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 mld_ifinfo *,
126 static void mld_v1_process_querier_timers(struct mld_ifinfo *);
127 static int mld_v1_transmit_report(struct in6_multi *, const int);
128 static void mld_v1_update_group(struct in6_multi *, const int);
129 static void mld_v2_cancel_link_timers(struct mld_ifinfo *);
130 static void mld_v2_dispatch_general_query(struct mld_ifinfo *);
132 mld_v2_encap_report(struct ifnet *, struct mbuf *);
133 static int mld_v2_enqueue_filter_change(struct ifqueue *,
135 static int mld_v2_enqueue_group_record(struct ifqueue *,
136 struct in6_multi *, const int, const int, const int,
138 static int mld_v2_input_query(struct ifnet *, const struct ip6_hdr *,
139 struct mbuf *, const int, const int);
140 static int mld_v2_merge_state_changes(struct in6_multi *,
142 static void mld_v2_process_group_timers(struct mld_ifinfo *,
143 struct ifqueue *, struct ifqueue *,
144 struct in6_multi *, const int);
145 static int mld_v2_process_group_query(struct in6_multi *,
146 struct mld_ifinfo *mli, int, struct mbuf *, const int);
147 static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS);
148 static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS);
151 * Normative references: RFC 2710, RFC 3590, RFC 3810.
154 * * The MLD subsystem lock ends up being system-wide for the moment,
155 * but could be per-VIMAGE later on.
156 * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK.
157 * Any may be taken independently; if any are held at the same
158 * time, the above lock order must be followed.
159 * * IN6_MULTI_LOCK covers in_multi.
160 * * MLD_LOCK covers per-link state and any global variables in this file.
161 * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of
162 * per-link state iterators.
165 * A special case for IPv6 is the in6_setscope() routine. ip6_output()
166 * will not accept an ifp; it wants an embedded scope ID, unlike
167 * ip_output(), which happily takes the ifp given to it. The embedded
168 * scope ID is only used by MLD to select the outgoing interface.
170 * During interface attach and detach, MLD will take MLD_LOCK *after*
171 * the IF_AFDATA_LOCK.
172 * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call
173 * it with MLD_LOCK held without triggering an LOR. A netisr with indirect
174 * dispatch could work around this, but we'd rather not do that, as it
175 * can introduce other races.
177 * As such, we exploit the fact that the scope ID is just the interface
178 * index, and embed it in the IPv6 destination address accordingly.
179 * This is potentially NOT VALID for MLDv1 reports, as they
180 * are always sent to the multicast group itself; as MLDv2
181 * reports are always sent to ff02::16, this is not an issue
182 * when MLDv2 is in use.
184 * This does not however eliminate the LOR when ip6_output() itself
185 * calls in6_setscope() internally whilst MLD_LOCK is held. This will
186 * trigger a LOR warning in WITNESS when the ifnet is detached.
188 * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given
189 * how it's used across the network stack. Here we're simply exploiting
190 * the fact that MLD runs at a similar layer in the stack to scope6.c.
193 * * Each in6_multi corresponds to an ifp, and each ifp corresponds
194 * to a vnet in ifp->if_vnet.
196 static struct mtx mld_mtx;
197 static MALLOC_DEFINE(M_MLD, "mld", "mld state");
199 #define MLD_EMBEDSCOPE(pin6, zoneid) \
200 if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \
201 IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \
202 (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \
205 * VIMAGE-wide globals.
207 static VNET_DEFINE(struct timeval, mld_gsrdelay) = {10, 0};
208 static VNET_DEFINE(LIST_HEAD(, mld_ifinfo), mli_head);
209 static VNET_DEFINE(int, interface_timers_running6);
210 static VNET_DEFINE(int, state_change_timers_running6);
211 static VNET_DEFINE(int, current_state_timers_running6);
213 #define V_mld_gsrdelay VNET(mld_gsrdelay)
214 #define V_mli_head VNET(mli_head)
215 #define V_interface_timers_running6 VNET(interface_timers_running6)
216 #define V_state_change_timers_running6 VNET(state_change_timers_running6)
217 #define V_current_state_timers_running6 VNET(current_state_timers_running6)
219 SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */
221 SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW, 0,
222 "IPv6 Multicast Listener Discovery");
225 * Virtualized sysctls.
227 SYSCTL_VNET_PROC(_net_inet6_mld, OID_AUTO, gsrdelay,
228 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
229 &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I",
230 "Rate limit for MLDv2 Group-and-Source queries in seconds");
233 * Non-virtualized sysctls.
235 static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo,
236 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo,
237 "Per-interface MLDv2 state");
239 static int mld_v1enable = 1;
240 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RW,
241 &mld_v1enable, 0, "Enable fallback to MLDv1");
242 TUNABLE_INT("net.inet6.mld.v1enable", &mld_v1enable);
244 static int mld_use_allow = 1;
245 SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RW,
246 &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves");
247 TUNABLE_INT("net.inet6.mld.use_allow", &mld_use_allow);
250 * Packed Router Alert option structure declaration.
255 struct ip6_opt_router ra;
259 * Router Alert hop-by-hop option header.
261 static struct mld_raopt mld_ra = {
263 .pad = { .ip6o_type = IP6OPT_PADN, 0 },
265 .ip6or_type = IP6OPT_ROUTER_ALERT,
266 .ip6or_len = IP6OPT_RTALERT_LEN - 2,
267 .ip6or_value[0] = ((IP6OPT_RTALERT_MLD >> 8) & 0xFF),
268 .ip6or_value[1] = (IP6OPT_RTALERT_MLD & 0xFF)
271 static struct ip6_pktopts mld_po;
274 mld_save_context(struct mbuf *m, struct ifnet *ifp)
278 m->m_pkthdr.header = ifp->if_vnet;
280 m->m_pkthdr.flowid = ifp->if_index;
284 mld_scrub_context(struct mbuf *m)
287 m->m_pkthdr.header = NULL;
288 m->m_pkthdr.flowid = 0;
292 * Restore context from a queued output chain.
293 * Return saved ifindex.
295 * VIMAGE: The assertion is there to make sure that we
296 * actually called CURVNET_SET() with what's in the mbuf chain.
298 static __inline uint32_t
299 mld_restore_context(struct mbuf *m)
302 #if defined(VIMAGE) && defined(INVARIANTS)
303 KASSERT(curvnet == m->m_pkthdr.header,
304 ("%s: called when curvnet was not restored", __func__));
306 return (m->m_pkthdr.flowid);
310 * Retrieve or set threshold between group-source queries in seconds.
312 * VIMAGE: Assume curvnet set by caller.
313 * SMPng: NOTE: Serialized by MLD lock.
316 sysctl_mld_gsr(SYSCTL_HANDLER_ARGS)
321 error = sysctl_wire_old_buffer(req, sizeof(int));
327 i = V_mld_gsrdelay.tv_sec;
329 error = sysctl_handle_int(oidp, &i, 0, req);
330 if (error || !req->newptr)
333 if (i < -1 || i >= 60) {
338 CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d",
339 V_mld_gsrdelay.tv_sec, i);
340 V_mld_gsrdelay.tv_sec = i;
348 * Expose struct mld_ifinfo to userland, keyed by ifindex.
349 * For use by ifmcstat(8).
351 * SMPng: NOTE: Does an unlocked ifindex space read.
352 * VIMAGE: Assume curvnet set by caller. The node handler itself
353 * is not directly virtualized.
356 sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS)
362 struct mld_ifinfo *mli;
367 if (req->newptr != NULL)
373 error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo));
380 if (name[0] <= 0 || name[0] > V_if_index) {
387 ifp = ifnet_byindex(name[0]);
391 LIST_FOREACH(mli, &V_mli_head, mli_link) {
392 if (ifp == mli->mli_ifp) {
393 error = SYSCTL_OUT(req, mli,
394 sizeof(struct mld_ifinfo));
406 * Dispatch an entire queue of pending packet chains.
407 * VIMAGE: Assumes the vnet pointer has been set.
410 mld_dispatch_queue(struct ifqueue *ifq, int limit)
418 CTR3(KTR_MLD, "%s: dispatch %p from %p", __func__, ifq, m);
419 mld_dispatch_packet(m);
426 * Filter outgoing MLD report state by group.
428 * Reports are ALWAYS suppressed for ALL-HOSTS (ff02::1)
429 * and node-local addresses. However, kernel and socket consumers
430 * always embed the KAME scope ID in the address provided, so strip it
431 * when performing comparison.
432 * Note: This is not the same as the *multicast* scope.
434 * Return zero if the given group is one for which MLD reports
435 * should be suppressed, or non-zero if reports should be issued.
438 mld_is_addr_reported(const struct in6_addr *addr)
441 KASSERT(IN6_IS_ADDR_MULTICAST(addr), ("%s: not multicast", __func__));
443 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_NODELOCAL)
446 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_LINKLOCAL) {
447 struct in6_addr tmp = *addr;
448 in6_clearscope(&tmp);
449 if (IN6_ARE_ADDR_EQUAL(&tmp, &in6addr_linklocal_allnodes))
457 * Attach MLD when PF_INET6 is attached to an interface.
459 * SMPng: Normally called with IF_AFDATA_LOCK held.
462 mld_domifattach(struct ifnet *ifp)
464 struct mld_ifinfo *mli;
466 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
467 __func__, ifp, ifp->if_xname);
471 mli = mli_alloc_locked(ifp);
472 if (!(ifp->if_flags & IFF_MULTICAST))
473 mli->mli_flags |= MLIF_SILENT;
475 mli->mli_flags |= MLIF_USEALLOW;
483 * VIMAGE: assume curvnet set by caller.
485 static struct mld_ifinfo *
486 mli_alloc_locked(/*const*/ struct ifnet *ifp)
488 struct mld_ifinfo *mli;
492 mli = malloc(sizeof(struct mld_ifinfo), M_MLD, M_NOWAIT|M_ZERO);
497 mli->mli_version = MLD_VERSION_2;
499 mli->mli_rv = MLD_RV_INIT;
500 mli->mli_qi = MLD_QI_INIT;
501 mli->mli_qri = MLD_QRI_INIT;
502 mli->mli_uri = MLD_URI_INIT;
504 SLIST_INIT(&mli->mli_relinmhead);
507 * Responses to general queries are subject to bounds.
509 IFQ_SET_MAXLEN(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS);
511 LIST_INSERT_HEAD(&V_mli_head, mli, mli_link);
513 CTR2(KTR_MLD, "allocate mld_ifinfo for ifp %p(%s)",
523 * NOTE: Some finalization tasks need to run before the protocol domain
524 * is detached, but also before the link layer does its cleanup.
525 * Run before link-layer cleanup; cleanup groups, but do not free MLD state.
527 * SMPng: Caller must hold IN6_MULTI_LOCK().
528 * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator.
529 * XXX This routine is also bitten by unlocked ifma_protospec access.
532 mld_ifdetach(struct ifnet *ifp)
534 struct mld_ifinfo *mli;
535 struct ifmultiaddr *ifma;
536 struct in6_multi *inm, *tinm;
538 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp,
541 IN6_MULTI_LOCK_ASSERT();
544 mli = MLD_IFINFO(ifp);
545 if (mli->mli_version == MLD_VERSION_2) {
547 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
548 if (ifma->ifma_addr->sa_family != AF_INET6 ||
549 ifma->ifma_protospec == NULL)
551 inm = (struct in6_multi *)ifma->ifma_protospec;
552 if (inm->in6m_state == MLD_LEAVING_MEMBER) {
553 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
556 in6m_clear_recorded(inm);
558 IF_ADDR_RUNLOCK(ifp);
559 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele,
561 SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele);
562 in6m_release_locked(inm);
570 * Hook for domifdetach.
571 * Runs after link-layer cleanup; free MLD state.
573 * SMPng: Normally called with IF_AFDATA_LOCK held.
576 mld_domifdetach(struct ifnet *ifp)
579 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
580 __func__, ifp, ifp->if_xname);
583 mli_delete_locked(ifp);
588 mli_delete_locked(const struct ifnet *ifp)
590 struct mld_ifinfo *mli, *tmli;
592 CTR3(KTR_MLD, "%s: freeing mld_ifinfo for ifp %p(%s)",
593 __func__, ifp, ifp->if_xname);
597 LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) {
598 if (mli->mli_ifp == ifp) {
600 * Free deferred General Query responses.
602 _IF_DRAIN(&mli->mli_gq);
604 LIST_REMOVE(mli, mli_link);
606 KASSERT(SLIST_EMPTY(&mli->mli_relinmhead),
607 ("%s: there are dangling in_multi references",
615 panic("%s: mld_ifinfo not found for ifp %p\n", __func__, ifp);
620 * Process a received MLDv1 general or address-specific query.
621 * Assumes that the query header has been pulled up to sizeof(mld_hdr).
623 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
624 * mld_addr. This is OK as we own the mbuf chain.
627 mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
628 /*const*/ struct mld_hdr *mld)
630 struct ifmultiaddr *ifma;
631 struct mld_ifinfo *mli;
632 struct in6_multi *inm;
633 int is_general_query;
636 char ip6tbuf[INET6_ADDRSTRLEN];
639 is_general_query = 0;
642 CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)",
643 ip6_sprintf(ip6tbuf, &mld->mld_addr),
649 * RFC3810 Section 6.2: MLD queries must originate from
650 * a router's link-local address.
652 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
653 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
654 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
660 * Do address field validation upfront before we accept
663 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
665 * MLDv1 General Query.
666 * If this was not sent to the all-nodes group, ignore it.
671 in6_clearscope(&dst);
672 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes))
674 is_general_query = 1;
677 * Embed scope ID of receiving interface in MLD query for
678 * lookup whilst we don't hold other locks.
680 in6_setscope(&mld->mld_addr, ifp, NULL);
687 * Switch to MLDv1 host compatibility mode.
689 mli = MLD_IFINFO(ifp);
690 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
691 mld_set_version(mli, MLD_VERSION_1);
693 timer = (ntohs(mld->mld_maxdelay) * PR_FASTHZ) / MLD_TIMER_SCALE;
698 if (is_general_query) {
700 * For each reporting group joined on this
701 * interface, kick the report timer.
703 CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)",
705 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
706 if (ifma->ifma_addr->sa_family != AF_INET6 ||
707 ifma->ifma_protospec == NULL)
709 inm = (struct in6_multi *)ifma->ifma_protospec;
710 mld_v1_update_group(inm, timer);
714 * MLDv1 Group-Specific Query.
715 * If this is a group-specific MLDv1 query, we need only
716 * look up the single group to process it.
718 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
720 CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)",
721 ip6_sprintf(ip6tbuf, &mld->mld_addr),
723 mld_v1_update_group(inm, timer);
725 /* XXX Clear embedded scope ID as userland won't expect it. */
726 in6_clearscope(&mld->mld_addr);
729 IF_ADDR_RUNLOCK(ifp);
737 * Update the report timer on a group in response to an MLDv1 query.
739 * If we are becoming the reporting member for this group, start the timer.
740 * If we already are the reporting member for this group, and timer is
741 * below the threshold, reset it.
743 * We may be updating the group for the first time since we switched
744 * to MLDv2. If we are, then we must clear any recorded source lists,
745 * and transition to REPORTING state; the group timer is overloaded
746 * for group and group-source query responses.
748 * Unlike MLDv2, the delay per group should be jittered
749 * to avoid bursts of MLDv1 reports.
752 mld_v1_update_group(struct in6_multi *inm, const int timer)
755 char ip6tbuf[INET6_ADDRSTRLEN];
758 CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__,
759 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
760 inm->in6m_ifp->if_xname, timer);
762 IN6_MULTI_LOCK_ASSERT();
764 switch (inm->in6m_state) {
766 case MLD_SILENT_MEMBER:
768 case MLD_REPORTING_MEMBER:
769 if (inm->in6m_timer != 0 &&
770 inm->in6m_timer <= timer) {
771 CTR1(KTR_MLD, "%s: REPORTING and timer running, "
772 "skipping.", __func__);
776 case MLD_SG_QUERY_PENDING_MEMBER:
777 case MLD_G_QUERY_PENDING_MEMBER:
778 case MLD_IDLE_MEMBER:
779 case MLD_LAZY_MEMBER:
780 case MLD_AWAKENING_MEMBER:
781 CTR1(KTR_MLD, "%s: ->REPORTING", __func__);
782 inm->in6m_state = MLD_REPORTING_MEMBER;
783 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
784 V_current_state_timers_running6 = 1;
786 case MLD_SLEEPING_MEMBER:
787 CTR1(KTR_MLD, "%s: ->AWAKENING", __func__);
788 inm->in6m_state = MLD_AWAKENING_MEMBER;
790 case MLD_LEAVING_MEMBER:
796 * Process a received MLDv2 general, group-specific or
797 * group-and-source-specific query.
799 * Assumes that the query header has been pulled up to sizeof(mldv2_query).
801 * Return 0 if successful, otherwise an appropriate error code is returned.
804 mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
805 struct mbuf *m, const int off, const int icmp6len)
807 struct mld_ifinfo *mli;
808 struct mldv2_query *mld;
809 struct in6_multi *inm;
810 uint32_t maxdelay, nsrc, qqi;
811 int is_general_query;
815 char ip6tbuf[INET6_ADDRSTRLEN];
818 is_general_query = 0;
821 * RFC3810 Section 6.2: MLD queries must originate from
822 * a router's link-local address.
824 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
825 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
826 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
831 CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, ifp->if_xname);
833 mld = (struct mldv2_query *)(mtod(m, uint8_t *) + off);
835 maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */
836 if (maxdelay >= 32678) {
837 maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) <<
838 (MLD_MRC_EXP(maxdelay) + 3);
840 timer = (maxdelay * PR_FASTHZ) / MLD_TIMER_SCALE;
844 qrv = MLD_QRV(mld->mld_misc);
846 CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__,
853 qqi = MLD_QQIC_MANT(mld->mld_qqi) <<
854 (MLD_QQIC_EXP(mld->mld_qqi) + 3);
857 nsrc = ntohs(mld->mld_numsrc);
858 if (nsrc > MLD_MAX_GS_SOURCES)
860 if (icmp6len < sizeof(struct mldv2_query) +
861 (nsrc * sizeof(struct in6_addr)))
865 * Do further input validation upfront to avoid resetting timers
866 * should we need to discard this query.
868 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
870 * General Queries SHOULD be directed to ff02::1.
871 * A general query with a source list has undefined
872 * behaviour; discard it.
877 in6_clearscope(&dst);
878 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes) ||
881 is_general_query = 1;
884 * Embed scope ID of receiving interface in MLD query for
885 * lookup whilst we don't hold other locks (due to KAME
886 * locking lameness). We own this mbuf chain just now.
888 in6_setscope(&mld->mld_addr, ifp, NULL);
894 mli = MLD_IFINFO(ifp);
895 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
898 * Discard the v2 query if we're in Compatibility Mode.
899 * The RFC is pretty clear that hosts need to stay in MLDv1 mode
900 * until the Old Version Querier Present timer expires.
902 if (mli->mli_version != MLD_VERSION_2)
905 mld_set_version(mli, MLD_VERSION_2);
908 mli->mli_qri = maxdelay;
910 CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi,
913 if (is_general_query) {
915 * MLDv2 General Query.
917 * Schedule a current-state report on this ifp for
918 * all groups, possibly containing source lists.
920 * If there is a pending General Query response
921 * scheduled earlier than the selected delay, do
922 * not schedule any other reports.
923 * Otherwise, reset the interface timer.
925 CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)",
927 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) {
928 mli->mli_v2_timer = MLD_RANDOM_DELAY(timer);
929 V_interface_timers_running6 = 1;
933 * MLDv2 Group-specific or Group-and-source-specific Query.
935 * Group-source-specific queries are throttled on
936 * a per-group basis to defeat denial-of-service attempts.
937 * Queries for groups we are not a member of on this
938 * link are simply ignored.
941 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
943 IF_ADDR_RUNLOCK(ifp);
947 if (!ratecheck(&inm->in6m_lastgsrtv,
949 CTR1(KTR_MLD, "%s: GS query throttled.",
951 IF_ADDR_RUNLOCK(ifp);
955 CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)",
958 * If there is a pending General Query response
959 * scheduled sooner than the selected delay, no
960 * further report need be scheduled.
961 * Otherwise, prepare to respond to the
962 * group-specific or group-and-source query.
964 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer)
965 mld_v2_process_group_query(inm, mli, timer, m, off);
967 /* XXX Clear embedded scope ID as userland won't expect it. */
968 in6_clearscope(&mld->mld_addr);
969 IF_ADDR_RUNLOCK(ifp);
980 * Process a recieved MLDv2 group-specific or group-and-source-specific
982 * Return <0 if any error occured. Currently this is ignored.
985 mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifinfo *mli,
986 int timer, struct mbuf *m0, const int off)
988 struct mldv2_query *mld;
992 IN6_MULTI_LOCK_ASSERT();
996 mld = (struct mldv2_query *)(mtod(m0, uint8_t *) + off);
998 switch (inm->in6m_state) {
1000 case MLD_SILENT_MEMBER:
1001 case MLD_SLEEPING_MEMBER:
1002 case MLD_LAZY_MEMBER:
1003 case MLD_AWAKENING_MEMBER:
1004 case MLD_IDLE_MEMBER:
1005 case MLD_LEAVING_MEMBER:
1008 case MLD_REPORTING_MEMBER:
1009 case MLD_G_QUERY_PENDING_MEMBER:
1010 case MLD_SG_QUERY_PENDING_MEMBER:
1014 nsrc = ntohs(mld->mld_numsrc);
1017 * Deal with group-specific queries upfront.
1018 * If any group query is already pending, purge any recorded
1019 * source-list state if it exists, and schedule a query response
1020 * for this group-specific query.
1023 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
1024 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) {
1025 in6m_clear_recorded(inm);
1026 timer = min(inm->in6m_timer, timer);
1028 inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER;
1029 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1030 V_current_state_timers_running6 = 1;
1035 * Deal with the case where a group-and-source-specific query has
1036 * been received but a group-specific query is already pending.
1038 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) {
1039 timer = min(inm->in6m_timer, timer);
1040 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1041 V_current_state_timers_running6 = 1;
1046 * Finally, deal with the case where a group-and-source-specific
1047 * query has been received, where a response to a previous g-s-r
1048 * query exists, or none exists.
1049 * In this case, we need to parse the source-list which the Querier
1050 * has provided us with and check if we have any source list filter
1051 * entries at T1 for these sources. If we do not, there is no need
1052 * schedule a report and the query may be dropped.
1053 * If we do, we must record them and schedule a current-state
1054 * report for those sources.
1056 if (inm->in6m_nsrc > 0) {
1063 soff = off + sizeof(struct mldv2_query);
1065 for (i = 0; i < nsrc; i++) {
1066 sp = mtod(m, uint8_t *) + soff;
1067 retval = in6m_record_source(inm,
1068 (const struct in6_addr *)sp);
1071 nrecorded += retval;
1072 soff += sizeof(struct in6_addr);
1073 if (soff >= m->m_len) {
1074 soff = soff - m->m_len;
1080 if (nrecorded > 0) {
1082 "%s: schedule response to SG query", __func__);
1083 inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER;
1084 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1085 V_current_state_timers_running6 = 1;
1093 * Process a received MLDv1 host membership report.
1094 * Assumes mld points to mld_hdr in pulled up mbuf chain.
1096 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
1097 * mld_addr. This is OK as we own the mbuf chain.
1100 mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6,
1101 /*const*/ struct mld_hdr *mld)
1103 struct in6_addr src, dst;
1104 struct in6_ifaddr *ia;
1105 struct in6_multi *inm;
1107 char ip6tbuf[INET6_ADDRSTRLEN];
1110 if (!mld_v1enable) {
1111 CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)",
1112 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1113 ifp, ifp->if_xname);
1117 if (ifp->if_flags & IFF_LOOPBACK)
1121 * MLDv1 reports must originate from a host's link-local address,
1122 * or the unspecified address (when booting).
1125 in6_clearscope(&src);
1126 if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) {
1127 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
1128 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
1129 ifp, ifp->if_xname);
1134 * RFC2710 Section 4: MLDv1 reports must pertain to a multicast
1135 * group, and must be directed to the group itself.
1138 in6_clearscope(&dst);
1139 if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) ||
1140 !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) {
1141 CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)",
1142 ip6_sprintf(ip6tbuf, &ip6->ip6_dst),
1143 ifp, ifp->if_xname);
1148 * Make sure we don't hear our own membership report, as fast
1149 * leave requires knowing that we are the only member of a
1150 * group. Assume we used the link-local address if available,
1151 * otherwise look for ::.
1153 * XXX Note that scope ID comparison is needed for the address
1154 * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be
1155 * performed for the on-wire address.
1157 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1158 if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) ||
1159 (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) {
1161 ifa_free(&ia->ia_ifa);
1165 ifa_free(&ia->ia_ifa);
1167 CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)",
1168 ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, ifp->if_xname);
1171 * Embed scope ID of receiving interface in MLD query for lookup
1172 * whilst we don't hold other locks (due to KAME locking lameness).
1174 if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr))
1175 in6_setscope(&mld->mld_addr, ifp, NULL);
1182 * MLDv1 report suppression.
1183 * If we are a member of this group, and our membership should be
1184 * reported, and our group timer is pending or about to be reset,
1185 * stop our group timer by transitioning to the 'lazy' state.
1187 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
1189 struct mld_ifinfo *mli;
1191 mli = inm->in6m_mli;
1192 KASSERT(mli != NULL,
1193 ("%s: no mli for ifp %p", __func__, ifp));
1196 * If we are in MLDv2 host mode, do not allow the
1197 * other host's MLDv1 report to suppress our reports.
1199 if (mli->mli_version == MLD_VERSION_2)
1202 inm->in6m_timer = 0;
1204 switch (inm->in6m_state) {
1205 case MLD_NOT_MEMBER:
1206 case MLD_SILENT_MEMBER:
1207 case MLD_SLEEPING_MEMBER:
1209 case MLD_REPORTING_MEMBER:
1210 case MLD_IDLE_MEMBER:
1211 case MLD_AWAKENING_MEMBER:
1213 "report suppressed for %s on ifp %p(%s)",
1214 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1215 ifp, ifp->if_xname);
1216 case MLD_LAZY_MEMBER:
1217 inm->in6m_state = MLD_LAZY_MEMBER;
1219 case MLD_G_QUERY_PENDING_MEMBER:
1220 case MLD_SG_QUERY_PENDING_MEMBER:
1221 case MLD_LEAVING_MEMBER:
1227 IF_ADDR_RUNLOCK(ifp);
1231 /* XXX Clear embedded scope ID as userland won't expect it. */
1232 in6_clearscope(&mld->mld_addr);
1240 * Assume query messages which fit in a single ICMPv6 message header
1241 * have been pulled up.
1242 * Assume that userland will want to see the message, even if it
1243 * otherwise fails kernel input validation; do not free it.
1244 * Pullup may however free the mbuf chain m if it fails.
1246 * Return IPPROTO_DONE if we freed m. Otherwise, return 0.
1249 mld_input(struct mbuf *m, int off, int icmp6len)
1252 struct ip6_hdr *ip6;
1253 struct mld_hdr *mld;
1256 CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off);
1258 ifp = m->m_pkthdr.rcvif;
1260 ip6 = mtod(m, struct ip6_hdr *);
1262 /* Pullup to appropriate size. */
1263 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1264 if (mld->mld_type == MLD_LISTENER_QUERY &&
1265 icmp6len >= sizeof(struct mldv2_query)) {
1266 mldlen = sizeof(struct mldv2_query);
1268 mldlen = sizeof(struct mld_hdr);
1270 IP6_EXTHDR_GET(mld, struct mld_hdr *, m, off, mldlen);
1272 ICMP6STAT_INC(icp6s_badlen);
1273 return (IPPROTO_DONE);
1277 * Userland needs to see all of this traffic for implementing
1278 * the endpoint discovery portion of multicast routing.
1280 switch (mld->mld_type) {
1281 case MLD_LISTENER_QUERY:
1282 icmp6_ifstat_inc(ifp, ifs6_in_mldquery);
1283 if (icmp6len == sizeof(struct mld_hdr)) {
1284 if (mld_v1_input_query(ifp, ip6, mld) != 0)
1286 } else if (icmp6len >= sizeof(struct mldv2_query)) {
1287 if (mld_v2_input_query(ifp, ip6, m, off,
1292 case MLD_LISTENER_REPORT:
1293 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1294 if (mld_v1_input_report(ifp, ip6, mld) != 0)
1297 case MLDV2_LISTENER_REPORT:
1298 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1300 case MLD_LISTENER_DONE:
1301 icmp6_ifstat_inc(ifp, ifs6_in_mlddone);
1311 * Fast timeout handler (global).
1312 * VIMAGE: Timeout handlers are expected to service all vimages.
1317 VNET_ITERATOR_DECL(vnet_iter);
1319 VNET_LIST_RLOCK_NOSLEEP();
1320 VNET_FOREACH(vnet_iter) {
1321 CURVNET_SET(vnet_iter);
1322 mld_fasttimo_vnet();
1325 VNET_LIST_RUNLOCK_NOSLEEP();
1329 * Fast timeout handler (per-vnet).
1331 * VIMAGE: Assume caller has set up our curvnet.
1334 mld_fasttimo_vnet(void)
1336 struct ifqueue scq; /* State-change packets */
1337 struct ifqueue qrq; /* Query response packets */
1339 struct mld_ifinfo *mli;
1340 struct ifmultiaddr *ifma;
1341 struct in6_multi *inm, *tinm;
1347 * Quick check to see if any work needs to be done, in order to
1348 * minimize the overhead of fasttimo processing.
1349 * SMPng: XXX Unlocked reads.
1351 if (!V_current_state_timers_running6 &&
1352 !V_interface_timers_running6 &&
1353 !V_state_change_timers_running6)
1360 * MLDv2 General Query response timer processing.
1362 if (V_interface_timers_running6) {
1363 CTR1(KTR_MLD, "%s: interface timers running", __func__);
1365 V_interface_timers_running6 = 0;
1366 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1367 if (mli->mli_v2_timer == 0) {
1369 } else if (--mli->mli_v2_timer == 0) {
1370 mld_v2_dispatch_general_query(mli);
1372 V_interface_timers_running6 = 1;
1377 if (!V_current_state_timers_running6 &&
1378 !V_state_change_timers_running6)
1381 V_current_state_timers_running6 = 0;
1382 V_state_change_timers_running6 = 0;
1384 CTR1(KTR_MLD, "%s: state change timers running", __func__);
1387 * MLD host report and state-change timer processing.
1388 * Note: Processing a v2 group timer may remove a node.
1390 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1393 if (mli->mli_version == MLD_VERSION_2) {
1394 uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri *
1397 memset(&qrq, 0, sizeof(struct ifqueue));
1398 IFQ_SET_MAXLEN(&qrq, MLD_MAX_G_GS_PACKETS);
1400 memset(&scq, 0, sizeof(struct ifqueue));
1401 IFQ_SET_MAXLEN(&scq, MLD_MAX_STATE_CHANGE_PACKETS);
1405 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1406 if (ifma->ifma_addr->sa_family != AF_INET6 ||
1407 ifma->ifma_protospec == NULL)
1409 inm = (struct in6_multi *)ifma->ifma_protospec;
1410 switch (mli->mli_version) {
1412 mld_v1_process_group_timer(mli, inm);
1415 mld_v2_process_group_timers(mli, &qrq,
1416 &scq, inm, uri_fasthz);
1420 IF_ADDR_RUNLOCK(ifp);
1422 switch (mli->mli_version) {
1425 * Transmit reports for this lifecycle. This
1426 * is done while not holding IF_ADDR_LOCK
1427 * since this can call
1428 * in6ifa_ifpforlinklocal() which locks
1429 * IF_ADDR_LOCK internally as well as
1430 * ip6_output() to transmit a packet.
1432 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead,
1434 SLIST_REMOVE_HEAD(&mli->mli_relinmhead,
1436 (void)mld_v1_transmit_report(inm,
1437 MLD_LISTENER_REPORT);
1441 mld_dispatch_queue(&qrq, 0);
1442 mld_dispatch_queue(&scq, 0);
1445 * Free the in_multi reference(s) for
1448 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead,
1450 SLIST_REMOVE_HEAD(&mli->mli_relinmhead,
1452 in6m_release_locked(inm);
1464 * Update host report group timer.
1465 * Will update the global pending timer flags.
1468 mld_v1_process_group_timer(struct mld_ifinfo *mli, struct in6_multi *inm)
1470 int report_timer_expired;
1472 IN6_MULTI_LOCK_ASSERT();
1475 if (inm->in6m_timer == 0) {
1476 report_timer_expired = 0;
1477 } else if (--inm->in6m_timer == 0) {
1478 report_timer_expired = 1;
1480 V_current_state_timers_running6 = 1;
1484 switch (inm->in6m_state) {
1485 case MLD_NOT_MEMBER:
1486 case MLD_SILENT_MEMBER:
1487 case MLD_IDLE_MEMBER:
1488 case MLD_LAZY_MEMBER:
1489 case MLD_SLEEPING_MEMBER:
1490 case MLD_AWAKENING_MEMBER:
1492 case MLD_REPORTING_MEMBER:
1493 if (report_timer_expired) {
1494 inm->in6m_state = MLD_IDLE_MEMBER;
1495 SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm,
1499 case MLD_G_QUERY_PENDING_MEMBER:
1500 case MLD_SG_QUERY_PENDING_MEMBER:
1501 case MLD_LEAVING_MEMBER:
1507 * Update a group's timers for MLDv2.
1508 * Will update the global pending timer flags.
1509 * Note: Unlocked read from mli.
1512 mld_v2_process_group_timers(struct mld_ifinfo *mli,
1513 struct ifqueue *qrq, struct ifqueue *scq,
1514 struct in6_multi *inm, const int uri_fasthz)
1516 int query_response_timer_expired;
1517 int state_change_retransmit_timer_expired;
1519 char ip6tbuf[INET6_ADDRSTRLEN];
1522 IN6_MULTI_LOCK_ASSERT();
1525 query_response_timer_expired = 0;
1526 state_change_retransmit_timer_expired = 0;
1529 * During a transition from compatibility mode back to MLDv2,
1530 * a group record in REPORTING state may still have its group
1531 * timer active. This is a no-op in this function; it is easier
1532 * to deal with it here than to complicate the slow-timeout path.
1534 if (inm->in6m_timer == 0) {
1535 query_response_timer_expired = 0;
1536 } else if (--inm->in6m_timer == 0) {
1537 query_response_timer_expired = 1;
1539 V_current_state_timers_running6 = 1;
1542 if (inm->in6m_sctimer == 0) {
1543 state_change_retransmit_timer_expired = 0;
1544 } else if (--inm->in6m_sctimer == 0) {
1545 state_change_retransmit_timer_expired = 1;
1547 V_state_change_timers_running6 = 1;
1550 /* We are in fasttimo, so be quick about it. */
1551 if (!state_change_retransmit_timer_expired &&
1552 !query_response_timer_expired)
1555 switch (inm->in6m_state) {
1556 case MLD_NOT_MEMBER:
1557 case MLD_SILENT_MEMBER:
1558 case MLD_SLEEPING_MEMBER:
1559 case MLD_LAZY_MEMBER:
1560 case MLD_AWAKENING_MEMBER:
1561 case MLD_IDLE_MEMBER:
1563 case MLD_G_QUERY_PENDING_MEMBER:
1564 case MLD_SG_QUERY_PENDING_MEMBER:
1566 * Respond to a previously pending Group-Specific
1567 * or Group-and-Source-Specific query by enqueueing
1568 * the appropriate Current-State report for
1569 * immediate transmission.
1571 if (query_response_timer_expired) {
1574 retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1,
1575 (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER),
1577 CTR2(KTR_MLD, "%s: enqueue record = %d",
1579 inm->in6m_state = MLD_REPORTING_MEMBER;
1580 in6m_clear_recorded(inm);
1583 case MLD_REPORTING_MEMBER:
1584 case MLD_LEAVING_MEMBER:
1585 if (state_change_retransmit_timer_expired) {
1587 * State-change retransmission timer fired.
1588 * If there are any further pending retransmissions,
1589 * set the global pending state-change flag, and
1592 if (--inm->in6m_scrv > 0) {
1593 inm->in6m_sctimer = uri_fasthz;
1594 V_state_change_timers_running6 = 1;
1597 * Retransmit the previously computed state-change
1598 * report. If there are no further pending
1599 * retransmissions, the mbuf queue will be consumed.
1600 * Update T0 state to T1 as we have now sent
1603 (void)mld_v2_merge_state_changes(inm, scq);
1606 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
1607 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1608 inm->in6m_ifp->if_xname);
1611 * If we are leaving the group for good, make sure
1612 * we release MLD's reference to it.
1613 * This release must be deferred using a SLIST,
1614 * as we are called from a loop which traverses
1615 * the in_ifmultiaddr TAILQ.
1617 if (inm->in6m_state == MLD_LEAVING_MEMBER &&
1618 inm->in6m_scrv == 0) {
1619 inm->in6m_state = MLD_NOT_MEMBER;
1620 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
1629 * Switch to a different version on the given interface,
1630 * as per Section 9.12.
1633 mld_set_version(struct mld_ifinfo *mli, const int version)
1635 int old_version_timer;
1639 CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__,
1640 version, mli->mli_ifp, mli->mli_ifp->if_xname);
1642 if (version == MLD_VERSION_1) {
1644 * Compute the "Older Version Querier Present" timer as per
1647 old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri;
1648 old_version_timer *= PR_SLOWHZ;
1649 mli->mli_v1_timer = old_version_timer;
1652 if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) {
1653 mli->mli_version = MLD_VERSION_1;
1654 mld_v2_cancel_link_timers(mli);
1659 * Cancel pending MLDv2 timers for the given link and all groups
1660 * joined on it; state-change, general-query, and group-query timers.
1663 mld_v2_cancel_link_timers(struct mld_ifinfo *mli)
1665 struct ifmultiaddr *ifma;
1667 struct in6_multi *inm, *tinm;
1669 CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__,
1670 mli->mli_ifp, mli->mli_ifp->if_xname);
1672 IN6_MULTI_LOCK_ASSERT();
1676 * Fast-track this potentially expensive operation
1677 * by checking all the global 'timer pending' flags.
1679 if (!V_interface_timers_running6 &&
1680 !V_state_change_timers_running6 &&
1681 !V_current_state_timers_running6)
1684 mli->mli_v2_timer = 0;
1689 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1690 if (ifma->ifma_addr->sa_family != AF_INET6)
1692 inm = (struct in6_multi *)ifma->ifma_protospec;
1693 switch (inm->in6m_state) {
1694 case MLD_NOT_MEMBER:
1695 case MLD_SILENT_MEMBER:
1696 case MLD_IDLE_MEMBER:
1697 case MLD_LAZY_MEMBER:
1698 case MLD_SLEEPING_MEMBER:
1699 case MLD_AWAKENING_MEMBER:
1701 case MLD_LEAVING_MEMBER:
1703 * If we are leaving the group and switching
1704 * version, we need to release the final
1705 * reference held for issuing the INCLUDE {}.
1707 SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm,
1710 case MLD_G_QUERY_PENDING_MEMBER:
1711 case MLD_SG_QUERY_PENDING_MEMBER:
1712 in6m_clear_recorded(inm);
1714 case MLD_REPORTING_MEMBER:
1715 inm->in6m_sctimer = 0;
1716 inm->in6m_timer = 0;
1717 inm->in6m_state = MLD_REPORTING_MEMBER;
1719 * Free any pending MLDv2 state-change records.
1721 _IF_DRAIN(&inm->in6m_scq);
1725 IF_ADDR_RUNLOCK(ifp);
1726 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) {
1727 SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele);
1728 in6m_release_locked(inm);
1733 * Global slowtimo handler.
1734 * VIMAGE: Timeout handlers are expected to service all vimages.
1739 VNET_ITERATOR_DECL(vnet_iter);
1741 VNET_LIST_RLOCK_NOSLEEP();
1742 VNET_FOREACH(vnet_iter) {
1743 CURVNET_SET(vnet_iter);
1744 mld_slowtimo_vnet();
1747 VNET_LIST_RUNLOCK_NOSLEEP();
1751 * Per-vnet slowtimo handler.
1754 mld_slowtimo_vnet(void)
1756 struct mld_ifinfo *mli;
1760 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1761 mld_v1_process_querier_timers(mli);
1768 * Update the Older Version Querier Present timers for a link.
1769 * See Section 9.12 of RFC 3810.
1772 mld_v1_process_querier_timers(struct mld_ifinfo *mli)
1777 if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) {
1779 * MLDv1 Querier Present timer expired; revert to MLDv2.
1782 "%s: transition from v%d -> v%d on %p(%s)",
1783 __func__, mli->mli_version, MLD_VERSION_2,
1784 mli->mli_ifp, mli->mli_ifp->if_xname);
1785 mli->mli_version = MLD_VERSION_2;
1790 * Transmit an MLDv1 report immediately.
1793 mld_v1_transmit_report(struct in6_multi *in6m, const int type)
1796 struct in6_ifaddr *ia;
1797 struct ip6_hdr *ip6;
1798 struct mbuf *mh, *md;
1799 struct mld_hdr *mld;
1801 IN6_MULTI_LOCK_ASSERT();
1804 ifp = in6m->in6m_ifp;
1805 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1806 /* ia may be NULL if link-local address is tentative. */
1808 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
1811 ifa_free(&ia->ia_ifa);
1814 MGET(md, M_DONTWAIT, MT_DATA);
1818 ifa_free(&ia->ia_ifa);
1824 * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so
1825 * that ether_output() does not need to allocate another mbuf
1826 * for the header in the most common case.
1828 MH_ALIGN(mh, sizeof(struct ip6_hdr));
1829 mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
1830 mh->m_len = sizeof(struct ip6_hdr);
1832 ip6 = mtod(mh, struct ip6_hdr *);
1834 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1835 ip6->ip6_vfc |= IPV6_VERSION;
1836 ip6->ip6_nxt = IPPROTO_ICMPV6;
1837 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
1838 ip6->ip6_dst = in6m->in6m_addr;
1840 md->m_len = sizeof(struct mld_hdr);
1841 mld = mtod(md, struct mld_hdr *);
1842 mld->mld_type = type;
1845 mld->mld_maxdelay = 0;
1846 mld->mld_reserved = 0;
1847 mld->mld_addr = in6m->in6m_addr;
1848 in6_clearscope(&mld->mld_addr);
1849 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
1850 sizeof(struct ip6_hdr), sizeof(struct mld_hdr));
1852 mld_save_context(mh, ifp);
1853 mh->m_flags |= M_MLDV1;
1855 mld_dispatch_packet(mh);
1858 ifa_free(&ia->ia_ifa);
1863 * Process a state change from the upper layer for the given IPv6 group.
1865 * Each socket holds a reference on the in_multi in its own ip_moptions.
1866 * The socket layer will have made the necessary updates to.the group
1867 * state, it is now up to MLD to issue a state change report if there
1868 * has been any change between T0 (when the last state-change was issued)
1871 * We use the MLDv2 state machine at group level. The MLd module
1872 * however makes the decision as to which MLD protocol version to speak.
1873 * A state change *from* INCLUDE {} always means an initial join.
1874 * A state change *to* INCLUDE {} always means a final leave.
1876 * If delay is non-zero, and the state change is an initial multicast
1877 * join, the state change report will be delayed by 'delay' ticks
1878 * in units of PR_FASTHZ if MLDv1 is active on the link; otherwise
1879 * the initial MLDv2 state change report will be delayed by whichever
1880 * is sooner, a pending state-change timer or delay itself.
1882 * VIMAGE: curvnet should have been set by caller, as this routine
1883 * is called from the socket option handlers.
1886 mld_change_state(struct in6_multi *inm, const int delay)
1888 struct mld_ifinfo *mli;
1892 IN6_MULTI_LOCK_ASSERT();
1897 * Try to detect if the upper layer just asked us to change state
1898 * for an interface which has now gone away.
1900 KASSERT(inm->in6m_ifma != NULL, ("%s: no ifma", __func__));
1901 ifp = inm->in6m_ifma->ifma_ifp;
1904 * Sanity check that netinet6's notion of ifp is the
1907 KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__));
1912 mli = MLD_IFINFO(ifp);
1913 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
1916 * If we detect a state transition to or from MCAST_UNDEFINED
1917 * for this group, then we are starting or finishing an MLD
1918 * life cycle for this group.
1920 if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) {
1921 CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__,
1922 inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode);
1923 if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) {
1924 CTR1(KTR_MLD, "%s: initial join", __func__);
1925 error = mld_initial_join(inm, mli, delay);
1927 } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) {
1928 CTR1(KTR_MLD, "%s: final leave", __func__);
1929 mld_final_leave(inm, mli);
1933 CTR1(KTR_MLD, "%s: filter set change", __func__);
1936 error = mld_handle_state_change(inm, mli);
1944 * Perform the initial join for an MLD group.
1946 * When joining a group:
1947 * If the group should have its MLD traffic suppressed, do nothing.
1948 * MLDv1 starts sending MLDv1 host membership reports.
1949 * MLDv2 will schedule an MLDv2 state-change report containing the
1950 * initial state of the membership.
1952 * If the delay argument is non-zero, then we must delay sending the
1953 * initial state change for delay ticks (in units of PR_FASTHZ).
1956 mld_initial_join(struct in6_multi *inm, struct mld_ifinfo *mli,
1960 struct ifqueue *ifq;
1961 int error, retval, syncstates;
1964 char ip6tbuf[INET6_ADDRSTRLEN];
1967 CTR4(KTR_MLD, "%s: initial join %s on ifp %p(%s)",
1968 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1969 inm->in6m_ifp, inm->in6m_ifp->if_xname);
1974 ifp = inm->in6m_ifp;
1976 IN6_MULTI_LOCK_ASSERT();
1979 KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__));
1982 * Groups joined on loopback or marked as 'not reported',
1983 * enter the MLD_SILENT_MEMBER state and
1984 * are never reported in any protocol exchanges.
1985 * All other groups enter the appropriate state machine
1986 * for the version in use on this link.
1987 * A link marked as MLIF_SILENT causes MLD to be completely
1988 * disabled for the link.
1990 if ((ifp->if_flags & IFF_LOOPBACK) ||
1991 (mli->mli_flags & MLIF_SILENT) ||
1992 !mld_is_addr_reported(&inm->in6m_addr)) {
1994 "%s: not kicking state machine for silent group", __func__);
1995 inm->in6m_state = MLD_SILENT_MEMBER;
1996 inm->in6m_timer = 0;
1999 * Deal with overlapping in_multi lifecycle.
2000 * If this group was LEAVING, then make sure
2001 * we drop the reference we picked up to keep the
2002 * group around for the final INCLUDE {} enqueue.
2004 if (mli->mli_version == MLD_VERSION_2 &&
2005 inm->in6m_state == MLD_LEAVING_MEMBER)
2006 in6m_release_locked(inm);
2008 inm->in6m_state = MLD_REPORTING_MEMBER;
2010 switch (mli->mli_version) {
2013 * If a delay was provided, only use it if
2014 * it is greater than the delay normally
2015 * used for an MLDv1 state change report,
2016 * and delay sending the initial MLDv1 report
2017 * by not transitioning to the IDLE state.
2019 odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * PR_FASTHZ);
2021 inm->in6m_timer = max(delay, odelay);
2022 V_current_state_timers_running6 = 1;
2024 inm->in6m_state = MLD_IDLE_MEMBER;
2025 error = mld_v1_transmit_report(inm,
2026 MLD_LISTENER_REPORT);
2028 inm->in6m_timer = odelay;
2029 V_current_state_timers_running6 = 1;
2036 * Defer update of T0 to T1, until the first copy
2037 * of the state change has been transmitted.
2042 * Immediately enqueue a State-Change Report for
2043 * this interface, freeing any previous reports.
2044 * Don't kick the timers if there is nothing to do,
2045 * or if an error occurred.
2047 ifq = &inm->in6m_scq;
2049 retval = mld_v2_enqueue_group_record(ifq, inm, 1,
2050 0, 0, (mli->mli_flags & MLIF_USEALLOW));
2051 CTR2(KTR_MLD, "%s: enqueue record = %d",
2054 error = retval * -1;
2059 * Schedule transmission of pending state-change
2060 * report up to RV times for this link. The timer
2061 * will fire at the next mld_fasttimo (~200ms),
2062 * giving us an opportunity to merge the reports.
2064 * If a delay was provided to this function, only
2065 * use this delay if sooner than the existing one.
2067 KASSERT(mli->mli_rv > 1,
2068 ("%s: invalid robustness %d", __func__,
2070 inm->in6m_scrv = mli->mli_rv;
2072 if (inm->in6m_sctimer > 1) {
2074 min(inm->in6m_sctimer, delay);
2076 inm->in6m_sctimer = delay;
2078 inm->in6m_sctimer = 1;
2079 V_state_change_timers_running6 = 1;
2087 * Only update the T0 state if state change is atomic,
2088 * i.e. we don't need to wait for a timer to fire before we
2089 * can consider the state change to have been communicated.
2093 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2094 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2095 inm->in6m_ifp->if_xname);
2102 * Issue an intermediate state change during the life-cycle.
2105 mld_handle_state_change(struct in6_multi *inm, struct mld_ifinfo *mli)
2110 char ip6tbuf[INET6_ADDRSTRLEN];
2113 CTR4(KTR_MLD, "%s: state change for %s on ifp %p(%s)",
2114 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2115 inm->in6m_ifp, inm->in6m_ifp->if_xname);
2117 ifp = inm->in6m_ifp;
2119 IN6_MULTI_LOCK_ASSERT();
2122 KASSERT(mli && mli->mli_ifp == ifp,
2123 ("%s: inconsistent ifp", __func__));
2125 if ((ifp->if_flags & IFF_LOOPBACK) ||
2126 (mli->mli_flags & MLIF_SILENT) ||
2127 !mld_is_addr_reported(&inm->in6m_addr) ||
2128 (mli->mli_version != MLD_VERSION_2)) {
2129 if (!mld_is_addr_reported(&inm->in6m_addr)) {
2131 "%s: not kicking state machine for silent group", __func__);
2133 CTR1(KTR_MLD, "%s: nothing to do", __func__);
2135 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2136 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2137 inm->in6m_ifp->if_xname);
2141 _IF_DRAIN(&inm->in6m_scq);
2143 retval = mld_v2_enqueue_group_record(&inm->in6m_scq, inm, 1, 0, 0,
2144 (mli->mli_flags & MLIF_USEALLOW));
2145 CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval);
2150 * If record(s) were enqueued, start the state-change
2151 * report timer for this group.
2153 inm->in6m_scrv = mli->mli_rv;
2154 inm->in6m_sctimer = 1;
2155 V_state_change_timers_running6 = 1;
2161 * Perform the final leave for a multicast address.
2163 * When leaving a group:
2164 * MLDv1 sends a DONE message, if and only if we are the reporter.
2165 * MLDv2 enqueues a state-change report containing a transition
2166 * to INCLUDE {} for immediate transmission.
2169 mld_final_leave(struct in6_multi *inm, struct mld_ifinfo *mli)
2173 char ip6tbuf[INET6_ADDRSTRLEN];
2178 CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)",
2179 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2180 inm->in6m_ifp, inm->in6m_ifp->if_xname);
2182 IN6_MULTI_LOCK_ASSERT();
2185 switch (inm->in6m_state) {
2186 case MLD_NOT_MEMBER:
2187 case MLD_SILENT_MEMBER:
2188 case MLD_LEAVING_MEMBER:
2189 /* Already leaving or left; do nothing. */
2191 "%s: not kicking state machine for silent group", __func__);
2193 case MLD_REPORTING_MEMBER:
2194 case MLD_IDLE_MEMBER:
2195 case MLD_G_QUERY_PENDING_MEMBER:
2196 case MLD_SG_QUERY_PENDING_MEMBER:
2197 if (mli->mli_version == MLD_VERSION_1) {
2199 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
2200 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER)
2201 panic("%s: MLDv2 state reached, not MLDv2 mode",
2204 mld_v1_transmit_report(inm, MLD_LISTENER_DONE);
2205 inm->in6m_state = MLD_NOT_MEMBER;
2206 } else if (mli->mli_version == MLD_VERSION_2) {
2208 * Stop group timer and all pending reports.
2209 * Immediately enqueue a state-change report
2210 * TO_IN {} to be sent on the next fast timeout,
2211 * giving us an opportunity to merge reports.
2213 _IF_DRAIN(&inm->in6m_scq);
2214 inm->in6m_timer = 0;
2215 inm->in6m_scrv = mli->mli_rv;
2216 CTR4(KTR_MLD, "%s: Leaving %s/%s with %d "
2217 "pending retransmissions.", __func__,
2218 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2219 inm->in6m_ifp->if_xname, inm->in6m_scrv);
2220 if (inm->in6m_scrv == 0) {
2221 inm->in6m_state = MLD_NOT_MEMBER;
2222 inm->in6m_sctimer = 0;
2226 in6m_acquire_locked(inm);
2228 retval = mld_v2_enqueue_group_record(
2229 &inm->in6m_scq, inm, 1, 0, 0,
2230 (mli->mli_flags & MLIF_USEALLOW));
2231 KASSERT(retval != 0,
2232 ("%s: enqueue record = %d", __func__,
2235 inm->in6m_state = MLD_LEAVING_MEMBER;
2236 inm->in6m_sctimer = 1;
2237 V_state_change_timers_running6 = 1;
2243 case MLD_LAZY_MEMBER:
2244 case MLD_SLEEPING_MEMBER:
2245 case MLD_AWAKENING_MEMBER:
2246 /* Our reports are suppressed; do nothing. */
2252 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2253 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2254 inm->in6m_ifp->if_xname);
2255 inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED;
2256 CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s",
2257 __func__, &inm->in6m_addr, inm->in6m_ifp->if_xname);
2262 * Enqueue an MLDv2 group record to the given output queue.
2264 * If is_state_change is zero, a current-state record is appended.
2265 * If is_state_change is non-zero, a state-change report is appended.
2267 * If is_group_query is non-zero, an mbuf packet chain is allocated.
2268 * If is_group_query is zero, and if there is a packet with free space
2269 * at the tail of the queue, it will be appended to providing there
2270 * is enough free space.
2271 * Otherwise a new mbuf packet chain is allocated.
2273 * If is_source_query is non-zero, each source is checked to see if
2274 * it was recorded for a Group-Source query, and will be omitted if
2275 * it is not both in-mode and recorded.
2277 * If use_block_allow is non-zero, state change reports for initial join
2278 * and final leave, on an inclusive mode group with a source list, will be
2279 * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively.
2281 * The function will attempt to allocate leading space in the packet
2282 * for the IPv6+ICMP headers to be prepended without fragmenting the chain.
2284 * If successful the size of all data appended to the queue is returned,
2285 * otherwise an error code less than zero is returned, or zero if
2286 * no record(s) were appended.
2289 mld_v2_enqueue_group_record(struct ifqueue *ifq, struct in6_multi *inm,
2290 const int is_state_change, const int is_group_query,
2291 const int is_source_query, const int use_block_allow)
2293 struct mldv2_record mr;
2294 struct mldv2_record *pmr;
2296 struct ip6_msource *ims, *nims;
2297 struct mbuf *m0, *m, *md;
2298 int error, is_filter_list_change;
2299 int minrec0len, m0srcs, msrcs, nbytes, off;
2300 int record_has_sources;
2305 char ip6tbuf[INET6_ADDRSTRLEN];
2308 IN6_MULTI_LOCK_ASSERT();
2311 ifp = inm->in6m_ifp;
2312 is_filter_list_change = 0;
2319 record_has_sources = 1;
2321 type = MLD_DO_NOTHING;
2322 mode = inm->in6m_st[1].iss_fmode;
2325 * If we did not transition out of ASM mode during t0->t1,
2326 * and there are no source nodes to process, we can skip
2327 * the generation of source records.
2329 if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 &&
2330 inm->in6m_nsrc == 0)
2331 record_has_sources = 0;
2333 if (is_state_change) {
2335 * Queue a state change record.
2336 * If the mode did not change, and there are non-ASM
2337 * listeners or source filters present,
2338 * we potentially need to issue two records for the group.
2339 * If there are ASM listeners, and there was no filter
2340 * mode transition of any kind, do nothing.
2342 * If we are transitioning to MCAST_UNDEFINED, we need
2343 * not send any sources. A transition to/from this state is
2344 * considered inclusive with some special treatment.
2346 * If we are rewriting initial joins/leaves to use
2347 * ALLOW/BLOCK, and the group's membership is inclusive,
2348 * we need to send sources in all cases.
2350 if (mode != inm->in6m_st[0].iss_fmode) {
2351 if (mode == MCAST_EXCLUDE) {
2352 CTR1(KTR_MLD, "%s: change to EXCLUDE",
2354 type = MLD_CHANGE_TO_EXCLUDE_MODE;
2356 CTR1(KTR_MLD, "%s: change to INCLUDE",
2358 if (use_block_allow) {
2361 * Here we're interested in state
2362 * edges either direction between
2363 * MCAST_UNDEFINED and MCAST_INCLUDE.
2364 * Perhaps we should just check
2365 * the group state, rather than
2368 if (mode == MCAST_UNDEFINED) {
2369 type = MLD_BLOCK_OLD_SOURCES;
2371 type = MLD_ALLOW_NEW_SOURCES;
2374 type = MLD_CHANGE_TO_INCLUDE_MODE;
2375 if (mode == MCAST_UNDEFINED)
2376 record_has_sources = 0;
2380 if (record_has_sources) {
2381 is_filter_list_change = 1;
2383 type = MLD_DO_NOTHING;
2388 * Queue a current state record.
2390 if (mode == MCAST_EXCLUDE) {
2391 type = MLD_MODE_IS_EXCLUDE;
2392 } else if (mode == MCAST_INCLUDE) {
2393 type = MLD_MODE_IS_INCLUDE;
2394 KASSERT(inm->in6m_st[1].iss_asm == 0,
2395 ("%s: inm %p is INCLUDE but ASM count is %d",
2396 __func__, inm, inm->in6m_st[1].iss_asm));
2401 * Generate the filter list changes using a separate function.
2403 if (is_filter_list_change)
2404 return (mld_v2_enqueue_filter_change(ifq, inm));
2406 if (type == MLD_DO_NOTHING) {
2407 CTR3(KTR_MLD, "%s: nothing to do for %s/%s",
2408 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2409 inm->in6m_ifp->if_xname);
2414 * If any sources are present, we must be able to fit at least
2415 * one in the trailing space of the tail packet's mbuf,
2418 minrec0len = sizeof(struct mldv2_record);
2419 if (record_has_sources)
2420 minrec0len += sizeof(struct in6_addr);
2422 CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__,
2423 mld_rec_type_to_str(type),
2424 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2425 inm->in6m_ifp->if_xname);
2428 * Check if we have a packet in the tail of the queue for this
2429 * group into which the first group record for this group will fit.
2430 * Otherwise allocate a new packet.
2431 * Always allocate leading space for IP6+RA+ICMPV6+REPORT.
2432 * Note: Group records for G/GSR query responses MUST be sent
2433 * in their own packet.
2436 if (!is_group_query &&
2438 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) &&
2439 (m0->m_pkthdr.len + minrec0len) <
2440 (ifp->if_mtu - MLD_MTUSPACE)) {
2441 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2442 sizeof(struct mldv2_record)) /
2443 sizeof(struct in6_addr);
2445 CTR1(KTR_MLD, "%s: use existing packet", __func__);
2447 if (_IF_QFULL(ifq)) {
2448 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2452 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2453 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2454 if (!is_state_change && !is_group_query)
2455 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2457 m = m_gethdr(M_DONTWAIT, MT_DATA);
2461 mld_save_context(m, ifp);
2463 CTR1(KTR_MLD, "%s: allocated first packet", __func__);
2467 * Append group record.
2468 * If we have sources, we don't know how many yet.
2473 mr.mr_addr = inm->in6m_addr;
2474 in6_clearscope(&mr.mr_addr);
2475 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2478 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2481 nbytes += sizeof(struct mldv2_record);
2484 * Append as many sources as will fit in the first packet.
2485 * If we are appending to a new packet, the chain allocation
2486 * may potentially use clusters; use m_getptr() in this case.
2487 * If we are appending to an existing packet, we need to obtain
2488 * a pointer to the group record after m_append(), in case a new
2489 * mbuf was allocated.
2491 * Only append sources which are in-mode at t1. If we are
2492 * transitioning to MCAST_UNDEFINED state on the group, and
2493 * use_block_allow is zero, do not include source entries.
2494 * Otherwise, we need to include this source in the report.
2496 * Only report recorded sources in our filter set when responding
2497 * to a group-source query.
2499 if (record_has_sources) {
2502 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2503 md->m_len - nbytes);
2505 md = m_getptr(m, 0, &off);
2506 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2510 RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs,
2512 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2513 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2514 now = im6s_get_mode(inm, ims, 1);
2515 CTR2(KTR_MLD, "%s: node is %d", __func__, now);
2516 if ((now != mode) ||
2518 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2519 CTR1(KTR_MLD, "%s: skip node", __func__);
2522 if (is_source_query && ims->im6s_stp == 0) {
2523 CTR1(KTR_MLD, "%s: skip unrecorded node",
2527 CTR1(KTR_MLD, "%s: append node", __func__);
2528 if (!m_append(m, sizeof(struct in6_addr),
2529 (void *)&ims->im6s_addr)) {
2532 CTR1(KTR_MLD, "%s: m_append() failed.",
2536 nbytes += sizeof(struct in6_addr);
2538 if (msrcs == m0srcs)
2541 CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__,
2543 pmr->mr_numsrc = htons(msrcs);
2544 nbytes += (msrcs * sizeof(struct in6_addr));
2547 if (is_source_query && msrcs == 0) {
2548 CTR1(KTR_MLD, "%s: no recorded sources to report", __func__);
2555 * We are good to go with first packet.
2558 CTR1(KTR_MLD, "%s: enqueueing first packet", __func__);
2559 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2560 _IF_ENQUEUE(ifq, m);
2562 m->m_pkthdr.PH_vt.vt_nrecs++;
2565 * No further work needed if no source list in packet(s).
2567 if (!record_has_sources)
2571 * Whilst sources remain to be announced, we need to allocate
2572 * a new packet and fill out as many sources as will fit.
2573 * Always try for a cluster first.
2575 while (nims != NULL) {
2576 if (_IF_QFULL(ifq)) {
2577 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2580 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2582 m = m_gethdr(M_DONTWAIT, MT_DATA);
2585 mld_save_context(m, ifp);
2586 md = m_getptr(m, 0, &off);
2587 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off);
2588 CTR1(KTR_MLD, "%s: allocated next packet", __func__);
2590 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2593 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2596 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2597 nbytes += sizeof(struct mldv2_record);
2599 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2600 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2603 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2604 CTR2(KTR_MLD, "%s: visit node %s",
2605 __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2606 now = im6s_get_mode(inm, ims, 1);
2607 if ((now != mode) ||
2609 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2610 CTR1(KTR_MLD, "%s: skip node", __func__);
2613 if (is_source_query && ims->im6s_stp == 0) {
2614 CTR1(KTR_MLD, "%s: skip unrecorded node",
2618 CTR1(KTR_MLD, "%s: append node", __func__);
2619 if (!m_append(m, sizeof(struct in6_addr),
2620 (void *)&ims->im6s_addr)) {
2623 CTR1(KTR_MLD, "%s: m_append() failed.",
2628 if (msrcs == m0srcs)
2631 pmr->mr_numsrc = htons(msrcs);
2632 nbytes += (msrcs * sizeof(struct in6_addr));
2634 CTR1(KTR_MLD, "%s: enqueueing next packet", __func__);
2635 _IF_ENQUEUE(ifq, m);
2642 * Type used to mark record pass completion.
2643 * We exploit the fact we can cast to this easily from the
2644 * current filter modes on each ip_msource node.
2647 REC_NONE = 0x00, /* MCAST_UNDEFINED */
2648 REC_ALLOW = 0x01, /* MCAST_INCLUDE */
2649 REC_BLOCK = 0x02, /* MCAST_EXCLUDE */
2650 REC_FULL = REC_ALLOW | REC_BLOCK
2654 * Enqueue an MLDv2 filter list change to the given output queue.
2656 * Source list filter state is held in an RB-tree. When the filter list
2657 * for a group is changed without changing its mode, we need to compute
2658 * the deltas between T0 and T1 for each source in the filter set,
2659 * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records.
2661 * As we may potentially queue two record types, and the entire R-B tree
2662 * needs to be walked at once, we break this out into its own function
2663 * so we can generate a tightly packed queue of packets.
2665 * XXX This could be written to only use one tree walk, although that makes
2666 * serializing into the mbuf chains a bit harder. For now we do two walks
2667 * which makes things easier on us, and it may or may not be harder on
2670 * If successful the size of all data appended to the queue is returned,
2671 * otherwise an error code less than zero is returned, or zero if
2672 * no record(s) were appended.
2675 mld_v2_enqueue_filter_change(struct ifqueue *ifq, struct in6_multi *inm)
2677 static const int MINRECLEN =
2678 sizeof(struct mldv2_record) + sizeof(struct in6_addr);
2680 struct mldv2_record mr;
2681 struct mldv2_record *pmr;
2682 struct ip6_msource *ims, *nims;
2683 struct mbuf *m, *m0, *md;
2684 int m0srcs, nbytes, npbytes, off, rsrcs, schanged;
2686 uint8_t mode, now, then;
2687 rectype_t crt, drt, nrt;
2689 char ip6tbuf[INET6_ADDRSTRLEN];
2692 IN6_MULTI_LOCK_ASSERT();
2694 if (inm->in6m_nsrc == 0 ||
2695 (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0))
2698 ifp = inm->in6m_ifp; /* interface */
2699 mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */
2700 crt = REC_NONE; /* current group record type */
2701 drt = REC_NONE; /* mask of completed group record types */
2702 nrt = REC_NONE; /* record type for current node */
2703 m0srcs = 0; /* # source which will fit in current mbuf chain */
2704 npbytes = 0; /* # of bytes appended this packet */
2705 nbytes = 0; /* # of bytes appended to group's state-change queue */
2706 rsrcs = 0; /* # sources encoded in current record */
2707 schanged = 0; /* # nodes encoded in overall filter change */
2708 nallow = 0; /* # of source entries in ALLOW_NEW */
2709 nblock = 0; /* # of source entries in BLOCK_OLD */
2710 nims = NULL; /* next tree node pointer */
2713 * For each possible filter record mode.
2714 * The first kind of source we encounter tells us which
2715 * is the first kind of record we start appending.
2716 * If a node transitioned to UNDEFINED at t1, its mode is treated
2717 * as the inverse of the group's filter mode.
2719 while (drt != REC_FULL) {
2723 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <=
2724 MLD_V2_REPORT_MAXRECS) &&
2725 (m0->m_pkthdr.len + MINRECLEN) <
2726 (ifp->if_mtu - MLD_MTUSPACE)) {
2728 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2729 sizeof(struct mldv2_record)) /
2730 sizeof(struct in6_addr);
2732 "%s: use previous packet", __func__);
2734 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
2736 m = m_gethdr(M_DONTWAIT, MT_DATA);
2739 "%s: m_get*() failed", __func__);
2742 m->m_pkthdr.PH_vt.vt_nrecs = 0;
2743 mld_save_context(m, ifp);
2744 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2745 sizeof(struct mldv2_record)) /
2746 sizeof(struct in6_addr);
2749 "%s: allocated new packet", __func__);
2752 * Append the MLD group record header to the
2753 * current packet's data area.
2754 * Recalculate pointer to free space for next
2755 * group record, in case m_append() allocated
2756 * a new mbuf or cluster.
2758 memset(&mr, 0, sizeof(mr));
2759 mr.mr_addr = inm->in6m_addr;
2760 in6_clearscope(&mr.mr_addr);
2761 if (!m_append(m, sizeof(mr), (void *)&mr)) {
2765 "%s: m_append() failed", __func__);
2768 npbytes += sizeof(struct mldv2_record);
2770 /* new packet; offset in chain */
2771 md = m_getptr(m, npbytes -
2772 sizeof(struct mldv2_record), &off);
2773 pmr = (struct mldv2_record *)(mtod(md,
2776 /* current packet; offset from last append */
2778 pmr = (struct mldv2_record *)(mtod(md,
2779 uint8_t *) + md->m_len -
2780 sizeof(struct mldv2_record));
2783 * Begin walking the tree for this record type
2784 * pass, or continue from where we left off
2785 * previously if we had to allocate a new packet.
2786 * Only report deltas in-mode at t1.
2787 * We need not report included sources as allowed
2788 * if we are in inclusive mode on the group,
2789 * however the converse is not true.
2793 nims = RB_MIN(ip6_msource_tree,
2796 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2797 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2798 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2799 now = im6s_get_mode(inm, ims, 1);
2800 then = im6s_get_mode(inm, ims, 0);
2801 CTR3(KTR_MLD, "%s: mode: t0 %d, t1 %d",
2802 __func__, then, now);
2805 "%s: skip unchanged", __func__);
2808 if (mode == MCAST_EXCLUDE &&
2809 now == MCAST_INCLUDE) {
2811 "%s: skip IN src on EX group",
2815 nrt = (rectype_t)now;
2816 if (nrt == REC_NONE)
2817 nrt = (rectype_t)(~mode & REC_FULL);
2818 if (schanged++ == 0) {
2820 } else if (crt != nrt)
2822 if (!m_append(m, sizeof(struct in6_addr),
2823 (void *)&ims->im6s_addr)) {
2827 "%s: m_append() failed", __func__);
2830 nallow += !!(crt == REC_ALLOW);
2831 nblock += !!(crt == REC_BLOCK);
2832 if (++rsrcs == m0srcs)
2836 * If we did not append any tree nodes on this
2837 * pass, back out of allocations.
2840 npbytes -= sizeof(struct mldv2_record);
2843 "%s: m_free(m)", __func__);
2847 "%s: m_adj(m, -mr)", __func__);
2848 m_adj(m, -((int)sizeof(
2849 struct mldv2_record)));
2853 npbytes += (rsrcs * sizeof(struct in6_addr));
2854 if (crt == REC_ALLOW)
2855 pmr->mr_type = MLD_ALLOW_NEW_SOURCES;
2856 else if (crt == REC_BLOCK)
2857 pmr->mr_type = MLD_BLOCK_OLD_SOURCES;
2858 pmr->mr_numsrc = htons(rsrcs);
2860 * Count the new group record, and enqueue this
2861 * packet if it wasn't already queued.
2863 m->m_pkthdr.PH_vt.vt_nrecs++;
2865 _IF_ENQUEUE(ifq, m);
2867 } while (nims != NULL);
2869 crt = (~crt & REC_FULL);
2872 CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__,
2879 mld_v2_merge_state_changes(struct in6_multi *inm, struct ifqueue *ifscq)
2882 struct mbuf *m; /* pending state-change */
2883 struct mbuf *m0; /* copy of pending state-change */
2884 struct mbuf *mt; /* last state-change in packet */
2885 int docopy, domerge;
2892 IN6_MULTI_LOCK_ASSERT();
2896 * If there are further pending retransmissions, make a writable
2897 * copy of each queued state-change message before merging.
2899 if (inm->in6m_scrv > 0)
2902 gq = &inm->in6m_scq;
2904 if (gq->ifq_head == NULL) {
2905 CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty",
2913 * Only merge the report into the current packet if
2914 * there is sufficient space to do so; an MLDv2 report
2915 * packet may only contain 65,535 group records.
2916 * Always use a simple mbuf chain concatentation to do this,
2917 * as large state changes for single groups may have
2918 * allocated clusters.
2921 mt = ifscq->ifq_tail;
2923 recslen = m_length(m, NULL);
2925 if ((mt->m_pkthdr.PH_vt.vt_nrecs +
2926 m->m_pkthdr.PH_vt.vt_nrecs <=
2927 MLD_V2_REPORT_MAXRECS) &&
2928 (mt->m_pkthdr.len + recslen <=
2929 (inm->in6m_ifp->if_mtu - MLD_MTUSPACE)))
2933 if (!domerge && _IF_QFULL(gq)) {
2935 "%s: outbound queue full, skipping whole packet %p",
2945 CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m);
2946 _IF_DEQUEUE(gq, m0);
2949 CTR2(KTR_MLD, "%s: copying %p", __func__, m);
2950 m0 = m_dup(m, M_NOWAIT);
2953 m0->m_nextpkt = NULL;
2958 CTR3(KTR_MLD, "%s: queueing %p to ifscq %p)",
2959 __func__, m0, ifscq);
2960 _IF_ENQUEUE(ifscq, m0);
2962 struct mbuf *mtl; /* last mbuf of packet mt */
2964 CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)",
2968 m0->m_flags &= ~M_PKTHDR;
2969 mt->m_pkthdr.len += recslen;
2970 mt->m_pkthdr.PH_vt.vt_nrecs +=
2971 m0->m_pkthdr.PH_vt.vt_nrecs;
2981 * Respond to a pending MLDv2 General Query.
2984 mld_v2_dispatch_general_query(struct mld_ifinfo *mli)
2986 struct ifmultiaddr *ifma;
2988 struct in6_multi *inm;
2991 IN6_MULTI_LOCK_ASSERT();
2994 KASSERT(mli->mli_version == MLD_VERSION_2,
2995 ("%s: called when version %d", __func__, mli->mli_version));
3000 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3001 if (ifma->ifma_addr->sa_family != AF_INET6 ||
3002 ifma->ifma_protospec == NULL)
3005 inm = (struct in6_multi *)ifma->ifma_protospec;
3006 KASSERT(ifp == inm->in6m_ifp,
3007 ("%s: inconsistent ifp", __func__));
3009 switch (inm->in6m_state) {
3010 case MLD_NOT_MEMBER:
3011 case MLD_SILENT_MEMBER:
3013 case MLD_REPORTING_MEMBER:
3014 case MLD_IDLE_MEMBER:
3015 case MLD_LAZY_MEMBER:
3016 case MLD_SLEEPING_MEMBER:
3017 case MLD_AWAKENING_MEMBER:
3018 inm->in6m_state = MLD_REPORTING_MEMBER;
3019 retval = mld_v2_enqueue_group_record(&mli->mli_gq,
3021 CTR2(KTR_MLD, "%s: enqueue record = %d",
3024 case MLD_G_QUERY_PENDING_MEMBER:
3025 case MLD_SG_QUERY_PENDING_MEMBER:
3026 case MLD_LEAVING_MEMBER:
3030 IF_ADDR_RUNLOCK(ifp);
3032 mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST);
3035 * Slew transmission of bursts over 500ms intervals.
3037 if (mli->mli_gq.ifq_head != NULL) {
3038 mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY(
3039 MLD_RESPONSE_BURST_INTERVAL);
3040 V_interface_timers_running6 = 1;
3045 * Transmit the next pending message in the output queue.
3047 * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis.
3048 * MRT: Nothing needs to be done, as MLD traffic is always local to
3049 * a link and uses a link-scope multicast address.
3052 mld_dispatch_packet(struct mbuf *m)
3054 struct ip6_moptions im6o;
3059 struct ip6_hdr *ip6;
3060 struct mld_hdr *mld;
3066 CTR2(KTR_MLD, "%s: transmit %p", __func__, m);
3069 * Set VNET image pointer from enqueued mbuf chain
3070 * before doing anything else. Whilst we use interface
3071 * indexes to guard against interface detach, they are
3072 * unique to each VIMAGE and must be retrieved.
3074 ifindex = mld_restore_context(m);
3077 * Check if the ifnet still exists. This limits the scope of
3078 * any race in the absence of a global ifp lock for low cost
3079 * (an array lookup).
3081 ifp = ifnet_byindex(ifindex);
3083 CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.",
3084 __func__, m, ifindex);
3086 IP6STAT_INC(ip6s_noroute);
3090 im6o.im6o_multicast_hlim = 1;
3091 im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL);
3092 im6o.im6o_multicast_ifp = ifp;
3094 if (m->m_flags & M_MLDV1) {
3097 m0 = mld_v2_encap_report(ifp, m);
3099 CTR2(KTR_MLD, "%s: dropped %p", __func__, m);
3100 IP6STAT_INC(ip6s_odropped);
3105 mld_scrub_context(m0);
3106 m->m_flags &= ~(M_PROTOFLAGS);
3107 m0->m_pkthdr.rcvif = V_loif;
3109 ip6 = mtod(m0, struct ip6_hdr *);
3111 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */
3114 * XXX XXX Break some KPI rules to prevent an LOR which would
3115 * occur if we called in6_setscope() at transmission.
3116 * See comments at top of file.
3118 MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index);
3122 * Retrieve the ICMPv6 type before handoff to ip6_output(),
3123 * so we can bump the stats.
3125 md = m_getptr(m0, sizeof(struct ip6_hdr), &off);
3126 mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off);
3127 type = mld->mld_type;
3129 error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o,
3132 CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error);
3135 ICMP6STAT_INC(icp6s_outhist[type]);
3137 icmp6_ifstat_inc(oifp, ifs6_out_msg);
3139 case MLD_LISTENER_REPORT:
3140 case MLDV2_LISTENER_REPORT:
3141 icmp6_ifstat_inc(oifp, ifs6_out_mldreport);
3143 case MLD_LISTENER_DONE:
3144 icmp6_ifstat_inc(oifp, ifs6_out_mlddone);
3153 * Encapsulate an MLDv2 report.
3155 * KAME IPv6 requires that hop-by-hop options be passed separately,
3156 * and that the IPv6 header be prepended in a separate mbuf.
3158 * Returns a pointer to the new mbuf chain head, or NULL if the
3159 * allocation failed.
3161 static struct mbuf *
3162 mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m)
3165 struct mldv2_report *mld;
3166 struct ip6_hdr *ip6;
3167 struct in6_ifaddr *ia;
3170 KASSERT(ifp != NULL, ("%s: null ifp", __func__));
3171 KASSERT((m->m_flags & M_PKTHDR),
3172 ("%s: mbuf chain %p is !M_PKTHDR", __func__, m));
3175 * RFC3590: OK to send as :: or tentative during DAD.
3177 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
3179 CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__);
3181 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3184 ifa_free(&ia->ia_ifa);
3188 MH_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report));
3190 mldreclen = m_length(m, NULL);
3191 CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen);
3193 mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report);
3194 mh->m_pkthdr.len = sizeof(struct ip6_hdr) +
3195 sizeof(struct mldv2_report) + mldreclen;
3197 ip6 = mtod(mh, struct ip6_hdr *);
3199 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3200 ip6->ip6_vfc |= IPV6_VERSION;
3201 ip6->ip6_nxt = IPPROTO_ICMPV6;
3202 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
3204 ifa_free(&ia->ia_ifa);
3205 ip6->ip6_dst = in6addr_linklocal_allv2routers;
3206 /* scope ID will be set in netisr */
3208 mld = (struct mldv2_report *)(ip6 + 1);
3209 mld->mld_type = MLDV2_LISTENER_REPORT;
3212 mld->mld_v2_reserved = 0;
3213 mld->mld_v2_numrecs = htons(m->m_pkthdr.PH_vt.vt_nrecs);
3214 m->m_pkthdr.PH_vt.vt_nrecs = 0;
3217 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
3218 sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen);
3224 mld_rec_type_to_str(const int type)
3228 case MLD_CHANGE_TO_EXCLUDE_MODE:
3231 case MLD_CHANGE_TO_INCLUDE_MODE:
3234 case MLD_MODE_IS_EXCLUDE:
3237 case MLD_MODE_IS_INCLUDE:
3240 case MLD_ALLOW_NEW_SOURCES:
3243 case MLD_BLOCK_OLD_SOURCES:
3254 mld_init(void *unused __unused)
3257 CTR1(KTR_MLD, "%s: initializing", __func__);
3260 ip6_initpktopts(&mld_po);
3261 mld_po.ip6po_hlim = 1;
3262 mld_po.ip6po_hbh = &mld_ra.hbh;
3263 mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
3264 mld_po.ip6po_flags = IP6PO_DONTFRAG;
3266 SYSINIT(mld_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_init, NULL);
3269 mld_uninit(void *unused __unused)
3272 CTR1(KTR_MLD, "%s: tearing down", __func__);
3275 SYSUNINIT(mld_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_uninit, NULL);
3278 vnet_mld_init(const void *unused __unused)
3281 CTR1(KTR_MLD, "%s: initializing", __func__);
3283 LIST_INIT(&V_mli_head);
3285 VNET_SYSINIT(vnet_mld_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_init,
3289 vnet_mld_uninit(const void *unused __unused)
3292 CTR1(KTR_MLD, "%s: tearing down", __func__);
3294 KASSERT(LIST_EMPTY(&V_mli_head),
3295 ("%s: mli list not empty; ifnets not detached?", __func__));
3297 VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_uninit,
3301 mld_modevent(module_t mod, int type, void *unused __unused)
3309 return (EOPNOTSUPP);
3314 static moduledata_t mld_mod = {
3319 DECLARE_MODULE(mld, mld_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);