;
grpentry_ptr = grpentry_ptr_next)
{
grpentry_ptr_next = grpentry_ptr->rpnext;
remap_grpentry(grpentry_ptr);
}
if (grp_rp_entry_ptr->rp_grp_prev != (rp_grp_entry_t *) NULL)
{
grp_rp_entry_ptr->rp_grp_prev->rp_grp_next =
grp_rp_entry_ptr->rp_grp_next;
}
else
{
grp_rp_entry_ptr->rp->rp_grp_next = grp_rp_entry_ptr->rp_grp_next;
}
if (grp_rp_entry_ptr->rp_grp_next != (rp_grp_entry_t *) NULL)
grp_rp_entry_ptr->rp_grp_next->rp_grp_prev =
grp_rp_entry_ptr->rp_grp_prev;
if (grp_rp_entry_ptr->rp->rp_grp_next == (rp_grp_entry_t *) NULL)
{
/* Delete the RP entry */
delete_rp_entry(used_cand_rp_list, used_grp_mask_list,
grp_rp_entry_ptr->rp);
}
free(grp_rp_entry_ptr);
}
}
/*
* TODO: currently not used.
*/
void
delete_rp(used_cand_rp_list, used_grp_mask_list, rp_addr)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
struct sockaddr_in6 *rp_addr;
{
cand_rp_t *cand_rp_ptr;
for (cand_rp_ptr = *used_cand_rp_list;
cand_rp_ptr != (cand_rp_t *) NULL;
cand_rp_ptr = cand_rp_ptr->next)
{
if (inet6_greaterthan(&cand_rp_ptr->rpentry->address, rp_addr))
continue;
if (inet6_equal(&cand_rp_ptr->rpentry->address,rp_addr))
break;
else
return; /* Not found */
}
if (cand_rp_ptr == (cand_rp_t *) NULL)
return; /* Not found */
delete_rp_entry(used_cand_rp_list, used_grp_mask_list, cand_rp_ptr);
}
static void
delete_rp_entry(used_cand_rp_list, used_grp_mask_list, cand_rp_delete)
cand_rp_t **used_cand_rp_list;
grp_mask_t **used_grp_mask_list;
cand_rp_t *cand_rp_delete;
{
rp_grp_entry_t *rp_grp_entry_ptr;
rp_grp_entry_t *rp_grp_entry_next;
grpentry_t *grpentry_ptr;
grpentry_t *grpentry_ptr_next;
if (cand_rp_delete == (cand_rp_t *) NULL)
return;
/* Remove from the cand-RP chain */
if (cand_rp_delete->prev != (cand_rp_t *) NULL)
cand_rp_delete->prev->next = cand_rp_delete->next;
else
*used_cand_rp_list = cand_rp_delete->next;
if (cand_rp_delete->next != (cand_rp_t *) NULL)
cand_rp_delete->next->prev = cand_rp_delete->prev;
if (cand_rp_delete->rpentry->mrtlink != (mrtentry_t *) NULL)
{
if (cand_rp_delete->rpentry->mrtlink->flags & MRTF_KERNEL_CACHE)
delete_mrtentry_all_kernel_cache(cand_rp_delete->rpentry->mrtlink);
FREE_MRTENTRY(cand_rp_delete->rpentry->mrtlink);
}
free((char *) cand_rp_delete->rpentry);
/* Remove all rp_grp entries for this RP */
for (rp_grp_entry_ptr = cand_rp_delete->rp_grp_next;
rp_grp_entry_ptr != (rp_grp_entry_t *) NULL;
rp_grp_entry_ptr = rp_grp_entry_next)
{
rp_grp_entry_next = rp_grp_entry_ptr->rp_grp_next;
rp_grp_entry_ptr->group->group_rp_number--;
/* First take care of the grp_rp chain */
if (rp_grp_entry_ptr->grp_rp_prev != (rp_grp_entry_t *) NULL)
{
rp_grp_entry_ptr->grp_rp_prev->grp_rp_next =
rp_grp_entry_ptr->grp_rp_next;
}
else
{
rp_grp_entry_ptr->group->grp_rp_next =
rp_grp_entry_ptr->grp_rp_next;
}
if (rp_grp_entry_ptr->grp_rp_next != (rp_grp_entry_t *) NULL)
{
rp_grp_entry_ptr->grp_rp_next->grp_rp_prev =
rp_grp_entry_ptr->grp_rp_prev;
}
if (rp_grp_entry_ptr->grp_rp_next == (rp_grp_entry_t *) NULL)
{
delete_grp_mask_entry(used_cand_rp_list, used_grp_mask_list,
rp_grp_entry_ptr->group);
}
/* Remap the related groups */
for (grpentry_ptr = rp_grp_entry_ptr->grplink;
grpentry_ptr != (grpentry_t *) NULL;
grpentry_ptr = grpentry_ptr_next)
{
grpentry_ptr_next = grpentry_ptr->rpnext;
remap_grpentry(grpentry_ptr);
}
free(rp_grp_entry_ptr);
}
free((char *) cand_rp_delete);
}
/*
* Rehash the RP for the group. XXX: currently, every time when
* remap_grpentry() is called, there has being a good reason to change the
* RP, so for performancy reasons no check is performed whether the RP will
* be really different one.
*/
int
remap_grpentry(grpentry_ptr)
grpentry_t *grpentry_ptr;
{
rpentry_t *rpentry_ptr;
rp_grp_entry_t *rp_grp_entry_ptr;
mrtentry_t *grp_route;
mrtentry_t *mrtentry_ptr;
if (grpentry_ptr == (grpentry_t *) NULL)
return (FALSE);
/* Remove from the list of all groups matching to the same RP */
if (grpentry_ptr->rpprev != (grpentry_t *) NULL)
grpentry_ptr->rpprev->rpnext = grpentry_ptr->rpnext;
else
{
if (grpentry_ptr->active_rp_grp != (rp_grp_entry_t *) NULL)
grpentry_ptr->active_rp_grp->grplink = grpentry_ptr->rpnext;
}
if (grpentry_ptr->rpnext != (grpentry_t *) NULL)
grpentry_ptr->rpnext->rpprev = grpentry_ptr->rpprev;
rp_grp_entry_ptr = rp_grp_match(&grpentry_ptr->group);
if (rp_grp_entry_ptr == (rp_grp_entry_t *) NULL)
{
/* If cannot remap, delete the group */
delete_grpentry(grpentry_ptr);
return (FALSE);
}
rpentry_ptr = rp_grp_entry_ptr->rp->rpentry;
/* Add to the new chain of all groups mapping to the same RP */
grpentry_ptr->rpaddr = rpentry_ptr->address;
grpentry_ptr->active_rp_grp = rp_grp_entry_ptr;
grpentry_ptr->rpnext = rp_grp_entry_ptr->grplink;
if (grpentry_ptr->rpnext != (grpentry_t *) NULL)
grpentry_ptr->rpnext->rpprev = grpentry_ptr;
grpentry_ptr->rpprev = (grpentry_t *) NULL;
rp_grp_entry_ptr->grplink = grpentry_ptr;
if ((grp_route = grpentry_ptr->grp_route) != (mrtentry_t *) NULL)
{
grp_route->upstream = rpentry_ptr->upstream;
grp_route->metric = rpentry_ptr->metric;
grp_route->preference = rpentry_ptr->preference;
change_interfaces(grp_route, rpentry_ptr->incoming,
&grp_route->joined_oifs,
&grp_route->pruned_oifs,
&grp_route->leaves,
&grp_route->asserted_oifs, MFC_UPDATE_FORCE);
}
for (mrtentry_ptr = grpentry_ptr->mrtlink;
mrtentry_ptr != (mrtentry_t *) NULL;
mrtentry_ptr = mrtentry_ptr->grpnext)
{
if (!(mrtentry_ptr->flags & MRTF_RP))
continue;
mrtentry_ptr->upstream = rpentry_ptr->upstream;
mrtentry_ptr->metric = rpentry_ptr->metric;
mrtentry_ptr->preference = rpentry_ptr->preference;
change_interfaces(mrtentry_ptr, rpentry_ptr->incoming,
&mrtentry_ptr->joined_oifs,
&mrtentry_ptr->pruned_oifs,
&mrtentry_ptr->leaves,
&mrtentry_ptr->asserted_oifs, MFC_UPDATE_FORCE);
}
return (TRUE);
}
rpentry_t *
rp_match(group)
struct sockaddr_in6 *group;
{
rp_grp_entry_t *rp_grp_entry_ptr;
rp_grp_entry_ptr = rp_grp_match(group);
if (rp_grp_entry_ptr != (rp_grp_entry_t *) NULL)
return (rp_grp_entry_ptr->rp->rpentry);
else
return (rpentry_t *) NULL;
}
rp_grp_entry_t *
rp_grp_match(group)
struct sockaddr_in6 *group;
{
grp_mask_t *grp_mask_ptr;
rp_grp_entry_t *grp_rp_entry_ptr;
rp_grp_entry_t *best_entry = (rp_grp_entry_t *) NULL;
u_int8 best_priority = ~0; /* Smaller is better */
u_int32 best_hash_value = 0; /* Bigger is better */
struct sockaddr_in6 best_address_h; /* Bigger is better */
u_int32 curr_hash_value = 0;
struct sockaddr_in6 curr_address_h;
struct sockaddr_in6 prefix_h;
struct sockaddr_in6 prefix_h2;
int i;
if (grp_mask_list == (grp_mask_t *) NULL)
return (rp_grp_entry_t *) NULL;
/* XXX: I compare on the adresses, inet6_equal use the scope too */
prefix_h.sin6_scope_id = prefix_h2.sin6_scope_id = 0;
for (grp_mask_ptr = grp_mask_list; grp_mask_ptr != (grp_mask_t *) NULL;
grp_mask_ptr = grp_mask_ptr->next)
{
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h2.sin6_addr.s6_addr[i] =
(grp_mask_ptr->group_addr.sin6_addr.s6_addr[i] &
grp_mask_ptr->group_mask.s6_addr[i]);
for (i = 0; i < sizeof(struct in6_addr); i++)
prefix_h.sin6_addr.s6_addr[i] =
(group->sin6_addr.s6_addr[i] &
grp_mask_ptr->group_mask.s6_addr[i]);
/* Search the grp_mask (group_prefix) list */
if (!inet6_equal(&prefix_h, &prefix_h2))
continue;
for (grp_rp_entry_ptr = grp_mask_ptr->grp_rp_next;
grp_rp_entry_ptr != (rp_grp_entry_t *) NULL;
grp_rp_entry_ptr = grp_rp_entry_ptr->grp_rp_next)
{
if (best_priority < grp_rp_entry_ptr->priority)
break;
curr_address_h = grp_rp_entry_ptr->rp->rpentry->address;
#if 0
curr_hash_value = RP_HASH_VALUE(crc((char *)&group->sin6_addr,
sizeof(struct in6_addr)),
crc((char *)&grp_mask_ptr->hash_mask,
sizeof(struct in6_addr)),
crc((char *)&curr_address_h.sin6_addr,
sizeof(struct in6_addr)));
#else
{
struct in6_addr masked_grp;
int i;
for (i = 0; i < sizeof(struct in6_addr); i++)
masked_grp.s6_addr[i] =
group->sin6_addr.s6_addr[i] &
grp_mask_ptr->hash_mask.s6_addr[i];
curr_hash_value = RP_HASH_VALUE2(crc((char *)&masked_grp,
sizeof(struct in6_addr)),
crc((char *)&curr_address_h.sin6_addr,
sizeof(struct in6_addr)));
}
#endif
if (best_priority == grp_rp_entry_ptr->priority)
{
/* Compare the hash_value and then the addresses */
if (curr_hash_value < best_hash_value)
continue;
if (curr_hash_value == best_hash_value)
if (inet6_lessthan(&curr_address_h ,&best_address_h))
continue;
}
/* The current entry in the loop is preferred */
best_entry = grp_rp_entry_ptr;
best_priority = best_entry->priority;
best_address_h = curr_address_h;
best_hash_value = curr_hash_value;
}
}
if (best_entry == (rp_grp_entry_t *) NULL)
return (rp_grp_entry_t *) NULL;
IF_DEBUG(DEBUG_PIM_CAND_RP)
log(LOG_DEBUG,0,"Rp_grp_match found %s for group %s",
inet6_fmt(&best_entry->rp->rpentry->address.sin6_addr),
inet6_fmt(&group->sin6_addr));
return (best_entry);
}
rpentry_t *
rp_find(rp_address)
struct sockaddr_in6 *rp_address;
{
cand_rp_t *cand_rp_ptr;
for (cand_rp_ptr = cand_rp_list; cand_rp_ptr != (cand_rp_t *) NULL;
cand_rp_ptr = cand_rp_ptr->next)
{
if( inet6_greaterthan(&cand_rp_ptr->rpentry->address,rp_address))
continue;
if( inet6_equal(&cand_rp_ptr->rpentry->address,rp_address))
return (cand_rp_ptr->rpentry);
return (rpentry_t *) NULL;
}
return (rpentry_t *) NULL;
}
/*
* Create a bootstrap message in "send_buff" and returns the data size
* (excluding the IP header and the PIM header) Can be used both by the
* Bootstrap router to multicast the RP-set or by the DR to unicast it to a
* new neighbor. It DOES NOT change any timers.
*/
int
create_pim6_bootstrap_message(send_buff)
char *send_buff;
{
u_int8 *data_ptr;
grp_mask_t *grp_mask_ptr;
rp_grp_entry_t *grp_rp_entry_ptr;
int datalen;
u_int8 masklen=0;
if (IN6_IS_ADDR_UNSPECIFIED(&curr_bsr_address.sin6_addr))
return (0);
data_ptr = (u_int8 *) (send_buff + sizeof(struct pim));
if( inet6_equal(&curr_bsr_address , &my_bsr_address ))
curr_bsr_fragment_tag++;
PUT_HOSTSHORT(curr_bsr_fragment_tag, data_ptr);
MASK_TO_MASKLEN6(curr_bsr_hash_mask, masklen);
PUT_BYTE(masklen, data_ptr);
PUT_BYTE(curr_bsr_priority, data_ptr);
PUT_EUADDR6(curr_bsr_address.sin6_addr, data_ptr);
/* TODO: XXX: No fragmentation support (yet) */
for (grp_mask_ptr = grp_mask_list; grp_mask_ptr != (grp_mask_t *) NULL;
grp_mask_ptr = grp_mask_ptr->next)
{
MASK_TO_MASKLEN6(grp_mask_ptr->group_mask, masklen);
PUT_EGADDR6(grp_mask_ptr->group_addr.sin6_addr, masklen, 0, data_ptr);
PUT_BYTE(grp_mask_ptr->group_rp_number, data_ptr);
PUT_BYTE(grp_mask_ptr->group_rp_number, data_ptr); /* TODO: if frag. */
PUT_HOSTSHORT(0, data_ptr);
for (grp_rp_entry_ptr = grp_mask_ptr->grp_rp_next;
grp_rp_entry_ptr != (rp_grp_entry_t *) NULL;
grp_rp_entry_ptr = grp_rp_entry_ptr->grp_rp_next)
{
PUT_EUADDR6(grp_rp_entry_ptr->rp->rpentry->address.sin6_addr, data_ptr);
PUT_HOSTSHORT(grp_rp_entry_ptr->advholdtime, data_ptr);
PUT_BYTE(grp_rp_entry_ptr->priority, data_ptr);
PUT_BYTE(0, data_ptr); /* The reserved field */
}
}
datalen = (data_ptr - (u_int8 *) send_buff) - sizeof(struct pim);
return (datalen);
}
/*
* Check if the rp_addr is the RP for the group corresponding to
* mrtentry_ptr. Return TRUE or FALSE.
*/
int
check_mrtentry_rp(mrtentry_ptr, rp_addr)
mrtentry_t *mrtentry_ptr;
struct sockaddr_in6 *rp_addr;
{
rp_grp_entry_t *rp_grp_entry_ptr;
if (mrtentry_ptr == (mrtentry_t *) NULL)
return (FALSE);
if (IN6_IS_ADDR_UNSPECIFIED(&rp_addr->sin6_addr))
return (FALSE);
rp_grp_entry_ptr = mrtentry_ptr->group->active_rp_grp;
if (rp_grp_entry_ptr == (rp_grp_entry_t *) NULL)
return (FALSE);
if (inet6_equal(&mrtentry_ptr->group->rpaddr,rp_addr))
return (TRUE);
return (FALSE);
}