2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2009 Bruce Simpson.
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 3. The name of the author may not be used to endorse or promote
15 * products derived from this software without specific prior written
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * $KAME: mld6.c,v 1.27 2001/04/04 05:17:30 itojun Exp $
34 * Copyright (c) 1988 Stephen Deering.
35 * Copyright (c) 1992, 1993
36 * The Regents of the University of California. All rights reserved.
38 * This code is derived from software contributed to Berkeley by
39 * Stephen Deering of Stanford University.
41 * Redistribution and use in source and binary forms, with or without
42 * modification, are permitted provided that the following conditions
44 * 1. Redistributions of source code must retain the above copyright
45 * notice, this list of conditions and the following disclaimer.
46 * 2. Redistributions in binary form must reproduce the above copyright
47 * notice, this list of conditions and the following disclaimer in the
48 * documentation and/or other materials provided with the distribution.
49 * 3. Neither the name of the University nor the names of its contributors
50 * may be used to endorse or promote products derived from this software
51 * without specific prior written permission.
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * @(#)igmp.c 8.1 (Berkeley) 7/19/93
68 #include <sys/cdefs.h>
69 __FBSDID("$FreeBSD$");
72 #include "opt_inet6.h"
74 #include <sys/param.h>
75 #include <sys/systm.h>
77 #include <sys/socket.h>
78 #include <sys/sysctl.h>
79 #include <sys/kernel.h>
80 #include <sys/callout.h>
81 #include <sys/malloc.h>
82 #include <sys/module.h>
86 #include <net/if_var.h>
87 #include <net/if_private.h>
88 #include <net/route.h>
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet6/in6_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/scope6_var.h>
97 #include <netinet/icmp6.h>
98 #include <netinet6/mld6.h>
99 #include <netinet6/mld6_var.h>
101 #include <security/mac/mac_framework.h>
104 #define KTR_MLD KTR_INET6
107 static void mli_delete_locked(struct ifnet *);
108 static void mld_dispatch_packet(struct mbuf *);
109 static void mld_dispatch_queue(struct mbufq *, int);
110 static void mld_final_leave(struct in6_multi *, struct mld_ifsoftc *);
111 static void mld_fasttimo_vnet(struct in6_multi_head *inmh);
112 static int mld_handle_state_change(struct in6_multi *,
113 struct mld_ifsoftc *);
114 static int mld_initial_join(struct in6_multi *, struct mld_ifsoftc *,
117 static char * mld_rec_type_to_str(const int);
119 static void mld_set_version(struct mld_ifsoftc *, const int);
120 static void mld_slowtimo_vnet(void);
121 static int mld_v1_input_query(struct ifnet *, const struct ip6_hdr *,
122 /*const*/ struct mld_hdr *);
123 static int mld_v1_input_report(struct ifnet *, const struct ip6_hdr *,
124 /*const*/ struct mld_hdr *);
125 static void mld_v1_process_group_timer(struct in6_multi_head *,
127 static void mld_v1_process_querier_timers(struct mld_ifsoftc *);
128 static int mld_v1_transmit_report(struct in6_multi *, const int);
129 static void mld_v1_update_group(struct in6_multi *, const int);
130 static void mld_v2_cancel_link_timers(struct mld_ifsoftc *);
131 static void mld_v2_dispatch_general_query(struct mld_ifsoftc *);
133 mld_v2_encap_report(struct ifnet *, struct mbuf *);
134 static int mld_v2_enqueue_filter_change(struct mbufq *,
136 static int mld_v2_enqueue_group_record(struct mbufq *,
137 struct in6_multi *, const int, const int, const int,
139 static int mld_v2_input_query(struct ifnet *, const struct ip6_hdr *,
140 struct mbuf *, struct mldv2_query *, const int, const int);
141 static int mld_v2_merge_state_changes(struct in6_multi *,
143 static void mld_v2_process_group_timers(struct in6_multi_head *,
144 struct mbufq *, struct mbufq *,
145 struct in6_multi *, const int);
146 static int mld_v2_process_group_query(struct in6_multi *,
147 struct mld_ifsoftc *mli, int, struct mbuf *,
148 struct mldv2_query *, const int);
149 static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS);
150 static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS);
153 * Normative references: RFC 2710, RFC 3590, RFC 3810.
156 * * The MLD subsystem lock ends up being system-wide for the moment,
157 * but could be per-VIMAGE later on.
158 * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK.
159 * Any may be taken independently; if any are held at the same
160 * time, the above lock order must be followed.
161 * * IN6_MULTI_LOCK covers in_multi.
162 * * MLD_LOCK covers per-link state and any global variables in this file.
163 * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of
164 * per-link state iterators.
167 * A special case for IPv6 is the in6_setscope() routine. ip6_output()
168 * will not accept an ifp; it wants an embedded scope ID, unlike
169 * ip_output(), which happily takes the ifp given to it. The embedded
170 * scope ID is only used by MLD to select the outgoing interface.
172 * During interface attach and detach, MLD will take MLD_LOCK *after*
173 * the IF_AFDATA_LOCK.
174 * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call
175 * it with MLD_LOCK held without triggering an LOR. A netisr with indirect
176 * dispatch could work around this, but we'd rather not do that, as it
177 * can introduce other races.
179 * As such, we exploit the fact that the scope ID is just the interface
180 * index, and embed it in the IPv6 destination address accordingly.
181 * This is potentially NOT VALID for MLDv1 reports, as they
182 * are always sent to the multicast group itself; as MLDv2
183 * reports are always sent to ff02::16, this is not an issue
184 * when MLDv2 is in use.
186 * This does not however eliminate the LOR when ip6_output() itself
187 * calls in6_setscope() internally whilst MLD_LOCK is held. This will
188 * trigger a LOR warning in WITNESS when the ifnet is detached.
190 * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given
191 * how it's used across the network stack. Here we're simply exploiting
192 * the fact that MLD runs at a similar layer in the stack to scope6.c.
195 * * Each in6_multi corresponds to an ifp, and each ifp corresponds
196 * to a vnet in ifp->if_vnet.
198 static struct mtx mld_mtx;
199 static MALLOC_DEFINE(M_MLD, "mld", "mld state");
201 #define MLD_EMBEDSCOPE(pin6, zoneid) \
202 if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \
203 IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \
204 (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \
207 * VIMAGE-wide globals.
209 VNET_DEFINE_STATIC(struct timeval, mld_gsrdelay) = {10, 0};
210 VNET_DEFINE_STATIC(LIST_HEAD(, mld_ifsoftc), mli_head);
211 VNET_DEFINE_STATIC(int, interface_timers_running6);
212 VNET_DEFINE_STATIC(int, state_change_timers_running6);
213 VNET_DEFINE_STATIC(int, current_state_timers_running6);
215 #define V_mld_gsrdelay VNET(mld_gsrdelay)
216 #define V_mli_head VNET(mli_head)
217 #define V_interface_timers_running6 VNET(interface_timers_running6)
218 #define V_state_change_timers_running6 VNET(state_change_timers_running6)
219 #define V_current_state_timers_running6 VNET(current_state_timers_running6)
221 SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */
223 SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
224 "IPv6 Multicast Listener Discovery");
227 * Virtualized sysctls.
229 SYSCTL_PROC(_net_inet6_mld, OID_AUTO, gsrdelay,
230 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
231 &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I",
232 "Rate limit for MLDv2 Group-and-Source queries in seconds");
235 * Non-virtualized sysctls.
237 static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo,
238 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo,
239 "Per-interface MLDv2 state");
241 static int mld_v1enable = 1;
242 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RWTUN,
243 &mld_v1enable, 0, "Enable fallback to MLDv1");
245 static int mld_v2enable = 1;
246 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v2enable, CTLFLAG_RWTUN,
247 &mld_v2enable, 0, "Enable MLDv2");
249 static int mld_use_allow = 1;
250 SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RWTUN,
251 &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves");
254 * Packed Router Alert option structure declaration.
259 struct ip6_opt_router ra;
263 * Router Alert hop-by-hop option header.
265 static struct mld_raopt mld_ra = {
267 .pad = { .ip6o_type = IP6OPT_PADN, 0 },
269 .ip6or_type = IP6OPT_ROUTER_ALERT,
270 .ip6or_len = IP6OPT_RTALERT_LEN - 2,
271 .ip6or_value[0] = ((IP6OPT_RTALERT_MLD >> 8) & 0xFF),
272 .ip6or_value[1] = (IP6OPT_RTALERT_MLD & 0xFF)
275 static struct ip6_pktopts mld_po;
278 mld_save_context(struct mbuf *m, struct ifnet *ifp)
282 m->m_pkthdr.PH_loc.ptr = ifp->if_vnet;
284 m->m_pkthdr.rcvif = ifp;
285 m->m_pkthdr.flowid = ifp->if_index;
289 mld_scrub_context(struct mbuf *m)
292 m->m_pkthdr.PH_loc.ptr = NULL;
293 m->m_pkthdr.flowid = 0;
297 * Restore context from a queued output chain.
298 * Return saved ifindex.
300 * VIMAGE: The assertion is there to make sure that we
301 * actually called CURVNET_SET() with what's in the mbuf chain.
303 static __inline uint32_t
304 mld_restore_context(struct mbuf *m)
307 #if defined(VIMAGE) && defined(INVARIANTS)
308 KASSERT(curvnet == m->m_pkthdr.PH_loc.ptr,
309 ("%s: called when curvnet was not restored: cuvnet %p m ptr %p",
310 __func__, curvnet, m->m_pkthdr.PH_loc.ptr));
312 return (m->m_pkthdr.flowid);
316 * Retrieve or set threshold between group-source queries in seconds.
318 * VIMAGE: Assume curvnet set by caller.
319 * SMPng: NOTE: Serialized by MLD lock.
322 sysctl_mld_gsr(SYSCTL_HANDLER_ARGS)
327 error = sysctl_wire_old_buffer(req, sizeof(int));
333 i = V_mld_gsrdelay.tv_sec;
335 error = sysctl_handle_int(oidp, &i, 0, req);
336 if (error || !req->newptr)
339 if (i < -1 || i >= 60) {
344 CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d",
345 V_mld_gsrdelay.tv_sec, i);
346 V_mld_gsrdelay.tv_sec = i;
354 * Expose struct mld_ifsoftc to userland, keyed by ifindex.
355 * For use by ifmcstat(8).
357 * VIMAGE: Assume curvnet set by caller. The node handler itself
358 * is not directly virtualized.
361 sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS)
363 struct epoch_tracker et;
368 struct mld_ifsoftc *mli;
373 if (req->newptr != NULL)
379 error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo));
384 IN6_MULTI_LIST_LOCK();
389 ifp = ifnet_byindex(name[0]);
393 LIST_FOREACH(mli, &V_mli_head, mli_link) {
394 if (ifp == mli->mli_ifp) {
395 struct mld_ifinfo info;
397 info.mli_version = mli->mli_version;
398 info.mli_v1_timer = mli->mli_v1_timer;
399 info.mli_v2_timer = mli->mli_v2_timer;
400 info.mli_flags = mli->mli_flags;
401 info.mli_rv = mli->mli_rv;
402 info.mli_qi = mli->mli_qi;
403 info.mli_qri = mli->mli_qri;
404 info.mli_uri = mli->mli_uri;
405 error = SYSCTL_OUT(req, &info, sizeof(info));
413 IN6_MULTI_LIST_UNLOCK();
419 * Dispatch an entire queue of pending packet chains.
420 * VIMAGE: Assumes the vnet pointer has been set.
423 mld_dispatch_queue(struct mbufq *mq, int limit)
427 while ((m = mbufq_dequeue(mq)) != NULL) {
428 CTR3(KTR_MLD, "%s: dispatch %p from %p", __func__, mq, m);
429 mld_dispatch_packet(m);
436 * Filter outgoing MLD report state by group.
438 * Reports are ALWAYS suppressed for ALL-HOSTS (ff02::1)
439 * and node-local addresses. However, kernel and socket consumers
440 * always embed the KAME scope ID in the address provided, so strip it
441 * when performing comparison.
442 * Note: This is not the same as the *multicast* scope.
444 * Return zero if the given group is one for which MLD reports
445 * should be suppressed, or non-zero if reports should be issued.
448 mld_is_addr_reported(const struct in6_addr *addr)
451 KASSERT(IN6_IS_ADDR_MULTICAST(addr), ("%s: not multicast", __func__));
453 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_NODELOCAL)
456 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_LINKLOCAL) {
457 struct in6_addr tmp = *addr;
458 in6_clearscope(&tmp);
459 if (IN6_ARE_ADDR_EQUAL(&tmp, &in6addr_linklocal_allnodes))
467 * Attach MLD when PF_INET6 is attached to an interface. Assumes that the
468 * current VNET is set by the caller.
471 mld_domifattach(struct ifnet *ifp)
473 struct mld_ifsoftc *mli;
475 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp, if_name(ifp));
477 mli = malloc(sizeof(struct mld_ifsoftc), M_MLD, M_WAITOK | M_ZERO);
479 mli->mli_version = MLD_VERSION_2;
481 mli->mli_rv = MLD_RV_INIT;
482 mli->mli_qi = MLD_QI_INIT;
483 mli->mli_qri = MLD_QRI_INIT;
484 mli->mli_uri = MLD_URI_INIT;
485 mbufq_init(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS);
486 if ((ifp->if_flags & IFF_MULTICAST) == 0)
487 mli->mli_flags |= MLIF_SILENT;
489 mli->mli_flags |= MLIF_USEALLOW;
492 LIST_INSERT_HEAD(&V_mli_head, mli, mli_link);
501 * NOTE: Some finalization tasks need to run before the protocol domain
502 * is detached, but also before the link layer does its cleanup.
503 * Run before link-layer cleanup; cleanup groups, but do not free MLD state.
505 * SMPng: Caller must hold IN6_MULTI_LOCK().
506 * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator.
507 * XXX This routine is also bitten by unlocked ifma_protospec access.
510 mld_ifdetach(struct ifnet *ifp, struct in6_multi_head *inmh)
512 struct epoch_tracker et;
513 struct mld_ifsoftc *mli;
514 struct ifmultiaddr *ifma;
515 struct in6_multi *inm;
517 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp,
520 IN6_MULTI_LIST_LOCK_ASSERT();
523 mli = MLD_IFINFO(ifp);
526 * Extract list of in6_multi associated with the detaching ifp
527 * which the PF_INET6 layer is about to release.
530 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
531 inm = in6m_ifmultiaddr_get_inm(ifma);
534 in6m_disconnect_locked(inmh, inm);
536 if (mli->mli_version == MLD_VERSION_2) {
537 in6m_clear_recorded(inm);
540 * We need to release the final reference held
541 * for issuing the INCLUDE {}.
543 if (inm->in6m_state == MLD_LEAVING_MEMBER) {
544 inm->in6m_state = MLD_NOT_MEMBER;
545 in6m_rele_locked(inmh, inm);
550 IF_ADDR_WUNLOCK(ifp);
555 * Hook for domifdetach.
556 * Runs after link-layer cleanup; free MLD state.
558 * SMPng: Normally called with IF_AFDATA_LOCK held.
561 mld_domifdetach(struct ifnet *ifp)
564 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
565 __func__, ifp, if_name(ifp));
568 mli_delete_locked(ifp);
573 mli_delete_locked(struct ifnet *ifp)
575 struct mld_ifsoftc *mli, *tmli;
577 CTR3(KTR_MLD, "%s: freeing mld_ifsoftc for ifp %p(%s)",
578 __func__, ifp, if_name(ifp));
582 LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) {
583 if (mli->mli_ifp == ifp) {
585 * Free deferred General Query responses.
587 mbufq_drain(&mli->mli_gq);
589 LIST_REMOVE(mli, mli_link);
598 * Process a received MLDv1 general or address-specific query.
599 * Assumes that the query header has been pulled up to sizeof(mld_hdr).
601 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
602 * mld_addr. This is OK as we own the mbuf chain.
605 mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
606 /*const*/ struct mld_hdr *mld)
608 struct ifmultiaddr *ifma;
609 struct mld_ifsoftc *mli;
610 struct in6_multi *inm;
611 int is_general_query;
614 char ip6tbuf[INET6_ADDRSTRLEN];
619 is_general_query = 0;
622 CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)",
623 ip6_sprintf(ip6tbuf, &mld->mld_addr),
629 * RFC3810 Section 6.2: MLD queries must originate from
630 * a router's link-local address.
632 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
633 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
634 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
640 * Do address field validation upfront before we accept
643 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
645 * MLDv1 General Query.
646 * If this was not sent to the all-nodes group, ignore it.
651 in6_clearscope(&dst);
652 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes))
654 is_general_query = 1;
657 * Embed scope ID of receiving interface in MLD query for
658 * lookup whilst we don't hold other locks.
660 in6_setscope(&mld->mld_addr, ifp, NULL);
663 IN6_MULTI_LIST_LOCK();
667 * Switch to MLDv1 host compatibility mode.
669 mli = MLD_IFINFO(ifp);
670 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
671 mld_set_version(mli, MLD_VERSION_1);
673 timer = (ntohs(mld->mld_maxdelay) * MLD_FASTHZ) / MLD_TIMER_SCALE;
677 if (is_general_query) {
679 * For each reporting group joined on this
680 * interface, kick the report timer.
682 CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)",
684 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
685 inm = in6m_ifmultiaddr_get_inm(ifma);
688 mld_v1_update_group(inm, timer);
692 * MLDv1 Group-Specific Query.
693 * If this is a group-specific MLDv1 query, we need only
694 * look up the single group to process it.
696 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
698 CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)",
699 ip6_sprintf(ip6tbuf, &mld->mld_addr),
701 mld_v1_update_group(inm, timer);
703 /* XXX Clear embedded scope ID as userland won't expect it. */
704 in6_clearscope(&mld->mld_addr);
708 IN6_MULTI_LIST_UNLOCK();
714 * Update the report timer on a group in response to an MLDv1 query.
716 * If we are becoming the reporting member for this group, start the timer.
717 * If we already are the reporting member for this group, and timer is
718 * below the threshold, reset it.
720 * We may be updating the group for the first time since we switched
721 * to MLDv2. If we are, then we must clear any recorded source lists,
722 * and transition to REPORTING state; the group timer is overloaded
723 * for group and group-source query responses.
725 * Unlike MLDv2, the delay per group should be jittered
726 * to avoid bursts of MLDv1 reports.
729 mld_v1_update_group(struct in6_multi *inm, const int timer)
732 char ip6tbuf[INET6_ADDRSTRLEN];
735 CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__,
736 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
737 if_name(inm->in6m_ifp), timer);
739 IN6_MULTI_LIST_LOCK_ASSERT();
741 switch (inm->in6m_state) {
743 case MLD_SILENT_MEMBER:
745 case MLD_REPORTING_MEMBER:
746 if (inm->in6m_timer != 0 &&
747 inm->in6m_timer <= timer) {
748 CTR1(KTR_MLD, "%s: REPORTING and timer running, "
749 "skipping.", __func__);
753 case MLD_SG_QUERY_PENDING_MEMBER:
754 case MLD_G_QUERY_PENDING_MEMBER:
755 case MLD_IDLE_MEMBER:
756 case MLD_LAZY_MEMBER:
757 case MLD_AWAKENING_MEMBER:
758 CTR1(KTR_MLD, "%s: ->REPORTING", __func__);
759 inm->in6m_state = MLD_REPORTING_MEMBER;
760 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
761 V_current_state_timers_running6 = 1;
763 case MLD_SLEEPING_MEMBER:
764 CTR1(KTR_MLD, "%s: ->AWAKENING", __func__);
765 inm->in6m_state = MLD_AWAKENING_MEMBER;
767 case MLD_LEAVING_MEMBER:
773 * Process a received MLDv2 general, group-specific or
774 * group-and-source-specific query.
776 * Assumes that mld points to a struct mldv2_query which is stored in
779 * Return 0 if successful, otherwise an appropriate error code is returned.
782 mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
783 struct mbuf *m, struct mldv2_query *mld, const int off, const int icmp6len)
785 struct mld_ifsoftc *mli;
786 struct in6_multi *inm;
787 uint32_t maxdelay, nsrc, qqi;
788 int is_general_query;
792 char ip6tbuf[INET6_ADDRSTRLEN];
798 CTR3(KTR_MLD, "ignore v2 query src %s on ifp %p(%s)",
799 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
805 * RFC3810 Section 6.2: MLD queries must originate from
806 * a router's link-local address.
808 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
809 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
810 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
815 is_general_query = 0;
817 CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, if_name(ifp));
819 maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */
820 if (maxdelay >= 32768) {
821 maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) <<
822 (MLD_MRC_EXP(maxdelay) + 3);
824 timer = (maxdelay * MLD_FASTHZ) / MLD_TIMER_SCALE;
828 qrv = MLD_QRV(mld->mld_misc);
830 CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__,
837 qqi = MLD_QQIC_MANT(mld->mld_qqi) <<
838 (MLD_QQIC_EXP(mld->mld_qqi) + 3);
841 nsrc = ntohs(mld->mld_numsrc);
842 if (nsrc > MLD_MAX_GS_SOURCES)
844 if (icmp6len < sizeof(struct mldv2_query) +
845 (nsrc * sizeof(struct in6_addr)))
849 * Do further input validation upfront to avoid resetting timers
850 * should we need to discard this query.
852 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
854 * A general query with a source list has undefined
855 * behaviour; discard it.
859 is_general_query = 1;
862 * Embed scope ID of receiving interface in MLD query for
863 * lookup whilst we don't hold other locks (due to KAME
864 * locking lameness). We own this mbuf chain just now.
866 in6_setscope(&mld->mld_addr, ifp, NULL);
869 IN6_MULTI_LIST_LOCK();
872 mli = MLD_IFINFO(ifp);
873 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
876 * Discard the v2 query if we're in Compatibility Mode.
877 * The RFC is pretty clear that hosts need to stay in MLDv1 mode
878 * until the Old Version Querier Present timer expires.
880 if (mli->mli_version != MLD_VERSION_2)
883 mld_set_version(mli, MLD_VERSION_2);
886 mli->mli_qri = maxdelay;
888 CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi,
891 if (is_general_query) {
893 * MLDv2 General Query.
895 * Schedule a current-state report on this ifp for
896 * all groups, possibly containing source lists.
898 * If there is a pending General Query response
899 * scheduled earlier than the selected delay, do
900 * not schedule any other reports.
901 * Otherwise, reset the interface timer.
903 CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)",
905 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) {
906 mli->mli_v2_timer = MLD_RANDOM_DELAY(timer);
907 V_interface_timers_running6 = 1;
911 * MLDv2 Group-specific or Group-and-source-specific Query.
913 * Group-source-specific queries are throttled on
914 * a per-group basis to defeat denial-of-service attempts.
915 * Queries for groups we are not a member of on this
916 * link are simply ignored.
918 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
922 if (!ratecheck(&inm->in6m_lastgsrtv,
924 CTR1(KTR_MLD, "%s: GS query throttled.",
929 CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)",
932 * If there is a pending General Query response
933 * scheduled sooner than the selected delay, no
934 * further report need be scheduled.
935 * Otherwise, prepare to respond to the
936 * group-specific or group-and-source query.
938 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer)
939 mld_v2_process_group_query(inm, mli, timer, m, mld, off);
941 /* XXX Clear embedded scope ID as userland won't expect it. */
942 in6_clearscope(&mld->mld_addr);
947 IN6_MULTI_LIST_UNLOCK();
953 * Process a received MLDv2 group-specific or group-and-source-specific
955 * Return <0 if any error occurred. Currently this is ignored.
958 mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifsoftc *mli,
959 int timer, struct mbuf *m0, struct mldv2_query *mld, const int off)
964 IN6_MULTI_LIST_LOCK_ASSERT();
969 switch (inm->in6m_state) {
971 case MLD_SILENT_MEMBER:
972 case MLD_SLEEPING_MEMBER:
973 case MLD_LAZY_MEMBER:
974 case MLD_AWAKENING_MEMBER:
975 case MLD_IDLE_MEMBER:
976 case MLD_LEAVING_MEMBER:
979 case MLD_REPORTING_MEMBER:
980 case MLD_G_QUERY_PENDING_MEMBER:
981 case MLD_SG_QUERY_PENDING_MEMBER:
985 nsrc = ntohs(mld->mld_numsrc);
987 /* Length should be checked by calling function. */
988 KASSERT((m0->m_flags & M_PKTHDR) == 0 ||
989 m0->m_pkthdr.len >= off + sizeof(struct mldv2_query) +
990 nsrc * sizeof(struct in6_addr),
991 ("mldv2 packet is too short: (%d bytes < %zd bytes, m=%p)",
992 m0->m_pkthdr.len, off + sizeof(struct mldv2_query) +
993 nsrc * sizeof(struct in6_addr), m0));
996 * Deal with group-specific queries upfront.
997 * If any group query is already pending, purge any recorded
998 * source-list state if it exists, and schedule a query response
999 * for this group-specific query.
1002 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
1003 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) {
1004 in6m_clear_recorded(inm);
1005 timer = min(inm->in6m_timer, timer);
1007 inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER;
1008 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1009 V_current_state_timers_running6 = 1;
1014 * Deal with the case where a group-and-source-specific query has
1015 * been received but a group-specific query is already pending.
1017 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) {
1018 timer = min(inm->in6m_timer, timer);
1019 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1020 V_current_state_timers_running6 = 1;
1025 * Finally, deal with the case where a group-and-source-specific
1026 * query has been received, where a response to a previous g-s-r
1027 * query exists, or none exists.
1028 * In this case, we need to parse the source-list which the Querier
1029 * has provided us with and check if we have any source list filter
1030 * entries at T1 for these sources. If we do not, there is no need
1031 * schedule a report and the query may be dropped.
1032 * If we do, we must record them and schedule a current-state
1033 * report for those sources.
1035 if (inm->in6m_nsrc > 0) {
1036 struct in6_addr srcaddr;
1040 soff = off + sizeof(struct mldv2_query);
1042 for (i = 0; i < nsrc; i++) {
1043 m_copydata(m0, soff, sizeof(struct in6_addr),
1045 retval = in6m_record_source(inm, &srcaddr);
1048 nrecorded += retval;
1049 soff += sizeof(struct in6_addr);
1051 if (nrecorded > 0) {
1053 "%s: schedule response to SG query", __func__);
1054 inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER;
1055 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1056 V_current_state_timers_running6 = 1;
1064 * Process a received MLDv1 host membership report.
1065 * Assumes mld points to mld_hdr in pulled up mbuf chain.
1067 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
1068 * mld_addr. This is OK as we own the mbuf chain.
1071 mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6,
1072 /*const*/ struct mld_hdr *mld)
1074 struct in6_addr src, dst;
1075 struct in6_ifaddr *ia;
1076 struct in6_multi *inm;
1078 char ip6tbuf[INET6_ADDRSTRLEN];
1083 if (!mld_v1enable) {
1084 CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)",
1085 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1090 if (ifp->if_flags & IFF_LOOPBACK)
1094 * MLDv1 reports must originate from a host's link-local address,
1095 * or the unspecified address (when booting).
1098 in6_clearscope(&src);
1099 if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) {
1100 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
1101 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
1107 * RFC2710 Section 4: MLDv1 reports must pertain to a multicast
1108 * group, and must be directed to the group itself.
1111 in6_clearscope(&dst);
1112 if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) ||
1113 !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) {
1114 CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)",
1115 ip6_sprintf(ip6tbuf, &ip6->ip6_dst),
1121 * Make sure we don't hear our own membership report, as fast
1122 * leave requires knowing that we are the only member of a
1123 * group. Assume we used the link-local address if available,
1124 * otherwise look for ::.
1126 * XXX Note that scope ID comparison is needed for the address
1127 * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be
1128 * performed for the on-wire address.
1130 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1131 if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) ||
1132 (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) {
1134 ifa_free(&ia->ia_ifa);
1138 ifa_free(&ia->ia_ifa);
1140 CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)",
1141 ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp));
1144 * Embed scope ID of receiving interface in MLD query for lookup
1145 * whilst we don't hold other locks (due to KAME locking lameness).
1147 if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr))
1148 in6_setscope(&mld->mld_addr, ifp, NULL);
1150 IN6_MULTI_LIST_LOCK();
1154 * MLDv1 report suppression.
1155 * If we are a member of this group, and our membership should be
1156 * reported, and our group timer is pending or about to be reset,
1157 * stop our group timer by transitioning to the 'lazy' state.
1159 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
1161 struct mld_ifsoftc *mli;
1163 mli = inm->in6m_mli;
1164 KASSERT(mli != NULL,
1165 ("%s: no mli for ifp %p", __func__, ifp));
1168 * If we are in MLDv2 host mode, do not allow the
1169 * other host's MLDv1 report to suppress our reports.
1171 if (mli->mli_version == MLD_VERSION_2)
1174 inm->in6m_timer = 0;
1176 switch (inm->in6m_state) {
1177 case MLD_NOT_MEMBER:
1178 case MLD_SILENT_MEMBER:
1179 case MLD_SLEEPING_MEMBER:
1181 case MLD_REPORTING_MEMBER:
1182 case MLD_IDLE_MEMBER:
1183 case MLD_AWAKENING_MEMBER:
1185 "report suppressed for %s on ifp %p(%s)",
1186 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1188 case MLD_LAZY_MEMBER:
1189 inm->in6m_state = MLD_LAZY_MEMBER;
1191 case MLD_G_QUERY_PENDING_MEMBER:
1192 case MLD_SG_QUERY_PENDING_MEMBER:
1193 case MLD_LEAVING_MEMBER:
1200 IN6_MULTI_LIST_UNLOCK();
1202 /* XXX Clear embedded scope ID as userland won't expect it. */
1203 in6_clearscope(&mld->mld_addr);
1211 * Assume query messages which fit in a single ICMPv6 message header
1212 * have been pulled up.
1213 * Assume that userland will want to see the message, even if it
1214 * otherwise fails kernel input validation; do not free it.
1215 * Pullup may however free the mbuf chain m if it fails.
1217 * Return IPPROTO_DONE if we freed m. Otherwise, return 0.
1220 mld_input(struct mbuf **mp, int off, int icmp6len)
1223 struct ip6_hdr *ip6;
1225 struct mld_hdr *mld;
1229 CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off);
1231 ifp = m->m_pkthdr.rcvif;
1233 /* Pullup to appropriate size. */
1234 if (m->m_len < off + sizeof(*mld)) {
1235 m = m_pullup(m, off + sizeof(*mld));
1237 ICMP6STAT_INC(icp6s_badlen);
1238 return (IPPROTO_DONE);
1241 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1242 if (mld->mld_type == MLD_LISTENER_QUERY &&
1243 icmp6len >= sizeof(struct mldv2_query)) {
1244 mldlen = sizeof(struct mldv2_query);
1246 mldlen = sizeof(struct mld_hdr);
1248 if (m->m_len < off + mldlen) {
1249 m = m_pullup(m, off + mldlen);
1251 ICMP6STAT_INC(icp6s_badlen);
1252 return (IPPROTO_DONE);
1256 ip6 = mtod(m, struct ip6_hdr *);
1257 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1260 * Userland needs to see all of this traffic for implementing
1261 * the endpoint discovery portion of multicast routing.
1263 switch (mld->mld_type) {
1264 case MLD_LISTENER_QUERY:
1265 icmp6_ifstat_inc(ifp, ifs6_in_mldquery);
1266 if (icmp6len == sizeof(struct mld_hdr)) {
1267 if (mld_v1_input_query(ifp, ip6, mld) != 0)
1269 } else if (icmp6len >= sizeof(struct mldv2_query)) {
1270 if (mld_v2_input_query(ifp, ip6, m,
1271 (struct mldv2_query *)mld, off, icmp6len) != 0)
1275 case MLD_LISTENER_REPORT:
1276 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1277 if (mld_v1_input_report(ifp, ip6, mld) != 0)
1280 case MLDV2_LISTENER_REPORT:
1281 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1283 case MLD_LISTENER_DONE:
1284 icmp6_ifstat_inc(ifp, ifs6_in_mlddone);
1294 * Fast timeout handler (global).
1295 * VIMAGE: Timeout handlers are expected to service all vimages.
1297 static struct callout mldfast_callout;
1299 mld_fasttimo(void *arg __unused)
1301 struct epoch_tracker et;
1302 struct in6_multi_head inmh;
1303 VNET_ITERATOR_DECL(vnet_iter);
1307 NET_EPOCH_ENTER(et);
1308 VNET_LIST_RLOCK_NOSLEEP();
1309 VNET_FOREACH(vnet_iter) {
1310 CURVNET_SET(vnet_iter);
1311 mld_fasttimo_vnet(&inmh);
1314 VNET_LIST_RUNLOCK_NOSLEEP();
1316 in6m_release_list_deferred(&inmh);
1318 callout_reset(&mldfast_callout, hz / MLD_FASTHZ, mld_fasttimo, NULL);
1322 * Fast timeout handler (per-vnet).
1324 * VIMAGE: Assume caller has set up our curvnet.
1327 mld_fasttimo_vnet(struct in6_multi_head *inmh)
1329 struct mbufq scq; /* State-change packets */
1330 struct mbufq qrq; /* Query response packets */
1332 struct mld_ifsoftc *mli;
1333 struct ifmultiaddr *ifma;
1334 struct in6_multi *inm;
1340 * Quick check to see if any work needs to be done, in order to
1341 * minimize the overhead of fasttimo processing.
1342 * SMPng: XXX Unlocked reads.
1344 if (!V_current_state_timers_running6 &&
1345 !V_interface_timers_running6 &&
1346 !V_state_change_timers_running6)
1349 IN6_MULTI_LIST_LOCK();
1353 * MLDv2 General Query response timer processing.
1355 if (V_interface_timers_running6) {
1356 CTR1(KTR_MLD, "%s: interface timers running", __func__);
1358 V_interface_timers_running6 = 0;
1359 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1360 if (mli->mli_v2_timer == 0) {
1362 } else if (--mli->mli_v2_timer == 0) {
1363 mld_v2_dispatch_general_query(mli);
1365 V_interface_timers_running6 = 1;
1370 if (!V_current_state_timers_running6 &&
1371 !V_state_change_timers_running6)
1374 V_current_state_timers_running6 = 0;
1375 V_state_change_timers_running6 = 0;
1377 CTR1(KTR_MLD, "%s: state change timers running", __func__);
1380 * MLD host report and state-change timer processing.
1381 * Note: Processing a v2 group timer may remove a node.
1383 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1386 if (mli->mli_version == MLD_VERSION_2) {
1387 uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri *
1389 mbufq_init(&qrq, MLD_MAX_G_GS_PACKETS);
1390 mbufq_init(&scq, MLD_MAX_STATE_CHANGE_PACKETS);
1394 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1395 inm = in6m_ifmultiaddr_get_inm(ifma);
1398 switch (mli->mli_version) {
1400 mld_v1_process_group_timer(inmh, inm);
1403 mld_v2_process_group_timers(inmh, &qrq,
1404 &scq, inm, uri_fasthz);
1408 IF_ADDR_WUNLOCK(ifp);
1410 switch (mli->mli_version) {
1413 * Transmit reports for this lifecycle. This
1414 * is done while not holding IF_ADDR_LOCK
1415 * since this can call
1416 * in6ifa_ifpforlinklocal() which locks
1417 * IF_ADDR_LOCK internally as well as
1418 * ip6_output() to transmit a packet.
1420 while ((inm = SLIST_FIRST(inmh)) != NULL) {
1421 SLIST_REMOVE_HEAD(inmh, in6m_defer);
1422 (void)mld_v1_transmit_report(inm,
1423 MLD_LISTENER_REPORT);
1427 mld_dispatch_queue(&qrq, 0);
1428 mld_dispatch_queue(&scq, 0);
1435 IN6_MULTI_LIST_UNLOCK();
1439 * Update host report group timer.
1440 * Will update the global pending timer flags.
1443 mld_v1_process_group_timer(struct in6_multi_head *inmh, struct in6_multi *inm)
1445 int report_timer_expired;
1447 IN6_MULTI_LIST_LOCK_ASSERT();
1450 if (inm->in6m_timer == 0) {
1451 report_timer_expired = 0;
1452 } else if (--inm->in6m_timer == 0) {
1453 report_timer_expired = 1;
1455 V_current_state_timers_running6 = 1;
1459 switch (inm->in6m_state) {
1460 case MLD_NOT_MEMBER:
1461 case MLD_SILENT_MEMBER:
1462 case MLD_IDLE_MEMBER:
1463 case MLD_LAZY_MEMBER:
1464 case MLD_SLEEPING_MEMBER:
1465 case MLD_AWAKENING_MEMBER:
1467 case MLD_REPORTING_MEMBER:
1468 if (report_timer_expired) {
1469 inm->in6m_state = MLD_IDLE_MEMBER;
1470 SLIST_INSERT_HEAD(inmh, inm, in6m_defer);
1473 case MLD_G_QUERY_PENDING_MEMBER:
1474 case MLD_SG_QUERY_PENDING_MEMBER:
1475 case MLD_LEAVING_MEMBER:
1481 * Update a group's timers for MLDv2.
1482 * Will update the global pending timer flags.
1483 * Note: Unlocked read from mli.
1486 mld_v2_process_group_timers(struct in6_multi_head *inmh,
1487 struct mbufq *qrq, struct mbufq *scq,
1488 struct in6_multi *inm, const int uri_fasthz)
1490 int query_response_timer_expired;
1491 int state_change_retransmit_timer_expired;
1493 char ip6tbuf[INET6_ADDRSTRLEN];
1496 IN6_MULTI_LIST_LOCK_ASSERT();
1499 query_response_timer_expired = 0;
1500 state_change_retransmit_timer_expired = 0;
1503 * During a transition from compatibility mode back to MLDv2,
1504 * a group record in REPORTING state may still have its group
1505 * timer active. This is a no-op in this function; it is easier
1506 * to deal with it here than to complicate the slow-timeout path.
1508 if (inm->in6m_timer == 0) {
1509 query_response_timer_expired = 0;
1510 } else if (--inm->in6m_timer == 0) {
1511 query_response_timer_expired = 1;
1513 V_current_state_timers_running6 = 1;
1516 if (inm->in6m_sctimer == 0) {
1517 state_change_retransmit_timer_expired = 0;
1518 } else if (--inm->in6m_sctimer == 0) {
1519 state_change_retransmit_timer_expired = 1;
1521 V_state_change_timers_running6 = 1;
1524 /* We are in fasttimo, so be quick about it. */
1525 if (!state_change_retransmit_timer_expired &&
1526 !query_response_timer_expired)
1529 switch (inm->in6m_state) {
1530 case MLD_NOT_MEMBER:
1531 case MLD_SILENT_MEMBER:
1532 case MLD_SLEEPING_MEMBER:
1533 case MLD_LAZY_MEMBER:
1534 case MLD_AWAKENING_MEMBER:
1535 case MLD_IDLE_MEMBER:
1537 case MLD_G_QUERY_PENDING_MEMBER:
1538 case MLD_SG_QUERY_PENDING_MEMBER:
1540 * Respond to a previously pending Group-Specific
1541 * or Group-and-Source-Specific query by enqueueing
1542 * the appropriate Current-State report for
1543 * immediate transmission.
1545 if (query_response_timer_expired) {
1546 int retval __unused;
1548 retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1,
1549 (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER),
1551 CTR2(KTR_MLD, "%s: enqueue record = %d",
1553 inm->in6m_state = MLD_REPORTING_MEMBER;
1554 in6m_clear_recorded(inm);
1557 case MLD_REPORTING_MEMBER:
1558 case MLD_LEAVING_MEMBER:
1559 if (state_change_retransmit_timer_expired) {
1561 * State-change retransmission timer fired.
1562 * If there are any further pending retransmissions,
1563 * set the global pending state-change flag, and
1566 if (--inm->in6m_scrv > 0) {
1567 inm->in6m_sctimer = uri_fasthz;
1568 V_state_change_timers_running6 = 1;
1571 * Retransmit the previously computed state-change
1572 * report. If there are no further pending
1573 * retransmissions, the mbuf queue will be consumed.
1574 * Update T0 state to T1 as we have now sent
1577 (void)mld_v2_merge_state_changes(inm, scq);
1580 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
1581 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1582 if_name(inm->in6m_ifp));
1585 * If we are leaving the group for good, make sure
1586 * we release MLD's reference to it.
1587 * This release must be deferred using a SLIST,
1588 * as we are called from a loop which traverses
1589 * the in_ifmultiaddr TAILQ.
1591 if (inm->in6m_state == MLD_LEAVING_MEMBER &&
1592 inm->in6m_scrv == 0) {
1593 inm->in6m_state = MLD_NOT_MEMBER;
1594 in6m_disconnect_locked(inmh, inm);
1595 in6m_rele_locked(inmh, inm);
1603 * Switch to a different version on the given interface,
1604 * as per Section 9.12.
1607 mld_set_version(struct mld_ifsoftc *mli, const int version)
1609 int old_version_timer;
1613 CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__,
1614 version, mli->mli_ifp, if_name(mli->mli_ifp));
1616 if (version == MLD_VERSION_1) {
1618 * Compute the "Older Version Querier Present" timer as per
1621 old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri;
1622 old_version_timer *= MLD_SLOWHZ;
1623 mli->mli_v1_timer = old_version_timer;
1626 if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) {
1627 mli->mli_version = MLD_VERSION_1;
1628 mld_v2_cancel_link_timers(mli);
1633 * Cancel pending MLDv2 timers for the given link and all groups
1634 * joined on it; state-change, general-query, and group-query timers.
1637 mld_v2_cancel_link_timers(struct mld_ifsoftc *mli)
1639 struct epoch_tracker et;
1640 struct in6_multi_head inmh;
1641 struct ifmultiaddr *ifma;
1643 struct in6_multi *inm;
1645 CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__,
1646 mli->mli_ifp, if_name(mli->mli_ifp));
1649 IN6_MULTI_LIST_LOCK_ASSERT();
1653 * Fast-track this potentially expensive operation
1654 * by checking all the global 'timer pending' flags.
1656 if (!V_interface_timers_running6 &&
1657 !V_state_change_timers_running6 &&
1658 !V_current_state_timers_running6)
1661 mli->mli_v2_timer = 0;
1666 NET_EPOCH_ENTER(et);
1667 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1668 inm = in6m_ifmultiaddr_get_inm(ifma);
1671 switch (inm->in6m_state) {
1672 case MLD_NOT_MEMBER:
1673 case MLD_SILENT_MEMBER:
1674 case MLD_IDLE_MEMBER:
1675 case MLD_LAZY_MEMBER:
1676 case MLD_SLEEPING_MEMBER:
1677 case MLD_AWAKENING_MEMBER:
1679 case MLD_LEAVING_MEMBER:
1681 * If we are leaving the group and switching
1682 * version, we need to release the final
1683 * reference held for issuing the INCLUDE {}.
1685 if (inm->in6m_refcount == 1)
1686 in6m_disconnect_locked(&inmh, inm);
1687 in6m_rele_locked(&inmh, inm);
1689 case MLD_G_QUERY_PENDING_MEMBER:
1690 case MLD_SG_QUERY_PENDING_MEMBER:
1691 in6m_clear_recorded(inm);
1693 case MLD_REPORTING_MEMBER:
1694 inm->in6m_sctimer = 0;
1695 inm->in6m_timer = 0;
1696 inm->in6m_state = MLD_REPORTING_MEMBER;
1698 * Free any pending MLDv2 state-change records.
1700 mbufq_drain(&inm->in6m_scq);
1705 IF_ADDR_WUNLOCK(ifp);
1706 in6m_release_list_deferred(&inmh);
1710 * Global slowtimo handler.
1711 * VIMAGE: Timeout handlers are expected to service all vimages.
1713 static struct callout mldslow_callout;
1715 mld_slowtimo(void *arg __unused)
1717 VNET_ITERATOR_DECL(vnet_iter);
1719 VNET_LIST_RLOCK_NOSLEEP();
1720 VNET_FOREACH(vnet_iter) {
1721 CURVNET_SET(vnet_iter);
1722 mld_slowtimo_vnet();
1725 VNET_LIST_RUNLOCK_NOSLEEP();
1727 callout_reset(&mldslow_callout, hz / MLD_SLOWHZ, mld_slowtimo, NULL);
1731 * Per-vnet slowtimo handler.
1734 mld_slowtimo_vnet(void)
1736 struct mld_ifsoftc *mli;
1740 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1741 mld_v1_process_querier_timers(mli);
1748 * Update the Older Version Querier Present timers for a link.
1749 * See Section 9.12 of RFC 3810.
1752 mld_v1_process_querier_timers(struct mld_ifsoftc *mli)
1757 if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) {
1759 * MLDv1 Querier Present timer expired; revert to MLDv2.
1762 "%s: transition from v%d -> v%d on %p(%s)",
1763 __func__, mli->mli_version, MLD_VERSION_2,
1764 mli->mli_ifp, if_name(mli->mli_ifp));
1765 mli->mli_version = MLD_VERSION_2;
1770 * Transmit an MLDv1 report immediately.
1773 mld_v1_transmit_report(struct in6_multi *in6m, const int type)
1776 struct in6_ifaddr *ia;
1777 struct ip6_hdr *ip6;
1778 struct mbuf *mh, *md;
1779 struct mld_hdr *mld;
1782 IN6_MULTI_LIST_LOCK_ASSERT();
1785 ifp = in6m->in6m_ifp;
1786 /* in process of being freed */
1789 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1790 /* ia may be NULL if link-local address is tentative. */
1792 mh = m_gethdr(M_NOWAIT, MT_DATA);
1795 ifa_free(&ia->ia_ifa);
1798 md = m_get(M_NOWAIT, MT_DATA);
1802 ifa_free(&ia->ia_ifa);
1808 * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so
1809 * that ether_output() does not need to allocate another mbuf
1810 * for the header in the most common case.
1812 M_ALIGN(mh, sizeof(struct ip6_hdr));
1813 mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
1814 mh->m_len = sizeof(struct ip6_hdr);
1816 ip6 = mtod(mh, struct ip6_hdr *);
1818 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1819 ip6->ip6_vfc |= IPV6_VERSION;
1820 ip6->ip6_nxt = IPPROTO_ICMPV6;
1821 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
1822 ip6->ip6_dst = in6m->in6m_addr;
1824 md->m_len = sizeof(struct mld_hdr);
1825 mld = mtod(md, struct mld_hdr *);
1826 mld->mld_type = type;
1829 mld->mld_maxdelay = 0;
1830 mld->mld_reserved = 0;
1831 mld->mld_addr = in6m->in6m_addr;
1832 in6_clearscope(&mld->mld_addr);
1833 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
1834 sizeof(struct ip6_hdr), sizeof(struct mld_hdr));
1836 mld_save_context(mh, ifp);
1837 mh->m_flags |= M_MLDV1;
1839 mld_dispatch_packet(mh);
1842 ifa_free(&ia->ia_ifa);
1847 * Process a state change from the upper layer for the given IPv6 group.
1849 * Each socket holds a reference on the in_multi in its own ip_moptions.
1850 * The socket layer will have made the necessary updates to.the group
1851 * state, it is now up to MLD to issue a state change report if there
1852 * has been any change between T0 (when the last state-change was issued)
1855 * We use the MLDv2 state machine at group level. The MLd module
1856 * however makes the decision as to which MLD protocol version to speak.
1857 * A state change *from* INCLUDE {} always means an initial join.
1858 * A state change *to* INCLUDE {} always means a final leave.
1860 * If delay is non-zero, and the state change is an initial multicast
1861 * join, the state change report will be delayed by 'delay' ticks
1862 * in units of MLD_FASTHZ if MLDv1 is active on the link; otherwise
1863 * the initial MLDv2 state change report will be delayed by whichever
1864 * is sooner, a pending state-change timer or delay itself.
1866 * VIMAGE: curvnet should have been set by caller, as this routine
1867 * is called from the socket option handlers.
1870 mld_change_state(struct in6_multi *inm, const int delay)
1872 struct mld_ifsoftc *mli;
1876 IN6_MULTI_LIST_LOCK_ASSERT();
1881 * Check if the in6_multi has already been disconnected.
1883 if (inm->in6m_ifp == NULL) {
1884 CTR1(KTR_MLD, "%s: inm is disconnected", __func__);
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;
1897 * Sanity check that netinet6's notion of ifp is the
1900 KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__));
1903 mli = MLD_IFINFO(ifp);
1904 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
1907 * If we detect a state transition to or from MCAST_UNDEFINED
1908 * for this group, then we are starting or finishing an MLD
1909 * life cycle for this group.
1911 if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) {
1912 CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__,
1913 inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode);
1914 if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) {
1915 CTR1(KTR_MLD, "%s: initial join", __func__);
1916 error = mld_initial_join(inm, mli, delay);
1918 } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) {
1919 CTR1(KTR_MLD, "%s: final leave", __func__);
1920 mld_final_leave(inm, mli);
1924 CTR1(KTR_MLD, "%s: filter set change", __func__);
1927 error = mld_handle_state_change(inm, mli);
1935 * Perform the initial join for an MLD group.
1937 * When joining a group:
1938 * If the group should have its MLD traffic suppressed, do nothing.
1939 * MLDv1 starts sending MLDv1 host membership reports.
1940 * MLDv2 will schedule an MLDv2 state-change report containing the
1941 * initial state of the membership.
1943 * If the delay argument is non-zero, then we must delay sending the
1944 * initial state change for delay ticks (in units of MLD_FASTHZ).
1947 mld_initial_join(struct in6_multi *inm, struct mld_ifsoftc *mli,
1950 struct epoch_tracker et;
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, if_name(inm->in6m_ifp));
1966 ifp = inm->in6m_ifp;
1968 IN6_MULTI_LIST_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 inm->in6m_refcount--;
1999 MPASS(inm->in6m_refcount > 0);
2001 inm->in6m_state = MLD_REPORTING_MEMBER;
2003 switch (mli->mli_version) {
2006 * If a delay was provided, only use it if
2007 * it is greater than the delay normally
2008 * used for an MLDv1 state change report,
2009 * and delay sending the initial MLDv1 report
2010 * by not transitioning to the IDLE state.
2012 odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * MLD_FASTHZ);
2014 inm->in6m_timer = max(delay, odelay);
2015 V_current_state_timers_running6 = 1;
2017 inm->in6m_state = MLD_IDLE_MEMBER;
2018 NET_EPOCH_ENTER(et);
2019 error = mld_v1_transmit_report(inm,
2020 MLD_LISTENER_REPORT);
2023 inm->in6m_timer = odelay;
2024 V_current_state_timers_running6 = 1;
2031 * Defer update of T0 to T1, until the first copy
2032 * of the state change has been transmitted.
2037 * Immediately enqueue a State-Change Report for
2038 * this interface, freeing any previous reports.
2039 * Don't kick the timers if there is nothing to do,
2040 * or if an error occurred.
2042 mq = &inm->in6m_scq;
2044 retval = mld_v2_enqueue_group_record(mq, inm, 1,
2045 0, 0, (mli->mli_flags & MLIF_USEALLOW));
2046 CTR2(KTR_MLD, "%s: enqueue record = %d",
2049 error = retval * -1;
2054 * Schedule transmission of pending state-change
2055 * report up to RV times for this link. The timer
2056 * will fire at the next mld_fasttimo (~200ms),
2057 * giving us an opportunity to merge the reports.
2059 * If a delay was provided to this function, only
2060 * use this delay if sooner than the existing one.
2062 KASSERT(mli->mli_rv > 1,
2063 ("%s: invalid robustness %d", __func__,
2065 inm->in6m_scrv = mli->mli_rv;
2067 if (inm->in6m_sctimer > 1) {
2069 min(inm->in6m_sctimer, delay);
2071 inm->in6m_sctimer = delay;
2073 inm->in6m_sctimer = 1;
2074 V_state_change_timers_running6 = 1;
2082 * Only update the T0 state if state change is atomic,
2083 * i.e. we don't need to wait for a timer to fire before we
2084 * can consider the state change to have been communicated.
2088 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2089 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2090 if_name(inm->in6m_ifp));
2097 * Issue an intermediate state change during the life-cycle.
2100 mld_handle_state_change(struct in6_multi *inm, struct mld_ifsoftc *mli)
2105 char ip6tbuf[INET6_ADDRSTRLEN];
2108 CTR4(KTR_MLD, "%s: state change for %s on ifp %p(%s)",
2109 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2110 inm->in6m_ifp, if_name(inm->in6m_ifp));
2112 ifp = inm->in6m_ifp;
2114 IN6_MULTI_LIST_LOCK_ASSERT();
2117 KASSERT(mli && mli->mli_ifp == ifp,
2118 ("%s: inconsistent ifp", __func__));
2120 if ((ifp->if_flags & IFF_LOOPBACK) ||
2121 (mli->mli_flags & MLIF_SILENT) ||
2122 !mld_is_addr_reported(&inm->in6m_addr) ||
2123 (mli->mli_version != MLD_VERSION_2)) {
2124 if (!mld_is_addr_reported(&inm->in6m_addr)) {
2126 "%s: not kicking state machine for silent group", __func__);
2128 CTR1(KTR_MLD, "%s: nothing to do", __func__);
2130 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2131 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2132 if_name(inm->in6m_ifp));
2136 mbufq_drain(&inm->in6m_scq);
2138 retval = mld_v2_enqueue_group_record(&inm->in6m_scq, inm, 1, 0, 0,
2139 (mli->mli_flags & MLIF_USEALLOW));
2140 CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval);
2145 * If record(s) were enqueued, start the state-change
2146 * report timer for this group.
2148 inm->in6m_scrv = mli->mli_rv;
2149 inm->in6m_sctimer = 1;
2150 V_state_change_timers_running6 = 1;
2156 * Perform the final leave for a multicast address.
2158 * When leaving a group:
2159 * MLDv1 sends a DONE message, if and only if we are the reporter.
2160 * MLDv2 enqueues a state-change report containing a transition
2161 * to INCLUDE {} for immediate transmission.
2164 mld_final_leave(struct in6_multi *inm, struct mld_ifsoftc *mli)
2166 struct epoch_tracker et;
2169 char ip6tbuf[INET6_ADDRSTRLEN];
2174 CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)",
2175 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2176 inm->in6m_ifp, if_name(inm->in6m_ifp));
2178 IN6_MULTI_LIST_LOCK_ASSERT();
2181 switch (inm->in6m_state) {
2182 case MLD_NOT_MEMBER:
2183 case MLD_SILENT_MEMBER:
2184 case MLD_LEAVING_MEMBER:
2185 /* Already leaving or left; do nothing. */
2187 "%s: not kicking state machine for silent group", __func__);
2189 case MLD_REPORTING_MEMBER:
2190 case MLD_IDLE_MEMBER:
2191 case MLD_G_QUERY_PENDING_MEMBER:
2192 case MLD_SG_QUERY_PENDING_MEMBER:
2193 if (mli->mli_version == MLD_VERSION_1) {
2195 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
2196 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER)
2197 panic("%s: MLDv2 state reached, not MLDv2 mode",
2200 NET_EPOCH_ENTER(et);
2201 mld_v1_transmit_report(inm, MLD_LISTENER_DONE);
2203 inm->in6m_state = MLD_NOT_MEMBER;
2204 V_current_state_timers_running6 = 1;
2205 } else if (mli->mli_version == MLD_VERSION_2) {
2207 * Stop group timer and all pending reports.
2208 * Immediately enqueue a state-change report
2209 * TO_IN {} to be sent on the next fast timeout,
2210 * giving us an opportunity to merge reports.
2212 mbufq_drain(&inm->in6m_scq);
2213 inm->in6m_timer = 0;
2214 inm->in6m_scrv = mli->mli_rv;
2215 CTR4(KTR_MLD, "%s: Leaving %s/%s with %d "
2216 "pending retransmissions.", __func__,
2217 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2218 if_name(inm->in6m_ifp), inm->in6m_scrv);
2219 if (inm->in6m_scrv == 0) {
2220 inm->in6m_state = MLD_NOT_MEMBER;
2221 inm->in6m_sctimer = 0;
2223 int retval __diagused;
2225 in6m_acquire_locked(inm);
2227 retval = mld_v2_enqueue_group_record(
2228 &inm->in6m_scq, inm, 1, 0, 0,
2229 (mli->mli_flags & MLIF_USEALLOW));
2230 KASSERT(retval != 0,
2231 ("%s: enqueue record = %d", __func__,
2234 inm->in6m_state = MLD_LEAVING_MEMBER;
2235 inm->in6m_sctimer = 1;
2236 V_state_change_timers_running6 = 1;
2242 case MLD_LAZY_MEMBER:
2243 case MLD_SLEEPING_MEMBER:
2244 case MLD_AWAKENING_MEMBER:
2245 /* Our reports are suppressed; do nothing. */
2251 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2252 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2253 if_name(inm->in6m_ifp));
2254 inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED;
2255 CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s",
2256 __func__, &inm->in6m_addr, if_name(inm->in6m_ifp));
2261 * Enqueue an MLDv2 group record to the given output queue.
2263 * If is_state_change is zero, a current-state record is appended.
2264 * If is_state_change is non-zero, a state-change report is appended.
2266 * If is_group_query is non-zero, an mbuf packet chain is allocated.
2267 * If is_group_query is zero, and if there is a packet with free space
2268 * at the tail of the queue, it will be appended to providing there
2269 * is enough free space.
2270 * Otherwise a new mbuf packet chain is allocated.
2272 * If is_source_query is non-zero, each source is checked to see if
2273 * it was recorded for a Group-Source query, and will be omitted if
2274 * it is not both in-mode and recorded.
2276 * If use_block_allow is non-zero, state change reports for initial join
2277 * and final leave, on an inclusive mode group with a source list, will be
2278 * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively.
2280 * The function will attempt to allocate leading space in the packet
2281 * for the IPv6+ICMP headers to be prepended without fragmenting the chain.
2283 * If successful the size of all data appended to the queue is returned,
2284 * otherwise an error code less than zero is returned, or zero if
2285 * no record(s) were appended.
2288 mld_v2_enqueue_group_record(struct mbufq *mq, struct in6_multi *inm,
2289 const int is_state_change, const int is_group_query,
2290 const int is_source_query, const int use_block_allow)
2292 struct mldv2_record mr;
2293 struct mldv2_record *pmr;
2295 struct ip6_msource *ims, *nims;
2296 struct mbuf *m0, *m, *md;
2297 int is_filter_list_change;
2298 int minrec0len, m0srcs, msrcs, nbytes, off;
2299 int record_has_sources;
2304 char ip6tbuf[INET6_ADDRSTRLEN];
2307 IN6_MULTI_LIST_LOCK_ASSERT();
2309 ifp = inm->in6m_ifp;
2310 is_filter_list_change = 0;
2317 record_has_sources = 1;
2319 type = MLD_DO_NOTHING;
2320 mode = inm->in6m_st[1].iss_fmode;
2323 * If we did not transition out of ASM mode during t0->t1,
2324 * and there are no source nodes to process, we can skip
2325 * the generation of source records.
2327 if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 &&
2328 inm->in6m_nsrc == 0)
2329 record_has_sources = 0;
2331 if (is_state_change) {
2333 * Queue a state change record.
2334 * If the mode did not change, and there are non-ASM
2335 * listeners or source filters present,
2336 * we potentially need to issue two records for the group.
2337 * If there are ASM listeners, and there was no filter
2338 * mode transition of any kind, do nothing.
2340 * If we are transitioning to MCAST_UNDEFINED, we need
2341 * not send any sources. A transition to/from this state is
2342 * considered inclusive with some special treatment.
2344 * If we are rewriting initial joins/leaves to use
2345 * ALLOW/BLOCK, and the group's membership is inclusive,
2346 * we need to send sources in all cases.
2348 if (mode != inm->in6m_st[0].iss_fmode) {
2349 if (mode == MCAST_EXCLUDE) {
2350 CTR1(KTR_MLD, "%s: change to EXCLUDE",
2352 type = MLD_CHANGE_TO_EXCLUDE_MODE;
2354 CTR1(KTR_MLD, "%s: change to INCLUDE",
2356 if (use_block_allow) {
2359 * Here we're interested in state
2360 * edges either direction between
2361 * MCAST_UNDEFINED and MCAST_INCLUDE.
2362 * Perhaps we should just check
2363 * the group state, rather than
2366 if (mode == MCAST_UNDEFINED) {
2367 type = MLD_BLOCK_OLD_SOURCES;
2369 type = MLD_ALLOW_NEW_SOURCES;
2372 type = MLD_CHANGE_TO_INCLUDE_MODE;
2373 if (mode == MCAST_UNDEFINED)
2374 record_has_sources = 0;
2378 if (record_has_sources) {
2379 is_filter_list_change = 1;
2381 type = MLD_DO_NOTHING;
2386 * Queue a current state record.
2388 if (mode == MCAST_EXCLUDE) {
2389 type = MLD_MODE_IS_EXCLUDE;
2390 } else if (mode == MCAST_INCLUDE) {
2391 type = MLD_MODE_IS_INCLUDE;
2392 KASSERT(inm->in6m_st[1].iss_asm == 0,
2393 ("%s: inm %p is INCLUDE but ASM count is %d",
2394 __func__, inm, inm->in6m_st[1].iss_asm));
2399 * Generate the filter list changes using a separate function.
2401 if (is_filter_list_change)
2402 return (mld_v2_enqueue_filter_change(mq, inm));
2404 if (type == MLD_DO_NOTHING) {
2405 CTR3(KTR_MLD, "%s: nothing to do for %s/%s",
2406 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2407 if_name(inm->in6m_ifp));
2412 * If any sources are present, we must be able to fit at least
2413 * one in the trailing space of the tail packet's mbuf,
2416 minrec0len = sizeof(struct mldv2_record);
2417 if (record_has_sources)
2418 minrec0len += sizeof(struct in6_addr);
2420 CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__,
2421 mld_rec_type_to_str(type),
2422 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2423 if_name(inm->in6m_ifp));
2426 * Check if we have a packet in the tail of the queue for this
2427 * group into which the first group record for this group will fit.
2428 * Otherwise allocate a new packet.
2429 * Always allocate leading space for IP6+RA+ICMPV6+REPORT.
2430 * Note: Group records for G/GSR query responses MUST be sent
2431 * in their own packet.
2433 m0 = mbufq_last(mq);
2434 if (!is_group_query &&
2436 (m0->m_pkthdr.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) &&
2437 (m0->m_pkthdr.len + minrec0len) <
2438 (ifp->if_mtu - MLD_MTUSPACE)) {
2439 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2440 sizeof(struct mldv2_record)) /
2441 sizeof(struct in6_addr);
2443 CTR1(KTR_MLD, "%s: use existing packet", __func__);
2445 if (mbufq_full(mq)) {
2446 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2450 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2451 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2452 if (!is_state_change && !is_group_query)
2453 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2455 m = m_gethdr(M_NOWAIT, MT_DATA);
2459 mld_save_context(m, ifp);
2461 CTR1(KTR_MLD, "%s: allocated first packet", __func__);
2465 * Append group record.
2466 * If we have sources, we don't know how many yet.
2471 mr.mr_addr = inm->in6m_addr;
2472 in6_clearscope(&mr.mr_addr);
2473 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2476 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2479 nbytes += sizeof(struct mldv2_record);
2482 * Append as many sources as will fit in the first packet.
2483 * If we are appending to a new packet, the chain allocation
2484 * may potentially use clusters; use m_getptr() in this case.
2485 * If we are appending to an existing packet, we need to obtain
2486 * a pointer to the group record after m_append(), in case a new
2487 * mbuf was allocated.
2489 * Only append sources which are in-mode at t1. If we are
2490 * transitioning to MCAST_UNDEFINED state on the group, and
2491 * use_block_allow is zero, do not include source entries.
2492 * Otherwise, we need to include this source in the report.
2494 * Only report recorded sources in our filter set when responding
2495 * to a group-source query.
2497 if (record_has_sources) {
2500 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2501 md->m_len - nbytes);
2503 md = m_getptr(m, 0, &off);
2504 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2508 RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs,
2510 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2511 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2512 now = im6s_get_mode(inm, ims, 1);
2513 CTR2(KTR_MLD, "%s: node is %d", __func__, now);
2514 if ((now != mode) ||
2516 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2517 CTR1(KTR_MLD, "%s: skip node", __func__);
2520 if (is_source_query && ims->im6s_stp == 0) {
2521 CTR1(KTR_MLD, "%s: skip unrecorded node",
2525 CTR1(KTR_MLD, "%s: append node", __func__);
2526 if (!m_append(m, sizeof(struct in6_addr),
2527 (void *)&ims->im6s_addr)) {
2530 CTR1(KTR_MLD, "%s: m_append() failed.",
2534 nbytes += sizeof(struct in6_addr);
2536 if (msrcs == m0srcs)
2539 CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__,
2541 pmr->mr_numsrc = htons(msrcs);
2542 nbytes += (msrcs * sizeof(struct in6_addr));
2545 if (is_source_query && msrcs == 0) {
2546 CTR1(KTR_MLD, "%s: no recorded sources to report", __func__);
2553 * We are good to go with first packet.
2556 CTR1(KTR_MLD, "%s: enqueueing first packet", __func__);
2557 m->m_pkthdr.vt_nrecs = 1;
2558 mbufq_enqueue(mq, m);
2560 m->m_pkthdr.vt_nrecs++;
2563 * No further work needed if no source list in packet(s).
2565 if (!record_has_sources)
2569 * Whilst sources remain to be announced, we need to allocate
2570 * a new packet and fill out as many sources as will fit.
2571 * Always try for a cluster first.
2573 while (nims != NULL) {
2574 if (mbufq_full(mq)) {
2575 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2578 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2580 m = m_gethdr(M_NOWAIT, MT_DATA);
2583 mld_save_context(m, ifp);
2584 md = m_getptr(m, 0, &off);
2585 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off);
2586 CTR1(KTR_MLD, "%s: allocated next packet", __func__);
2588 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2591 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2594 m->m_pkthdr.vt_nrecs = 1;
2595 nbytes += sizeof(struct mldv2_record);
2597 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2598 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2601 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2602 CTR2(KTR_MLD, "%s: visit node %s",
2603 __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2604 now = im6s_get_mode(inm, ims, 1);
2605 if ((now != mode) ||
2607 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2608 CTR1(KTR_MLD, "%s: skip node", __func__);
2611 if (is_source_query && ims->im6s_stp == 0) {
2612 CTR1(KTR_MLD, "%s: skip unrecorded node",
2616 CTR1(KTR_MLD, "%s: append node", __func__);
2617 if (!m_append(m, sizeof(struct in6_addr),
2618 (void *)&ims->im6s_addr)) {
2621 CTR1(KTR_MLD, "%s: m_append() failed.",
2626 if (msrcs == m0srcs)
2629 pmr->mr_numsrc = htons(msrcs);
2630 nbytes += (msrcs * sizeof(struct in6_addr));
2632 CTR1(KTR_MLD, "%s: enqueueing next packet", __func__);
2633 mbufq_enqueue(mq, m);
2640 * Type used to mark record pass completion.
2641 * We exploit the fact we can cast to this easily from the
2642 * current filter modes on each ip_msource node.
2645 REC_NONE = 0x00, /* MCAST_UNDEFINED */
2646 REC_ALLOW = 0x01, /* MCAST_INCLUDE */
2647 REC_BLOCK = 0x02, /* MCAST_EXCLUDE */
2648 REC_FULL = REC_ALLOW | REC_BLOCK
2652 * Enqueue an MLDv2 filter list change to the given output queue.
2654 * Source list filter state is held in an RB-tree. When the filter list
2655 * for a group is changed without changing its mode, we need to compute
2656 * the deltas between T0 and T1 for each source in the filter set,
2657 * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records.
2659 * As we may potentially queue two record types, and the entire R-B tree
2660 * needs to be walked at once, we break this out into its own function
2661 * so we can generate a tightly packed queue of packets.
2663 * XXX This could be written to only use one tree walk, although that makes
2664 * serializing into the mbuf chains a bit harder. For now we do two walks
2665 * which makes things easier on us, and it may or may not be harder on
2668 * If successful the size of all data appended to the queue is returned,
2669 * otherwise an error code less than zero is returned, or zero if
2670 * no record(s) were appended.
2673 mld_v2_enqueue_filter_change(struct mbufq *mq, struct in6_multi *inm)
2675 static const int MINRECLEN =
2676 sizeof(struct mldv2_record) + sizeof(struct in6_addr);
2678 struct mldv2_record mr;
2679 struct mldv2_record *pmr;
2680 struct ip6_msource *ims, *nims;
2681 struct mbuf *m, *m0, *md;
2682 int m0srcs, nbytes, npbytes, off, rsrcs, schanged;
2683 uint8_t mode, now, then;
2684 rectype_t crt, drt, nrt;
2687 char ip6tbuf[INET6_ADDRSTRLEN];
2690 IN6_MULTI_LIST_LOCK_ASSERT();
2692 if (inm->in6m_nsrc == 0 ||
2693 (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0))
2696 ifp = inm->in6m_ifp; /* interface */
2697 mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */
2698 crt = REC_NONE; /* current group record type */
2699 drt = REC_NONE; /* mask of completed group record types */
2700 nrt = REC_NONE; /* record type for current node */
2701 m0srcs = 0; /* # source which will fit in current mbuf chain */
2702 npbytes = 0; /* # of bytes appended this packet */
2703 nbytes = 0; /* # of bytes appended to group's state-change queue */
2704 rsrcs = 0; /* # sources encoded in current record */
2705 schanged = 0; /* # nodes encoded in overall filter change */
2707 nallow = 0; /* # of source entries in ALLOW_NEW */
2708 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) {
2721 m0 = mbufq_last(mq);
2723 (m0->m_pkthdr.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_NOWAIT, MT_DATA, M_PKTHDR);
2736 m = m_gethdr(M_NOWAIT, MT_DATA);
2739 "%s: m_get*() failed", __func__);
2742 m->m_pkthdr.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__);
2831 nallow += !!(crt == REC_ALLOW);
2832 nblock += !!(crt == REC_BLOCK);
2834 if (++rsrcs == m0srcs)
2838 * If we did not append any tree nodes on this
2839 * pass, back out of allocations.
2842 npbytes -= sizeof(struct mldv2_record);
2845 "%s: m_free(m)", __func__);
2849 "%s: m_adj(m, -mr)", __func__);
2850 m_adj(m, -((int)sizeof(
2851 struct mldv2_record)));
2855 npbytes += (rsrcs * sizeof(struct in6_addr));
2856 if (crt == REC_ALLOW)
2857 pmr->mr_type = MLD_ALLOW_NEW_SOURCES;
2858 else if (crt == REC_BLOCK)
2859 pmr->mr_type = MLD_BLOCK_OLD_SOURCES;
2860 pmr->mr_numsrc = htons(rsrcs);
2862 * Count the new group record, and enqueue this
2863 * packet if it wasn't already queued.
2865 m->m_pkthdr.vt_nrecs++;
2867 mbufq_enqueue(mq, m);
2869 } while (nims != NULL);
2871 crt = (~crt & REC_FULL);
2874 CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__,
2881 mld_v2_merge_state_changes(struct in6_multi *inm, struct mbufq *scq)
2884 struct mbuf *m; /* pending state-change */
2885 struct mbuf *m0; /* copy of pending state-change */
2886 struct mbuf *mt; /* last state-change in packet */
2887 int docopy, domerge;
2894 IN6_MULTI_LIST_LOCK_ASSERT();
2898 * If there are further pending retransmissions, make a writable
2899 * copy of each queued state-change message before merging.
2901 if (inm->in6m_scrv > 0)
2904 gq = &inm->in6m_scq;
2906 if (mbufq_first(gq) == NULL) {
2907 CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty",
2912 m = mbufq_first(gq);
2915 * Only merge the report into the current packet if
2916 * there is sufficient space to do so; an MLDv2 report
2917 * packet may only contain 65,535 group records.
2918 * Always use a simple mbuf chain concatentation to do this,
2919 * as large state changes for single groups may have
2920 * allocated clusters.
2923 mt = mbufq_last(scq);
2925 recslen = m_length(m, NULL);
2927 if ((mt->m_pkthdr.vt_nrecs +
2928 m->m_pkthdr.vt_nrecs <=
2929 MLD_V2_REPORT_MAXRECS) &&
2930 (mt->m_pkthdr.len + recslen <=
2931 (inm->in6m_ifp->if_mtu - MLD_MTUSPACE)))
2935 if (!domerge && mbufq_full(gq)) {
2937 "%s: outbound queue full, skipping whole packet %p",
2947 CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m);
2948 m0 = mbufq_dequeue(gq);
2951 CTR2(KTR_MLD, "%s: copying %p", __func__, m);
2952 m0 = m_dup(m, M_NOWAIT);
2955 m0->m_nextpkt = NULL;
2960 CTR3(KTR_MLD, "%s: queueing %p to scq %p)",
2962 mbufq_enqueue(scq, m0);
2964 struct mbuf *mtl; /* last mbuf of packet mt */
2966 CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)",
2970 m0->m_flags &= ~M_PKTHDR;
2971 mt->m_pkthdr.len += recslen;
2972 mt->m_pkthdr.vt_nrecs +=
2973 m0->m_pkthdr.vt_nrecs;
2983 * Respond to a pending MLDv2 General Query.
2986 mld_v2_dispatch_general_query(struct mld_ifsoftc *mli)
2988 struct ifmultiaddr *ifma;
2990 struct in6_multi *inm;
2991 int retval __unused;
2994 IN6_MULTI_LIST_LOCK_ASSERT();
2997 KASSERT(mli->mli_version == MLD_VERSION_2,
2998 ("%s: called when version %d", __func__, mli->mli_version));
3001 * Check that there are some packets queued. If so, send them first.
3002 * For large number of groups the reply to general query can take
3003 * many packets, we should finish sending them before starting of
3004 * queuing the new reply.
3006 if (mbufq_len(&mli->mli_gq) != 0)
3011 CK_STAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
3012 inm = in6m_ifmultiaddr_get_inm(ifma);
3015 KASSERT(ifp == inm->in6m_ifp,
3016 ("%s: inconsistent ifp", __func__));
3018 switch (inm->in6m_state) {
3019 case MLD_NOT_MEMBER:
3020 case MLD_SILENT_MEMBER:
3022 case MLD_REPORTING_MEMBER:
3023 case MLD_IDLE_MEMBER:
3024 case MLD_LAZY_MEMBER:
3025 case MLD_SLEEPING_MEMBER:
3026 case MLD_AWAKENING_MEMBER:
3027 inm->in6m_state = MLD_REPORTING_MEMBER;
3028 retval = mld_v2_enqueue_group_record(&mli->mli_gq,
3030 CTR2(KTR_MLD, "%s: enqueue record = %d",
3033 case MLD_G_QUERY_PENDING_MEMBER:
3034 case MLD_SG_QUERY_PENDING_MEMBER:
3035 case MLD_LEAVING_MEMBER:
3041 mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST);
3044 * Slew transmission of bursts over 500ms intervals.
3046 if (mbufq_first(&mli->mli_gq) != NULL) {
3047 mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY(
3048 MLD_RESPONSE_BURST_INTERVAL);
3049 V_interface_timers_running6 = 1;
3054 * Transmit the next pending message in the output queue.
3056 * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis.
3057 * MRT: Nothing needs to be done, as MLD traffic is always local to
3058 * a link and uses a link-scope multicast address.
3061 mld_dispatch_packet(struct mbuf *m)
3063 struct ip6_moptions im6o;
3068 struct ip6_hdr *ip6;
3069 struct mld_hdr *mld;
3075 CTR2(KTR_MLD, "%s: transmit %p", __func__, m);
3079 * Set VNET image pointer from enqueued mbuf chain
3080 * before doing anything else. Whilst we use interface
3081 * indexes to guard against interface detach, they are
3082 * unique to each VIMAGE and must be retrieved.
3084 ifindex = mld_restore_context(m);
3087 * Check if the ifnet still exists. This limits the scope of
3088 * any race in the absence of a global ifp lock for low cost
3089 * (an array lookup).
3091 ifp = ifnet_byindex(ifindex);
3093 CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.",
3094 __func__, m, ifindex);
3096 IP6STAT_INC(ip6s_noroute);
3100 im6o.im6o_multicast_hlim = 1;
3101 im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL);
3102 im6o.im6o_multicast_ifp = ifp;
3104 if (m->m_flags & M_MLDV1) {
3107 m0 = mld_v2_encap_report(ifp, m);
3109 CTR2(KTR_MLD, "%s: dropped %p", __func__, m);
3110 IP6STAT_INC(ip6s_odropped);
3115 mld_scrub_context(m0);
3117 m0->m_pkthdr.rcvif = V_loif;
3119 ip6 = mtod(m0, struct ip6_hdr *);
3121 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */
3124 * XXX XXX Break some KPI rules to prevent an LOR which would
3125 * occur if we called in6_setscope() at transmission.
3126 * See comments at top of file.
3128 MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index);
3132 * Retrieve the ICMPv6 type before handoff to ip6_output(),
3133 * so we can bump the stats.
3135 md = m_getptr(m0, sizeof(struct ip6_hdr), &off);
3136 mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off);
3137 type = mld->mld_type;
3140 error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o,
3143 CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error);
3146 ICMP6STAT_INC(icp6s_outhist[type]);
3148 icmp6_ifstat_inc(oifp, ifs6_out_msg);
3150 case MLD_LISTENER_REPORT:
3151 case MLDV2_LISTENER_REPORT:
3152 icmp6_ifstat_inc(oifp, ifs6_out_mldreport);
3154 case MLD_LISTENER_DONE:
3155 icmp6_ifstat_inc(oifp, ifs6_out_mlddone);
3164 * Encapsulate an MLDv2 report.
3166 * KAME IPv6 requires that hop-by-hop options be passed separately,
3167 * and that the IPv6 header be prepended in a separate mbuf.
3169 * Returns a pointer to the new mbuf chain head, or NULL if the
3170 * allocation failed.
3172 static struct mbuf *
3173 mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m)
3176 struct mldv2_report *mld;
3177 struct ip6_hdr *ip6;
3178 struct in6_ifaddr *ia;
3181 KASSERT(ifp != NULL, ("%s: null ifp", __func__));
3182 KASSERT((m->m_flags & M_PKTHDR),
3183 ("%s: mbuf chain %p is !M_PKTHDR", __func__, m));
3186 * RFC3590: OK to send as :: or tentative during DAD.
3189 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
3191 CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__);
3193 mh = m_gethdr(M_NOWAIT, MT_DATA);
3196 ifa_free(&ia->ia_ifa);
3200 M_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report));
3202 mldreclen = m_length(m, NULL);
3203 CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen);
3205 mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report);
3206 mh->m_pkthdr.len = sizeof(struct ip6_hdr) +
3207 sizeof(struct mldv2_report) + mldreclen;
3209 ip6 = mtod(mh, struct ip6_hdr *);
3211 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3212 ip6->ip6_vfc |= IPV6_VERSION;
3213 ip6->ip6_nxt = IPPROTO_ICMPV6;
3214 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
3216 ifa_free(&ia->ia_ifa);
3217 ip6->ip6_dst = in6addr_linklocal_allv2routers;
3218 /* scope ID will be set in netisr */
3220 mld = (struct mldv2_report *)(ip6 + 1);
3221 mld->mld_type = MLDV2_LISTENER_REPORT;
3224 mld->mld_v2_reserved = 0;
3225 mld->mld_v2_numrecs = htons(m->m_pkthdr.vt_nrecs);
3226 m->m_pkthdr.vt_nrecs = 0;
3229 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
3230 sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen);
3236 mld_rec_type_to_str(const int type)
3240 case MLD_CHANGE_TO_EXCLUDE_MODE:
3243 case MLD_CHANGE_TO_INCLUDE_MODE:
3246 case MLD_MODE_IS_EXCLUDE:
3249 case MLD_MODE_IS_INCLUDE:
3252 case MLD_ALLOW_NEW_SOURCES:
3255 case MLD_BLOCK_OLD_SOURCES:
3266 mld_init(void *unused __unused)
3269 CTR1(KTR_MLD, "%s: initializing", __func__);
3272 ip6_initpktopts(&mld_po);
3273 mld_po.ip6po_hlim = 1;
3274 mld_po.ip6po_hbh = &mld_ra.hbh;
3275 mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
3276 mld_po.ip6po_flags = IP6PO_DONTFRAG;
3278 callout_init(&mldslow_callout, 1);
3279 callout_reset(&mldslow_callout, hz / MLD_SLOWHZ, mld_slowtimo, NULL);
3280 callout_init(&mldfast_callout, 1);
3281 callout_reset(&mldfast_callout, hz / MLD_FASTHZ, mld_fasttimo, NULL);
3283 SYSINIT(mld_init, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_init, NULL);
3286 mld_uninit(void *unused __unused)
3289 CTR1(KTR_MLD, "%s: tearing down", __func__);
3290 callout_drain(&mldslow_callout);
3291 callout_drain(&mldfast_callout);
3294 SYSUNINIT(mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_uninit, NULL);
3297 vnet_mld_init(const void *unused __unused)
3300 CTR1(KTR_MLD, "%s: initializing", __func__);
3302 LIST_INIT(&V_mli_head);
3304 VNET_SYSINIT(vnet_mld_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_init,
3308 vnet_mld_uninit(const void *unused __unused)
3311 /* This can happen if we shutdown the network stack. */
3312 CTR1(KTR_MLD, "%s: tearing down", __func__);
3314 VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_uninit,
3318 mld_modevent(module_t mod, int type, void *unused __unused)
3326 return (EOPNOTSUPP);
3331 static moduledata_t mld_mod = {
3336 DECLARE_MODULE(mld, mld_mod, SI_SUB_PROTO_MC, SI_ORDER_ANY);