2 * Copyright (c) 2009 Bruce Simpson.
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
12 * 3. The name of the author may not be used to endorse or promote
13 * products derived from this software without specific prior written
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * $KAME: mld6.c,v 1.27 2001/04/04 05:17:30 itojun Exp $
32 * Copyright (c) 1988 Stephen Deering.
33 * Copyright (c) 1992, 1993
34 * The Regents of the University of California. All rights reserved.
36 * This code is derived from software contributed to Berkeley by
37 * Stephen Deering of Stanford University.
39 * Redistribution and use in source and binary forms, with or without
40 * modification, are permitted provided that the following conditions
42 * 1. Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 * notice, this list of conditions and the following disclaimer in the
46 * documentation and/or other materials provided with the distribution.
47 * 4. Neither the name of the University nor the names of its contributors
48 * may be used to endorse or promote products derived from this software
49 * without specific prior written permission.
51 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * @(#)igmp.c 8.1 (Berkeley) 7/19/93
66 #include <sys/cdefs.h>
67 __FBSDID("$FreeBSD$");
70 #include "opt_inet6.h"
72 #include <sys/param.h>
73 #include <sys/systm.h>
75 #include <sys/socket.h>
76 #include <sys/protosw.h>
77 #include <sys/sysctl.h>
78 #include <sys/kernel.h>
79 #include <sys/callout.h>
80 #include <sys/malloc.h>
81 #include <sys/module.h>
85 #include <net/route.h>
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet6/scope6_var.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/mld6.h>
96 #include <netinet6/mld6_var.h>
98 #include <security/mac/mac_framework.h>
101 #define KTR_MLD KTR_INET6
104 static struct mld_ifinfo *
105 mli_alloc_locked(struct ifnet *);
106 static void mli_delete_locked(const struct ifnet *);
107 static void mld_dispatch_packet(struct mbuf *);
108 static void mld_dispatch_queue(struct ifqueue *, int);
109 static void mld_final_leave(struct in6_multi *, struct mld_ifinfo *);
110 static void mld_fasttimo_vnet(void);
111 static int mld_handle_state_change(struct in6_multi *,
112 struct mld_ifinfo *);
113 static int mld_initial_join(struct in6_multi *, struct mld_ifinfo *,
116 static char * mld_rec_type_to_str(const int);
118 static void mld_set_version(struct mld_ifinfo *, const int);
119 static void mld_slowtimo_vnet(void);
120 static int mld_v1_input_query(struct ifnet *, const struct ip6_hdr *,
121 /*const*/ struct mld_hdr *);
122 static int mld_v1_input_report(struct ifnet *, const struct ip6_hdr *,
123 /*const*/ struct mld_hdr *);
124 static void mld_v1_process_group_timer(struct in6_multi *, const int);
125 static void mld_v1_process_querier_timers(struct mld_ifinfo *);
126 static int mld_v1_transmit_report(struct in6_multi *, const int);
127 static void mld_v1_update_group(struct in6_multi *, const int);
128 static void mld_v2_cancel_link_timers(struct mld_ifinfo *);
129 static void mld_v2_dispatch_general_query(struct mld_ifinfo *);
131 mld_v2_encap_report(struct ifnet *, struct mbuf *);
132 static int mld_v2_enqueue_filter_change(struct ifqueue *,
134 static int mld_v2_enqueue_group_record(struct ifqueue *,
135 struct in6_multi *, const int, const int, const int,
137 static int mld_v2_input_query(struct ifnet *, const struct ip6_hdr *,
138 struct mbuf *, const int, const int);
139 static int mld_v2_merge_state_changes(struct in6_multi *,
141 static void mld_v2_process_group_timers(struct mld_ifinfo *,
142 struct ifqueue *, struct ifqueue *,
143 struct in6_multi *, const int);
144 static int mld_v2_process_group_query(struct in6_multi *,
145 struct mld_ifinfo *mli, int, struct mbuf *, const int);
146 static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS);
147 static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS);
150 * Normative references: RFC 2710, RFC 3590, RFC 3810.
153 * * The MLD subsystem lock ends up being system-wide for the moment,
154 * but could be per-VIMAGE later on.
155 * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK.
156 * Any may be taken independently; if any are held at the same
157 * time, the above lock order must be followed.
158 * * IN6_MULTI_LOCK covers in_multi.
159 * * MLD_LOCK covers per-link state and any global variables in this file.
160 * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of
161 * per-link state iterators.
164 * A special case for IPv6 is the in6_setscope() routine. ip6_output()
165 * will not accept an ifp; it wants an embedded scope ID, unlike
166 * ip_output(), which happily takes the ifp given to it. The embedded
167 * scope ID is only used by MLD to select the outgoing interface.
169 * During interface attach and detach, MLD will take MLD_LOCK *after*
170 * the IF_AFDATA_LOCK.
171 * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call
172 * it with MLD_LOCK held without triggering an LOR. A netisr with indirect
173 * dispatch could work around this, but we'd rather not do that, as it
174 * can introduce other races.
176 * As such, we exploit the fact that the scope ID is just the interface
177 * index, and embed it in the IPv6 destination address accordingly.
178 * This is potentially NOT VALID for MLDv1 reports, as they
179 * are always sent to the multicast group itself; as MLDv2
180 * reports are always sent to ff02::16, this is not an issue
181 * when MLDv2 is in use.
183 * This does not however eliminate the LOR when ip6_output() itself
184 * calls in6_setscope() internally whilst MLD_LOCK is held. This will
185 * trigger a LOR warning in WITNESS when the ifnet is detached.
187 * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given
188 * how it's used across the network stack. Here we're simply exploiting
189 * the fact that MLD runs at a similar layer in the stack to scope6.c.
192 * * Each in6_multi corresponds to an ifp, and each ifp corresponds
193 * to a vnet in ifp->if_vnet.
195 static struct mtx mld_mtx;
196 MALLOC_DEFINE(M_MLD, "mld", "mld state");
198 #define MLD_EMBEDSCOPE(pin6, zoneid) \
199 if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \
200 IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \
201 (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \
204 * VIMAGE-wide globals.
206 static VNET_DEFINE(struct timeval, mld_gsrdelay) = {10, 0};
207 static VNET_DEFINE(LIST_HEAD(, mld_ifinfo), mli_head);
208 static VNET_DEFINE(int, interface_timers_running6);
209 static VNET_DEFINE(int, state_change_timers_running6);
210 static VNET_DEFINE(int, current_state_timers_running6);
212 #define V_mld_gsrdelay VNET(mld_gsrdelay)
213 #define V_mli_head VNET(mli_head)
214 #define V_interface_timers_running6 VNET(interface_timers_running6)
215 #define V_state_change_timers_running6 VNET(state_change_timers_running6)
216 #define V_current_state_timers_running6 VNET(current_state_timers_running6)
218 SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */
220 SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW, 0,
221 "IPv6 Multicast Listener Discovery");
224 * Virtualized sysctls.
226 SYSCTL_VNET_PROC(_net_inet6_mld, OID_AUTO, gsrdelay,
227 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
228 &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I",
229 "Rate limit for MLDv2 Group-and-Source queries in seconds");
232 * Non-virtualized sysctls.
234 SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo, CTLFLAG_RD | CTLFLAG_MPSAFE,
235 sysctl_mld_ifinfo, "Per-interface MLDv2 state");
237 static int mld_v1enable = 1;
238 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RW,
239 &mld_v1enable, 0, "Enable fallback to MLDv1");
240 TUNABLE_INT("net.inet6.mld.v1enable", &mld_v1enable);
242 static int mld_use_allow = 1;
243 SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RW,
244 &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves");
245 TUNABLE_INT("net.inet6.mld.use_allow", &mld_use_allow);
248 * Packed Router Alert option structure declaration.
253 struct ip6_opt_router ra;
257 * Router Alert hop-by-hop option header.
259 static struct mld_raopt mld_ra = {
261 .pad = { .ip6o_type = IP6OPT_PADN, 0 },
263 .ip6or_type = IP6OPT_ROUTER_ALERT,
264 .ip6or_len = IP6OPT_RTALERT_LEN - 2,
265 .ip6or_value[0] = ((IP6OPT_RTALERT_MLD >> 8) & 0xFF),
266 .ip6or_value[1] = (IP6OPT_RTALERT_MLD & 0xFF)
269 static struct ip6_pktopts mld_po;
272 mld_save_context(struct mbuf *m, struct ifnet *ifp)
276 m->m_pkthdr.header = ifp->if_vnet;
278 m->m_pkthdr.flowid = ifp->if_index;
282 mld_scrub_context(struct mbuf *m)
285 m->m_pkthdr.header = NULL;
286 m->m_pkthdr.flowid = 0;
290 * Restore context from a queued output chain.
291 * Return saved ifindex.
293 * VIMAGE: The assertion is there to make sure that we
294 * actually called CURVNET_SET() with what's in the mbuf chain.
296 static __inline uint32_t
297 mld_restore_context(struct mbuf *m)
300 #if defined(VIMAGE) && defined(INVARIANTS)
301 KASSERT(curvnet == m->m_pkthdr.header,
302 ("%s: called when curvnet was not restored", __func__));
304 return (m->m_pkthdr.flowid);
308 * Retrieve or set threshold between group-source queries in seconds.
310 * VIMAGE: Assume curvnet set by caller.
311 * SMPng: NOTE: Serialized by MLD lock.
314 sysctl_mld_gsr(SYSCTL_HANDLER_ARGS)
319 error = sysctl_wire_old_buffer(req, sizeof(int));
325 i = V_mld_gsrdelay.tv_sec;
327 error = sysctl_handle_int(oidp, &i, 0, req);
328 if (error || !req->newptr)
331 if (i < -1 || i >= 60) {
336 CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d",
337 V_mld_gsrdelay.tv_sec, i);
338 V_mld_gsrdelay.tv_sec = i;
346 * Expose struct mld_ifinfo to userland, keyed by ifindex.
347 * For use by ifmcstat(8).
349 * SMPng: NOTE: Does an unlocked ifindex space read.
350 * VIMAGE: Assume curvnet set by caller. The node handler itself
351 * is not directly virtualized.
354 sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS)
360 struct mld_ifinfo *mli;
365 if (req->newptr != NULL)
371 error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo));
378 if (name[0] <= 0 || name[0] > V_if_index) {
385 ifp = ifnet_byindex(name[0]);
389 LIST_FOREACH(mli, &V_mli_head, mli_link) {
390 if (ifp == mli->mli_ifp) {
391 error = SYSCTL_OUT(req, mli,
392 sizeof(struct mld_ifinfo));
404 * Dispatch an entire queue of pending packet chains.
405 * VIMAGE: Assumes the vnet pointer has been set.
408 mld_dispatch_queue(struct ifqueue *ifq, int limit)
416 CTR3(KTR_MLD, "%s: dispatch %p from %p", __func__, ifq, m);
417 mld_dispatch_packet(m);
424 * Filter outgoing MLD report state by group.
426 * Reports are ALWAYS suppressed for ALL-HOSTS (ff02::1)
427 * and node-local addresses. However, kernel and socket consumers
428 * always embed the KAME scope ID in the address provided, so strip it
429 * when performing comparison.
430 * Note: This is not the same as the *multicast* scope.
432 * Return zero if the given group is one for which MLD reports
433 * should be suppressed, or non-zero if reports should be issued.
436 mld_is_addr_reported(const struct in6_addr *addr)
439 KASSERT(IN6_IS_ADDR_MULTICAST(addr), ("%s: not multicast", __func__));
441 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_NODELOCAL)
444 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_LINKLOCAL) {
445 struct in6_addr tmp = *addr;
446 in6_clearscope(&tmp);
447 if (IN6_ARE_ADDR_EQUAL(&tmp, &in6addr_linklocal_allnodes))
455 * Attach MLD when PF_INET6 is attached to an interface.
457 * SMPng: Normally called with IF_AFDATA_LOCK held.
460 mld_domifattach(struct ifnet *ifp)
462 struct mld_ifinfo *mli;
464 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
465 __func__, ifp, ifp->if_xname);
469 mli = mli_alloc_locked(ifp);
470 if (!(ifp->if_flags & IFF_MULTICAST))
471 mli->mli_flags |= MLIF_SILENT;
473 mli->mli_flags |= MLIF_USEALLOW;
481 * VIMAGE: assume curvnet set by caller.
483 static struct mld_ifinfo *
484 mli_alloc_locked(/*const*/ struct ifnet *ifp)
486 struct mld_ifinfo *mli;
490 mli = malloc(sizeof(struct mld_ifinfo), M_MLD, M_NOWAIT|M_ZERO);
495 mli->mli_version = MLD_VERSION_2;
497 mli->mli_rv = MLD_RV_INIT;
498 mli->mli_qi = MLD_QI_INIT;
499 mli->mli_qri = MLD_QRI_INIT;
500 mli->mli_uri = MLD_URI_INIT;
502 SLIST_INIT(&mli->mli_relinmhead);
505 * Responses to general queries are subject to bounds.
507 IFQ_SET_MAXLEN(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS);
509 LIST_INSERT_HEAD(&V_mli_head, mli, mli_link);
511 CTR2(KTR_MLD, "allocate mld_ifinfo for ifp %p(%s)",
521 * NOTE: Some finalization tasks need to run before the protocol domain
522 * is detached, but also before the link layer does its cleanup.
523 * Run before link-layer cleanup; cleanup groups, but do not free MLD state.
525 * SMPng: Caller must hold IN6_MULTI_LOCK().
526 * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator.
527 * XXX This routine is also bitten by unlocked ifma_protospec access.
530 mld_ifdetach(struct ifnet *ifp)
532 struct mld_ifinfo *mli;
533 struct ifmultiaddr *ifma;
534 struct in6_multi *inm, *tinm;
536 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp,
539 IN6_MULTI_LOCK_ASSERT();
542 mli = MLD_IFINFO(ifp);
543 if (mli->mli_version == MLD_VERSION_2) {
545 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
546 if (ifma->ifma_addr->sa_family != AF_INET6 ||
547 ifma->ifma_protospec == NULL)
549 inm = (struct in6_multi *)ifma->ifma_protospec;
550 if (inm->in6m_state == MLD_LEAVING_MEMBER) {
551 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
554 in6m_clear_recorded(inm);
557 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele,
559 SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele);
560 in6m_release_locked(inm);
568 * Hook for domifdetach.
569 * Runs after link-layer cleanup; free MLD state.
571 * SMPng: Normally called with IF_AFDATA_LOCK held.
574 mld_domifdetach(struct ifnet *ifp)
577 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
578 __func__, ifp, ifp->if_xname);
581 mli_delete_locked(ifp);
586 mli_delete_locked(const struct ifnet *ifp)
588 struct mld_ifinfo *mli, *tmli;
590 CTR3(KTR_MLD, "%s: freeing mld_ifinfo for ifp %p(%s)",
591 __func__, ifp, ifp->if_xname);
595 LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) {
596 if (mli->mli_ifp == ifp) {
598 * Free deferred General Query responses.
600 _IF_DRAIN(&mli->mli_gq);
602 LIST_REMOVE(mli, mli_link);
604 KASSERT(SLIST_EMPTY(&mli->mli_relinmhead),
605 ("%s: there are dangling in_multi references",
613 panic("%s: mld_ifinfo not found for ifp %p\n", __func__, ifp);
618 * Process a received MLDv1 general or address-specific query.
619 * Assumes that the query header has been pulled up to sizeof(mld_hdr).
621 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
622 * mld_addr. This is OK as we own the mbuf chain.
625 mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
626 /*const*/ struct mld_hdr *mld)
628 struct ifmultiaddr *ifma;
629 struct mld_ifinfo *mli;
630 struct in6_multi *inm;
631 int is_general_query;
634 char ip6tbuf[INET6_ADDRSTRLEN];
637 is_general_query = 0;
640 CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)",
641 ip6_sprintf(ip6tbuf, &mld->mld_addr),
647 * RFC3810 Section 6.2: MLD queries must originate from
648 * a router's link-local address.
650 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
651 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
652 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
658 * Do address field validation upfront before we accept
661 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
663 * MLDv1 General Query.
664 * If this was not sent to the all-nodes group, ignore it.
669 in6_clearscope(&dst);
670 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes))
672 is_general_query = 1;
675 * Embed scope ID of receiving interface in MLD query for
676 * lookup whilst we don't hold other locks.
678 in6_setscope(&mld->mld_addr, ifp, NULL);
685 * Switch to MLDv1 host compatibility mode.
687 mli = MLD_IFINFO(ifp);
688 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
689 mld_set_version(mli, MLD_VERSION_1);
691 timer = (ntohs(mld->mld_maxdelay) * PR_FASTHZ) / MLD_TIMER_SCALE;
696 if (is_general_query) {
698 * For each reporting group joined on this
699 * interface, kick the report timer.
701 CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)",
703 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
704 if (ifma->ifma_addr->sa_family != AF_INET6 ||
705 ifma->ifma_protospec == NULL)
707 inm = (struct in6_multi *)ifma->ifma_protospec;
708 mld_v1_update_group(inm, timer);
712 * MLDv1 Group-Specific Query.
713 * If this is a group-specific MLDv1 query, we need only
714 * look up the single group to process it.
716 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
718 CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)",
719 ip6_sprintf(ip6tbuf, &mld->mld_addr),
721 mld_v1_update_group(inm, timer);
723 /* XXX Clear embedded scope ID as userland won't expect it. */
724 in6_clearscope(&mld->mld_addr);
735 * Update the report timer on a group in response to an MLDv1 query.
737 * If we are becoming the reporting member for this group, start the timer.
738 * If we already are the reporting member for this group, and timer is
739 * below the threshold, reset it.
741 * We may be updating the group for the first time since we switched
742 * to MLDv2. If we are, then we must clear any recorded source lists,
743 * and transition to REPORTING state; the group timer is overloaded
744 * for group and group-source query responses.
746 * Unlike MLDv2, the delay per group should be jittered
747 * to avoid bursts of MLDv1 reports.
750 mld_v1_update_group(struct in6_multi *inm, const int timer)
753 char ip6tbuf[INET6_ADDRSTRLEN];
756 CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__,
757 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
758 inm->in6m_ifp->if_xname, timer);
760 IN6_MULTI_LOCK_ASSERT();
762 switch (inm->in6m_state) {
764 case MLD_SILENT_MEMBER:
766 case MLD_REPORTING_MEMBER:
767 if (inm->in6m_timer != 0 &&
768 inm->in6m_timer <= timer) {
769 CTR1(KTR_MLD, "%s: REPORTING and timer running, "
770 "skipping.", __func__);
774 case MLD_SG_QUERY_PENDING_MEMBER:
775 case MLD_G_QUERY_PENDING_MEMBER:
776 case MLD_IDLE_MEMBER:
777 case MLD_LAZY_MEMBER:
778 case MLD_AWAKENING_MEMBER:
779 CTR1(KTR_MLD, "%s: ->REPORTING", __func__);
780 inm->in6m_state = MLD_REPORTING_MEMBER;
781 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
782 V_current_state_timers_running6 = 1;
784 case MLD_SLEEPING_MEMBER:
785 CTR1(KTR_MLD, "%s: ->AWAKENING", __func__);
786 inm->in6m_state = MLD_AWAKENING_MEMBER;
788 case MLD_LEAVING_MEMBER:
794 * Process a received MLDv2 general, group-specific or
795 * group-and-source-specific query.
797 * Assumes that the query header has been pulled up to sizeof(mldv2_query).
799 * Return 0 if successful, otherwise an appropriate error code is returned.
802 mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
803 struct mbuf *m, const int off, const int icmp6len)
805 struct mld_ifinfo *mli;
806 struct mldv2_query *mld;
807 struct in6_multi *inm;
808 uint32_t maxdelay, nsrc, qqi;
809 int is_general_query;
813 char ip6tbuf[INET6_ADDRSTRLEN];
816 is_general_query = 0;
819 * RFC3810 Section 6.2: MLD queries must originate from
820 * a router's link-local address.
822 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
823 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
824 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
829 CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, ifp->if_xname);
831 mld = (struct mldv2_query *)(mtod(m, uint8_t *) + off);
833 maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */
834 if (maxdelay >= 32678) {
835 maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) <<
836 (MLD_MRC_EXP(maxdelay) + 3);
838 timer = (maxdelay * PR_FASTHZ) / MLD_TIMER_SCALE;
842 qrv = MLD_QRV(mld->mld_misc);
844 CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__,
851 qqi = MLD_QQIC_MANT(mld->mld_qqi) <<
852 (MLD_QQIC_EXP(mld->mld_qqi) + 3);
855 nsrc = ntohs(mld->mld_numsrc);
856 if (nsrc > MLD_MAX_GS_SOURCES)
858 if (icmp6len < sizeof(struct mldv2_query) +
859 (nsrc * sizeof(struct in6_addr)))
863 * Do further input validation upfront to avoid resetting timers
864 * should we need to discard this query.
866 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
868 * General Queries SHOULD be directed to ff02::1.
869 * A general query with a source list has undefined
870 * behaviour; discard it.
875 in6_clearscope(&dst);
876 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes) ||
879 is_general_query = 1;
882 * Embed scope ID of receiving interface in MLD query for
883 * lookup whilst we don't hold other locks (due to KAME
884 * locking lameness). We own this mbuf chain just now.
886 in6_setscope(&mld->mld_addr, ifp, NULL);
892 mli = MLD_IFINFO(ifp);
893 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
896 * Discard the v2 query if we're in Compatibility Mode.
897 * The RFC is pretty clear that hosts need to stay in MLDv1 mode
898 * until the Old Version Querier Present timer expires.
900 if (mli->mli_version != MLD_VERSION_2)
903 mld_set_version(mli, MLD_VERSION_2);
906 mli->mli_qri = maxdelay;
908 CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi,
911 if (is_general_query) {
913 * MLDv2 General Query.
915 * Schedule a current-state report on this ifp for
916 * all groups, possibly containing source lists.
918 * If there is a pending General Query response
919 * scheduled earlier than the selected delay, do
920 * not schedule any other reports.
921 * Otherwise, reset the interface timer.
923 CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)",
925 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) {
926 mli->mli_v2_timer = MLD_RANDOM_DELAY(timer);
927 V_interface_timers_running6 = 1;
931 * MLDv2 Group-specific or Group-and-source-specific Query.
933 * Group-source-specific queries are throttled on
934 * a per-group basis to defeat denial-of-service attempts.
935 * Queries for groups we are not a member of on this
936 * link are simply ignored.
939 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
945 if (!ratecheck(&inm->in6m_lastgsrtv,
947 CTR1(KTR_MLD, "%s: GS query throttled.",
953 CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)",
956 * If there is a pending General Query response
957 * scheduled sooner than the selected delay, no
958 * further report need be scheduled.
959 * Otherwise, prepare to respond to the
960 * group-specific or group-and-source query.
962 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer)
963 mld_v2_process_group_query(inm, mli, timer, m, off);
965 /* XXX Clear embedded scope ID as userland won't expect it. */
966 in6_clearscope(&mld->mld_addr);
978 * Process a recieved MLDv2 group-specific or group-and-source-specific
980 * Return <0 if any error occured. Currently this is ignored.
983 mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifinfo *mli,
984 int timer, struct mbuf *m0, const int off)
986 struct mldv2_query *mld;
990 IN6_MULTI_LOCK_ASSERT();
994 mld = (struct mldv2_query *)(mtod(m0, uint8_t *) + off);
996 switch (inm->in6m_state) {
998 case MLD_SILENT_MEMBER:
999 case MLD_SLEEPING_MEMBER:
1000 case MLD_LAZY_MEMBER:
1001 case MLD_AWAKENING_MEMBER:
1002 case MLD_IDLE_MEMBER:
1003 case MLD_LEAVING_MEMBER:
1006 case MLD_REPORTING_MEMBER:
1007 case MLD_G_QUERY_PENDING_MEMBER:
1008 case MLD_SG_QUERY_PENDING_MEMBER:
1012 nsrc = ntohs(mld->mld_numsrc);
1015 * Deal with group-specific queries upfront.
1016 * If any group query is already pending, purge any recorded
1017 * source-list state if it exists, and schedule a query response
1018 * for this group-specific query.
1021 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
1022 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) {
1023 in6m_clear_recorded(inm);
1024 timer = min(inm->in6m_timer, timer);
1026 inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER;
1027 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1028 V_current_state_timers_running6 = 1;
1033 * Deal with the case where a group-and-source-specific query has
1034 * been received but a group-specific query is already pending.
1036 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) {
1037 timer = min(inm->in6m_timer, timer);
1038 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1039 V_current_state_timers_running6 = 1;
1044 * Finally, deal with the case where a group-and-source-specific
1045 * query has been received, where a response to a previous g-s-r
1046 * query exists, or none exists.
1047 * In this case, we need to parse the source-list which the Querier
1048 * has provided us with and check if we have any source list filter
1049 * entries at T1 for these sources. If we do not, there is no need
1050 * schedule a report and the query may be dropped.
1051 * If we do, we must record them and schedule a current-state
1052 * report for those sources.
1054 if (inm->in6m_nsrc > 0) {
1061 soff = off + sizeof(struct mldv2_query);
1063 for (i = 0; i < nsrc; i++) {
1064 sp = mtod(m, uint8_t *) + soff;
1065 retval = in6m_record_source(inm,
1066 (const struct in6_addr *)sp);
1069 nrecorded += retval;
1070 soff += sizeof(struct in6_addr);
1071 if (soff >= m->m_len) {
1072 soff = soff - m->m_len;
1078 if (nrecorded > 0) {
1080 "%s: schedule response to SG query", __func__);
1081 inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER;
1082 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1083 V_current_state_timers_running6 = 1;
1091 * Process a received MLDv1 host membership report.
1092 * Assumes mld points to mld_hdr in pulled up mbuf chain.
1094 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
1095 * mld_addr. This is OK as we own the mbuf chain.
1098 mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6,
1099 /*const*/ struct mld_hdr *mld)
1101 struct in6_addr src, dst;
1102 struct in6_ifaddr *ia;
1103 struct in6_multi *inm;
1105 char ip6tbuf[INET6_ADDRSTRLEN];
1108 if (!mld_v1enable) {
1109 CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)",
1110 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1111 ifp, ifp->if_xname);
1115 if (ifp->if_flags & IFF_LOOPBACK)
1119 * MLDv1 reports must originate from a host's link-local address,
1120 * or the unspecified address (when booting).
1123 in6_clearscope(&src);
1124 if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) {
1125 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
1126 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
1127 ifp, ifp->if_xname);
1132 * RFC2710 Section 4: MLDv1 reports must pertain to a multicast
1133 * group, and must be directed to the group itself.
1136 in6_clearscope(&dst);
1137 if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) ||
1138 !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) {
1139 CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)",
1140 ip6_sprintf(ip6tbuf, &ip6->ip6_dst),
1141 ifp, ifp->if_xname);
1146 * Make sure we don't hear our own membership report, as fast
1147 * leave requires knowing that we are the only member of a
1148 * group. Assume we used the link-local address if available,
1149 * otherwise look for ::.
1151 * XXX Note that scope ID comparison is needed for the address
1152 * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be
1153 * performed for the on-wire address.
1155 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1156 if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) ||
1157 (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) {
1159 ifa_free(&ia->ia_ifa);
1163 ifa_free(&ia->ia_ifa);
1165 CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)",
1166 ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, ifp->if_xname);
1169 * Embed scope ID of receiving interface in MLD query for lookup
1170 * whilst we don't hold other locks (due to KAME locking lameness).
1172 if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr))
1173 in6_setscope(&mld->mld_addr, ifp, NULL);
1180 * MLDv1 report suppression.
1181 * If we are a member of this group, and our membership should be
1182 * reported, and our group timer is pending or about to be reset,
1183 * stop our group timer by transitioning to the 'lazy' state.
1185 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
1187 struct mld_ifinfo *mli;
1189 mli = inm->in6m_mli;
1190 KASSERT(mli != NULL,
1191 ("%s: no mli for ifp %p", __func__, ifp));
1194 * If we are in MLDv2 host mode, do not allow the
1195 * other host's MLDv1 report to suppress our reports.
1197 if (mli->mli_version == MLD_VERSION_2)
1200 inm->in6m_timer = 0;
1202 switch (inm->in6m_state) {
1203 case MLD_NOT_MEMBER:
1204 case MLD_SILENT_MEMBER:
1205 case MLD_SLEEPING_MEMBER:
1207 case MLD_REPORTING_MEMBER:
1208 case MLD_IDLE_MEMBER:
1209 case MLD_AWAKENING_MEMBER:
1211 "report suppressed for %s on ifp %p(%s)",
1212 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1213 ifp, ifp->if_xname);
1214 case MLD_LAZY_MEMBER:
1215 inm->in6m_state = MLD_LAZY_MEMBER;
1217 case MLD_G_QUERY_PENDING_MEMBER:
1218 case MLD_SG_QUERY_PENDING_MEMBER:
1219 case MLD_LEAVING_MEMBER:
1226 IF_ADDR_UNLOCK(ifp);
1229 /* XXX Clear embedded scope ID as userland won't expect it. */
1230 in6_clearscope(&mld->mld_addr);
1238 * Assume query messages which fit in a single ICMPv6 message header
1239 * have been pulled up.
1240 * Assume that userland will want to see the message, even if it
1241 * otherwise fails kernel input validation; do not free it.
1242 * Pullup may however free the mbuf chain m if it fails.
1244 * Return IPPROTO_DONE if we freed m. Otherwise, return 0.
1247 mld_input(struct mbuf *m, int off, int icmp6len)
1250 struct ip6_hdr *ip6;
1251 struct mld_hdr *mld;
1254 CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off);
1256 ifp = m->m_pkthdr.rcvif;
1258 ip6 = mtod(m, struct ip6_hdr *);
1260 /* Pullup to appropriate size. */
1261 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1262 if (mld->mld_type == MLD_LISTENER_QUERY &&
1263 icmp6len >= sizeof(struct mldv2_query)) {
1264 mldlen = sizeof(struct mldv2_query);
1266 mldlen = sizeof(struct mld_hdr);
1268 IP6_EXTHDR_GET(mld, struct mld_hdr *, m, off, mldlen);
1270 ICMP6STAT_INC(icp6s_badlen);
1271 return (IPPROTO_DONE);
1275 * Userland needs to see all of this traffic for implementing
1276 * the endpoint discovery portion of multicast routing.
1278 switch (mld->mld_type) {
1279 case MLD_LISTENER_QUERY:
1280 icmp6_ifstat_inc(ifp, ifs6_in_mldquery);
1281 if (icmp6len == sizeof(struct mld_hdr)) {
1282 if (mld_v1_input_query(ifp, ip6, mld) != 0)
1284 } else if (icmp6len >= sizeof(struct mldv2_query)) {
1285 if (mld_v2_input_query(ifp, ip6, m, off,
1290 case MLD_LISTENER_REPORT:
1291 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1292 if (mld_v1_input_report(ifp, ip6, mld) != 0)
1295 case MLDV2_LISTENER_REPORT:
1296 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1298 case MLD_LISTENER_DONE:
1299 icmp6_ifstat_inc(ifp, ifs6_in_mlddone);
1309 * Fast timeout handler (global).
1310 * VIMAGE: Timeout handlers are expected to service all vimages.
1315 VNET_ITERATOR_DECL(vnet_iter);
1317 VNET_LIST_RLOCK_NOSLEEP();
1318 VNET_FOREACH(vnet_iter) {
1319 CURVNET_SET(vnet_iter);
1320 mld_fasttimo_vnet();
1323 VNET_LIST_RUNLOCK_NOSLEEP();
1327 * Fast timeout handler (per-vnet).
1329 * VIMAGE: Assume caller has set up our curvnet.
1332 mld_fasttimo_vnet(void)
1334 struct ifqueue scq; /* State-change packets */
1335 struct ifqueue qrq; /* Query response packets */
1337 struct mld_ifinfo *mli;
1338 struct ifmultiaddr *ifma, *tifma;
1339 struct in6_multi *inm;
1345 * Quick check to see if any work needs to be done, in order to
1346 * minimize the overhead of fasttimo processing.
1347 * SMPng: XXX Unlocked reads.
1349 if (!V_current_state_timers_running6 &&
1350 !V_interface_timers_running6 &&
1351 !V_state_change_timers_running6)
1358 * MLDv2 General Query response timer processing.
1360 if (V_interface_timers_running6) {
1361 CTR1(KTR_MLD, "%s: interface timers running", __func__);
1363 V_interface_timers_running6 = 0;
1364 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1365 if (mli->mli_v2_timer == 0) {
1367 } else if (--mli->mli_v2_timer == 0) {
1368 mld_v2_dispatch_general_query(mli);
1370 V_interface_timers_running6 = 1;
1375 if (!V_current_state_timers_running6 &&
1376 !V_state_change_timers_running6)
1379 V_current_state_timers_running6 = 0;
1380 V_state_change_timers_running6 = 0;
1382 CTR1(KTR_MLD, "%s: state change timers running", __func__);
1385 * MLD host report and state-change timer processing.
1386 * Note: Processing a v2 group timer may remove a node.
1388 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1391 if (mli->mli_version == MLD_VERSION_2) {
1392 uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri *
1395 memset(&qrq, 0, sizeof(struct ifqueue));
1396 IFQ_SET_MAXLEN(&qrq, MLD_MAX_G_GS_PACKETS);
1398 memset(&scq, 0, sizeof(struct ifqueue));
1399 IFQ_SET_MAXLEN(&scq, MLD_MAX_STATE_CHANGE_PACKETS);
1403 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link,
1405 if (ifma->ifma_addr->sa_family != AF_INET6 ||
1406 ifma->ifma_protospec == NULL)
1408 inm = (struct in6_multi *)ifma->ifma_protospec;
1409 switch (mli->mli_version) {
1412 * XXX Drop IF_ADDR lock temporarily to
1413 * avoid recursion caused by a potential
1414 * call by in6ifa_ifpforlinklocal().
1417 IF_ADDR_UNLOCK(ifp);
1418 mld_v1_process_group_timer(inm,
1423 mld_v2_process_group_timers(mli, &qrq,
1424 &scq, inm, uri_fasthz);
1428 IF_ADDR_UNLOCK(ifp);
1430 if (mli->mli_version == MLD_VERSION_2) {
1431 struct in6_multi *tinm;
1433 mld_dispatch_queue(&qrq, 0);
1434 mld_dispatch_queue(&scq, 0);
1437 * Free the in_multi reference(s) for
1440 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead,
1442 SLIST_REMOVE_HEAD(&mli->mli_relinmhead,
1444 in6m_release_locked(inm);
1455 * Update host report group timer.
1456 * Will update the global pending timer flags.
1459 mld_v1_process_group_timer(struct in6_multi *inm, const int version)
1461 int report_timer_expired;
1463 IN6_MULTI_LOCK_ASSERT();
1466 if (inm->in6m_timer == 0) {
1467 report_timer_expired = 0;
1468 } else if (--inm->in6m_timer == 0) {
1469 report_timer_expired = 1;
1471 V_current_state_timers_running6 = 1;
1475 switch (inm->in6m_state) {
1476 case MLD_NOT_MEMBER:
1477 case MLD_SILENT_MEMBER:
1478 case MLD_IDLE_MEMBER:
1479 case MLD_LAZY_MEMBER:
1480 case MLD_SLEEPING_MEMBER:
1481 case MLD_AWAKENING_MEMBER:
1483 case MLD_REPORTING_MEMBER:
1484 if (report_timer_expired) {
1485 inm->in6m_state = MLD_IDLE_MEMBER;
1486 (void)mld_v1_transmit_report(inm,
1487 MLD_LISTENER_REPORT);
1490 case MLD_G_QUERY_PENDING_MEMBER:
1491 case MLD_SG_QUERY_PENDING_MEMBER:
1492 case MLD_LEAVING_MEMBER:
1498 * Update a group's timers for MLDv2.
1499 * Will update the global pending timer flags.
1500 * Note: Unlocked read from mli.
1503 mld_v2_process_group_timers(struct mld_ifinfo *mli,
1504 struct ifqueue *qrq, struct ifqueue *scq,
1505 struct in6_multi *inm, const int uri_fasthz)
1507 int query_response_timer_expired;
1508 int state_change_retransmit_timer_expired;
1510 char ip6tbuf[INET6_ADDRSTRLEN];
1513 IN6_MULTI_LOCK_ASSERT();
1516 query_response_timer_expired = 0;
1517 state_change_retransmit_timer_expired = 0;
1520 * During a transition from compatibility mode back to MLDv2,
1521 * a group record in REPORTING state may still have its group
1522 * timer active. This is a no-op in this function; it is easier
1523 * to deal with it here than to complicate the slow-timeout path.
1525 if (inm->in6m_timer == 0) {
1526 query_response_timer_expired = 0;
1527 } else if (--inm->in6m_timer == 0) {
1528 query_response_timer_expired = 1;
1530 V_current_state_timers_running6 = 1;
1533 if (inm->in6m_sctimer == 0) {
1534 state_change_retransmit_timer_expired = 0;
1535 } else if (--inm->in6m_sctimer == 0) {
1536 state_change_retransmit_timer_expired = 1;
1538 V_state_change_timers_running6 = 1;
1541 /* We are in fasttimo, so be quick about it. */
1542 if (!state_change_retransmit_timer_expired &&
1543 !query_response_timer_expired)
1546 switch (inm->in6m_state) {
1547 case MLD_NOT_MEMBER:
1548 case MLD_SILENT_MEMBER:
1549 case MLD_SLEEPING_MEMBER:
1550 case MLD_LAZY_MEMBER:
1551 case MLD_AWAKENING_MEMBER:
1552 case MLD_IDLE_MEMBER:
1554 case MLD_G_QUERY_PENDING_MEMBER:
1555 case MLD_SG_QUERY_PENDING_MEMBER:
1557 * Respond to a previously pending Group-Specific
1558 * or Group-and-Source-Specific query by enqueueing
1559 * the appropriate Current-State report for
1560 * immediate transmission.
1562 if (query_response_timer_expired) {
1565 retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1,
1566 (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER),
1568 CTR2(KTR_MLD, "%s: enqueue record = %d",
1570 inm->in6m_state = MLD_REPORTING_MEMBER;
1571 in6m_clear_recorded(inm);
1574 case MLD_REPORTING_MEMBER:
1575 case MLD_LEAVING_MEMBER:
1576 if (state_change_retransmit_timer_expired) {
1578 * State-change retransmission timer fired.
1579 * If there are any further pending retransmissions,
1580 * set the global pending state-change flag, and
1583 if (--inm->in6m_scrv > 0) {
1584 inm->in6m_sctimer = uri_fasthz;
1585 V_state_change_timers_running6 = 1;
1588 * Retransmit the previously computed state-change
1589 * report. If there are no further pending
1590 * retransmissions, the mbuf queue will be consumed.
1591 * Update T0 state to T1 as we have now sent
1594 (void)mld_v2_merge_state_changes(inm, scq);
1597 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
1598 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1599 inm->in6m_ifp->if_xname);
1602 * If we are leaving the group for good, make sure
1603 * we release MLD's reference to it.
1604 * This release must be deferred using a SLIST,
1605 * as we are called from a loop which traverses
1606 * the in_ifmultiaddr TAILQ.
1608 if (inm->in6m_state == MLD_LEAVING_MEMBER &&
1609 inm->in6m_scrv == 0) {
1610 inm->in6m_state = MLD_NOT_MEMBER;
1611 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
1620 * Switch to a different version on the given interface,
1621 * as per Section 9.12.
1624 mld_set_version(struct mld_ifinfo *mli, const int version)
1626 int old_version_timer;
1630 CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__,
1631 version, mli->mli_ifp, mli->mli_ifp->if_xname);
1633 if (version == MLD_VERSION_1) {
1635 * Compute the "Older Version Querier Present" timer as per
1638 old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri;
1639 old_version_timer *= PR_SLOWHZ;
1640 mli->mli_v1_timer = old_version_timer;
1643 if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) {
1644 mli->mli_version = MLD_VERSION_1;
1645 mld_v2_cancel_link_timers(mli);
1650 * Cancel pending MLDv2 timers for the given link and all groups
1651 * joined on it; state-change, general-query, and group-query timers.
1654 mld_v2_cancel_link_timers(struct mld_ifinfo *mli)
1656 struct ifmultiaddr *ifma;
1658 struct in6_multi *inm;
1660 CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__,
1661 mli->mli_ifp, mli->mli_ifp->if_xname);
1663 IN6_MULTI_LOCK_ASSERT();
1667 * Fast-track this potentially expensive operation
1668 * by checking all the global 'timer pending' flags.
1670 if (!V_interface_timers_running6 &&
1671 !V_state_change_timers_running6 &&
1672 !V_current_state_timers_running6)
1675 mli->mli_v2_timer = 0;
1680 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1681 if (ifma->ifma_addr->sa_family != AF_INET6)
1683 inm = (struct in6_multi *)ifma->ifma_protospec;
1684 switch (inm->in6m_state) {
1685 case MLD_NOT_MEMBER:
1686 case MLD_SILENT_MEMBER:
1687 case MLD_IDLE_MEMBER:
1688 case MLD_LAZY_MEMBER:
1689 case MLD_SLEEPING_MEMBER:
1690 case MLD_AWAKENING_MEMBER:
1692 case MLD_LEAVING_MEMBER:
1694 * If we are leaving the group and switching
1695 * version, we need to release the final
1696 * reference held for issuing the INCLUDE {}.
1698 * SMPNG: Must drop and re-acquire IF_ADDR_LOCK
1699 * around in6m_release_locked(), as it is not
1700 * a recursive mutex.
1702 IF_ADDR_UNLOCK(ifp);
1703 in6m_release_locked(inm);
1706 case MLD_G_QUERY_PENDING_MEMBER:
1707 case MLD_SG_QUERY_PENDING_MEMBER:
1708 in6m_clear_recorded(inm);
1710 case MLD_REPORTING_MEMBER:
1711 inm->in6m_sctimer = 0;
1712 inm->in6m_timer = 0;
1713 inm->in6m_state = MLD_REPORTING_MEMBER;
1715 * Free any pending MLDv2 state-change records.
1717 _IF_DRAIN(&inm->in6m_scq);
1721 IF_ADDR_UNLOCK(ifp);
1725 * Global slowtimo handler.
1726 * VIMAGE: Timeout handlers are expected to service all vimages.
1731 VNET_ITERATOR_DECL(vnet_iter);
1733 VNET_LIST_RLOCK_NOSLEEP();
1734 VNET_FOREACH(vnet_iter) {
1735 CURVNET_SET(vnet_iter);
1736 mld_slowtimo_vnet();
1739 VNET_LIST_RUNLOCK_NOSLEEP();
1743 * Per-vnet slowtimo handler.
1746 mld_slowtimo_vnet(void)
1748 struct mld_ifinfo *mli;
1752 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1753 mld_v1_process_querier_timers(mli);
1760 * Update the Older Version Querier Present timers for a link.
1761 * See Section 9.12 of RFC 3810.
1764 mld_v1_process_querier_timers(struct mld_ifinfo *mli)
1769 if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) {
1771 * MLDv1 Querier Present timer expired; revert to MLDv2.
1774 "%s: transition from v%d -> v%d on %p(%s)",
1775 __func__, mli->mli_version, MLD_VERSION_2,
1776 mli->mli_ifp, mli->mli_ifp->if_xname);
1777 mli->mli_version = MLD_VERSION_2;
1782 * Transmit an MLDv1 report immediately.
1785 mld_v1_transmit_report(struct in6_multi *in6m, const int type)
1788 struct in6_ifaddr *ia;
1789 struct ip6_hdr *ip6;
1790 struct mbuf *mh, *md;
1791 struct mld_hdr *mld;
1793 IN6_MULTI_LOCK_ASSERT();
1796 ifp = in6m->in6m_ifp;
1797 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1798 /* ia may be NULL if link-local address is tentative. */
1800 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
1803 ifa_free(&ia->ia_ifa);
1806 MGET(md, M_DONTWAIT, MT_DATA);
1810 ifa_free(&ia->ia_ifa);
1816 * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so
1817 * that ether_output() does not need to allocate another mbuf
1818 * for the header in the most common case.
1820 MH_ALIGN(mh, sizeof(struct ip6_hdr));
1821 mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
1822 mh->m_len = sizeof(struct ip6_hdr);
1824 ip6 = mtod(mh, struct ip6_hdr *);
1826 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1827 ip6->ip6_vfc |= IPV6_VERSION;
1828 ip6->ip6_nxt = IPPROTO_ICMPV6;
1829 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
1830 ip6->ip6_dst = in6m->in6m_addr;
1832 md->m_len = sizeof(struct mld_hdr);
1833 mld = mtod(md, struct mld_hdr *);
1834 mld->mld_type = type;
1837 mld->mld_maxdelay = 0;
1838 mld->mld_reserved = 0;
1839 mld->mld_addr = in6m->in6m_addr;
1840 in6_clearscope(&mld->mld_addr);
1841 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
1842 sizeof(struct ip6_hdr), sizeof(struct mld_hdr));
1844 mld_save_context(mh, ifp);
1845 mh->m_flags |= M_MLDV1;
1847 mld_dispatch_packet(mh);
1850 ifa_free(&ia->ia_ifa);
1855 * Process a state change from the upper layer for the given IPv6 group.
1857 * Each socket holds a reference on the in_multi in its own ip_moptions.
1858 * The socket layer will have made the necessary updates to.the group
1859 * state, it is now up to MLD to issue a state change report if there
1860 * has been any change between T0 (when the last state-change was issued)
1863 * We use the MLDv2 state machine at group level. The MLd module
1864 * however makes the decision as to which MLD protocol version to speak.
1865 * A state change *from* INCLUDE {} always means an initial join.
1866 * A state change *to* INCLUDE {} always means a final leave.
1868 * If delay is non-zero, and the state change is an initial multicast
1869 * join, the state change report will be delayed by 'delay' ticks
1870 * in units of PR_FASTHZ if MLDv1 is active on the link; otherwise
1871 * the initial MLDv2 state change report will be delayed by whichever
1872 * is sooner, a pending state-change timer or delay itself.
1874 * VIMAGE: curvnet should have been set by caller, as this routine
1875 * is called from the socket option handlers.
1878 mld_change_state(struct in6_multi *inm, const int delay)
1880 struct mld_ifinfo *mli;
1884 IN6_MULTI_LOCK_ASSERT();
1889 * Try to detect if the upper layer just asked us to change state
1890 * for an interface which has now gone away.
1892 KASSERT(inm->in6m_ifma != NULL, ("%s: no ifma", __func__));
1893 ifp = inm->in6m_ifma->ifma_ifp;
1896 * Sanity check that netinet6's notion of ifp is the
1899 KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__));
1904 mli = MLD_IFINFO(ifp);
1905 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
1908 * If we detect a state transition to or from MCAST_UNDEFINED
1909 * for this group, then we are starting or finishing an MLD
1910 * life cycle for this group.
1912 if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) {
1913 CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__,
1914 inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode);
1915 if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) {
1916 CTR1(KTR_MLD, "%s: initial join", __func__);
1917 error = mld_initial_join(inm, mli, delay);
1919 } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) {
1920 CTR1(KTR_MLD, "%s: final leave", __func__);
1921 mld_final_leave(inm, mli);
1925 CTR1(KTR_MLD, "%s: filter set change", __func__);
1928 error = mld_handle_state_change(inm, mli);
1936 * Perform the initial join for an MLD group.
1938 * When joining a group:
1939 * If the group should have its MLD traffic suppressed, do nothing.
1940 * MLDv1 starts sending MLDv1 host membership reports.
1941 * MLDv2 will schedule an MLDv2 state-change report containing the
1942 * initial state of the membership.
1944 * If the delay argument is non-zero, then we must delay sending the
1945 * initial state change for delay ticks (in units of PR_FASTHZ).
1948 mld_initial_join(struct in6_multi *inm, struct mld_ifinfo *mli,
1952 struct ifqueue *ifq;
1953 int error, retval, syncstates;
1956 char ip6tbuf[INET6_ADDRSTRLEN];
1959 CTR4(KTR_MLD, "%s: initial join %s on ifp %p(%s)",
1960 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1961 inm->in6m_ifp, inm->in6m_ifp->if_xname);
1966 ifp = inm->in6m_ifp;
1968 IN6_MULTI_LOCK_ASSERT();
1971 KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__));
1974 * Groups joined on loopback or marked as 'not reported',
1975 * enter the MLD_SILENT_MEMBER state and
1976 * are never reported in any protocol exchanges.
1977 * All other groups enter the appropriate state machine
1978 * for the version in use on this link.
1979 * A link marked as MLIF_SILENT causes MLD to be completely
1980 * disabled for the link.
1982 if ((ifp->if_flags & IFF_LOOPBACK) ||
1983 (mli->mli_flags & MLIF_SILENT) ||
1984 !mld_is_addr_reported(&inm->in6m_addr)) {
1986 "%s: not kicking state machine for silent group", __func__);
1987 inm->in6m_state = MLD_SILENT_MEMBER;
1988 inm->in6m_timer = 0;
1991 * Deal with overlapping in_multi lifecycle.
1992 * If this group was LEAVING, then make sure
1993 * we drop the reference we picked up to keep the
1994 * group around for the final INCLUDE {} enqueue.
1996 if (mli->mli_version == MLD_VERSION_2 &&
1997 inm->in6m_state == MLD_LEAVING_MEMBER)
1998 in6m_release_locked(inm);
2000 inm->in6m_state = MLD_REPORTING_MEMBER;
2002 switch (mli->mli_version) {
2005 * If a delay was provided, only use it if
2006 * it is greater than the delay normally
2007 * used for an MLDv1 state change report,
2008 * and delay sending the initial MLDv1 report
2009 * by not transitioning to the IDLE state.
2011 odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * PR_FASTHZ);
2013 inm->in6m_timer = max(delay, odelay);
2014 V_current_state_timers_running6 = 1;
2016 inm->in6m_state = MLD_IDLE_MEMBER;
2017 error = mld_v1_transmit_report(inm,
2018 MLD_LISTENER_REPORT);
2020 inm->in6m_timer = odelay;
2021 V_current_state_timers_running6 = 1;
2028 * Defer update of T0 to T1, until the first copy
2029 * of the state change has been transmitted.
2034 * Immediately enqueue a State-Change Report for
2035 * this interface, freeing any previous reports.
2036 * Don't kick the timers if there is nothing to do,
2037 * or if an error occurred.
2039 ifq = &inm->in6m_scq;
2041 retval = mld_v2_enqueue_group_record(ifq, inm, 1,
2042 0, 0, (mli->mli_flags & MLIF_USEALLOW));
2043 CTR2(KTR_MLD, "%s: enqueue record = %d",
2046 error = retval * -1;
2051 * Schedule transmission of pending state-change
2052 * report up to RV times for this link. The timer
2053 * will fire at the next mld_fasttimo (~200ms),
2054 * giving us an opportunity to merge the reports.
2056 * If a delay was provided to this function, only
2057 * use this delay if sooner than the existing one.
2059 KASSERT(mli->mli_rv > 1,
2060 ("%s: invalid robustness %d", __func__,
2062 inm->in6m_scrv = mli->mli_rv;
2064 if (inm->in6m_sctimer > 1) {
2066 min(inm->in6m_sctimer, delay);
2068 inm->in6m_sctimer = delay;
2070 inm->in6m_sctimer = 1;
2071 V_state_change_timers_running6 = 1;
2079 * Only update the T0 state if state change is atomic,
2080 * i.e. we don't need to wait for a timer to fire before we
2081 * can consider the state change to have been communicated.
2085 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2086 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2087 inm->in6m_ifp->if_xname);
2094 * Issue an intermediate state change during the life-cycle.
2097 mld_handle_state_change(struct in6_multi *inm, struct mld_ifinfo *mli)
2102 char ip6tbuf[INET6_ADDRSTRLEN];
2105 CTR4(KTR_MLD, "%s: state change for %s on ifp %p(%s)",
2106 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2107 inm->in6m_ifp, inm->in6m_ifp->if_xname);
2109 ifp = inm->in6m_ifp;
2111 IN6_MULTI_LOCK_ASSERT();
2114 KASSERT(mli && mli->mli_ifp == ifp,
2115 ("%s: inconsistent ifp", __func__));
2117 if ((ifp->if_flags & IFF_LOOPBACK) ||
2118 (mli->mli_flags & MLIF_SILENT) ||
2119 !mld_is_addr_reported(&inm->in6m_addr) ||
2120 (mli->mli_version != MLD_VERSION_2)) {
2121 if (!mld_is_addr_reported(&inm->in6m_addr)) {
2123 "%s: not kicking state machine for silent group", __func__);
2125 CTR1(KTR_MLD, "%s: nothing to do", __func__);
2127 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2128 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2129 inm->in6m_ifp->if_xname);
2133 _IF_DRAIN(&inm->in6m_scq);
2135 retval = mld_v2_enqueue_group_record(&inm->in6m_scq, inm, 1, 0, 0,
2136 (mli->mli_flags & MLIF_USEALLOW));
2137 CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval);
2142 * If record(s) were enqueued, start the state-change
2143 * report timer for this group.
2145 inm->in6m_scrv = mli->mli_rv;
2146 inm->in6m_sctimer = 1;
2147 V_state_change_timers_running6 = 1;
2153 * Perform the final leave for a multicast address.
2155 * When leaving a group:
2156 * MLDv1 sends a DONE message, if and only if we are the reporter.
2157 * MLDv2 enqueues a state-change report containing a transition
2158 * to INCLUDE {} for immediate transmission.
2161 mld_final_leave(struct in6_multi *inm, struct mld_ifinfo *mli)
2165 char ip6tbuf[INET6_ADDRSTRLEN];
2170 CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)",
2171 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2172 inm->in6m_ifp, inm->in6m_ifp->if_xname);
2174 IN6_MULTI_LOCK_ASSERT();
2177 switch (inm->in6m_state) {
2178 case MLD_NOT_MEMBER:
2179 case MLD_SILENT_MEMBER:
2180 case MLD_LEAVING_MEMBER:
2181 /* Already leaving or left; do nothing. */
2183 "%s: not kicking state machine for silent group", __func__);
2185 case MLD_REPORTING_MEMBER:
2186 case MLD_IDLE_MEMBER:
2187 case MLD_G_QUERY_PENDING_MEMBER:
2188 case MLD_SG_QUERY_PENDING_MEMBER:
2189 if (mli->mli_version == MLD_VERSION_1) {
2191 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
2192 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER)
2193 panic("%s: MLDv2 state reached, not MLDv2 mode",
2196 mld_v1_transmit_report(inm, MLD_LISTENER_DONE);
2197 inm->in6m_state = MLD_NOT_MEMBER;
2198 } else if (mli->mli_version == MLD_VERSION_2) {
2200 * Stop group timer and all pending reports.
2201 * Immediately enqueue a state-change report
2202 * TO_IN {} to be sent on the next fast timeout,
2203 * giving us an opportunity to merge reports.
2205 _IF_DRAIN(&inm->in6m_scq);
2206 inm->in6m_timer = 0;
2207 inm->in6m_scrv = mli->mli_rv;
2208 CTR4(KTR_MLD, "%s: Leaving %s/%s with %d "
2209 "pending retransmissions.", __func__,
2210 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2211 inm->in6m_ifp->if_xname, inm->in6m_scrv);
2212 if (inm->in6m_scrv == 0) {
2213 inm->in6m_state = MLD_NOT_MEMBER;
2214 inm->in6m_sctimer = 0;
2218 in6m_acquire_locked(inm);
2220 retval = mld_v2_enqueue_group_record(
2221 &inm->in6m_scq, inm, 1, 0, 0,
2222 (mli->mli_flags & MLIF_USEALLOW));
2223 KASSERT(retval != 0,
2224 ("%s: enqueue record = %d", __func__,
2227 inm->in6m_state = MLD_LEAVING_MEMBER;
2228 inm->in6m_sctimer = 1;
2229 V_state_change_timers_running6 = 1;
2235 case MLD_LAZY_MEMBER:
2236 case MLD_SLEEPING_MEMBER:
2237 case MLD_AWAKENING_MEMBER:
2238 /* Our reports are suppressed; do nothing. */
2244 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2245 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2246 inm->in6m_ifp->if_xname);
2247 inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED;
2248 CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s",
2249 __func__, &inm->in6m_addr, inm->in6m_ifp->if_xname);
2254 * Enqueue an MLDv2 group record to the given output queue.
2256 * If is_state_change is zero, a current-state record is appended.
2257 * If is_state_change is non-zero, a state-change report is appended.
2259 * If is_group_query is non-zero, an mbuf packet chain is allocated.
2260 * If is_group_query is zero, and if there is a packet with free space
2261 * at the tail of the queue, it will be appended to providing there
2262 * is enough free space.
2263 * Otherwise a new mbuf packet chain is allocated.
2265 * If is_source_query is non-zero, each source is checked to see if
2266 * it was recorded for a Group-Source query, and will be omitted if
2267 * it is not both in-mode and recorded.
2269 * If use_block_allow is non-zero, state change reports for initial join
2270 * and final leave, on an inclusive mode group with a source list, will be
2271 * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively.
2273 * The function will attempt to allocate leading space in the packet
2274 * for the IPv6+ICMP headers to be prepended without fragmenting the chain.
2276 * If successful the size of all data appended to the queue is returned,
2277 * otherwise an error code less than zero is returned, or zero if
2278 * no record(s) were appended.
2281 mld_v2_enqueue_group_record(struct ifqueue *ifq, struct in6_multi *inm,
2282 const int is_state_change, const int is_group_query,
2283 const int is_source_query, const int use_block_allow)
2285 struct mldv2_record mr;
2286 struct mldv2_record *pmr;
2288 struct ip6_msource *ims, *nims;
2289 struct mbuf *m0, *m, *md;
2290 int error, is_filter_list_change;
2291 int minrec0len, m0srcs, msrcs, nbytes, off;
2292 int record_has_sources;
2297 char ip6tbuf[INET6_ADDRSTRLEN];
2300 IN6_MULTI_LOCK_ASSERT();
2303 ifp = inm->in6m_ifp;
2304 is_filter_list_change = 0;
2311 record_has_sources = 1;
2313 type = MLD_DO_NOTHING;
2314 mode = inm->in6m_st[1].iss_fmode;
2317 * If we did not transition out of ASM mode during t0->t1,
2318 * and there are no source nodes to process, we can skip
2319 * the generation of source records.
2321 if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 &&
2322 inm->in6m_nsrc == 0)
2323 record_has_sources = 0;
2325 if (is_state_change) {
2327 * Queue a state change record.
2328 * If the mode did not change, and there are non-ASM
2329 * listeners or source filters present,
2330 * we potentially need to issue two records for the group.
2331 * If there are ASM listeners, and there was no filter
2332 * mode transition of any kind, do nothing.
2334 * If we are transitioning to MCAST_UNDEFINED, we need
2335 * not send any sources. A transition to/from this state is
2336 * considered inclusive with some special treatment.
2338 * If we are rewriting initial joins/leaves to use
2339 * ALLOW/BLOCK, and the group's membership is inclusive,
2340 * we need to send sources in all cases.
2342 if (mode != inm->in6m_st[0].iss_fmode) {
2343 if (mode == MCAST_EXCLUDE) {
2344 CTR1(KTR_MLD, "%s: change to EXCLUDE",
2346 type = MLD_CHANGE_TO_EXCLUDE_MODE;
2348 CTR1(KTR_MLD, "%s: change to INCLUDE",
2350 if (use_block_allow) {
2353 * Here we're interested in state
2354 * edges either direction between
2355 * MCAST_UNDEFINED and MCAST_INCLUDE.
2356 * Perhaps we should just check
2357 * the group state, rather than
2360 if (mode == MCAST_UNDEFINED) {
2361 type = MLD_BLOCK_OLD_SOURCES;
2363 type = MLD_ALLOW_NEW_SOURCES;
2366 type = MLD_CHANGE_TO_INCLUDE_MODE;
2367 if (mode == MCAST_UNDEFINED)
2368 record_has_sources = 0;
2372 if (record_has_sources) {
2373 is_filter_list_change = 1;
2375 type = MLD_DO_NOTHING;
2380 * Queue a current state record.
2382 if (mode == MCAST_EXCLUDE) {
2383 type = MLD_MODE_IS_EXCLUDE;
2384 } else if (mode == MCAST_INCLUDE) {
2385 type = MLD_MODE_IS_INCLUDE;
2386 KASSERT(inm->in6m_st[1].iss_asm == 0,
2387 ("%s: inm %p is INCLUDE but ASM count is %d",
2388 __func__, inm, inm->in6m_st[1].iss_asm));
2393 * Generate the filter list changes using a separate function.
2395 if (is_filter_list_change)
2396 return (mld_v2_enqueue_filter_change(ifq, inm));
2398 if (type == MLD_DO_NOTHING) {
2399 CTR3(KTR_MLD, "%s: nothing to do for %s/%s",
2400 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2401 inm->in6m_ifp->if_xname);
2406 * If any sources are present, we must be able to fit at least
2407 * one in the trailing space of the tail packet's mbuf,
2410 minrec0len = sizeof(struct mldv2_record);
2411 if (record_has_sources)
2412 minrec0len += sizeof(struct in6_addr);
2414 CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__,
2415 mld_rec_type_to_str(type),
2416 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2417 inm->in6m_ifp->if_xname);
2420 * Check if we have a packet in the tail of the queue for this
2421 * group into which the first group record for this group will fit.
2422 * Otherwise allocate a new packet.
2423 * Always allocate leading space for IP6+RA+ICMPV6+REPORT.
2424 * Note: Group records for G/GSR query responses MUST be sent
2425 * in their own packet.
2428 if (!is_group_query &&
2430 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) &&
2431 (m0->m_pkthdr.len + minrec0len) <
2432 (ifp->if_mtu - MLD_MTUSPACE)) {
2433 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2434 sizeof(struct mldv2_record)) /
2435 sizeof(struct in6_addr);
2437 CTR1(KTR_MLD, "%s: use existing packet", __func__);
2439 if (_IF_QFULL(ifq)) {
2440 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2444 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2445 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2446 if (!is_state_change && !is_group_query)
2447 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2449 m = m_gethdr(M_DONTWAIT, MT_DATA);
2453 mld_save_context(m, ifp);
2455 CTR1(KTR_MLD, "%s: allocated first packet", __func__);
2459 * Append group record.
2460 * If we have sources, we don't know how many yet.
2465 mr.mr_addr = inm->in6m_addr;
2466 in6_clearscope(&mr.mr_addr);
2467 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2470 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2473 nbytes += sizeof(struct mldv2_record);
2476 * Append as many sources as will fit in the first packet.
2477 * If we are appending to a new packet, the chain allocation
2478 * may potentially use clusters; use m_getptr() in this case.
2479 * If we are appending to an existing packet, we need to obtain
2480 * a pointer to the group record after m_append(), in case a new
2481 * mbuf was allocated.
2483 * Only append sources which are in-mode at t1. If we are
2484 * transitioning to MCAST_UNDEFINED state on the group, and
2485 * use_block_allow is zero, do not include source entries.
2486 * Otherwise, we need to include this source in the report.
2488 * Only report recorded sources in our filter set when responding
2489 * to a group-source query.
2491 if (record_has_sources) {
2494 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2495 md->m_len - nbytes);
2497 md = m_getptr(m, 0, &off);
2498 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2502 RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs,
2504 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2505 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2506 now = im6s_get_mode(inm, ims, 1);
2507 CTR2(KTR_MLD, "%s: node is %d", __func__, now);
2508 if ((now != mode) ||
2510 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2511 CTR1(KTR_MLD, "%s: skip node", __func__);
2514 if (is_source_query && ims->im6s_stp == 0) {
2515 CTR1(KTR_MLD, "%s: skip unrecorded node",
2519 CTR1(KTR_MLD, "%s: append node", __func__);
2520 if (!m_append(m, sizeof(struct in6_addr),
2521 (void *)&ims->im6s_addr)) {
2524 CTR1(KTR_MLD, "%s: m_append() failed.",
2528 nbytes += sizeof(struct in6_addr);
2530 if (msrcs == m0srcs)
2533 CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__,
2535 pmr->mr_numsrc = htons(msrcs);
2536 nbytes += (msrcs * sizeof(struct in6_addr));
2539 if (is_source_query && msrcs == 0) {
2540 CTR1(KTR_MLD, "%s: no recorded sources to report", __func__);
2547 * We are good to go with first packet.
2550 CTR1(KTR_MLD, "%s: enqueueing first packet", __func__);
2551 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2552 _IF_ENQUEUE(ifq, m);
2554 m->m_pkthdr.PH_vt.vt_nrecs++;
2557 * No further work needed if no source list in packet(s).
2559 if (!record_has_sources)
2563 * Whilst sources remain to be announced, we need to allocate
2564 * a new packet and fill out as many sources as will fit.
2565 * Always try for a cluster first.
2567 while (nims != NULL) {
2568 if (_IF_QFULL(ifq)) {
2569 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2572 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2574 m = m_gethdr(M_DONTWAIT, MT_DATA);
2577 mld_save_context(m, ifp);
2578 md = m_getptr(m, 0, &off);
2579 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off);
2580 CTR1(KTR_MLD, "%s: allocated next packet", __func__);
2582 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2585 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2588 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2589 nbytes += sizeof(struct mldv2_record);
2591 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2592 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2595 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2596 CTR2(KTR_MLD, "%s: visit node %s",
2597 __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2598 now = im6s_get_mode(inm, ims, 1);
2599 if ((now != mode) ||
2601 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2602 CTR1(KTR_MLD, "%s: skip node", __func__);
2605 if (is_source_query && ims->im6s_stp == 0) {
2606 CTR1(KTR_MLD, "%s: skip unrecorded node",
2610 CTR1(KTR_MLD, "%s: append node", __func__);
2611 if (!m_append(m, sizeof(struct in6_addr),
2612 (void *)&ims->im6s_addr)) {
2615 CTR1(KTR_MLD, "%s: m_append() failed.",
2620 if (msrcs == m0srcs)
2623 pmr->mr_numsrc = htons(msrcs);
2624 nbytes += (msrcs * sizeof(struct in6_addr));
2626 CTR1(KTR_MLD, "%s: enqueueing next packet", __func__);
2627 _IF_ENQUEUE(ifq, m);
2634 * Type used to mark record pass completion.
2635 * We exploit the fact we can cast to this easily from the
2636 * current filter modes on each ip_msource node.
2639 REC_NONE = 0x00, /* MCAST_UNDEFINED */
2640 REC_ALLOW = 0x01, /* MCAST_INCLUDE */
2641 REC_BLOCK = 0x02, /* MCAST_EXCLUDE */
2642 REC_FULL = REC_ALLOW | REC_BLOCK
2646 * Enqueue an MLDv2 filter list change to the given output queue.
2648 * Source list filter state is held in an RB-tree. When the filter list
2649 * for a group is changed without changing its mode, we need to compute
2650 * the deltas between T0 and T1 for each source in the filter set,
2651 * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records.
2653 * As we may potentially queue two record types, and the entire R-B tree
2654 * needs to be walked at once, we break this out into its own function
2655 * so we can generate a tightly packed queue of packets.
2657 * XXX This could be written to only use one tree walk, although that makes
2658 * serializing into the mbuf chains a bit harder. For now we do two walks
2659 * which makes things easier on us, and it may or may not be harder on
2662 * If successful the size of all data appended to the queue is returned,
2663 * otherwise an error code less than zero is returned, or zero if
2664 * no record(s) were appended.
2667 mld_v2_enqueue_filter_change(struct ifqueue *ifq, struct in6_multi *inm)
2669 static const int MINRECLEN =
2670 sizeof(struct mldv2_record) + sizeof(struct in6_addr);
2672 struct mldv2_record mr;
2673 struct mldv2_record *pmr;
2674 struct ip6_msource *ims, *nims;
2675 struct mbuf *m, *m0, *md;
2676 int m0srcs, nbytes, npbytes, off, rsrcs, schanged;
2678 uint8_t mode, now, then;
2679 rectype_t crt, drt, nrt;
2681 char ip6tbuf[INET6_ADDRSTRLEN];
2684 IN6_MULTI_LOCK_ASSERT();
2686 if (inm->in6m_nsrc == 0 ||
2687 (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0))
2690 ifp = inm->in6m_ifp; /* interface */
2691 mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */
2692 crt = REC_NONE; /* current group record type */
2693 drt = REC_NONE; /* mask of completed group record types */
2694 nrt = REC_NONE; /* record type for current node */
2695 m0srcs = 0; /* # source which will fit in current mbuf chain */
2696 npbytes = 0; /* # of bytes appended this packet */
2697 nbytes = 0; /* # of bytes appended to group's state-change queue */
2698 rsrcs = 0; /* # sources encoded in current record */
2699 schanged = 0; /* # nodes encoded in overall filter change */
2700 nallow = 0; /* # of source entries in ALLOW_NEW */
2701 nblock = 0; /* # of source entries in BLOCK_OLD */
2702 nims = NULL; /* next tree node pointer */
2705 * For each possible filter record mode.
2706 * The first kind of source we encounter tells us which
2707 * is the first kind of record we start appending.
2708 * If a node transitioned to UNDEFINED at t1, its mode is treated
2709 * as the inverse of the group's filter mode.
2711 while (drt != REC_FULL) {
2715 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <=
2716 MLD_V2_REPORT_MAXRECS) &&
2717 (m0->m_pkthdr.len + MINRECLEN) <
2718 (ifp->if_mtu - MLD_MTUSPACE)) {
2720 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2721 sizeof(struct mldv2_record)) /
2722 sizeof(struct in6_addr);
2724 "%s: use previous packet", __func__);
2726 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2728 m = m_gethdr(M_DONTWAIT, MT_DATA);
2731 "%s: m_get*() failed", __func__);
2734 m->m_pkthdr.PH_vt.vt_nrecs = 0;
2735 mld_save_context(m, ifp);
2736 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2737 sizeof(struct mldv2_record)) /
2738 sizeof(struct in6_addr);
2741 "%s: allocated new packet", __func__);
2744 * Append the MLD group record header to the
2745 * current packet's data area.
2746 * Recalculate pointer to free space for next
2747 * group record, in case m_append() allocated
2748 * a new mbuf or cluster.
2750 memset(&mr, 0, sizeof(mr));
2751 mr.mr_addr = inm->in6m_addr;
2752 in6_clearscope(&mr.mr_addr);
2753 if (!m_append(m, sizeof(mr), (void *)&mr)) {
2757 "%s: m_append() failed", __func__);
2760 npbytes += sizeof(struct mldv2_record);
2762 /* new packet; offset in chain */
2763 md = m_getptr(m, npbytes -
2764 sizeof(struct mldv2_record), &off);
2765 pmr = (struct mldv2_record *)(mtod(md,
2768 /* current packet; offset from last append */
2770 pmr = (struct mldv2_record *)(mtod(md,
2771 uint8_t *) + md->m_len -
2772 sizeof(struct mldv2_record));
2775 * Begin walking the tree for this record type
2776 * pass, or continue from where we left off
2777 * previously if we had to allocate a new packet.
2778 * Only report deltas in-mode at t1.
2779 * We need not report included sources as allowed
2780 * if we are in inclusive mode on the group,
2781 * however the converse is not true.
2785 nims = RB_MIN(ip6_msource_tree,
2788 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2789 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2790 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2791 now = im6s_get_mode(inm, ims, 1);
2792 then = im6s_get_mode(inm, ims, 0);
2793 CTR3(KTR_MLD, "%s: mode: t0 %d, t1 %d",
2794 __func__, then, now);
2797 "%s: skip unchanged", __func__);
2800 if (mode == MCAST_EXCLUDE &&
2801 now == MCAST_INCLUDE) {
2803 "%s: skip IN src on EX group",
2807 nrt = (rectype_t)now;
2808 if (nrt == REC_NONE)
2809 nrt = (rectype_t)(~mode & REC_FULL);
2810 if (schanged++ == 0) {
2812 } else if (crt != nrt)
2814 if (!m_append(m, sizeof(struct in6_addr),
2815 (void *)&ims->im6s_addr)) {
2819 "%s: m_append() failed", __func__);
2822 nallow += !!(crt == REC_ALLOW);
2823 nblock += !!(crt == REC_BLOCK);
2824 if (++rsrcs == m0srcs)
2828 * If we did not append any tree nodes on this
2829 * pass, back out of allocations.
2832 npbytes -= sizeof(struct mldv2_record);
2835 "%s: m_free(m)", __func__);
2839 "%s: m_adj(m, -mr)", __func__);
2840 m_adj(m, -((int)sizeof(
2841 struct mldv2_record)));
2845 npbytes += (rsrcs * sizeof(struct in6_addr));
2846 if (crt == REC_ALLOW)
2847 pmr->mr_type = MLD_ALLOW_NEW_SOURCES;
2848 else if (crt == REC_BLOCK)
2849 pmr->mr_type = MLD_BLOCK_OLD_SOURCES;
2850 pmr->mr_numsrc = htons(rsrcs);
2852 * Count the new group record, and enqueue this
2853 * packet if it wasn't already queued.
2855 m->m_pkthdr.PH_vt.vt_nrecs++;
2857 _IF_ENQUEUE(ifq, m);
2859 } while (nims != NULL);
2861 crt = (~crt & REC_FULL);
2864 CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__,
2871 mld_v2_merge_state_changes(struct in6_multi *inm, struct ifqueue *ifscq)
2874 struct mbuf *m; /* pending state-change */
2875 struct mbuf *m0; /* copy of pending state-change */
2876 struct mbuf *mt; /* last state-change in packet */
2877 int docopy, domerge;
2884 IN6_MULTI_LOCK_ASSERT();
2888 * If there are further pending retransmissions, make a writable
2889 * copy of each queued state-change message before merging.
2891 if (inm->in6m_scrv > 0)
2894 gq = &inm->in6m_scq;
2896 if (gq->ifq_head == NULL) {
2897 CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty",
2905 * Only merge the report into the current packet if
2906 * there is sufficient space to do so; an MLDv2 report
2907 * packet may only contain 65,535 group records.
2908 * Always use a simple mbuf chain concatentation to do this,
2909 * as large state changes for single groups may have
2910 * allocated clusters.
2913 mt = ifscq->ifq_tail;
2915 recslen = m_length(m, NULL);
2917 if ((mt->m_pkthdr.PH_vt.vt_nrecs +
2918 m->m_pkthdr.PH_vt.vt_nrecs <=
2919 MLD_V2_REPORT_MAXRECS) &&
2920 (mt->m_pkthdr.len + recslen <=
2921 (inm->in6m_ifp->if_mtu - MLD_MTUSPACE)))
2925 if (!domerge && _IF_QFULL(gq)) {
2927 "%s: outbound queue full, skipping whole packet %p",
2937 CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m);
2938 _IF_DEQUEUE(gq, m0);
2941 CTR2(KTR_MLD, "%s: copying %p", __func__, m);
2942 m0 = m_dup(m, M_NOWAIT);
2945 m0->m_nextpkt = NULL;
2950 CTR3(KTR_MLD, "%s: queueing %p to ifscq %p)",
2951 __func__, m0, ifscq);
2952 _IF_ENQUEUE(ifscq, m0);
2954 struct mbuf *mtl; /* last mbuf of packet mt */
2956 CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)",
2960 m0->m_flags &= ~M_PKTHDR;
2961 mt->m_pkthdr.len += recslen;
2962 mt->m_pkthdr.PH_vt.vt_nrecs +=
2963 m0->m_pkthdr.PH_vt.vt_nrecs;
2973 * Respond to a pending MLDv2 General Query.
2976 mld_v2_dispatch_general_query(struct mld_ifinfo *mli)
2978 struct ifmultiaddr *ifma, *tifma;
2980 struct in6_multi *inm;
2983 IN6_MULTI_LOCK_ASSERT();
2986 KASSERT(mli->mli_version == MLD_VERSION_2,
2987 ("%s: called when version %d", __func__, mli->mli_version));
2992 TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, tifma) {
2993 if (ifma->ifma_addr->sa_family != AF_INET6 ||
2994 ifma->ifma_protospec == NULL)
2997 inm = (struct in6_multi *)ifma->ifma_protospec;
2998 KASSERT(ifp == inm->in6m_ifp,
2999 ("%s: inconsistent ifp", __func__));
3001 switch (inm->in6m_state) {
3002 case MLD_NOT_MEMBER:
3003 case MLD_SILENT_MEMBER:
3005 case MLD_REPORTING_MEMBER:
3006 case MLD_IDLE_MEMBER:
3007 case MLD_LAZY_MEMBER:
3008 case MLD_SLEEPING_MEMBER:
3009 case MLD_AWAKENING_MEMBER:
3010 inm->in6m_state = MLD_REPORTING_MEMBER;
3011 retval = mld_v2_enqueue_group_record(&mli->mli_gq,
3013 CTR2(KTR_MLD, "%s: enqueue record = %d",
3016 case MLD_G_QUERY_PENDING_MEMBER:
3017 case MLD_SG_QUERY_PENDING_MEMBER:
3018 case MLD_LEAVING_MEMBER:
3022 IF_ADDR_UNLOCK(ifp);
3024 mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST);
3027 * Slew transmission of bursts over 500ms intervals.
3029 if (mli->mli_gq.ifq_head != NULL) {
3030 mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY(
3031 MLD_RESPONSE_BURST_INTERVAL);
3032 V_interface_timers_running6 = 1;
3037 * Transmit the next pending message in the output queue.
3039 * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis.
3040 * MRT: Nothing needs to be done, as MLD traffic is always local to
3041 * a link and uses a link-scope multicast address.
3044 mld_dispatch_packet(struct mbuf *m)
3046 struct ip6_moptions im6o;
3051 struct ip6_hdr *ip6;
3052 struct mld_hdr *mld;
3058 CTR2(KTR_MLD, "%s: transmit %p", __func__, m);
3061 * Set VNET image pointer from enqueued mbuf chain
3062 * before doing anything else. Whilst we use interface
3063 * indexes to guard against interface detach, they are
3064 * unique to each VIMAGE and must be retrieved.
3066 ifindex = mld_restore_context(m);
3069 * Check if the ifnet still exists. This limits the scope of
3070 * any race in the absence of a global ifp lock for low cost
3071 * (an array lookup).
3073 ifp = ifnet_byindex(ifindex);
3075 CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.",
3076 __func__, m, ifindex);
3078 IP6STAT_INC(ip6s_noroute);
3082 im6o.im6o_multicast_hlim = 1;
3083 im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL);
3084 im6o.im6o_multicast_ifp = ifp;
3086 if (m->m_flags & M_MLDV1) {
3089 m0 = mld_v2_encap_report(ifp, m);
3091 CTR2(KTR_MLD, "%s: dropped %p", __func__, m);
3093 IP6STAT_INC(ip6s_odropped);
3098 mld_scrub_context(m0);
3099 m->m_flags &= ~(M_PROTOFLAGS);
3100 m0->m_pkthdr.rcvif = V_loif;
3102 ip6 = mtod(m0, struct ip6_hdr *);
3104 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */
3107 * XXX XXX Break some KPI rules to prevent an LOR which would
3108 * occur if we called in6_setscope() at transmission.
3109 * See comments at top of file.
3111 MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index);
3115 * Retrieve the ICMPv6 type before handoff to ip6_output(),
3116 * so we can bump the stats.
3118 md = m_getptr(m0, sizeof(struct ip6_hdr), &off);
3119 mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off);
3120 type = mld->mld_type;
3122 error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o,
3125 CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error);
3128 ICMP6STAT_INC(icp6s_outhist[type]);
3130 icmp6_ifstat_inc(oifp, ifs6_out_msg);
3132 case MLD_LISTENER_REPORT:
3133 case MLDV2_LISTENER_REPORT:
3134 icmp6_ifstat_inc(oifp, ifs6_out_mldreport);
3136 case MLD_LISTENER_DONE:
3137 icmp6_ifstat_inc(oifp, ifs6_out_mlddone);
3146 * Encapsulate an MLDv2 report.
3148 * KAME IPv6 requires that hop-by-hop options be passed separately,
3149 * and that the IPv6 header be prepended in a separate mbuf.
3151 * Returns a pointer to the new mbuf chain head, or NULL if the
3152 * allocation failed.
3154 static struct mbuf *
3155 mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m)
3158 struct mldv2_report *mld;
3159 struct ip6_hdr *ip6;
3160 struct in6_ifaddr *ia;
3163 KASSERT(ifp != NULL, ("%s: null ifp", __func__));
3164 KASSERT((m->m_flags & M_PKTHDR),
3165 ("%s: mbuf chain %p is !M_PKTHDR", __func__, m));
3168 * RFC3590: OK to send as :: or tentative during DAD.
3170 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
3172 CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__);
3174 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3177 ifa_free(&ia->ia_ifa);
3181 MH_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report));
3183 mldreclen = m_length(m, NULL);
3184 CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen);
3186 mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report);
3187 mh->m_pkthdr.len = sizeof(struct ip6_hdr) +
3188 sizeof(struct mldv2_report) + mldreclen;
3190 ip6 = mtod(mh, struct ip6_hdr *);
3192 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3193 ip6->ip6_vfc |= IPV6_VERSION;
3194 ip6->ip6_nxt = IPPROTO_ICMPV6;
3195 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
3197 ifa_free(&ia->ia_ifa);
3198 ip6->ip6_dst = in6addr_linklocal_allv2routers;
3199 /* scope ID will be set in netisr */
3201 mld = (struct mldv2_report *)(ip6 + 1);
3202 mld->mld_type = MLDV2_LISTENER_REPORT;
3205 mld->mld_v2_reserved = 0;
3206 mld->mld_v2_numrecs = htons(m->m_pkthdr.PH_vt.vt_nrecs);
3207 m->m_pkthdr.PH_vt.vt_nrecs = 0;
3210 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
3211 sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen);
3217 mld_rec_type_to_str(const int type)
3221 case MLD_CHANGE_TO_EXCLUDE_MODE:
3224 case MLD_CHANGE_TO_INCLUDE_MODE:
3227 case MLD_MODE_IS_EXCLUDE:
3230 case MLD_MODE_IS_INCLUDE:
3233 case MLD_ALLOW_NEW_SOURCES:
3236 case MLD_BLOCK_OLD_SOURCES:
3247 mld_init(void *unused __unused)
3250 CTR1(KTR_MLD, "%s: initializing", __func__);
3253 ip6_initpktopts(&mld_po);
3254 mld_po.ip6po_hlim = 1;
3255 mld_po.ip6po_hbh = &mld_ra.hbh;
3256 mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
3257 mld_po.ip6po_flags = IP6PO_DONTFRAG;
3259 SYSINIT(mld_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_init, NULL);
3262 mld_uninit(void *unused __unused)
3265 CTR1(KTR_MLD, "%s: tearing down", __func__);
3268 SYSUNINIT(mld_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_uninit, NULL);
3271 vnet_mld_init(const void *unused __unused)
3274 CTR1(KTR_MLD, "%s: initializing", __func__);
3276 LIST_INIT(&V_mli_head);
3278 VNET_SYSINIT(vnet_mld_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_init,
3282 vnet_mld_uninit(const void *unused __unused)
3285 CTR1(KTR_MLD, "%s: tearing down", __func__);
3287 KASSERT(LIST_EMPTY(&V_mli_head),
3288 ("%s: mli list not empty; ifnets not detached?", __func__));
3290 VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_uninit,
3294 mld_modevent(module_t mod, int type, void *unused __unused)
3302 return (EOPNOTSUPP);
3307 static moduledata_t mld_mod = {
3312 DECLARE_MODULE(mld, mld_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);