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/protosw.h>
79 #include <sys/sysctl.h>
80 #include <sys/kernel.h>
81 #include <sys/callout.h>
82 #include <sys/malloc.h>
83 #include <sys/module.h>
87 #include <net/if_var.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 struct mld_ifsoftc *
108 mli_alloc_locked(struct ifnet *);
109 static void mli_delete_locked(const struct ifnet *);
110 static void mld_dispatch_packet(struct mbuf *);
111 static void mld_dispatch_queue(struct mbufq *, int);
112 static void mld_final_leave(struct in6_multi *, struct mld_ifsoftc *);
113 static void mld_fasttimo_vnet(void);
114 static int mld_handle_state_change(struct in6_multi *,
115 struct mld_ifsoftc *);
116 static int mld_initial_join(struct in6_multi *, struct mld_ifsoftc *,
119 static char * mld_rec_type_to_str(const int);
121 static void mld_set_version(struct mld_ifsoftc *, const int);
122 static void mld_slowtimo_vnet(void);
123 static int mld_v1_input_query(struct ifnet *, const struct ip6_hdr *,
124 /*const*/ struct mld_hdr *);
125 static int mld_v1_input_report(struct ifnet *, const struct ip6_hdr *,
126 /*const*/ struct mld_hdr *);
127 static void mld_v1_process_group_timer(struct in6_multi_head *,
129 static void mld_v1_process_querier_timers(struct mld_ifsoftc *);
130 static int mld_v1_transmit_report(struct in6_multi *, const int);
131 static void mld_v1_update_group(struct in6_multi *, const int);
132 static void mld_v2_cancel_link_timers(struct mld_ifsoftc *);
133 static void mld_v2_dispatch_general_query(struct mld_ifsoftc *);
135 mld_v2_encap_report(struct ifnet *, struct mbuf *);
136 static int mld_v2_enqueue_filter_change(struct mbufq *,
138 static int mld_v2_enqueue_group_record(struct mbufq *,
139 struct in6_multi *, const int, const int, const int,
141 static int mld_v2_input_query(struct ifnet *, const struct ip6_hdr *,
142 struct mbuf *, const int, const int);
143 static int mld_v2_merge_state_changes(struct in6_multi *,
145 static void mld_v2_process_group_timers(struct in6_multi_head *,
146 struct mbufq *, struct mbufq *,
147 struct in6_multi *, const int);
148 static int mld_v2_process_group_query(struct in6_multi *,
149 struct mld_ifsoftc *mli, int, struct mbuf *, const int);
150 static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS);
151 static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS);
154 * Normative references: RFC 2710, RFC 3590, RFC 3810.
157 * * The MLD subsystem lock ends up being system-wide for the moment,
158 * but could be per-VIMAGE later on.
159 * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK.
160 * Any may be taken independently; if any are held at the same
161 * time, the above lock order must be followed.
162 * * IN6_MULTI_LOCK covers in_multi.
163 * * MLD_LOCK covers per-link state and any global variables in this file.
164 * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of
165 * per-link state iterators.
168 * A special case for IPv6 is the in6_setscope() routine. ip6_output()
169 * will not accept an ifp; it wants an embedded scope ID, unlike
170 * ip_output(), which happily takes the ifp given to it. The embedded
171 * scope ID is only used by MLD to select the outgoing interface.
173 * During interface attach and detach, MLD will take MLD_LOCK *after*
174 * the IF_AFDATA_LOCK.
175 * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call
176 * it with MLD_LOCK held without triggering an LOR. A netisr with indirect
177 * dispatch could work around this, but we'd rather not do that, as it
178 * can introduce other races.
180 * As such, we exploit the fact that the scope ID is just the interface
181 * index, and embed it in the IPv6 destination address accordingly.
182 * This is potentially NOT VALID for MLDv1 reports, as they
183 * are always sent to the multicast group itself; as MLDv2
184 * reports are always sent to ff02::16, this is not an issue
185 * when MLDv2 is in use.
187 * This does not however eliminate the LOR when ip6_output() itself
188 * calls in6_setscope() internally whilst MLD_LOCK is held. This will
189 * trigger a LOR warning in WITNESS when the ifnet is detached.
191 * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given
192 * how it's used across the network stack. Here we're simply exploiting
193 * the fact that MLD runs at a similar layer in the stack to scope6.c.
196 * * Each in6_multi corresponds to an ifp, and each ifp corresponds
197 * to a vnet in ifp->if_vnet.
199 static struct mtx mld_mtx;
200 static MALLOC_DEFINE(M_MLD, "mld", "mld state");
202 #define MLD_EMBEDSCOPE(pin6, zoneid) \
203 if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \
204 IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \
205 (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \
208 * VIMAGE-wide globals.
210 static VNET_DEFINE(struct timeval, mld_gsrdelay) = {10, 0};
211 static VNET_DEFINE(LIST_HEAD(, mld_ifsoftc), mli_head);
212 static VNET_DEFINE(int, interface_timers_running6);
213 static VNET_DEFINE(int, state_change_timers_running6);
214 static VNET_DEFINE(int, current_state_timers_running6);
216 #define V_mld_gsrdelay VNET(mld_gsrdelay)
217 #define V_mli_head VNET(mli_head)
218 #define V_interface_timers_running6 VNET(interface_timers_running6)
219 #define V_state_change_timers_running6 VNET(state_change_timers_running6)
220 #define V_current_state_timers_running6 VNET(current_state_timers_running6)
222 SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */
224 SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW, 0,
225 "IPv6 Multicast Listener Discovery");
228 * Virtualized sysctls.
230 SYSCTL_PROC(_net_inet6_mld, OID_AUTO, gsrdelay,
231 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
232 &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I",
233 "Rate limit for MLDv2 Group-and-Source queries in seconds");
236 * Non-virtualized sysctls.
238 static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo,
239 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo,
240 "Per-interface MLDv2 state");
242 static int mld_v1enable = 1;
243 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RWTUN,
244 &mld_v1enable, 0, "Enable fallback to MLDv1");
246 static int mld_use_allow = 1;
247 SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RWTUN,
248 &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves");
251 * Packed Router Alert option structure declaration.
256 struct ip6_opt_router ra;
260 * Router Alert hop-by-hop option header.
262 static struct mld_raopt mld_ra = {
264 .pad = { .ip6o_type = IP6OPT_PADN, 0 },
266 .ip6or_type = IP6OPT_ROUTER_ALERT,
267 .ip6or_len = IP6OPT_RTALERT_LEN - 2,
268 .ip6or_value[0] = ((IP6OPT_RTALERT_MLD >> 8) & 0xFF),
269 .ip6or_value[1] = (IP6OPT_RTALERT_MLD & 0xFF)
272 static struct ip6_pktopts mld_po;
275 mld_save_context(struct mbuf *m, struct ifnet *ifp)
279 m->m_pkthdr.PH_loc.ptr = ifp->if_vnet;
281 m->m_pkthdr.flowid = ifp->if_index;
285 mld_scrub_context(struct mbuf *m)
288 m->m_pkthdr.PH_loc.ptr = NULL;
289 m->m_pkthdr.flowid = 0;
293 * Restore context from a queued output chain.
294 * Return saved ifindex.
296 * VIMAGE: The assertion is there to make sure that we
297 * actually called CURVNET_SET() with what's in the mbuf chain.
299 static __inline uint32_t
300 mld_restore_context(struct mbuf *m)
303 #if defined(VIMAGE) && defined(INVARIANTS)
304 KASSERT(curvnet == m->m_pkthdr.PH_loc.ptr,
305 ("%s: called when curvnet was not restored: cuvnet %p m ptr %p",
306 __func__, curvnet, m->m_pkthdr.PH_loc.ptr));
308 return (m->m_pkthdr.flowid);
312 * Retrieve or set threshold between group-source queries in seconds.
314 * VIMAGE: Assume curvnet set by caller.
315 * SMPng: NOTE: Serialized by MLD lock.
318 sysctl_mld_gsr(SYSCTL_HANDLER_ARGS)
323 error = sysctl_wire_old_buffer(req, sizeof(int));
329 i = V_mld_gsrdelay.tv_sec;
331 error = sysctl_handle_int(oidp, &i, 0, req);
332 if (error || !req->newptr)
335 if (i < -1 || i >= 60) {
340 CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d",
341 V_mld_gsrdelay.tv_sec, i);
342 V_mld_gsrdelay.tv_sec = i;
350 * Expose struct mld_ifsoftc to userland, keyed by ifindex.
351 * For use by ifmcstat(8).
353 * SMPng: NOTE: Does an unlocked ifindex space read.
354 * VIMAGE: Assume curvnet set by caller. The node handler itself
355 * is not directly virtualized.
358 sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS)
364 struct mld_ifsoftc *mli;
369 if (req->newptr != NULL)
375 error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo));
380 IN6_MULTI_LIST_LOCK();
383 if (name[0] <= 0 || name[0] > V_if_index) {
390 ifp = ifnet_byindex(name[0]);
394 LIST_FOREACH(mli, &V_mli_head, mli_link) {
395 if (ifp == mli->mli_ifp) {
396 struct mld_ifinfo info;
398 info.mli_version = mli->mli_version;
399 info.mli_v1_timer = mli->mli_v1_timer;
400 info.mli_v2_timer = mli->mli_v2_timer;
401 info.mli_flags = mli->mli_flags;
402 info.mli_rv = mli->mli_rv;
403 info.mli_qi = mli->mli_qi;
404 info.mli_qri = mli->mli_qri;
405 info.mli_uri = mli->mli_uri;
406 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.
469 * SMPng: Normally called with IF_AFDATA_LOCK held.
472 mld_domifattach(struct ifnet *ifp)
474 struct mld_ifsoftc *mli;
476 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
477 __func__, ifp, if_name(ifp));
481 mli = mli_alloc_locked(ifp);
482 if (!(ifp->if_flags & IFF_MULTICAST))
483 mli->mli_flags |= MLIF_SILENT;
485 mli->mli_flags |= MLIF_USEALLOW;
493 * VIMAGE: assume curvnet set by caller.
495 static struct mld_ifsoftc *
496 mli_alloc_locked(/*const*/ struct ifnet *ifp)
498 struct mld_ifsoftc *mli;
502 mli = malloc(sizeof(struct mld_ifsoftc), M_MLD, M_NOWAIT|M_ZERO);
507 mli->mli_version = MLD_VERSION_2;
509 mli->mli_rv = MLD_RV_INIT;
510 mli->mli_qi = MLD_QI_INIT;
511 mli->mli_qri = MLD_QRI_INIT;
512 mli->mli_uri = MLD_URI_INIT;
513 mbufq_init(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS);
515 LIST_INSERT_HEAD(&V_mli_head, mli, mli_link);
517 CTR2(KTR_MLD, "allocate mld_ifsoftc for ifp %p(%s)",
527 * NOTE: Some finalization tasks need to run before the protocol domain
528 * is detached, but also before the link layer does its cleanup.
529 * Run before link-layer cleanup; cleanup groups, but do not free MLD state.
531 * SMPng: Caller must hold IN6_MULTI_LOCK().
532 * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator.
533 * XXX This routine is also bitten by unlocked ifma_protospec access.
536 mld_ifdetach(struct ifnet *ifp)
538 struct mld_ifsoftc *mli;
539 struct ifmultiaddr *ifma;
540 struct in6_multi *inm;
541 struct in6_multi_head inmh;
543 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp,
547 IN6_MULTI_LIST_LOCK_ASSERT();
550 mli = MLD_IFINFO(ifp);
551 if (mli->mli_version == MLD_VERSION_2) {
553 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
554 if (ifma->ifma_addr->sa_family != AF_INET6 ||
555 ifma->ifma_protospec == NULL)
557 inm = (struct in6_multi *)ifma->ifma_protospec;
558 if (inm->in6m_state == MLD_LEAVING_MEMBER) {
559 in6m_rele_locked(&inmh, inm);
560 ifma->ifma_protospec = NULL;
562 in6m_clear_recorded(inm);
564 IF_ADDR_RUNLOCK(ifp);
568 in6m_release_list_deferred(&inmh);
572 * Hook for domifdetach.
573 * Runs after link-layer cleanup; free MLD state.
575 * SMPng: Normally called with IF_AFDATA_LOCK held.
578 mld_domifdetach(struct ifnet *ifp)
581 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
582 __func__, ifp, if_name(ifp));
585 mli_delete_locked(ifp);
590 mli_delete_locked(const struct ifnet *ifp)
592 struct mld_ifsoftc *mli, *tmli;
594 CTR3(KTR_MLD, "%s: freeing mld_ifsoftc for ifp %p(%s)",
595 __func__, ifp, if_name(ifp));
599 LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) {
600 if (mli->mli_ifp == ifp) {
602 * Free deferred General Query responses.
604 mbufq_drain(&mli->mli_gq);
606 LIST_REMOVE(mli, mli_link);
615 * Process a received MLDv1 general or address-specific query.
616 * Assumes that the query header has been pulled up to sizeof(mld_hdr).
618 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
619 * mld_addr. This is OK as we own the mbuf chain.
622 mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
623 /*const*/ struct mld_hdr *mld)
625 struct ifmultiaddr *ifma;
626 struct mld_ifsoftc *mli;
627 struct in6_multi *inm;
628 int is_general_query;
631 char ip6tbuf[INET6_ADDRSTRLEN];
634 is_general_query = 0;
637 CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)",
638 ip6_sprintf(ip6tbuf, &mld->mld_addr),
644 * RFC3810 Section 6.2: MLD queries must originate from
645 * a router's link-local address.
647 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
648 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
649 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
655 * Do address field validation upfront before we accept
658 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
660 * MLDv1 General Query.
661 * If this was not sent to the all-nodes group, ignore it.
666 in6_clearscope(&dst);
667 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes))
669 is_general_query = 1;
672 * Embed scope ID of receiving interface in MLD query for
673 * lookup whilst we don't hold other locks.
675 in6_setscope(&mld->mld_addr, ifp, NULL);
678 IN6_MULTI_LIST_LOCK();
682 * Switch to MLDv1 host compatibility mode.
684 mli = MLD_IFINFO(ifp);
685 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
686 mld_set_version(mli, MLD_VERSION_1);
688 timer = (ntohs(mld->mld_maxdelay) * PR_FASTHZ) / MLD_TIMER_SCALE;
693 if (is_general_query) {
695 * For each reporting group joined on this
696 * interface, kick the report timer.
698 CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)",
700 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
701 if (ifma->ifma_addr->sa_family != AF_INET6 ||
702 ifma->ifma_protospec == NULL)
704 inm = (struct in6_multi *)ifma->ifma_protospec;
705 mld_v1_update_group(inm, timer);
709 * MLDv1 Group-Specific Query.
710 * If this is a group-specific MLDv1 query, we need only
711 * look up the single group to process it.
713 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
715 CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)",
716 ip6_sprintf(ip6tbuf, &mld->mld_addr),
718 mld_v1_update_group(inm, timer);
720 /* XXX Clear embedded scope ID as userland won't expect it. */
721 in6_clearscope(&mld->mld_addr);
724 IF_ADDR_RUNLOCK(ifp);
726 IN6_MULTI_LIST_UNLOCK();
732 * Update the report timer on a group in response to an MLDv1 query.
734 * If we are becoming the reporting member for this group, start the timer.
735 * If we already are the reporting member for this group, and timer is
736 * below the threshold, reset it.
738 * We may be updating the group for the first time since we switched
739 * to MLDv2. If we are, then we must clear any recorded source lists,
740 * and transition to REPORTING state; the group timer is overloaded
741 * for group and group-source query responses.
743 * Unlike MLDv2, the delay per group should be jittered
744 * to avoid bursts of MLDv1 reports.
747 mld_v1_update_group(struct in6_multi *inm, const int timer)
750 char ip6tbuf[INET6_ADDRSTRLEN];
753 CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__,
754 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
755 if_name(inm->in6m_ifp), timer);
757 IN6_MULTI_LIST_LOCK_ASSERT();
759 switch (inm->in6m_state) {
761 case MLD_SILENT_MEMBER:
763 case MLD_REPORTING_MEMBER:
764 if (inm->in6m_timer != 0 &&
765 inm->in6m_timer <= timer) {
766 CTR1(KTR_MLD, "%s: REPORTING and timer running, "
767 "skipping.", __func__);
771 case MLD_SG_QUERY_PENDING_MEMBER:
772 case MLD_G_QUERY_PENDING_MEMBER:
773 case MLD_IDLE_MEMBER:
774 case MLD_LAZY_MEMBER:
775 case MLD_AWAKENING_MEMBER:
776 CTR1(KTR_MLD, "%s: ->REPORTING", __func__);
777 inm->in6m_state = MLD_REPORTING_MEMBER;
778 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
779 V_current_state_timers_running6 = 1;
781 case MLD_SLEEPING_MEMBER:
782 CTR1(KTR_MLD, "%s: ->AWAKENING", __func__);
783 inm->in6m_state = MLD_AWAKENING_MEMBER;
785 case MLD_LEAVING_MEMBER:
791 * Process a received MLDv2 general, group-specific or
792 * group-and-source-specific query.
794 * Assumes that the query header has been pulled up to sizeof(mldv2_query).
796 * Return 0 if successful, otherwise an appropriate error code is returned.
799 mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
800 struct mbuf *m, const int off, const int icmp6len)
802 struct mld_ifsoftc *mli;
803 struct mldv2_query *mld;
804 struct in6_multi *inm;
805 uint32_t maxdelay, nsrc, qqi;
806 int is_general_query;
810 char ip6tbuf[INET6_ADDRSTRLEN];
813 is_general_query = 0;
816 * RFC3810 Section 6.2: MLD queries must originate from
817 * a router's link-local address.
819 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
820 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
821 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
826 CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, if_name(ifp));
828 mld = (struct mldv2_query *)(mtod(m, uint8_t *) + off);
830 maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */
831 if (maxdelay >= 32768) {
832 maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) <<
833 (MLD_MRC_EXP(maxdelay) + 3);
835 timer = (maxdelay * PR_FASTHZ) / MLD_TIMER_SCALE;
839 qrv = MLD_QRV(mld->mld_misc);
841 CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__,
848 qqi = MLD_QQIC_MANT(mld->mld_qqi) <<
849 (MLD_QQIC_EXP(mld->mld_qqi) + 3);
852 nsrc = ntohs(mld->mld_numsrc);
853 if (nsrc > MLD_MAX_GS_SOURCES)
855 if (icmp6len < sizeof(struct mldv2_query) +
856 (nsrc * sizeof(struct in6_addr)))
860 * Do further input validation upfront to avoid resetting timers
861 * should we need to discard this query.
863 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
865 * A general query with a source list has undefined
866 * behaviour; discard it.
870 is_general_query = 1;
873 * Embed scope ID of receiving interface in MLD query for
874 * lookup whilst we don't hold other locks (due to KAME
875 * locking lameness). We own this mbuf chain just now.
877 in6_setscope(&mld->mld_addr, ifp, NULL);
880 IN6_MULTI_LIST_LOCK();
883 mli = MLD_IFINFO(ifp);
884 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
887 * Discard the v2 query if we're in Compatibility Mode.
888 * The RFC is pretty clear that hosts need to stay in MLDv1 mode
889 * until the Old Version Querier Present timer expires.
891 if (mli->mli_version != MLD_VERSION_2)
894 mld_set_version(mli, MLD_VERSION_2);
897 mli->mli_qri = maxdelay;
899 CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi,
902 if (is_general_query) {
904 * MLDv2 General Query.
906 * Schedule a current-state report on this ifp for
907 * all groups, possibly containing source lists.
909 * If there is a pending General Query response
910 * scheduled earlier than the selected delay, do
911 * not schedule any other reports.
912 * Otherwise, reset the interface timer.
914 CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)",
916 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) {
917 mli->mli_v2_timer = MLD_RANDOM_DELAY(timer);
918 V_interface_timers_running6 = 1;
922 * MLDv2 Group-specific or Group-and-source-specific Query.
924 * Group-source-specific queries are throttled on
925 * a per-group basis to defeat denial-of-service attempts.
926 * Queries for groups we are not a member of on this
927 * link are simply ignored.
930 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
932 IF_ADDR_RUNLOCK(ifp);
936 if (!ratecheck(&inm->in6m_lastgsrtv,
938 CTR1(KTR_MLD, "%s: GS query throttled.",
940 IF_ADDR_RUNLOCK(ifp);
944 CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)",
947 * If there is a pending General Query response
948 * scheduled sooner than the selected delay, no
949 * further report need be scheduled.
950 * Otherwise, prepare to respond to the
951 * group-specific or group-and-source query.
953 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer)
954 mld_v2_process_group_query(inm, mli, timer, m, off);
956 /* XXX Clear embedded scope ID as userland won't expect it. */
957 in6_clearscope(&mld->mld_addr);
958 IF_ADDR_RUNLOCK(ifp);
963 IN6_MULTI_LIST_UNLOCK();
969 * Process a received MLDv2 group-specific or group-and-source-specific
971 * Return <0 if any error occurred. Currently this is ignored.
974 mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifsoftc *mli,
975 int timer, struct mbuf *m0, const int off)
977 struct mldv2_query *mld;
981 IN6_MULTI_LIST_LOCK_ASSERT();
985 mld = (struct mldv2_query *)(mtod(m0, uint8_t *) + off);
987 switch (inm->in6m_state) {
989 case MLD_SILENT_MEMBER:
990 case MLD_SLEEPING_MEMBER:
991 case MLD_LAZY_MEMBER:
992 case MLD_AWAKENING_MEMBER:
993 case MLD_IDLE_MEMBER:
994 case MLD_LEAVING_MEMBER:
997 case MLD_REPORTING_MEMBER:
998 case MLD_G_QUERY_PENDING_MEMBER:
999 case MLD_SG_QUERY_PENDING_MEMBER:
1003 nsrc = ntohs(mld->mld_numsrc);
1006 * Deal with group-specific queries upfront.
1007 * If any group query is already pending, purge any recorded
1008 * source-list state if it exists, and schedule a query response
1009 * for this group-specific query.
1012 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
1013 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) {
1014 in6m_clear_recorded(inm);
1015 timer = min(inm->in6m_timer, timer);
1017 inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER;
1018 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1019 V_current_state_timers_running6 = 1;
1024 * Deal with the case where a group-and-source-specific query has
1025 * been received but a group-specific query is already pending.
1027 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) {
1028 timer = min(inm->in6m_timer, timer);
1029 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1030 V_current_state_timers_running6 = 1;
1035 * Finally, deal with the case where a group-and-source-specific
1036 * query has been received, where a response to a previous g-s-r
1037 * query exists, or none exists.
1038 * In this case, we need to parse the source-list which the Querier
1039 * has provided us with and check if we have any source list filter
1040 * entries at T1 for these sources. If we do not, there is no need
1041 * schedule a report and the query may be dropped.
1042 * If we do, we must record them and schedule a current-state
1043 * report for those sources.
1045 if (inm->in6m_nsrc > 0) {
1052 soff = off + sizeof(struct mldv2_query);
1054 for (i = 0; i < nsrc; i++) {
1055 sp = mtod(m, uint8_t *) + soff;
1056 retval = in6m_record_source(inm,
1057 (const struct in6_addr *)sp);
1060 nrecorded += retval;
1061 soff += sizeof(struct in6_addr);
1062 if (soff >= m->m_len) {
1063 soff = soff - m->m_len;
1069 if (nrecorded > 0) {
1071 "%s: schedule response to SG query", __func__);
1072 inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER;
1073 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1074 V_current_state_timers_running6 = 1;
1082 * Process a received MLDv1 host membership report.
1083 * Assumes mld points to mld_hdr in pulled up mbuf chain.
1085 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
1086 * mld_addr. This is OK as we own the mbuf chain.
1089 mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6,
1090 /*const*/ struct mld_hdr *mld)
1092 struct in6_addr src, dst;
1093 struct in6_ifaddr *ia;
1094 struct in6_multi *inm;
1096 char ip6tbuf[INET6_ADDRSTRLEN];
1099 if (!mld_v1enable) {
1100 CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)",
1101 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1106 if (ifp->if_flags & IFF_LOOPBACK)
1110 * MLDv1 reports must originate from a host's link-local address,
1111 * or the unspecified address (when booting).
1114 in6_clearscope(&src);
1115 if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) {
1116 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
1117 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
1123 * RFC2710 Section 4: MLDv1 reports must pertain to a multicast
1124 * group, and must be directed to the group itself.
1127 in6_clearscope(&dst);
1128 if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) ||
1129 !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) {
1130 CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)",
1131 ip6_sprintf(ip6tbuf, &ip6->ip6_dst),
1137 * Make sure we don't hear our own membership report, as fast
1138 * leave requires knowing that we are the only member of a
1139 * group. Assume we used the link-local address if available,
1140 * otherwise look for ::.
1142 * XXX Note that scope ID comparison is needed for the address
1143 * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be
1144 * performed for the on-wire address.
1146 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1147 if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) ||
1148 (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) {
1150 ifa_free(&ia->ia_ifa);
1154 ifa_free(&ia->ia_ifa);
1156 CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)",
1157 ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp));
1160 * Embed scope ID of receiving interface in MLD query for lookup
1161 * whilst we don't hold other locks (due to KAME locking lameness).
1163 if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr))
1164 in6_setscope(&mld->mld_addr, ifp, NULL);
1166 IN6_MULTI_LIST_LOCK();
1171 * MLDv1 report suppression.
1172 * If we are a member of this group, and our membership should be
1173 * reported, and our group timer is pending or about to be reset,
1174 * stop our group timer by transitioning to the 'lazy' state.
1176 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
1178 struct mld_ifsoftc *mli;
1180 mli = inm->in6m_mli;
1181 KASSERT(mli != NULL,
1182 ("%s: no mli for ifp %p", __func__, ifp));
1185 * If we are in MLDv2 host mode, do not allow the
1186 * other host's MLDv1 report to suppress our reports.
1188 if (mli->mli_version == MLD_VERSION_2)
1191 inm->in6m_timer = 0;
1193 switch (inm->in6m_state) {
1194 case MLD_NOT_MEMBER:
1195 case MLD_SILENT_MEMBER:
1196 case MLD_SLEEPING_MEMBER:
1198 case MLD_REPORTING_MEMBER:
1199 case MLD_IDLE_MEMBER:
1200 case MLD_AWAKENING_MEMBER:
1202 "report suppressed for %s on ifp %p(%s)",
1203 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1205 case MLD_LAZY_MEMBER:
1206 inm->in6m_state = MLD_LAZY_MEMBER;
1208 case MLD_G_QUERY_PENDING_MEMBER:
1209 case MLD_SG_QUERY_PENDING_MEMBER:
1210 case MLD_LEAVING_MEMBER:
1216 IF_ADDR_RUNLOCK(ifp);
1218 IN6_MULTI_LIST_UNLOCK();
1220 /* XXX Clear embedded scope ID as userland won't expect it. */
1221 in6_clearscope(&mld->mld_addr);
1229 * Assume query messages which fit in a single ICMPv6 message header
1230 * have been pulled up.
1231 * Assume that userland will want to see the message, even if it
1232 * otherwise fails kernel input validation; do not free it.
1233 * Pullup may however free the mbuf chain m if it fails.
1235 * Return IPPROTO_DONE if we freed m. Otherwise, return 0.
1238 mld_input(struct mbuf *m, int off, int icmp6len)
1241 struct ip6_hdr *ip6;
1242 struct mld_hdr *mld;
1245 CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off);
1247 ifp = m->m_pkthdr.rcvif;
1249 ip6 = mtod(m, struct ip6_hdr *);
1251 /* Pullup to appropriate size. */
1252 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1253 if (mld->mld_type == MLD_LISTENER_QUERY &&
1254 icmp6len >= sizeof(struct mldv2_query)) {
1255 mldlen = sizeof(struct mldv2_query);
1257 mldlen = sizeof(struct mld_hdr);
1259 IP6_EXTHDR_GET(mld, struct mld_hdr *, m, off, mldlen);
1261 ICMP6STAT_INC(icp6s_badlen);
1262 return (IPPROTO_DONE);
1266 * Userland needs to see all of this traffic for implementing
1267 * the endpoint discovery portion of multicast routing.
1269 switch (mld->mld_type) {
1270 case MLD_LISTENER_QUERY:
1271 icmp6_ifstat_inc(ifp, ifs6_in_mldquery);
1272 if (icmp6len == sizeof(struct mld_hdr)) {
1273 if (mld_v1_input_query(ifp, ip6, mld) != 0)
1275 } else if (icmp6len >= sizeof(struct mldv2_query)) {
1276 if (mld_v2_input_query(ifp, ip6, m, off,
1281 case MLD_LISTENER_REPORT:
1282 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1283 if (mld_v1_input_report(ifp, ip6, mld) != 0)
1286 case MLDV2_LISTENER_REPORT:
1287 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1289 case MLD_LISTENER_DONE:
1290 icmp6_ifstat_inc(ifp, ifs6_in_mlddone);
1300 * Fast timeout handler (global).
1301 * VIMAGE: Timeout handlers are expected to service all vimages.
1306 VNET_ITERATOR_DECL(vnet_iter);
1308 VNET_LIST_RLOCK_NOSLEEP();
1309 VNET_FOREACH(vnet_iter) {
1310 CURVNET_SET(vnet_iter);
1311 mld_fasttimo_vnet();
1314 VNET_LIST_RUNLOCK_NOSLEEP();
1318 * Fast timeout handler (per-vnet).
1320 * VIMAGE: Assume caller has set up our curvnet.
1323 mld_fasttimo_vnet(void)
1325 struct mbufq scq; /* State-change packets */
1326 struct mbufq qrq; /* Query response packets */
1328 struct mld_ifsoftc *mli;
1329 struct ifmultiaddr *ifma;
1330 struct in6_multi *inm, *tinm;
1331 struct in6_multi_head inmh;
1337 * Quick check to see if any work needs to be done, in order to
1338 * minimize the overhead of fasttimo processing.
1339 * SMPng: XXX Unlocked reads.
1341 if (!V_current_state_timers_running6 &&
1342 !V_interface_timers_running6 &&
1343 !V_state_change_timers_running6)
1347 IN6_MULTI_LIST_LOCK();
1351 * MLDv2 General Query response timer processing.
1353 if (V_interface_timers_running6) {
1354 CTR1(KTR_MLD, "%s: interface timers running", __func__);
1356 V_interface_timers_running6 = 0;
1357 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1358 if (mli->mli_v2_timer == 0) {
1360 } else if (--mli->mli_v2_timer == 0) {
1361 mld_v2_dispatch_general_query(mli);
1363 V_interface_timers_running6 = 1;
1368 if (!V_current_state_timers_running6 &&
1369 !V_state_change_timers_running6)
1372 V_current_state_timers_running6 = 0;
1373 V_state_change_timers_running6 = 0;
1375 CTR1(KTR_MLD, "%s: state change timers running", __func__);
1378 * MLD host report and state-change timer processing.
1379 * Note: Processing a v2 group timer may remove a node.
1381 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1384 if (mli->mli_version == MLD_VERSION_2) {
1385 uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri *
1387 mbufq_init(&qrq, MLD_MAX_G_GS_PACKETS);
1388 mbufq_init(&scq, MLD_MAX_STATE_CHANGE_PACKETS);
1392 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1393 if (ifma->ifma_addr->sa_family != AF_INET6 ||
1394 ifma->ifma_protospec == NULL)
1396 inm = (struct in6_multi *)ifma->ifma_protospec;
1397 switch (mli->mli_version) {
1399 mld_v1_process_group_timer(&inmh, inm);
1402 mld_v2_process_group_timers(&inmh, &qrq,
1403 &scq, inm, uri_fasthz);
1407 IF_ADDR_RUNLOCK(ifp);
1409 switch (mli->mli_version) {
1412 * Transmit reports for this lifecycle. This
1413 * is done while not holding IF_ADDR_LOCK
1414 * since this can call
1415 * in6ifa_ifpforlinklocal() which locks
1416 * IF_ADDR_LOCK internally as well as
1417 * ip6_output() to transmit a packet.
1419 SLIST_FOREACH_SAFE(inm, &inmh, in6m_nrele, tinm) {
1420 SLIST_REMOVE_HEAD(&inmh,
1422 (void)mld_v1_transmit_report(inm,
1423 MLD_LISTENER_REPORT);
1427 mld_dispatch_queue(&qrq, 0);
1428 mld_dispatch_queue(&scq, 0);
1431 * Free the in_multi reference(s) for
1434 in6m_release_list_deferred(&inmh);
1441 IN6_MULTI_LIST_UNLOCK();
1445 * Update host report group timer.
1446 * Will update the global pending timer flags.
1449 mld_v1_process_group_timer(struct in6_multi_head *inmh, struct in6_multi *inm)
1451 int report_timer_expired;
1453 IN6_MULTI_LIST_LOCK_ASSERT();
1456 if (inm->in6m_timer == 0) {
1457 report_timer_expired = 0;
1458 } else if (--inm->in6m_timer == 0) {
1459 report_timer_expired = 1;
1461 V_current_state_timers_running6 = 1;
1465 switch (inm->in6m_state) {
1466 case MLD_NOT_MEMBER:
1467 case MLD_SILENT_MEMBER:
1468 case MLD_IDLE_MEMBER:
1469 case MLD_LAZY_MEMBER:
1470 case MLD_SLEEPING_MEMBER:
1471 case MLD_AWAKENING_MEMBER:
1473 case MLD_REPORTING_MEMBER:
1474 if (report_timer_expired) {
1475 inm->in6m_state = MLD_IDLE_MEMBER;
1476 in6m_rele_locked(inmh, inm);
1479 case MLD_G_QUERY_PENDING_MEMBER:
1480 case MLD_SG_QUERY_PENDING_MEMBER:
1481 case MLD_LEAVING_MEMBER:
1487 * Update a group's timers for MLDv2.
1488 * Will update the global pending timer flags.
1489 * Note: Unlocked read from mli.
1492 mld_v2_process_group_timers(struct in6_multi_head *inmh,
1493 struct mbufq *qrq, struct mbufq *scq,
1494 struct in6_multi *inm, const int uri_fasthz)
1496 int query_response_timer_expired;
1497 int state_change_retransmit_timer_expired;
1499 char ip6tbuf[INET6_ADDRSTRLEN];
1502 IN6_MULTI_LIST_LOCK_ASSERT();
1505 query_response_timer_expired = 0;
1506 state_change_retransmit_timer_expired = 0;
1509 * During a transition from compatibility mode back to MLDv2,
1510 * a group record in REPORTING state may still have its group
1511 * timer active. This is a no-op in this function; it is easier
1512 * to deal with it here than to complicate the slow-timeout path.
1514 if (inm->in6m_timer == 0) {
1515 query_response_timer_expired = 0;
1516 } else if (--inm->in6m_timer == 0) {
1517 query_response_timer_expired = 1;
1519 V_current_state_timers_running6 = 1;
1522 if (inm->in6m_sctimer == 0) {
1523 state_change_retransmit_timer_expired = 0;
1524 } else if (--inm->in6m_sctimer == 0) {
1525 state_change_retransmit_timer_expired = 1;
1527 V_state_change_timers_running6 = 1;
1530 /* We are in fasttimo, so be quick about it. */
1531 if (!state_change_retransmit_timer_expired &&
1532 !query_response_timer_expired)
1535 switch (inm->in6m_state) {
1536 case MLD_NOT_MEMBER:
1537 case MLD_SILENT_MEMBER:
1538 case MLD_SLEEPING_MEMBER:
1539 case MLD_LAZY_MEMBER:
1540 case MLD_AWAKENING_MEMBER:
1541 case MLD_IDLE_MEMBER:
1543 case MLD_G_QUERY_PENDING_MEMBER:
1544 case MLD_SG_QUERY_PENDING_MEMBER:
1546 * Respond to a previously pending Group-Specific
1547 * or Group-and-Source-Specific query by enqueueing
1548 * the appropriate Current-State report for
1549 * immediate transmission.
1551 if (query_response_timer_expired) {
1554 retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1,
1555 (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER),
1557 CTR2(KTR_MLD, "%s: enqueue record = %d",
1559 inm->in6m_state = MLD_REPORTING_MEMBER;
1560 in6m_clear_recorded(inm);
1563 case MLD_REPORTING_MEMBER:
1564 case MLD_LEAVING_MEMBER:
1565 if (state_change_retransmit_timer_expired) {
1567 * State-change retransmission timer fired.
1568 * If there are any further pending retransmissions,
1569 * set the global pending state-change flag, and
1572 if (--inm->in6m_scrv > 0) {
1573 inm->in6m_sctimer = uri_fasthz;
1574 V_state_change_timers_running6 = 1;
1577 * Retransmit the previously computed state-change
1578 * report. If there are no further pending
1579 * retransmissions, the mbuf queue will be consumed.
1580 * Update T0 state to T1 as we have now sent
1583 (void)mld_v2_merge_state_changes(inm, scq);
1586 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
1587 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1588 if_name(inm->in6m_ifp));
1591 * If we are leaving the group for good, make sure
1592 * we release MLD's reference to it.
1593 * This release must be deferred using a SLIST,
1594 * as we are called from a loop which traverses
1595 * the in_ifmultiaddr TAILQ.
1597 if (inm->in6m_state == MLD_LEAVING_MEMBER &&
1598 inm->in6m_scrv == 0) {
1599 inm->in6m_state = MLD_NOT_MEMBER;
1600 in6m_rele_locked(inmh, inm);
1608 * Switch to a different version on the given interface,
1609 * as per Section 9.12.
1612 mld_set_version(struct mld_ifsoftc *mli, const int version)
1614 int old_version_timer;
1618 CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__,
1619 version, mli->mli_ifp, if_name(mli->mli_ifp));
1621 if (version == MLD_VERSION_1) {
1623 * Compute the "Older Version Querier Present" timer as per
1626 old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri;
1627 old_version_timer *= PR_SLOWHZ;
1628 mli->mli_v1_timer = old_version_timer;
1631 if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) {
1632 mli->mli_version = MLD_VERSION_1;
1633 mld_v2_cancel_link_timers(mli);
1638 * Cancel pending MLDv2 timers for the given link and all groups
1639 * joined on it; state-change, general-query, and group-query timers.
1642 mld_v2_cancel_link_timers(struct mld_ifsoftc *mli)
1644 struct ifmultiaddr *ifma;
1646 struct in6_multi *inm;
1647 struct in6_multi_head inmh;
1649 CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__,
1650 mli->mli_ifp, if_name(mli->mli_ifp));
1653 IN6_MULTI_LIST_LOCK_ASSERT();
1657 * Fast-track this potentially expensive operation
1658 * by checking all the global 'timer pending' flags.
1660 if (!V_interface_timers_running6 &&
1661 !V_state_change_timers_running6 &&
1662 !V_current_state_timers_running6)
1665 mli->mli_v2_timer = 0;
1670 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1671 if (ifma->ifma_addr->sa_family != AF_INET6)
1673 inm = (struct in6_multi *)ifma->ifma_protospec;
1674 switch (inm->in6m_state) {
1675 case MLD_NOT_MEMBER:
1676 case MLD_SILENT_MEMBER:
1677 case MLD_IDLE_MEMBER:
1678 case MLD_LAZY_MEMBER:
1679 case MLD_SLEEPING_MEMBER:
1680 case MLD_AWAKENING_MEMBER:
1682 case MLD_LEAVING_MEMBER:
1684 * If we are leaving the group and switching
1685 * version, we need to release the final
1686 * reference held for issuing the INCLUDE {}.
1688 in6m_rele_locked(&inmh, inm);
1689 ifma->ifma_protospec = NULL;
1691 case MLD_G_QUERY_PENDING_MEMBER:
1692 case MLD_SG_QUERY_PENDING_MEMBER:
1693 in6m_clear_recorded(inm);
1695 case MLD_REPORTING_MEMBER:
1696 inm->in6m_sctimer = 0;
1697 inm->in6m_timer = 0;
1698 inm->in6m_state = MLD_REPORTING_MEMBER;
1700 * Free any pending MLDv2 state-change records.
1702 mbufq_drain(&inm->in6m_scq);
1706 IF_ADDR_RUNLOCK(ifp);
1707 in6m_release_list_deferred(&inmh);
1711 * Global slowtimo handler.
1712 * VIMAGE: Timeout handlers are expected to service all vimages.
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();
1729 * Per-vnet slowtimo handler.
1732 mld_slowtimo_vnet(void)
1734 struct mld_ifsoftc *mli;
1738 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1739 mld_v1_process_querier_timers(mli);
1746 * Update the Older Version Querier Present timers for a link.
1747 * See Section 9.12 of RFC 3810.
1750 mld_v1_process_querier_timers(struct mld_ifsoftc *mli)
1755 if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) {
1757 * MLDv1 Querier Present timer expired; revert to MLDv2.
1760 "%s: transition from v%d -> v%d on %p(%s)",
1761 __func__, mli->mli_version, MLD_VERSION_2,
1762 mli->mli_ifp, if_name(mli->mli_ifp));
1763 mli->mli_version = MLD_VERSION_2;
1768 * Transmit an MLDv1 report immediately.
1771 mld_v1_transmit_report(struct in6_multi *in6m, const int type)
1774 struct in6_ifaddr *ia;
1775 struct ip6_hdr *ip6;
1776 struct mbuf *mh, *md;
1777 struct mld_hdr *mld;
1779 IN6_MULTI_LIST_LOCK_ASSERT();
1782 ifp = in6m->in6m_ifp;
1783 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1784 /* ia may be NULL if link-local address is tentative. */
1786 mh = m_gethdr(M_NOWAIT, MT_DATA);
1789 ifa_free(&ia->ia_ifa);
1792 md = m_get(M_NOWAIT, MT_DATA);
1796 ifa_free(&ia->ia_ifa);
1802 * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so
1803 * that ether_output() does not need to allocate another mbuf
1804 * for the header in the most common case.
1806 M_ALIGN(mh, sizeof(struct ip6_hdr));
1807 mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
1808 mh->m_len = sizeof(struct ip6_hdr);
1810 ip6 = mtod(mh, struct ip6_hdr *);
1812 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1813 ip6->ip6_vfc |= IPV6_VERSION;
1814 ip6->ip6_nxt = IPPROTO_ICMPV6;
1815 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
1816 ip6->ip6_dst = in6m->in6m_addr;
1818 md->m_len = sizeof(struct mld_hdr);
1819 mld = mtod(md, struct mld_hdr *);
1820 mld->mld_type = type;
1823 mld->mld_maxdelay = 0;
1824 mld->mld_reserved = 0;
1825 mld->mld_addr = in6m->in6m_addr;
1826 in6_clearscope(&mld->mld_addr);
1827 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
1828 sizeof(struct ip6_hdr), sizeof(struct mld_hdr));
1830 mld_save_context(mh, ifp);
1831 mh->m_flags |= M_MLDV1;
1833 mld_dispatch_packet(mh);
1836 ifa_free(&ia->ia_ifa);
1841 * Process a state change from the upper layer for the given IPv6 group.
1843 * Each socket holds a reference on the in_multi in its own ip_moptions.
1844 * The socket layer will have made the necessary updates to.the group
1845 * state, it is now up to MLD to issue a state change report if there
1846 * has been any change between T0 (when the last state-change was issued)
1849 * We use the MLDv2 state machine at group level. The MLd module
1850 * however makes the decision as to which MLD protocol version to speak.
1851 * A state change *from* INCLUDE {} always means an initial join.
1852 * A state change *to* INCLUDE {} always means a final leave.
1854 * If delay is non-zero, and the state change is an initial multicast
1855 * join, the state change report will be delayed by 'delay' ticks
1856 * in units of PR_FASTHZ if MLDv1 is active on the link; otherwise
1857 * the initial MLDv2 state change report will be delayed by whichever
1858 * is sooner, a pending state-change timer or delay itself.
1860 * VIMAGE: curvnet should have been set by caller, as this routine
1861 * is called from the socket option handlers.
1864 mld_change_state(struct in6_multi *inm, const int delay)
1866 struct mld_ifsoftc *mli;
1870 IN6_MULTI_LIST_LOCK_ASSERT();
1875 * Try to detect if the upper layer just asked us to change state
1876 * for an interface which has now gone away.
1878 KASSERT(inm->in6m_ifma != NULL, ("%s: no ifma", __func__));
1879 ifp = inm->in6m_ifma->ifma_ifp;
1882 * Sanity check that netinet6's notion of ifp is the
1885 KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__));
1890 mli = MLD_IFINFO(ifp);
1891 KASSERT(mli != NULL, ("%s: no mld_ifsoftc for ifp %p", __func__, ifp));
1894 * If we detect a state transition to or from MCAST_UNDEFINED
1895 * for this group, then we are starting or finishing an MLD
1896 * life cycle for this group.
1898 if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) {
1899 CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__,
1900 inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode);
1901 if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) {
1902 CTR1(KTR_MLD, "%s: initial join", __func__);
1903 error = mld_initial_join(inm, mli, delay);
1905 } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) {
1906 CTR1(KTR_MLD, "%s: final leave", __func__);
1907 mld_final_leave(inm, mli);
1911 CTR1(KTR_MLD, "%s: filter set change", __func__);
1914 error = mld_handle_state_change(inm, mli);
1922 * Perform the initial join for an MLD group.
1924 * When joining a group:
1925 * If the group should have its MLD traffic suppressed, do nothing.
1926 * MLDv1 starts sending MLDv1 host membership reports.
1927 * MLDv2 will schedule an MLDv2 state-change report containing the
1928 * initial state of the membership.
1930 * If the delay argument is non-zero, then we must delay sending the
1931 * initial state change for delay ticks (in units of PR_FASTHZ).
1934 mld_initial_join(struct in6_multi *inm, struct mld_ifsoftc *mli,
1939 int error, retval, syncstates;
1942 char ip6tbuf[INET6_ADDRSTRLEN];
1945 CTR4(KTR_MLD, "%s: initial join %s on ifp %p(%s)",
1946 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1947 inm->in6m_ifp, if_name(inm->in6m_ifp));
1952 ifp = inm->in6m_ifp;
1954 IN6_MULTI_LIST_LOCK_ASSERT();
1957 KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__));
1960 * Groups joined on loopback or marked as 'not reported',
1961 * enter the MLD_SILENT_MEMBER state and
1962 * are never reported in any protocol exchanges.
1963 * All other groups enter the appropriate state machine
1964 * for the version in use on this link.
1965 * A link marked as MLIF_SILENT causes MLD to be completely
1966 * disabled for the link.
1968 if ((ifp->if_flags & IFF_LOOPBACK) ||
1969 (mli->mli_flags & MLIF_SILENT) ||
1970 !mld_is_addr_reported(&inm->in6m_addr)) {
1972 "%s: not kicking state machine for silent group", __func__);
1973 inm->in6m_state = MLD_SILENT_MEMBER;
1974 inm->in6m_timer = 0;
1977 * Deal with overlapping in_multi lifecycle.
1978 * If this group was LEAVING, then make sure
1979 * we drop the reference we picked up to keep the
1980 * group around for the final INCLUDE {} enqueue.
1982 if (mli->mli_version == MLD_VERSION_2 &&
1983 inm->in6m_state == MLD_LEAVING_MEMBER)
1984 in6m_release_deferred(inm);
1986 inm->in6m_state = MLD_REPORTING_MEMBER;
1988 switch (mli->mli_version) {
1991 * If a delay was provided, only use it if
1992 * it is greater than the delay normally
1993 * used for an MLDv1 state change report,
1994 * and delay sending the initial MLDv1 report
1995 * by not transitioning to the IDLE state.
1997 odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * PR_FASTHZ);
1999 inm->in6m_timer = max(delay, odelay);
2000 V_current_state_timers_running6 = 1;
2002 inm->in6m_state = MLD_IDLE_MEMBER;
2003 error = mld_v1_transmit_report(inm,
2004 MLD_LISTENER_REPORT);
2006 inm->in6m_timer = odelay;
2007 V_current_state_timers_running6 = 1;
2014 * Defer update of T0 to T1, until the first copy
2015 * of the state change has been transmitted.
2020 * Immediately enqueue a State-Change Report for
2021 * this interface, freeing any previous reports.
2022 * Don't kick the timers if there is nothing to do,
2023 * or if an error occurred.
2025 mq = &inm->in6m_scq;
2027 retval = mld_v2_enqueue_group_record(mq, inm, 1,
2028 0, 0, (mli->mli_flags & MLIF_USEALLOW));
2029 CTR2(KTR_MLD, "%s: enqueue record = %d",
2032 error = retval * -1;
2037 * Schedule transmission of pending state-change
2038 * report up to RV times for this link. The timer
2039 * will fire at the next mld_fasttimo (~200ms),
2040 * giving us an opportunity to merge the reports.
2042 * If a delay was provided to this function, only
2043 * use this delay if sooner than the existing one.
2045 KASSERT(mli->mli_rv > 1,
2046 ("%s: invalid robustness %d", __func__,
2048 inm->in6m_scrv = mli->mli_rv;
2050 if (inm->in6m_sctimer > 1) {
2052 min(inm->in6m_sctimer, delay);
2054 inm->in6m_sctimer = delay;
2056 inm->in6m_sctimer = 1;
2057 V_state_change_timers_running6 = 1;
2065 * Only update the T0 state if state change is atomic,
2066 * i.e. we don't need to wait for a timer to fire before we
2067 * can consider the state change to have been communicated.
2071 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2072 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2073 if_name(inm->in6m_ifp));
2080 * Issue an intermediate state change during the life-cycle.
2083 mld_handle_state_change(struct in6_multi *inm, struct mld_ifsoftc *mli)
2088 char ip6tbuf[INET6_ADDRSTRLEN];
2091 CTR4(KTR_MLD, "%s: state change for %s on ifp %p(%s)",
2092 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2093 inm->in6m_ifp, if_name(inm->in6m_ifp));
2095 ifp = inm->in6m_ifp;
2097 IN6_MULTI_LIST_LOCK_ASSERT();
2100 KASSERT(mli && mli->mli_ifp == ifp,
2101 ("%s: inconsistent ifp", __func__));
2103 if ((ifp->if_flags & IFF_LOOPBACK) ||
2104 (mli->mli_flags & MLIF_SILENT) ||
2105 !mld_is_addr_reported(&inm->in6m_addr) ||
2106 (mli->mli_version != MLD_VERSION_2)) {
2107 if (!mld_is_addr_reported(&inm->in6m_addr)) {
2109 "%s: not kicking state machine for silent group", __func__);
2111 CTR1(KTR_MLD, "%s: nothing to do", __func__);
2113 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2114 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2115 if_name(inm->in6m_ifp));
2119 mbufq_drain(&inm->in6m_scq);
2121 retval = mld_v2_enqueue_group_record(&inm->in6m_scq, inm, 1, 0, 0,
2122 (mli->mli_flags & MLIF_USEALLOW));
2123 CTR2(KTR_MLD, "%s: enqueue record = %d", __func__, retval);
2128 * If record(s) were enqueued, start the state-change
2129 * report timer for this group.
2131 inm->in6m_scrv = mli->mli_rv;
2132 inm->in6m_sctimer = 1;
2133 V_state_change_timers_running6 = 1;
2139 * Perform the final leave for a multicast address.
2141 * When leaving a group:
2142 * MLDv1 sends a DONE message, if and only if we are the reporter.
2143 * MLDv2 enqueues a state-change report containing a transition
2144 * to INCLUDE {} for immediate transmission.
2147 mld_final_leave(struct in6_multi *inm, struct mld_ifsoftc *mli)
2151 char ip6tbuf[INET6_ADDRSTRLEN];
2156 CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)",
2157 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2158 inm->in6m_ifp, if_name(inm->in6m_ifp));
2160 IN6_MULTI_LIST_LOCK_ASSERT();
2163 switch (inm->in6m_state) {
2164 case MLD_NOT_MEMBER:
2165 case MLD_SILENT_MEMBER:
2166 case MLD_LEAVING_MEMBER:
2167 /* Already leaving or left; do nothing. */
2169 "%s: not kicking state machine for silent group", __func__);
2171 case MLD_REPORTING_MEMBER:
2172 case MLD_IDLE_MEMBER:
2173 case MLD_G_QUERY_PENDING_MEMBER:
2174 case MLD_SG_QUERY_PENDING_MEMBER:
2175 if (mli->mli_version == MLD_VERSION_1) {
2177 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
2178 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER)
2179 panic("%s: MLDv2 state reached, not MLDv2 mode",
2182 mld_v1_transmit_report(inm, MLD_LISTENER_DONE);
2183 inm->in6m_state = MLD_NOT_MEMBER;
2184 V_current_state_timers_running6 = 1;
2185 } else if (mli->mli_version == MLD_VERSION_2) {
2187 * Stop group timer and all pending reports.
2188 * Immediately enqueue a state-change report
2189 * TO_IN {} to be sent on the next fast timeout,
2190 * giving us an opportunity to merge reports.
2192 mbufq_drain(&inm->in6m_scq);
2193 inm->in6m_timer = 0;
2194 inm->in6m_scrv = mli->mli_rv;
2195 CTR4(KTR_MLD, "%s: Leaving %s/%s with %d "
2196 "pending retransmissions.", __func__,
2197 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2198 if_name(inm->in6m_ifp), inm->in6m_scrv);
2199 if (inm->in6m_scrv == 0) {
2200 inm->in6m_state = MLD_NOT_MEMBER;
2201 inm->in6m_sctimer = 0;
2205 in6m_acquire_locked(inm);
2207 retval = mld_v2_enqueue_group_record(
2208 &inm->in6m_scq, inm, 1, 0, 0,
2209 (mli->mli_flags & MLIF_USEALLOW));
2210 KASSERT(retval != 0,
2211 ("%s: enqueue record = %d", __func__,
2214 inm->in6m_state = MLD_LEAVING_MEMBER;
2215 inm->in6m_sctimer = 1;
2216 V_state_change_timers_running6 = 1;
2222 case MLD_LAZY_MEMBER:
2223 case MLD_SLEEPING_MEMBER:
2224 case MLD_AWAKENING_MEMBER:
2225 /* Our reports are suppressed; do nothing. */
2231 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2232 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2233 if_name(inm->in6m_ifp));
2234 inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED;
2235 CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s",
2236 __func__, &inm->in6m_addr, if_name(inm->in6m_ifp));
2241 * Enqueue an MLDv2 group record to the given output queue.
2243 * If is_state_change is zero, a current-state record is appended.
2244 * If is_state_change is non-zero, a state-change report is appended.
2246 * If is_group_query is non-zero, an mbuf packet chain is allocated.
2247 * If is_group_query is zero, and if there is a packet with free space
2248 * at the tail of the queue, it will be appended to providing there
2249 * is enough free space.
2250 * Otherwise a new mbuf packet chain is allocated.
2252 * If is_source_query is non-zero, each source is checked to see if
2253 * it was recorded for a Group-Source query, and will be omitted if
2254 * it is not both in-mode and recorded.
2256 * If use_block_allow is non-zero, state change reports for initial join
2257 * and final leave, on an inclusive mode group with a source list, will be
2258 * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively.
2260 * The function will attempt to allocate leading space in the packet
2261 * for the IPv6+ICMP headers to be prepended without fragmenting the chain.
2263 * If successful the size of all data appended to the queue is returned,
2264 * otherwise an error code less than zero is returned, or zero if
2265 * no record(s) were appended.
2268 mld_v2_enqueue_group_record(struct mbufq *mq, struct in6_multi *inm,
2269 const int is_state_change, const int is_group_query,
2270 const int is_source_query, const int use_block_allow)
2272 struct mldv2_record mr;
2273 struct mldv2_record *pmr;
2275 struct ip6_msource *ims, *nims;
2276 struct mbuf *m0, *m, *md;
2277 int is_filter_list_change;
2278 int minrec0len, m0srcs, msrcs, nbytes, off;
2279 int record_has_sources;
2284 char ip6tbuf[INET6_ADDRSTRLEN];
2287 IN6_MULTI_LIST_LOCK_ASSERT();
2289 ifp = inm->in6m_ifp;
2290 is_filter_list_change = 0;
2297 record_has_sources = 1;
2299 type = MLD_DO_NOTHING;
2300 mode = inm->in6m_st[1].iss_fmode;
2303 * If we did not transition out of ASM mode during t0->t1,
2304 * and there are no source nodes to process, we can skip
2305 * the generation of source records.
2307 if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 &&
2308 inm->in6m_nsrc == 0)
2309 record_has_sources = 0;
2311 if (is_state_change) {
2313 * Queue a state change record.
2314 * If the mode did not change, and there are non-ASM
2315 * listeners or source filters present,
2316 * we potentially need to issue two records for the group.
2317 * If there are ASM listeners, and there was no filter
2318 * mode transition of any kind, do nothing.
2320 * If we are transitioning to MCAST_UNDEFINED, we need
2321 * not send any sources. A transition to/from this state is
2322 * considered inclusive with some special treatment.
2324 * If we are rewriting initial joins/leaves to use
2325 * ALLOW/BLOCK, and the group's membership is inclusive,
2326 * we need to send sources in all cases.
2328 if (mode != inm->in6m_st[0].iss_fmode) {
2329 if (mode == MCAST_EXCLUDE) {
2330 CTR1(KTR_MLD, "%s: change to EXCLUDE",
2332 type = MLD_CHANGE_TO_EXCLUDE_MODE;
2334 CTR1(KTR_MLD, "%s: change to INCLUDE",
2336 if (use_block_allow) {
2339 * Here we're interested in state
2340 * edges either direction between
2341 * MCAST_UNDEFINED and MCAST_INCLUDE.
2342 * Perhaps we should just check
2343 * the group state, rather than
2346 if (mode == MCAST_UNDEFINED) {
2347 type = MLD_BLOCK_OLD_SOURCES;
2349 type = MLD_ALLOW_NEW_SOURCES;
2352 type = MLD_CHANGE_TO_INCLUDE_MODE;
2353 if (mode == MCAST_UNDEFINED)
2354 record_has_sources = 0;
2358 if (record_has_sources) {
2359 is_filter_list_change = 1;
2361 type = MLD_DO_NOTHING;
2366 * Queue a current state record.
2368 if (mode == MCAST_EXCLUDE) {
2369 type = MLD_MODE_IS_EXCLUDE;
2370 } else if (mode == MCAST_INCLUDE) {
2371 type = MLD_MODE_IS_INCLUDE;
2372 KASSERT(inm->in6m_st[1].iss_asm == 0,
2373 ("%s: inm %p is INCLUDE but ASM count is %d",
2374 __func__, inm, inm->in6m_st[1].iss_asm));
2379 * Generate the filter list changes using a separate function.
2381 if (is_filter_list_change)
2382 return (mld_v2_enqueue_filter_change(mq, inm));
2384 if (type == MLD_DO_NOTHING) {
2385 CTR3(KTR_MLD, "%s: nothing to do for %s/%s",
2386 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2387 if_name(inm->in6m_ifp));
2392 * If any sources are present, we must be able to fit at least
2393 * one in the trailing space of the tail packet's mbuf,
2396 minrec0len = sizeof(struct mldv2_record);
2397 if (record_has_sources)
2398 minrec0len += sizeof(struct in6_addr);
2400 CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__,
2401 mld_rec_type_to_str(type),
2402 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2403 if_name(inm->in6m_ifp));
2406 * Check if we have a packet in the tail of the queue for this
2407 * group into which the first group record for this group will fit.
2408 * Otherwise allocate a new packet.
2409 * Always allocate leading space for IP6+RA+ICMPV6+REPORT.
2410 * Note: Group records for G/GSR query responses MUST be sent
2411 * in their own packet.
2413 m0 = mbufq_last(mq);
2414 if (!is_group_query &&
2416 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) &&
2417 (m0->m_pkthdr.len + minrec0len) <
2418 (ifp->if_mtu - MLD_MTUSPACE)) {
2419 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2420 sizeof(struct mldv2_record)) /
2421 sizeof(struct in6_addr);
2423 CTR1(KTR_MLD, "%s: use existing packet", __func__);
2425 if (mbufq_full(mq)) {
2426 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2430 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2431 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2432 if (!is_state_change && !is_group_query)
2433 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2435 m = m_gethdr(M_NOWAIT, MT_DATA);
2439 mld_save_context(m, ifp);
2441 CTR1(KTR_MLD, "%s: allocated first packet", __func__);
2445 * Append group record.
2446 * If we have sources, we don't know how many yet.
2451 mr.mr_addr = inm->in6m_addr;
2452 in6_clearscope(&mr.mr_addr);
2453 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2456 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2459 nbytes += sizeof(struct mldv2_record);
2462 * Append as many sources as will fit in the first packet.
2463 * If we are appending to a new packet, the chain allocation
2464 * may potentially use clusters; use m_getptr() in this case.
2465 * If we are appending to an existing packet, we need to obtain
2466 * a pointer to the group record after m_append(), in case a new
2467 * mbuf was allocated.
2469 * Only append sources which are in-mode at t1. If we are
2470 * transitioning to MCAST_UNDEFINED state on the group, and
2471 * use_block_allow is zero, do not include source entries.
2472 * Otherwise, we need to include this source in the report.
2474 * Only report recorded sources in our filter set when responding
2475 * to a group-source query.
2477 if (record_has_sources) {
2480 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2481 md->m_len - nbytes);
2483 md = m_getptr(m, 0, &off);
2484 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2488 RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs,
2490 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2491 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2492 now = im6s_get_mode(inm, ims, 1);
2493 CTR2(KTR_MLD, "%s: node is %d", __func__, now);
2494 if ((now != mode) ||
2496 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2497 CTR1(KTR_MLD, "%s: skip node", __func__);
2500 if (is_source_query && ims->im6s_stp == 0) {
2501 CTR1(KTR_MLD, "%s: skip unrecorded node",
2505 CTR1(KTR_MLD, "%s: append node", __func__);
2506 if (!m_append(m, sizeof(struct in6_addr),
2507 (void *)&ims->im6s_addr)) {
2510 CTR1(KTR_MLD, "%s: m_append() failed.",
2514 nbytes += sizeof(struct in6_addr);
2516 if (msrcs == m0srcs)
2519 CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__,
2521 pmr->mr_numsrc = htons(msrcs);
2522 nbytes += (msrcs * sizeof(struct in6_addr));
2525 if (is_source_query && msrcs == 0) {
2526 CTR1(KTR_MLD, "%s: no recorded sources to report", __func__);
2533 * We are good to go with first packet.
2536 CTR1(KTR_MLD, "%s: enqueueing first packet", __func__);
2537 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2538 mbufq_enqueue(mq, m);
2540 m->m_pkthdr.PH_vt.vt_nrecs++;
2543 * No further work needed if no source list in packet(s).
2545 if (!record_has_sources)
2549 * Whilst sources remain to be announced, we need to allocate
2550 * a new packet and fill out as many sources as will fit.
2551 * Always try for a cluster first.
2553 while (nims != NULL) {
2554 if (mbufq_full(mq)) {
2555 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2558 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2560 m = m_gethdr(M_NOWAIT, MT_DATA);
2563 mld_save_context(m, ifp);
2564 md = m_getptr(m, 0, &off);
2565 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off);
2566 CTR1(KTR_MLD, "%s: allocated next packet", __func__);
2568 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2571 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2574 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2575 nbytes += sizeof(struct mldv2_record);
2577 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2578 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2581 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2582 CTR2(KTR_MLD, "%s: visit node %s",
2583 __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2584 now = im6s_get_mode(inm, ims, 1);
2585 if ((now != mode) ||
2587 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2588 CTR1(KTR_MLD, "%s: skip node", __func__);
2591 if (is_source_query && ims->im6s_stp == 0) {
2592 CTR1(KTR_MLD, "%s: skip unrecorded node",
2596 CTR1(KTR_MLD, "%s: append node", __func__);
2597 if (!m_append(m, sizeof(struct in6_addr),
2598 (void *)&ims->im6s_addr)) {
2601 CTR1(KTR_MLD, "%s: m_append() failed.",
2606 if (msrcs == m0srcs)
2609 pmr->mr_numsrc = htons(msrcs);
2610 nbytes += (msrcs * sizeof(struct in6_addr));
2612 CTR1(KTR_MLD, "%s: enqueueing next packet", __func__);
2613 mbufq_enqueue(mq, m);
2620 * Type used to mark record pass completion.
2621 * We exploit the fact we can cast to this easily from the
2622 * current filter modes on each ip_msource node.
2625 REC_NONE = 0x00, /* MCAST_UNDEFINED */
2626 REC_ALLOW = 0x01, /* MCAST_INCLUDE */
2627 REC_BLOCK = 0x02, /* MCAST_EXCLUDE */
2628 REC_FULL = REC_ALLOW | REC_BLOCK
2632 * Enqueue an MLDv2 filter list change to the given output queue.
2634 * Source list filter state is held in an RB-tree. When the filter list
2635 * for a group is changed without changing its mode, we need to compute
2636 * the deltas between T0 and T1 for each source in the filter set,
2637 * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records.
2639 * As we may potentially queue two record types, and the entire R-B tree
2640 * needs to be walked at once, we break this out into its own function
2641 * so we can generate a tightly packed queue of packets.
2643 * XXX This could be written to only use one tree walk, although that makes
2644 * serializing into the mbuf chains a bit harder. For now we do two walks
2645 * which makes things easier on us, and it may or may not be harder on
2648 * If successful the size of all data appended to the queue is returned,
2649 * otherwise an error code less than zero is returned, or zero if
2650 * no record(s) were appended.
2653 mld_v2_enqueue_filter_change(struct mbufq *mq, struct in6_multi *inm)
2655 static const int MINRECLEN =
2656 sizeof(struct mldv2_record) + sizeof(struct in6_addr);
2658 struct mldv2_record mr;
2659 struct mldv2_record *pmr;
2660 struct ip6_msource *ims, *nims;
2661 struct mbuf *m, *m0, *md;
2662 int m0srcs, nbytes, npbytes, off, rsrcs, schanged;
2664 uint8_t mode, now, then;
2665 rectype_t crt, drt, nrt;
2667 char ip6tbuf[INET6_ADDRSTRLEN];
2670 IN6_MULTI_LIST_LOCK_ASSERT();
2672 if (inm->in6m_nsrc == 0 ||
2673 (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0))
2676 ifp = inm->in6m_ifp; /* interface */
2677 mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */
2678 crt = REC_NONE; /* current group record type */
2679 drt = REC_NONE; /* mask of completed group record types */
2680 nrt = REC_NONE; /* record type for current node */
2681 m0srcs = 0; /* # source which will fit in current mbuf chain */
2682 npbytes = 0; /* # of bytes appended this packet */
2683 nbytes = 0; /* # of bytes appended to group's state-change queue */
2684 rsrcs = 0; /* # sources encoded in current record */
2685 schanged = 0; /* # nodes encoded in overall filter change */
2686 nallow = 0; /* # of source entries in ALLOW_NEW */
2687 nblock = 0; /* # of source entries in BLOCK_OLD */
2688 nims = NULL; /* next tree node pointer */
2691 * For each possible filter record mode.
2692 * The first kind of source we encounter tells us which
2693 * is the first kind of record we start appending.
2694 * If a node transitioned to UNDEFINED at t1, its mode is treated
2695 * as the inverse of the group's filter mode.
2697 while (drt != REC_FULL) {
2699 m0 = mbufq_last(mq);
2701 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <=
2702 MLD_V2_REPORT_MAXRECS) &&
2703 (m0->m_pkthdr.len + MINRECLEN) <
2704 (ifp->if_mtu - MLD_MTUSPACE)) {
2706 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2707 sizeof(struct mldv2_record)) /
2708 sizeof(struct in6_addr);
2710 "%s: use previous packet", __func__);
2712 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2714 m = m_gethdr(M_NOWAIT, MT_DATA);
2717 "%s: m_get*() failed", __func__);
2720 m->m_pkthdr.PH_vt.vt_nrecs = 0;
2721 mld_save_context(m, ifp);
2722 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2723 sizeof(struct mldv2_record)) /
2724 sizeof(struct in6_addr);
2727 "%s: allocated new packet", __func__);
2730 * Append the MLD group record header to the
2731 * current packet's data area.
2732 * Recalculate pointer to free space for next
2733 * group record, in case m_append() allocated
2734 * a new mbuf or cluster.
2736 memset(&mr, 0, sizeof(mr));
2737 mr.mr_addr = inm->in6m_addr;
2738 in6_clearscope(&mr.mr_addr);
2739 if (!m_append(m, sizeof(mr), (void *)&mr)) {
2743 "%s: m_append() failed", __func__);
2746 npbytes += sizeof(struct mldv2_record);
2748 /* new packet; offset in chain */
2749 md = m_getptr(m, npbytes -
2750 sizeof(struct mldv2_record), &off);
2751 pmr = (struct mldv2_record *)(mtod(md,
2754 /* current packet; offset from last append */
2756 pmr = (struct mldv2_record *)(mtod(md,
2757 uint8_t *) + md->m_len -
2758 sizeof(struct mldv2_record));
2761 * Begin walking the tree for this record type
2762 * pass, or continue from where we left off
2763 * previously if we had to allocate a new packet.
2764 * Only report deltas in-mode at t1.
2765 * We need not report included sources as allowed
2766 * if we are in inclusive mode on the group,
2767 * however the converse is not true.
2771 nims = RB_MIN(ip6_msource_tree,
2774 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2775 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2776 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2777 now = im6s_get_mode(inm, ims, 1);
2778 then = im6s_get_mode(inm, ims, 0);
2779 CTR3(KTR_MLD, "%s: mode: t0 %d, t1 %d",
2780 __func__, then, now);
2783 "%s: skip unchanged", __func__);
2786 if (mode == MCAST_EXCLUDE &&
2787 now == MCAST_INCLUDE) {
2789 "%s: skip IN src on EX group",
2793 nrt = (rectype_t)now;
2794 if (nrt == REC_NONE)
2795 nrt = (rectype_t)(~mode & REC_FULL);
2796 if (schanged++ == 0) {
2798 } else if (crt != nrt)
2800 if (!m_append(m, sizeof(struct in6_addr),
2801 (void *)&ims->im6s_addr)) {
2805 "%s: m_append() failed", __func__);
2808 nallow += !!(crt == REC_ALLOW);
2809 nblock += !!(crt == REC_BLOCK);
2810 if (++rsrcs == m0srcs)
2814 * If we did not append any tree nodes on this
2815 * pass, back out of allocations.
2818 npbytes -= sizeof(struct mldv2_record);
2821 "%s: m_free(m)", __func__);
2825 "%s: m_adj(m, -mr)", __func__);
2826 m_adj(m, -((int)sizeof(
2827 struct mldv2_record)));
2831 npbytes += (rsrcs * sizeof(struct in6_addr));
2832 if (crt == REC_ALLOW)
2833 pmr->mr_type = MLD_ALLOW_NEW_SOURCES;
2834 else if (crt == REC_BLOCK)
2835 pmr->mr_type = MLD_BLOCK_OLD_SOURCES;
2836 pmr->mr_numsrc = htons(rsrcs);
2838 * Count the new group record, and enqueue this
2839 * packet if it wasn't already queued.
2841 m->m_pkthdr.PH_vt.vt_nrecs++;
2843 mbufq_enqueue(mq, m);
2845 } while (nims != NULL);
2847 crt = (~crt & REC_FULL);
2850 CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__,
2857 mld_v2_merge_state_changes(struct in6_multi *inm, struct mbufq *scq)
2860 struct mbuf *m; /* pending state-change */
2861 struct mbuf *m0; /* copy of pending state-change */
2862 struct mbuf *mt; /* last state-change in packet */
2863 int docopy, domerge;
2870 IN6_MULTI_LIST_LOCK_ASSERT();
2874 * If there are further pending retransmissions, make a writable
2875 * copy of each queued state-change message before merging.
2877 if (inm->in6m_scrv > 0)
2880 gq = &inm->in6m_scq;
2882 if (mbufq_first(gq) == NULL) {
2883 CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty",
2888 m = mbufq_first(gq);
2891 * Only merge the report into the current packet if
2892 * there is sufficient space to do so; an MLDv2 report
2893 * packet may only contain 65,535 group records.
2894 * Always use a simple mbuf chain concatentation to do this,
2895 * as large state changes for single groups may have
2896 * allocated clusters.
2899 mt = mbufq_last(scq);
2901 recslen = m_length(m, NULL);
2903 if ((mt->m_pkthdr.PH_vt.vt_nrecs +
2904 m->m_pkthdr.PH_vt.vt_nrecs <=
2905 MLD_V2_REPORT_MAXRECS) &&
2906 (mt->m_pkthdr.len + recslen <=
2907 (inm->in6m_ifp->if_mtu - MLD_MTUSPACE)))
2911 if (!domerge && mbufq_full(gq)) {
2913 "%s: outbound queue full, skipping whole packet %p",
2923 CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m);
2924 m0 = mbufq_dequeue(gq);
2927 CTR2(KTR_MLD, "%s: copying %p", __func__, m);
2928 m0 = m_dup(m, M_NOWAIT);
2931 m0->m_nextpkt = NULL;
2936 CTR3(KTR_MLD, "%s: queueing %p to scq %p)",
2938 mbufq_enqueue(scq, m0);
2940 struct mbuf *mtl; /* last mbuf of packet mt */
2942 CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)",
2946 m0->m_flags &= ~M_PKTHDR;
2947 mt->m_pkthdr.len += recslen;
2948 mt->m_pkthdr.PH_vt.vt_nrecs +=
2949 m0->m_pkthdr.PH_vt.vt_nrecs;
2959 * Respond to a pending MLDv2 General Query.
2962 mld_v2_dispatch_general_query(struct mld_ifsoftc *mli)
2964 struct ifmultiaddr *ifma;
2966 struct in6_multi *inm;
2969 IN6_MULTI_LIST_LOCK_ASSERT();
2972 KASSERT(mli->mli_version == MLD_VERSION_2,
2973 ("%s: called when version %d", __func__, mli->mli_version));
2976 * Check that there are some packets queued. If so, send them first.
2977 * For large number of groups the reply to general query can take
2978 * many packets, we should finish sending them before starting of
2979 * queuing the new reply.
2981 if (mbufq_len(&mli->mli_gq) != 0)
2987 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2988 if (ifma->ifma_addr->sa_family != AF_INET6 ||
2989 ifma->ifma_protospec == NULL)
2992 inm = (struct in6_multi *)ifma->ifma_protospec;
2993 KASSERT(ifp == inm->in6m_ifp,
2994 ("%s: inconsistent ifp", __func__));
2996 switch (inm->in6m_state) {
2997 case MLD_NOT_MEMBER:
2998 case MLD_SILENT_MEMBER:
3000 case MLD_REPORTING_MEMBER:
3001 case MLD_IDLE_MEMBER:
3002 case MLD_LAZY_MEMBER:
3003 case MLD_SLEEPING_MEMBER:
3004 case MLD_AWAKENING_MEMBER:
3005 inm->in6m_state = MLD_REPORTING_MEMBER;
3006 retval = mld_v2_enqueue_group_record(&mli->mli_gq,
3008 CTR2(KTR_MLD, "%s: enqueue record = %d",
3011 case MLD_G_QUERY_PENDING_MEMBER:
3012 case MLD_SG_QUERY_PENDING_MEMBER:
3013 case MLD_LEAVING_MEMBER:
3017 IF_ADDR_RUNLOCK(ifp);
3020 mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST);
3023 * Slew transmission of bursts over 500ms intervals.
3025 if (mbufq_first(&mli->mli_gq) != NULL) {
3026 mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY(
3027 MLD_RESPONSE_BURST_INTERVAL);
3028 V_interface_timers_running6 = 1;
3033 * Transmit the next pending message in the output queue.
3035 * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis.
3036 * MRT: Nothing needs to be done, as MLD traffic is always local to
3037 * a link and uses a link-scope multicast address.
3040 mld_dispatch_packet(struct mbuf *m)
3042 struct ip6_moptions im6o;
3047 struct ip6_hdr *ip6;
3048 struct mld_hdr *mld;
3054 CTR2(KTR_MLD, "%s: transmit %p", __func__, m);
3057 * Set VNET image pointer from enqueued mbuf chain
3058 * before doing anything else. Whilst we use interface
3059 * indexes to guard against interface detach, they are
3060 * unique to each VIMAGE and must be retrieved.
3062 ifindex = mld_restore_context(m);
3065 * Check if the ifnet still exists. This limits the scope of
3066 * any race in the absence of a global ifp lock for low cost
3067 * (an array lookup).
3069 ifp = ifnet_byindex(ifindex);
3071 CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.",
3072 __func__, m, ifindex);
3074 IP6STAT_INC(ip6s_noroute);
3078 im6o.im6o_multicast_hlim = 1;
3079 im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL);
3080 im6o.im6o_multicast_ifp = ifp;
3082 if (m->m_flags & M_MLDV1) {
3085 m0 = mld_v2_encap_report(ifp, m);
3087 CTR2(KTR_MLD, "%s: dropped %p", __func__, m);
3088 IP6STAT_INC(ip6s_odropped);
3093 mld_scrub_context(m0);
3095 m0->m_pkthdr.rcvif = V_loif;
3097 ip6 = mtod(m0, struct ip6_hdr *);
3099 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */
3102 * XXX XXX Break some KPI rules to prevent an LOR which would
3103 * occur if we called in6_setscope() at transmission.
3104 * See comments at top of file.
3106 MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index);
3110 * Retrieve the ICMPv6 type before handoff to ip6_output(),
3111 * so we can bump the stats.
3113 md = m_getptr(m0, sizeof(struct ip6_hdr), &off);
3114 mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off);
3115 type = mld->mld_type;
3117 error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o,
3120 CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error);
3123 ICMP6STAT_INC(icp6s_outhist[type]);
3125 icmp6_ifstat_inc(oifp, ifs6_out_msg);
3127 case MLD_LISTENER_REPORT:
3128 case MLDV2_LISTENER_REPORT:
3129 icmp6_ifstat_inc(oifp, ifs6_out_mldreport);
3131 case MLD_LISTENER_DONE:
3132 icmp6_ifstat_inc(oifp, ifs6_out_mlddone);
3141 * Encapsulate an MLDv2 report.
3143 * KAME IPv6 requires that hop-by-hop options be passed separately,
3144 * and that the IPv6 header be prepended in a separate mbuf.
3146 * Returns a pointer to the new mbuf chain head, or NULL if the
3147 * allocation failed.
3149 static struct mbuf *
3150 mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m)
3153 struct mldv2_report *mld;
3154 struct ip6_hdr *ip6;
3155 struct in6_ifaddr *ia;
3158 KASSERT(ifp != NULL, ("%s: null ifp", __func__));
3159 KASSERT((m->m_flags & M_PKTHDR),
3160 ("%s: mbuf chain %p is !M_PKTHDR", __func__, m));
3163 * RFC3590: OK to send as :: or tentative during DAD.
3165 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
3167 CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__);
3169 mh = m_gethdr(M_NOWAIT, MT_DATA);
3172 ifa_free(&ia->ia_ifa);
3176 M_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report));
3178 mldreclen = m_length(m, NULL);
3179 CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen);
3181 mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report);
3182 mh->m_pkthdr.len = sizeof(struct ip6_hdr) +
3183 sizeof(struct mldv2_report) + mldreclen;
3185 ip6 = mtod(mh, struct ip6_hdr *);
3187 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3188 ip6->ip6_vfc |= IPV6_VERSION;
3189 ip6->ip6_nxt = IPPROTO_ICMPV6;
3190 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
3192 ifa_free(&ia->ia_ifa);
3193 ip6->ip6_dst = in6addr_linklocal_allv2routers;
3194 /* scope ID will be set in netisr */
3196 mld = (struct mldv2_report *)(ip6 + 1);
3197 mld->mld_type = MLDV2_LISTENER_REPORT;
3200 mld->mld_v2_reserved = 0;
3201 mld->mld_v2_numrecs = htons(m->m_pkthdr.PH_vt.vt_nrecs);
3202 m->m_pkthdr.PH_vt.vt_nrecs = 0;
3205 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
3206 sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen);
3212 mld_rec_type_to_str(const int type)
3216 case MLD_CHANGE_TO_EXCLUDE_MODE:
3219 case MLD_CHANGE_TO_INCLUDE_MODE:
3222 case MLD_MODE_IS_EXCLUDE:
3225 case MLD_MODE_IS_INCLUDE:
3228 case MLD_ALLOW_NEW_SOURCES:
3231 case MLD_BLOCK_OLD_SOURCES:
3242 mld_init(void *unused __unused)
3245 CTR1(KTR_MLD, "%s: initializing", __func__);
3248 ip6_initpktopts(&mld_po);
3249 mld_po.ip6po_hlim = 1;
3250 mld_po.ip6po_hbh = &mld_ra.hbh;
3251 mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
3252 mld_po.ip6po_flags = IP6PO_DONTFRAG;
3254 SYSINIT(mld_init, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_init, NULL);
3257 mld_uninit(void *unused __unused)
3260 CTR1(KTR_MLD, "%s: tearing down", __func__);
3263 SYSUNINIT(mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, mld_uninit, NULL);
3266 vnet_mld_init(const void *unused __unused)
3269 CTR1(KTR_MLD, "%s: initializing", __func__);
3271 LIST_INIT(&V_mli_head);
3273 VNET_SYSINIT(vnet_mld_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_init,
3277 vnet_mld_uninit(const void *unused __unused)
3280 /* This can happen if we shutdown the network stack. */
3281 CTR1(KTR_MLD, "%s: tearing down", __func__);
3283 VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mld_uninit,
3287 mld_modevent(module_t mod, int type, void *unused __unused)
3295 return (EOPNOTSUPP);
3300 static moduledata_t mld_mod = {
3305 DECLARE_MODULE(mld, mld_mod, SI_SUB_PROTO_MC, SI_ORDER_ANY);