2 * Copyright (c) 2009 Bruce Simpson.
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
12 * 3. The name of the author may not be used to endorse or promote
13 * products derived from this software without specific prior written
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * $KAME: mld6.c,v 1.27 2001/04/04 05:17:30 itojun Exp $
32 * Copyright (c) 1988 Stephen Deering.
33 * Copyright (c) 1992, 1993
34 * The Regents of the University of California. All rights reserved.
36 * This code is derived from software contributed to Berkeley by
37 * Stephen Deering of Stanford University.
39 * Redistribution and use in source and binary forms, with or without
40 * modification, are permitted provided that the following conditions
42 * 1. Redistributions of source code must retain the above copyright
43 * notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 * notice, this list of conditions and the following disclaimer in the
46 * documentation and/or other materials provided with the distribution.
47 * 4. Neither the name of the University nor the names of its contributors
48 * may be used to endorse or promote products derived from this software
49 * without specific prior written permission.
51 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * @(#)igmp.c 8.1 (Berkeley) 7/19/93
66 #include <sys/cdefs.h>
67 __FBSDID("$FreeBSD$");
70 #include "opt_inet6.h"
72 #include <sys/param.h>
73 #include <sys/systm.h>
75 #include <sys/socket.h>
76 #include <sys/protosw.h>
77 #include <sys/sysctl.h>
78 #include <sys/kernel.h>
79 #include <sys/callout.h>
80 #include <sys/malloc.h>
81 #include <sys/module.h>
85 #include <net/if_var.h>
86 #include <net/route.h>
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/scope6_var.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/mld6.h>
97 #include <netinet6/mld6_var.h>
99 #include <security/mac/mac_framework.h>
102 #define KTR_MLD KTR_INET6
105 static struct mld_ifinfo *
106 mli_alloc_locked(struct ifnet *);
107 static void mli_delete_locked(const struct ifnet *);
108 static void mld_dispatch_packet(struct mbuf *);
109 static void mld_dispatch_queue(struct ifqueue *, int);
110 static void mld_final_leave(struct in6_multi *, struct mld_ifinfo *);
111 static void mld_fasttimo_vnet(void);
112 static int mld_handle_state_change(struct in6_multi *,
113 struct mld_ifinfo *);
114 static int mld_initial_join(struct in6_multi *, struct mld_ifinfo *,
117 static char * mld_rec_type_to_str(const int);
119 static void mld_set_version(struct mld_ifinfo *, 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 mld_ifinfo *,
127 static void mld_v1_process_querier_timers(struct mld_ifinfo *);
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_ifinfo *);
131 static void mld_v2_dispatch_general_query(struct mld_ifinfo *);
133 mld_v2_encap_report(struct ifnet *, struct mbuf *);
134 static int mld_v2_enqueue_filter_change(struct ifqueue *,
136 static int mld_v2_enqueue_group_record(struct ifqueue *,
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 *, const int, const int);
141 static int mld_v2_merge_state_changes(struct in6_multi *,
143 static void mld_v2_process_group_timers(struct mld_ifinfo *,
144 struct ifqueue *, struct ifqueue *,
145 struct in6_multi *, const int);
146 static int mld_v2_process_group_query(struct in6_multi *,
147 struct mld_ifinfo *mli, int, struct mbuf *, const int);
148 static int sysctl_mld_gsr(SYSCTL_HANDLER_ARGS);
149 static int sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS);
152 * Normative references: RFC 2710, RFC 3590, RFC 3810.
155 * * The MLD subsystem lock ends up being system-wide for the moment,
156 * but could be per-VIMAGE later on.
157 * * The permitted lock order is: IN6_MULTI_LOCK, MLD_LOCK, IF_ADDR_LOCK.
158 * Any may be taken independently; if any are held at the same
159 * time, the above lock order must be followed.
160 * * IN6_MULTI_LOCK covers in_multi.
161 * * MLD_LOCK covers per-link state and any global variables in this file.
162 * * IF_ADDR_LOCK covers if_multiaddrs, which is used for a variety of
163 * per-link state iterators.
166 * A special case for IPv6 is the in6_setscope() routine. ip6_output()
167 * will not accept an ifp; it wants an embedded scope ID, unlike
168 * ip_output(), which happily takes the ifp given to it. The embedded
169 * scope ID is only used by MLD to select the outgoing interface.
171 * During interface attach and detach, MLD will take MLD_LOCK *after*
172 * the IF_AFDATA_LOCK.
173 * As in6_setscope() takes IF_AFDATA_LOCK then SCOPE_LOCK, we can't call
174 * it with MLD_LOCK held without triggering an LOR. A netisr with indirect
175 * dispatch could work around this, but we'd rather not do that, as it
176 * can introduce other races.
178 * As such, we exploit the fact that the scope ID is just the interface
179 * index, and embed it in the IPv6 destination address accordingly.
180 * This is potentially NOT VALID for MLDv1 reports, as they
181 * are always sent to the multicast group itself; as MLDv2
182 * reports are always sent to ff02::16, this is not an issue
183 * when MLDv2 is in use.
185 * This does not however eliminate the LOR when ip6_output() itself
186 * calls in6_setscope() internally whilst MLD_LOCK is held. This will
187 * trigger a LOR warning in WITNESS when the ifnet is detached.
189 * The right answer is probably to make IF_AFDATA_LOCK an rwlock, given
190 * how it's used across the network stack. Here we're simply exploiting
191 * the fact that MLD runs at a similar layer in the stack to scope6.c.
194 * * Each in6_multi corresponds to an ifp, and each ifp corresponds
195 * to a vnet in ifp->if_vnet.
197 static struct mtx mld_mtx;
198 static MALLOC_DEFINE(M_MLD, "mld", "mld state");
200 #define MLD_EMBEDSCOPE(pin6, zoneid) \
201 if (IN6_IS_SCOPE_LINKLOCAL(pin6) || \
202 IN6_IS_ADDR_MC_INTFACELOCAL(pin6)) \
203 (pin6)->s6_addr16[1] = htons((zoneid) & 0xFFFF) \
206 * VIMAGE-wide globals.
208 static VNET_DEFINE(struct timeval, mld_gsrdelay) = {10, 0};
209 static VNET_DEFINE(LIST_HEAD(, mld_ifinfo), mli_head);
210 static VNET_DEFINE(int, interface_timers_running6);
211 static VNET_DEFINE(int, state_change_timers_running6);
212 static VNET_DEFINE(int, current_state_timers_running6);
214 #define V_mld_gsrdelay VNET(mld_gsrdelay)
215 #define V_mli_head VNET(mli_head)
216 #define V_interface_timers_running6 VNET(interface_timers_running6)
217 #define V_state_change_timers_running6 VNET(state_change_timers_running6)
218 #define V_current_state_timers_running6 VNET(current_state_timers_running6)
220 SYSCTL_DECL(_net_inet6); /* Note: Not in any common header. */
222 SYSCTL_NODE(_net_inet6, OID_AUTO, mld, CTLFLAG_RW, 0,
223 "IPv6 Multicast Listener Discovery");
226 * Virtualized sysctls.
228 SYSCTL_VNET_PROC(_net_inet6_mld, OID_AUTO, gsrdelay,
229 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
230 &VNET_NAME(mld_gsrdelay.tv_sec), 0, sysctl_mld_gsr, "I",
231 "Rate limit for MLDv2 Group-and-Source queries in seconds");
234 * Non-virtualized sysctls.
236 static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo,
237 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo,
238 "Per-interface MLDv2 state");
240 static int mld_v1enable = 1;
241 SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RW,
242 &mld_v1enable, 0, "Enable fallback to MLDv1");
243 TUNABLE_INT("net.inet6.mld.v1enable", &mld_v1enable);
245 static int mld_use_allow = 1;
246 SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RW,
247 &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves");
248 TUNABLE_INT("net.inet6.mld.use_allow", &mld_use_allow);
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", __func__));
307 return (m->m_pkthdr.flowid);
311 * Retrieve or set threshold between group-source queries in seconds.
313 * VIMAGE: Assume curvnet set by caller.
314 * SMPng: NOTE: Serialized by MLD lock.
317 sysctl_mld_gsr(SYSCTL_HANDLER_ARGS)
322 error = sysctl_wire_old_buffer(req, sizeof(int));
328 i = V_mld_gsrdelay.tv_sec;
330 error = sysctl_handle_int(oidp, &i, 0, req);
331 if (error || !req->newptr)
334 if (i < -1 || i >= 60) {
339 CTR2(KTR_MLD, "change mld_gsrdelay from %d to %d",
340 V_mld_gsrdelay.tv_sec, i);
341 V_mld_gsrdelay.tv_sec = i;
349 * Expose struct mld_ifinfo to userland, keyed by ifindex.
350 * For use by ifmcstat(8).
352 * SMPng: NOTE: Does an unlocked ifindex space read.
353 * VIMAGE: Assume curvnet set by caller. The node handler itself
354 * is not directly virtualized.
357 sysctl_mld_ifinfo(SYSCTL_HANDLER_ARGS)
363 struct mld_ifinfo *mli;
368 if (req->newptr != NULL)
374 error = sysctl_wire_old_buffer(req, sizeof(struct mld_ifinfo));
381 if (name[0] <= 0 || name[0] > V_if_index) {
388 ifp = ifnet_byindex(name[0]);
392 LIST_FOREACH(mli, &V_mli_head, mli_link) {
393 if (ifp == mli->mli_ifp) {
394 error = SYSCTL_OUT(req, mli,
395 sizeof(struct mld_ifinfo));
407 * Dispatch an entire queue of pending packet chains.
408 * VIMAGE: Assumes the vnet pointer has been set.
411 mld_dispatch_queue(struct ifqueue *ifq, int limit)
419 CTR3(KTR_MLD, "%s: dispatch %p from %p", __func__, ifq, m);
420 mld_dispatch_packet(m);
427 * Filter outgoing MLD report state by group.
429 * Reports are ALWAYS suppressed for ALL-HOSTS (ff02::1)
430 * and node-local addresses. However, kernel and socket consumers
431 * always embed the KAME scope ID in the address provided, so strip it
432 * when performing comparison.
433 * Note: This is not the same as the *multicast* scope.
435 * Return zero if the given group is one for which MLD reports
436 * should be suppressed, or non-zero if reports should be issued.
439 mld_is_addr_reported(const struct in6_addr *addr)
442 KASSERT(IN6_IS_ADDR_MULTICAST(addr), ("%s: not multicast", __func__));
444 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_NODELOCAL)
447 if (IPV6_ADDR_MC_SCOPE(addr) == IPV6_ADDR_SCOPE_LINKLOCAL) {
448 struct in6_addr tmp = *addr;
449 in6_clearscope(&tmp);
450 if (IN6_ARE_ADDR_EQUAL(&tmp, &in6addr_linklocal_allnodes))
458 * Attach MLD when PF_INET6 is attached to an interface.
460 * SMPng: Normally called with IF_AFDATA_LOCK held.
463 mld_domifattach(struct ifnet *ifp)
465 struct mld_ifinfo *mli;
467 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
468 __func__, ifp, if_name(ifp));
472 mli = mli_alloc_locked(ifp);
473 if (!(ifp->if_flags & IFF_MULTICAST))
474 mli->mli_flags |= MLIF_SILENT;
476 mli->mli_flags |= MLIF_USEALLOW;
484 * VIMAGE: assume curvnet set by caller.
486 static struct mld_ifinfo *
487 mli_alloc_locked(/*const*/ struct ifnet *ifp)
489 struct mld_ifinfo *mli;
493 mli = malloc(sizeof(struct mld_ifinfo), M_MLD, M_NOWAIT|M_ZERO);
498 mli->mli_version = MLD_VERSION_2;
500 mli->mli_rv = MLD_RV_INIT;
501 mli->mli_qi = MLD_QI_INIT;
502 mli->mli_qri = MLD_QRI_INIT;
503 mli->mli_uri = MLD_URI_INIT;
505 SLIST_INIT(&mli->mli_relinmhead);
508 * Responses to general queries are subject to bounds.
510 IFQ_SET_MAXLEN(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS);
512 LIST_INSERT_HEAD(&V_mli_head, mli, mli_link);
514 CTR2(KTR_MLD, "allocate mld_ifinfo for ifp %p(%s)",
524 * NOTE: Some finalization tasks need to run before the protocol domain
525 * is detached, but also before the link layer does its cleanup.
526 * Run before link-layer cleanup; cleanup groups, but do not free MLD state.
528 * SMPng: Caller must hold IN6_MULTI_LOCK().
529 * Must take IF_ADDR_LOCK() to cover if_multiaddrs iterator.
530 * XXX This routine is also bitten by unlocked ifma_protospec access.
533 mld_ifdetach(struct ifnet *ifp)
535 struct mld_ifinfo *mli;
536 struct ifmultiaddr *ifma;
537 struct in6_multi *inm, *tinm;
539 CTR3(KTR_MLD, "%s: called for ifp %p(%s)", __func__, ifp,
542 IN6_MULTI_LOCK_ASSERT();
545 mli = MLD_IFINFO(ifp);
546 if (mli->mli_version == MLD_VERSION_2) {
548 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
549 if (ifma->ifma_addr->sa_family != AF_INET6 ||
550 ifma->ifma_protospec == NULL)
552 inm = (struct in6_multi *)ifma->ifma_protospec;
553 if (inm->in6m_state == MLD_LEAVING_MEMBER) {
554 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
557 in6m_clear_recorded(inm);
559 IF_ADDR_RUNLOCK(ifp);
560 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele,
562 SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele);
563 in6m_release_locked(inm);
571 * Hook for domifdetach.
572 * Runs after link-layer cleanup; free MLD state.
574 * SMPng: Normally called with IF_AFDATA_LOCK held.
577 mld_domifdetach(struct ifnet *ifp)
580 CTR3(KTR_MLD, "%s: called for ifp %p(%s)",
581 __func__, ifp, if_name(ifp));
584 mli_delete_locked(ifp);
589 mli_delete_locked(const struct ifnet *ifp)
591 struct mld_ifinfo *mli, *tmli;
593 CTR3(KTR_MLD, "%s: freeing mld_ifinfo for ifp %p(%s)",
594 __func__, ifp, if_name(ifp));
598 LIST_FOREACH_SAFE(mli, &V_mli_head, mli_link, tmli) {
599 if (mli->mli_ifp == ifp) {
601 * Free deferred General Query responses.
603 _IF_DRAIN(&mli->mli_gq);
605 LIST_REMOVE(mli, mli_link);
607 KASSERT(SLIST_EMPTY(&mli->mli_relinmhead),
608 ("%s: there are dangling in_multi references",
616 panic("%s: mld_ifinfo not found for ifp %p\n", __func__, ifp);
621 * Process a received MLDv1 general or address-specific query.
622 * Assumes that the query header has been pulled up to sizeof(mld_hdr).
624 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
625 * mld_addr. This is OK as we own the mbuf chain.
628 mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
629 /*const*/ struct mld_hdr *mld)
631 struct ifmultiaddr *ifma;
632 struct mld_ifinfo *mli;
633 struct in6_multi *inm;
634 int is_general_query;
637 char ip6tbuf[INET6_ADDRSTRLEN];
640 is_general_query = 0;
643 CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)",
644 ip6_sprintf(ip6tbuf, &mld->mld_addr),
650 * RFC3810 Section 6.2: MLD queries must originate from
651 * a router's link-local address.
653 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
654 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
655 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
661 * Do address field validation upfront before we accept
664 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
666 * MLDv1 General Query.
667 * If this was not sent to the all-nodes group, ignore it.
672 in6_clearscope(&dst);
673 if (!IN6_ARE_ADDR_EQUAL(&dst, &in6addr_linklocal_allnodes))
675 is_general_query = 1;
678 * Embed scope ID of receiving interface in MLD query for
679 * lookup whilst we don't hold other locks.
681 in6_setscope(&mld->mld_addr, ifp, NULL);
688 * Switch to MLDv1 host compatibility mode.
690 mli = MLD_IFINFO(ifp);
691 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
692 mld_set_version(mli, MLD_VERSION_1);
694 timer = (ntohs(mld->mld_maxdelay) * PR_FASTHZ) / MLD_TIMER_SCALE;
699 if (is_general_query) {
701 * For each reporting group joined on this
702 * interface, kick the report timer.
704 CTR2(KTR_MLD, "process v1 general query on ifp %p(%s)",
706 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
707 if (ifma->ifma_addr->sa_family != AF_INET6 ||
708 ifma->ifma_protospec == NULL)
710 inm = (struct in6_multi *)ifma->ifma_protospec;
711 mld_v1_update_group(inm, timer);
715 * MLDv1 Group-Specific Query.
716 * If this is a group-specific MLDv1 query, we need only
717 * look up the single group to process it.
719 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
721 CTR3(KTR_MLD, "process v1 query %s on ifp %p(%s)",
722 ip6_sprintf(ip6tbuf, &mld->mld_addr),
724 mld_v1_update_group(inm, timer);
726 /* XXX Clear embedded scope ID as userland won't expect it. */
727 in6_clearscope(&mld->mld_addr);
730 IF_ADDR_RUNLOCK(ifp);
738 * Update the report timer on a group in response to an MLDv1 query.
740 * If we are becoming the reporting member for this group, start the timer.
741 * If we already are the reporting member for this group, and timer is
742 * below the threshold, reset it.
744 * We may be updating the group for the first time since we switched
745 * to MLDv2. If we are, then we must clear any recorded source lists,
746 * and transition to REPORTING state; the group timer is overloaded
747 * for group and group-source query responses.
749 * Unlike MLDv2, the delay per group should be jittered
750 * to avoid bursts of MLDv1 reports.
753 mld_v1_update_group(struct in6_multi *inm, const int timer)
756 char ip6tbuf[INET6_ADDRSTRLEN];
759 CTR4(KTR_MLD, "%s: %s/%s timer=%d", __func__,
760 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
761 if_name(inm->in6m_ifp), timer);
763 IN6_MULTI_LOCK_ASSERT();
765 switch (inm->in6m_state) {
767 case MLD_SILENT_MEMBER:
769 case MLD_REPORTING_MEMBER:
770 if (inm->in6m_timer != 0 &&
771 inm->in6m_timer <= timer) {
772 CTR1(KTR_MLD, "%s: REPORTING and timer running, "
773 "skipping.", __func__);
777 case MLD_SG_QUERY_PENDING_MEMBER:
778 case MLD_G_QUERY_PENDING_MEMBER:
779 case MLD_IDLE_MEMBER:
780 case MLD_LAZY_MEMBER:
781 case MLD_AWAKENING_MEMBER:
782 CTR1(KTR_MLD, "%s: ->REPORTING", __func__);
783 inm->in6m_state = MLD_REPORTING_MEMBER;
784 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
785 V_current_state_timers_running6 = 1;
787 case MLD_SLEEPING_MEMBER:
788 CTR1(KTR_MLD, "%s: ->AWAKENING", __func__);
789 inm->in6m_state = MLD_AWAKENING_MEMBER;
791 case MLD_LEAVING_MEMBER:
797 * Process a received MLDv2 general, group-specific or
798 * group-and-source-specific query.
800 * Assumes that the query header has been pulled up to sizeof(mldv2_query).
802 * Return 0 if successful, otherwise an appropriate error code is returned.
805 mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6,
806 struct mbuf *m, const int off, const int icmp6len)
808 struct mld_ifinfo *mli;
809 struct mldv2_query *mld;
810 struct in6_multi *inm;
811 uint32_t maxdelay, nsrc, qqi;
812 int is_general_query;
816 char ip6tbuf[INET6_ADDRSTRLEN];
819 is_general_query = 0;
822 * RFC3810 Section 6.2: MLD queries must originate from
823 * a router's link-local address.
825 if (!IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
826 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
827 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
832 CTR2(KTR_MLD, "input v2 query on ifp %p(%s)", ifp, if_name(ifp));
834 mld = (struct mldv2_query *)(mtod(m, uint8_t *) + off);
836 maxdelay = ntohs(mld->mld_maxdelay); /* in 1/10ths of a second */
837 if (maxdelay >= 32768) {
838 maxdelay = (MLD_MRC_MANT(maxdelay) | 0x1000) <<
839 (MLD_MRC_EXP(maxdelay) + 3);
841 timer = (maxdelay * PR_FASTHZ) / MLD_TIMER_SCALE;
845 qrv = MLD_QRV(mld->mld_misc);
847 CTR3(KTR_MLD, "%s: clamping qrv %d to %d", __func__,
854 qqi = MLD_QQIC_MANT(mld->mld_qqi) <<
855 (MLD_QQIC_EXP(mld->mld_qqi) + 3);
858 nsrc = ntohs(mld->mld_numsrc);
859 if (nsrc > MLD_MAX_GS_SOURCES)
861 if (icmp6len < sizeof(struct mldv2_query) +
862 (nsrc * sizeof(struct in6_addr)))
866 * Do further input validation upfront to avoid resetting timers
867 * should we need to discard this query.
869 if (IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr)) {
871 * A general query with a source list has undefined
872 * behaviour; discard it.
876 is_general_query = 1;
879 * Embed scope ID of receiving interface in MLD query for
880 * lookup whilst we don't hold other locks (due to KAME
881 * locking lameness). We own this mbuf chain just now.
883 in6_setscope(&mld->mld_addr, ifp, NULL);
889 mli = MLD_IFINFO(ifp);
890 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
893 * Discard the v2 query if we're in Compatibility Mode.
894 * The RFC is pretty clear that hosts need to stay in MLDv1 mode
895 * until the Old Version Querier Present timer expires.
897 if (mli->mli_version != MLD_VERSION_2)
900 mld_set_version(mli, MLD_VERSION_2);
903 mli->mli_qri = maxdelay;
905 CTR4(KTR_MLD, "%s: qrv %d qi %d maxdelay %d", __func__, qrv, qqi,
908 if (is_general_query) {
910 * MLDv2 General Query.
912 * Schedule a current-state report on this ifp for
913 * all groups, possibly containing source lists.
915 * If there is a pending General Query response
916 * scheduled earlier than the selected delay, do
917 * not schedule any other reports.
918 * Otherwise, reset the interface timer.
920 CTR2(KTR_MLD, "process v2 general query on ifp %p(%s)",
922 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer) {
923 mli->mli_v2_timer = MLD_RANDOM_DELAY(timer);
924 V_interface_timers_running6 = 1;
928 * MLDv2 Group-specific or Group-and-source-specific Query.
930 * Group-source-specific queries are throttled on
931 * a per-group basis to defeat denial-of-service attempts.
932 * Queries for groups we are not a member of on this
933 * link are simply ignored.
936 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
938 IF_ADDR_RUNLOCK(ifp);
942 if (!ratecheck(&inm->in6m_lastgsrtv,
944 CTR1(KTR_MLD, "%s: GS query throttled.",
946 IF_ADDR_RUNLOCK(ifp);
950 CTR2(KTR_MLD, "process v2 group query on ifp %p(%s)",
953 * If there is a pending General Query response
954 * scheduled sooner than the selected delay, no
955 * further report need be scheduled.
956 * Otherwise, prepare to respond to the
957 * group-specific or group-and-source query.
959 if (mli->mli_v2_timer == 0 || mli->mli_v2_timer >= timer)
960 mld_v2_process_group_query(inm, mli, timer, m, off);
962 /* XXX Clear embedded scope ID as userland won't expect it. */
963 in6_clearscope(&mld->mld_addr);
964 IF_ADDR_RUNLOCK(ifp);
975 * Process a recieved MLDv2 group-specific or group-and-source-specific
977 * Return <0 if any error occured. Currently this is ignored.
980 mld_v2_process_group_query(struct in6_multi *inm, struct mld_ifinfo *mli,
981 int timer, struct mbuf *m0, const int off)
983 struct mldv2_query *mld;
987 IN6_MULTI_LOCK_ASSERT();
991 mld = (struct mldv2_query *)(mtod(m0, uint8_t *) + off);
993 switch (inm->in6m_state) {
995 case MLD_SILENT_MEMBER:
996 case MLD_SLEEPING_MEMBER:
997 case MLD_LAZY_MEMBER:
998 case MLD_AWAKENING_MEMBER:
999 case MLD_IDLE_MEMBER:
1000 case MLD_LEAVING_MEMBER:
1003 case MLD_REPORTING_MEMBER:
1004 case MLD_G_QUERY_PENDING_MEMBER:
1005 case MLD_SG_QUERY_PENDING_MEMBER:
1009 nsrc = ntohs(mld->mld_numsrc);
1012 * Deal with group-specific queries upfront.
1013 * If any group query is already pending, purge any recorded
1014 * source-list state if it exists, and schedule a query response
1015 * for this group-specific query.
1018 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
1019 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER) {
1020 in6m_clear_recorded(inm);
1021 timer = min(inm->in6m_timer, timer);
1023 inm->in6m_state = MLD_G_QUERY_PENDING_MEMBER;
1024 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1025 V_current_state_timers_running6 = 1;
1030 * Deal with the case where a group-and-source-specific query has
1031 * been received but a group-specific query is already pending.
1033 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER) {
1034 timer = min(inm->in6m_timer, timer);
1035 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1036 V_current_state_timers_running6 = 1;
1041 * Finally, deal with the case where a group-and-source-specific
1042 * query has been received, where a response to a previous g-s-r
1043 * query exists, or none exists.
1044 * In this case, we need to parse the source-list which the Querier
1045 * has provided us with and check if we have any source list filter
1046 * entries at T1 for these sources. If we do not, there is no need
1047 * schedule a report and the query may be dropped.
1048 * If we do, we must record them and schedule a current-state
1049 * report for those sources.
1051 if (inm->in6m_nsrc > 0) {
1058 soff = off + sizeof(struct mldv2_query);
1060 for (i = 0; i < nsrc; i++) {
1061 sp = mtod(m, uint8_t *) + soff;
1062 retval = in6m_record_source(inm,
1063 (const struct in6_addr *)sp);
1066 nrecorded += retval;
1067 soff += sizeof(struct in6_addr);
1068 if (soff >= m->m_len) {
1069 soff = soff - m->m_len;
1075 if (nrecorded > 0) {
1077 "%s: schedule response to SG query", __func__);
1078 inm->in6m_state = MLD_SG_QUERY_PENDING_MEMBER;
1079 inm->in6m_timer = MLD_RANDOM_DELAY(timer);
1080 V_current_state_timers_running6 = 1;
1088 * Process a received MLDv1 host membership report.
1089 * Assumes mld points to mld_hdr in pulled up mbuf chain.
1091 * NOTE: Can't be fully const correct as we temporarily embed scope ID in
1092 * mld_addr. This is OK as we own the mbuf chain.
1095 mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6,
1096 /*const*/ struct mld_hdr *mld)
1098 struct in6_addr src, dst;
1099 struct in6_ifaddr *ia;
1100 struct in6_multi *inm;
1102 char ip6tbuf[INET6_ADDRSTRLEN];
1105 if (!mld_v1enable) {
1106 CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)",
1107 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1112 if (ifp->if_flags & IFF_LOOPBACK)
1116 * MLDv1 reports must originate from a host's link-local address,
1117 * or the unspecified address (when booting).
1120 in6_clearscope(&src);
1121 if (!IN6_IS_SCOPE_LINKLOCAL(&src) && !IN6_IS_ADDR_UNSPECIFIED(&src)) {
1122 CTR3(KTR_MLD, "ignore v1 query src %s on ifp %p(%s)",
1123 ip6_sprintf(ip6tbuf, &ip6->ip6_src),
1129 * RFC2710 Section 4: MLDv1 reports must pertain to a multicast
1130 * group, and must be directed to the group itself.
1133 in6_clearscope(&dst);
1134 if (!IN6_IS_ADDR_MULTICAST(&mld->mld_addr) ||
1135 !IN6_ARE_ADDR_EQUAL(&mld->mld_addr, &dst)) {
1136 CTR3(KTR_MLD, "ignore v1 query dst %s on ifp %p(%s)",
1137 ip6_sprintf(ip6tbuf, &ip6->ip6_dst),
1143 * Make sure we don't hear our own membership report, as fast
1144 * leave requires knowing that we are the only member of a
1145 * group. Assume we used the link-local address if available,
1146 * otherwise look for ::.
1148 * XXX Note that scope ID comparison is needed for the address
1149 * returned by in6ifa_ifpforlinklocal(), but SHOULD NOT be
1150 * performed for the on-wire address.
1152 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1153 if ((ia && IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, IA6_IN6(ia))) ||
1154 (ia == NULL && IN6_IS_ADDR_UNSPECIFIED(&src))) {
1156 ifa_free(&ia->ia_ifa);
1160 ifa_free(&ia->ia_ifa);
1162 CTR3(KTR_MLD, "process v1 report %s on ifp %p(%s)",
1163 ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp));
1166 * Embed scope ID of receiving interface in MLD query for lookup
1167 * whilst we don't hold other locks (due to KAME locking lameness).
1169 if (!IN6_IS_ADDR_UNSPECIFIED(&mld->mld_addr))
1170 in6_setscope(&mld->mld_addr, ifp, NULL);
1177 * MLDv1 report suppression.
1178 * If we are a member of this group, and our membership should be
1179 * reported, and our group timer is pending or about to be reset,
1180 * stop our group timer by transitioning to the 'lazy' state.
1182 inm = in6m_lookup_locked(ifp, &mld->mld_addr);
1184 struct mld_ifinfo *mli;
1186 mli = inm->in6m_mli;
1187 KASSERT(mli != NULL,
1188 ("%s: no mli for ifp %p", __func__, ifp));
1191 * If we are in MLDv2 host mode, do not allow the
1192 * other host's MLDv1 report to suppress our reports.
1194 if (mli->mli_version == MLD_VERSION_2)
1197 inm->in6m_timer = 0;
1199 switch (inm->in6m_state) {
1200 case MLD_NOT_MEMBER:
1201 case MLD_SILENT_MEMBER:
1202 case MLD_SLEEPING_MEMBER:
1204 case MLD_REPORTING_MEMBER:
1205 case MLD_IDLE_MEMBER:
1206 case MLD_AWAKENING_MEMBER:
1208 "report suppressed for %s on ifp %p(%s)",
1209 ip6_sprintf(ip6tbuf, &mld->mld_addr),
1211 case MLD_LAZY_MEMBER:
1212 inm->in6m_state = MLD_LAZY_MEMBER;
1214 case MLD_G_QUERY_PENDING_MEMBER:
1215 case MLD_SG_QUERY_PENDING_MEMBER:
1216 case MLD_LEAVING_MEMBER:
1222 IF_ADDR_RUNLOCK(ifp);
1226 /* XXX Clear embedded scope ID as userland won't expect it. */
1227 in6_clearscope(&mld->mld_addr);
1235 * Assume query messages which fit in a single ICMPv6 message header
1236 * have been pulled up.
1237 * Assume that userland will want to see the message, even if it
1238 * otherwise fails kernel input validation; do not free it.
1239 * Pullup may however free the mbuf chain m if it fails.
1241 * Return IPPROTO_DONE if we freed m. Otherwise, return 0.
1244 mld_input(struct mbuf *m, int off, int icmp6len)
1247 struct ip6_hdr *ip6;
1248 struct mld_hdr *mld;
1251 CTR3(KTR_MLD, "%s: called w/mbuf (%p,%d)", __func__, m, off);
1253 ifp = m->m_pkthdr.rcvif;
1255 ip6 = mtod(m, struct ip6_hdr *);
1257 /* Pullup to appropriate size. */
1258 mld = (struct mld_hdr *)(mtod(m, uint8_t *) + off);
1259 if (mld->mld_type == MLD_LISTENER_QUERY &&
1260 icmp6len >= sizeof(struct mldv2_query)) {
1261 mldlen = sizeof(struct mldv2_query);
1263 mldlen = sizeof(struct mld_hdr);
1265 IP6_EXTHDR_GET(mld, struct mld_hdr *, m, off, mldlen);
1267 ICMP6STAT_INC(icp6s_badlen);
1268 return (IPPROTO_DONE);
1272 * Userland needs to see all of this traffic for implementing
1273 * the endpoint discovery portion of multicast routing.
1275 switch (mld->mld_type) {
1276 case MLD_LISTENER_QUERY:
1277 icmp6_ifstat_inc(ifp, ifs6_in_mldquery);
1278 if (icmp6len == sizeof(struct mld_hdr)) {
1279 if (mld_v1_input_query(ifp, ip6, mld) != 0)
1281 } else if (icmp6len >= sizeof(struct mldv2_query)) {
1282 if (mld_v2_input_query(ifp, ip6, m, off,
1287 case MLD_LISTENER_REPORT:
1288 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1289 if (mld_v1_input_report(ifp, ip6, mld) != 0)
1292 case MLDV2_LISTENER_REPORT:
1293 icmp6_ifstat_inc(ifp, ifs6_in_mldreport);
1295 case MLD_LISTENER_DONE:
1296 icmp6_ifstat_inc(ifp, ifs6_in_mlddone);
1306 * Fast timeout handler (global).
1307 * VIMAGE: Timeout handlers are expected to service all vimages.
1312 VNET_ITERATOR_DECL(vnet_iter);
1314 VNET_LIST_RLOCK_NOSLEEP();
1315 VNET_FOREACH(vnet_iter) {
1316 CURVNET_SET(vnet_iter);
1317 mld_fasttimo_vnet();
1320 VNET_LIST_RUNLOCK_NOSLEEP();
1324 * Fast timeout handler (per-vnet).
1326 * VIMAGE: Assume caller has set up our curvnet.
1329 mld_fasttimo_vnet(void)
1331 struct ifqueue scq; /* State-change packets */
1332 struct ifqueue qrq; /* Query response packets */
1334 struct mld_ifinfo *mli;
1335 struct ifmultiaddr *ifma;
1336 struct in6_multi *inm, *tinm;
1342 * Quick check to see if any work needs to be done, in order to
1343 * minimize the overhead of fasttimo processing.
1344 * SMPng: XXX Unlocked reads.
1346 if (!V_current_state_timers_running6 &&
1347 !V_interface_timers_running6 &&
1348 !V_state_change_timers_running6)
1355 * MLDv2 General Query response timer processing.
1357 if (V_interface_timers_running6) {
1358 CTR1(KTR_MLD, "%s: interface timers running", __func__);
1360 V_interface_timers_running6 = 0;
1361 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1362 if (mli->mli_v2_timer == 0) {
1364 } else if (--mli->mli_v2_timer == 0) {
1365 mld_v2_dispatch_general_query(mli);
1367 V_interface_timers_running6 = 1;
1372 if (!V_current_state_timers_running6 &&
1373 !V_state_change_timers_running6)
1376 V_current_state_timers_running6 = 0;
1377 V_state_change_timers_running6 = 0;
1379 CTR1(KTR_MLD, "%s: state change timers running", __func__);
1382 * MLD host report and state-change timer processing.
1383 * Note: Processing a v2 group timer may remove a node.
1385 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1388 if (mli->mli_version == MLD_VERSION_2) {
1389 uri_fasthz = MLD_RANDOM_DELAY(mli->mli_uri *
1392 memset(&qrq, 0, sizeof(struct ifqueue));
1393 IFQ_SET_MAXLEN(&qrq, MLD_MAX_G_GS_PACKETS);
1395 memset(&scq, 0, sizeof(struct ifqueue));
1396 IFQ_SET_MAXLEN(&scq, MLD_MAX_STATE_CHANGE_PACKETS);
1400 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1401 if (ifma->ifma_addr->sa_family != AF_INET6 ||
1402 ifma->ifma_protospec == NULL)
1404 inm = (struct in6_multi *)ifma->ifma_protospec;
1405 switch (mli->mli_version) {
1407 mld_v1_process_group_timer(mli, inm);
1410 mld_v2_process_group_timers(mli, &qrq,
1411 &scq, inm, uri_fasthz);
1415 IF_ADDR_RUNLOCK(ifp);
1417 switch (mli->mli_version) {
1420 * Transmit reports for this lifecycle. This
1421 * is done while not holding IF_ADDR_LOCK
1422 * since this can call
1423 * in6ifa_ifpforlinklocal() which locks
1424 * IF_ADDR_LOCK internally as well as
1425 * ip6_output() to transmit a packet.
1427 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead,
1429 SLIST_REMOVE_HEAD(&mli->mli_relinmhead,
1431 (void)mld_v1_transmit_report(inm,
1432 MLD_LISTENER_REPORT);
1436 mld_dispatch_queue(&qrq, 0);
1437 mld_dispatch_queue(&scq, 0);
1440 * Free the in_multi reference(s) for
1443 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead,
1445 SLIST_REMOVE_HEAD(&mli->mli_relinmhead,
1447 in6m_release_locked(inm);
1459 * Update host report group timer.
1460 * Will update the global pending timer flags.
1463 mld_v1_process_group_timer(struct mld_ifinfo *mli, struct in6_multi *inm)
1465 int report_timer_expired;
1467 IN6_MULTI_LOCK_ASSERT();
1470 if (inm->in6m_timer == 0) {
1471 report_timer_expired = 0;
1472 } else if (--inm->in6m_timer == 0) {
1473 report_timer_expired = 1;
1475 V_current_state_timers_running6 = 1;
1479 switch (inm->in6m_state) {
1480 case MLD_NOT_MEMBER:
1481 case MLD_SILENT_MEMBER:
1482 case MLD_IDLE_MEMBER:
1483 case MLD_LAZY_MEMBER:
1484 case MLD_SLEEPING_MEMBER:
1485 case MLD_AWAKENING_MEMBER:
1487 case MLD_REPORTING_MEMBER:
1488 if (report_timer_expired) {
1489 inm->in6m_state = MLD_IDLE_MEMBER;
1490 SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm,
1494 case MLD_G_QUERY_PENDING_MEMBER:
1495 case MLD_SG_QUERY_PENDING_MEMBER:
1496 case MLD_LEAVING_MEMBER:
1502 * Update a group's timers for MLDv2.
1503 * Will update the global pending timer flags.
1504 * Note: Unlocked read from mli.
1507 mld_v2_process_group_timers(struct mld_ifinfo *mli,
1508 struct ifqueue *qrq, struct ifqueue *scq,
1509 struct in6_multi *inm, const int uri_fasthz)
1511 int query_response_timer_expired;
1512 int state_change_retransmit_timer_expired;
1514 char ip6tbuf[INET6_ADDRSTRLEN];
1517 IN6_MULTI_LOCK_ASSERT();
1520 query_response_timer_expired = 0;
1521 state_change_retransmit_timer_expired = 0;
1524 * During a transition from compatibility mode back to MLDv2,
1525 * a group record in REPORTING state may still have its group
1526 * timer active. This is a no-op in this function; it is easier
1527 * to deal with it here than to complicate the slow-timeout path.
1529 if (inm->in6m_timer == 0) {
1530 query_response_timer_expired = 0;
1531 } else if (--inm->in6m_timer == 0) {
1532 query_response_timer_expired = 1;
1534 V_current_state_timers_running6 = 1;
1537 if (inm->in6m_sctimer == 0) {
1538 state_change_retransmit_timer_expired = 0;
1539 } else if (--inm->in6m_sctimer == 0) {
1540 state_change_retransmit_timer_expired = 1;
1542 V_state_change_timers_running6 = 1;
1545 /* We are in fasttimo, so be quick about it. */
1546 if (!state_change_retransmit_timer_expired &&
1547 !query_response_timer_expired)
1550 switch (inm->in6m_state) {
1551 case MLD_NOT_MEMBER:
1552 case MLD_SILENT_MEMBER:
1553 case MLD_SLEEPING_MEMBER:
1554 case MLD_LAZY_MEMBER:
1555 case MLD_AWAKENING_MEMBER:
1556 case MLD_IDLE_MEMBER:
1558 case MLD_G_QUERY_PENDING_MEMBER:
1559 case MLD_SG_QUERY_PENDING_MEMBER:
1561 * Respond to a previously pending Group-Specific
1562 * or Group-and-Source-Specific query by enqueueing
1563 * the appropriate Current-State report for
1564 * immediate transmission.
1566 if (query_response_timer_expired) {
1569 retval = mld_v2_enqueue_group_record(qrq, inm, 0, 1,
1570 (inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER),
1572 CTR2(KTR_MLD, "%s: enqueue record = %d",
1574 inm->in6m_state = MLD_REPORTING_MEMBER;
1575 in6m_clear_recorded(inm);
1578 case MLD_REPORTING_MEMBER:
1579 case MLD_LEAVING_MEMBER:
1580 if (state_change_retransmit_timer_expired) {
1582 * State-change retransmission timer fired.
1583 * If there are any further pending retransmissions,
1584 * set the global pending state-change flag, and
1587 if (--inm->in6m_scrv > 0) {
1588 inm->in6m_sctimer = uri_fasthz;
1589 V_state_change_timers_running6 = 1;
1592 * Retransmit the previously computed state-change
1593 * report. If there are no further pending
1594 * retransmissions, the mbuf queue will be consumed.
1595 * Update T0 state to T1 as we have now sent
1598 (void)mld_v2_merge_state_changes(inm, scq);
1601 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
1602 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1603 if_name(inm->in6m_ifp));
1606 * If we are leaving the group for good, make sure
1607 * we release MLD's reference to it.
1608 * This release must be deferred using a SLIST,
1609 * as we are called from a loop which traverses
1610 * the in_ifmultiaddr TAILQ.
1612 if (inm->in6m_state == MLD_LEAVING_MEMBER &&
1613 inm->in6m_scrv == 0) {
1614 inm->in6m_state = MLD_NOT_MEMBER;
1615 SLIST_INSERT_HEAD(&mli->mli_relinmhead,
1624 * Switch to a different version on the given interface,
1625 * as per Section 9.12.
1628 mld_set_version(struct mld_ifinfo *mli, const int version)
1630 int old_version_timer;
1634 CTR4(KTR_MLD, "%s: switching to v%d on ifp %p(%s)", __func__,
1635 version, mli->mli_ifp, if_name(mli->mli_ifp));
1637 if (version == MLD_VERSION_1) {
1639 * Compute the "Older Version Querier Present" timer as per
1642 old_version_timer = (mli->mli_rv * mli->mli_qi) + mli->mli_qri;
1643 old_version_timer *= PR_SLOWHZ;
1644 mli->mli_v1_timer = old_version_timer;
1647 if (mli->mli_v1_timer > 0 && mli->mli_version != MLD_VERSION_1) {
1648 mli->mli_version = MLD_VERSION_1;
1649 mld_v2_cancel_link_timers(mli);
1654 * Cancel pending MLDv2 timers for the given link and all groups
1655 * joined on it; state-change, general-query, and group-query timers.
1658 mld_v2_cancel_link_timers(struct mld_ifinfo *mli)
1660 struct ifmultiaddr *ifma;
1662 struct in6_multi *inm, *tinm;
1664 CTR3(KTR_MLD, "%s: cancel v2 timers on ifp %p(%s)", __func__,
1665 mli->mli_ifp, if_name(mli->mli_ifp));
1667 IN6_MULTI_LOCK_ASSERT();
1671 * Fast-track this potentially expensive operation
1672 * by checking all the global 'timer pending' flags.
1674 if (!V_interface_timers_running6 &&
1675 !V_state_change_timers_running6 &&
1676 !V_current_state_timers_running6)
1679 mli->mli_v2_timer = 0;
1684 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1685 if (ifma->ifma_addr->sa_family != AF_INET6)
1687 inm = (struct in6_multi *)ifma->ifma_protospec;
1688 switch (inm->in6m_state) {
1689 case MLD_NOT_MEMBER:
1690 case MLD_SILENT_MEMBER:
1691 case MLD_IDLE_MEMBER:
1692 case MLD_LAZY_MEMBER:
1693 case MLD_SLEEPING_MEMBER:
1694 case MLD_AWAKENING_MEMBER:
1696 case MLD_LEAVING_MEMBER:
1698 * If we are leaving the group and switching
1699 * version, we need to release the final
1700 * reference held for issuing the INCLUDE {}.
1702 SLIST_INSERT_HEAD(&mli->mli_relinmhead, inm,
1705 case MLD_G_QUERY_PENDING_MEMBER:
1706 case MLD_SG_QUERY_PENDING_MEMBER:
1707 in6m_clear_recorded(inm);
1709 case MLD_REPORTING_MEMBER:
1710 inm->in6m_sctimer = 0;
1711 inm->in6m_timer = 0;
1712 inm->in6m_state = MLD_REPORTING_MEMBER;
1714 * Free any pending MLDv2 state-change records.
1716 _IF_DRAIN(&inm->in6m_scq);
1720 IF_ADDR_RUNLOCK(ifp);
1721 SLIST_FOREACH_SAFE(inm, &mli->mli_relinmhead, in6m_nrele, tinm) {
1722 SLIST_REMOVE_HEAD(&mli->mli_relinmhead, in6m_nrele);
1723 in6m_release_locked(inm);
1728 * Global slowtimo handler.
1729 * VIMAGE: Timeout handlers are expected to service all vimages.
1734 VNET_ITERATOR_DECL(vnet_iter);
1736 VNET_LIST_RLOCK_NOSLEEP();
1737 VNET_FOREACH(vnet_iter) {
1738 CURVNET_SET(vnet_iter);
1739 mld_slowtimo_vnet();
1742 VNET_LIST_RUNLOCK_NOSLEEP();
1746 * Per-vnet slowtimo handler.
1749 mld_slowtimo_vnet(void)
1751 struct mld_ifinfo *mli;
1755 LIST_FOREACH(mli, &V_mli_head, mli_link) {
1756 mld_v1_process_querier_timers(mli);
1763 * Update the Older Version Querier Present timers for a link.
1764 * See Section 9.12 of RFC 3810.
1767 mld_v1_process_querier_timers(struct mld_ifinfo *mli)
1772 if (mli->mli_version != MLD_VERSION_2 && --mli->mli_v1_timer == 0) {
1774 * MLDv1 Querier Present timer expired; revert to MLDv2.
1777 "%s: transition from v%d -> v%d on %p(%s)",
1778 __func__, mli->mli_version, MLD_VERSION_2,
1779 mli->mli_ifp, if_name(mli->mli_ifp));
1780 mli->mli_version = MLD_VERSION_2;
1785 * Transmit an MLDv1 report immediately.
1788 mld_v1_transmit_report(struct in6_multi *in6m, const int type)
1791 struct in6_ifaddr *ia;
1792 struct ip6_hdr *ip6;
1793 struct mbuf *mh, *md;
1794 struct mld_hdr *mld;
1796 IN6_MULTI_LOCK_ASSERT();
1799 ifp = in6m->in6m_ifp;
1800 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
1801 /* ia may be NULL if link-local address is tentative. */
1803 mh = m_gethdr(M_NOWAIT, MT_DATA);
1806 ifa_free(&ia->ia_ifa);
1809 md = m_get(M_NOWAIT, MT_DATA);
1813 ifa_free(&ia->ia_ifa);
1819 * FUTURE: Consider increasing alignment by ETHER_HDR_LEN, so
1820 * that ether_output() does not need to allocate another mbuf
1821 * for the header in the most common case.
1823 MH_ALIGN(mh, sizeof(struct ip6_hdr));
1824 mh->m_pkthdr.len = sizeof(struct ip6_hdr) + sizeof(struct mld_hdr);
1825 mh->m_len = sizeof(struct ip6_hdr);
1827 ip6 = mtod(mh, struct ip6_hdr *);
1829 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1830 ip6->ip6_vfc |= IPV6_VERSION;
1831 ip6->ip6_nxt = IPPROTO_ICMPV6;
1832 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
1833 ip6->ip6_dst = in6m->in6m_addr;
1835 md->m_len = sizeof(struct mld_hdr);
1836 mld = mtod(md, struct mld_hdr *);
1837 mld->mld_type = type;
1840 mld->mld_maxdelay = 0;
1841 mld->mld_reserved = 0;
1842 mld->mld_addr = in6m->in6m_addr;
1843 in6_clearscope(&mld->mld_addr);
1844 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
1845 sizeof(struct ip6_hdr), sizeof(struct mld_hdr));
1847 mld_save_context(mh, ifp);
1848 mh->m_flags |= M_MLDV1;
1850 mld_dispatch_packet(mh);
1853 ifa_free(&ia->ia_ifa);
1858 * Process a state change from the upper layer for the given IPv6 group.
1860 * Each socket holds a reference on the in_multi in its own ip_moptions.
1861 * The socket layer will have made the necessary updates to.the group
1862 * state, it is now up to MLD to issue a state change report if there
1863 * has been any change between T0 (when the last state-change was issued)
1866 * We use the MLDv2 state machine at group level. The MLd module
1867 * however makes the decision as to which MLD protocol version to speak.
1868 * A state change *from* INCLUDE {} always means an initial join.
1869 * A state change *to* INCLUDE {} always means a final leave.
1871 * If delay is non-zero, and the state change is an initial multicast
1872 * join, the state change report will be delayed by 'delay' ticks
1873 * in units of PR_FASTHZ if MLDv1 is active on the link; otherwise
1874 * the initial MLDv2 state change report will be delayed by whichever
1875 * is sooner, a pending state-change timer or delay itself.
1877 * VIMAGE: curvnet should have been set by caller, as this routine
1878 * is called from the socket option handlers.
1881 mld_change_state(struct in6_multi *inm, const int delay)
1883 struct mld_ifinfo *mli;
1887 IN6_MULTI_LOCK_ASSERT();
1892 * Try to detect if the upper layer just asked us to change state
1893 * for an interface which has now gone away.
1895 KASSERT(inm->in6m_ifma != NULL, ("%s: no ifma", __func__));
1896 ifp = inm->in6m_ifma->ifma_ifp;
1899 * Sanity check that netinet6's notion of ifp is the
1902 KASSERT(inm->in6m_ifp == ifp, ("%s: bad ifp", __func__));
1907 mli = MLD_IFINFO(ifp);
1908 KASSERT(mli != NULL, ("%s: no mld_ifinfo for ifp %p", __func__, ifp));
1911 * If we detect a state transition to or from MCAST_UNDEFINED
1912 * for this group, then we are starting or finishing an MLD
1913 * life cycle for this group.
1915 if (inm->in6m_st[1].iss_fmode != inm->in6m_st[0].iss_fmode) {
1916 CTR3(KTR_MLD, "%s: inm transition %d -> %d", __func__,
1917 inm->in6m_st[0].iss_fmode, inm->in6m_st[1].iss_fmode);
1918 if (inm->in6m_st[0].iss_fmode == MCAST_UNDEFINED) {
1919 CTR1(KTR_MLD, "%s: initial join", __func__);
1920 error = mld_initial_join(inm, mli, delay);
1922 } else if (inm->in6m_st[1].iss_fmode == MCAST_UNDEFINED) {
1923 CTR1(KTR_MLD, "%s: final leave", __func__);
1924 mld_final_leave(inm, mli);
1928 CTR1(KTR_MLD, "%s: filter set change", __func__);
1931 error = mld_handle_state_change(inm, mli);
1939 * Perform the initial join for an MLD group.
1941 * When joining a group:
1942 * If the group should have its MLD traffic suppressed, do nothing.
1943 * MLDv1 starts sending MLDv1 host membership reports.
1944 * MLDv2 will schedule an MLDv2 state-change report containing the
1945 * initial state of the membership.
1947 * If the delay argument is non-zero, then we must delay sending the
1948 * initial state change for delay ticks (in units of PR_FASTHZ).
1951 mld_initial_join(struct in6_multi *inm, struct mld_ifinfo *mli,
1955 struct ifqueue *ifq;
1956 int error, retval, syncstates;
1959 char ip6tbuf[INET6_ADDRSTRLEN];
1962 CTR4(KTR_MLD, "%s: initial join %s on ifp %p(%s)",
1963 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
1964 inm->in6m_ifp, if_name(inm->in6m_ifp));
1969 ifp = inm->in6m_ifp;
1971 IN6_MULTI_LOCK_ASSERT();
1974 KASSERT(mli && mli->mli_ifp == ifp, ("%s: inconsistent ifp", __func__));
1977 * Groups joined on loopback or marked as 'not reported',
1978 * enter the MLD_SILENT_MEMBER state and
1979 * are never reported in any protocol exchanges.
1980 * All other groups enter the appropriate state machine
1981 * for the version in use on this link.
1982 * A link marked as MLIF_SILENT causes MLD to be completely
1983 * disabled for the link.
1985 if ((ifp->if_flags & IFF_LOOPBACK) ||
1986 (mli->mli_flags & MLIF_SILENT) ||
1987 !mld_is_addr_reported(&inm->in6m_addr)) {
1989 "%s: not kicking state machine for silent group", __func__);
1990 inm->in6m_state = MLD_SILENT_MEMBER;
1991 inm->in6m_timer = 0;
1994 * Deal with overlapping in_multi lifecycle.
1995 * If this group was LEAVING, then make sure
1996 * we drop the reference we picked up to keep the
1997 * group around for the final INCLUDE {} enqueue.
1999 if (mli->mli_version == MLD_VERSION_2 &&
2000 inm->in6m_state == MLD_LEAVING_MEMBER)
2001 in6m_release_locked(inm);
2003 inm->in6m_state = MLD_REPORTING_MEMBER;
2005 switch (mli->mli_version) {
2008 * If a delay was provided, only use it if
2009 * it is greater than the delay normally
2010 * used for an MLDv1 state change report,
2011 * and delay sending the initial MLDv1 report
2012 * by not transitioning to the IDLE state.
2014 odelay = MLD_RANDOM_DELAY(MLD_V1_MAX_RI * PR_FASTHZ);
2016 inm->in6m_timer = max(delay, odelay);
2017 V_current_state_timers_running6 = 1;
2019 inm->in6m_state = MLD_IDLE_MEMBER;
2020 error = mld_v1_transmit_report(inm,
2021 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 ifq = &inm->in6m_scq;
2044 retval = mld_v2_enqueue_group_record(ifq, 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_ifinfo *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_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 _IF_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_ifinfo *mli)
2168 char ip6tbuf[INET6_ADDRSTRLEN];
2173 CTR4(KTR_MLD, "%s: final leave %s on ifp %p(%s)",
2174 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2175 inm->in6m_ifp, if_name(inm->in6m_ifp));
2177 IN6_MULTI_LOCK_ASSERT();
2180 switch (inm->in6m_state) {
2181 case MLD_NOT_MEMBER:
2182 case MLD_SILENT_MEMBER:
2183 case MLD_LEAVING_MEMBER:
2184 /* Already leaving or left; do nothing. */
2186 "%s: not kicking state machine for silent group", __func__);
2188 case MLD_REPORTING_MEMBER:
2189 case MLD_IDLE_MEMBER:
2190 case MLD_G_QUERY_PENDING_MEMBER:
2191 case MLD_SG_QUERY_PENDING_MEMBER:
2192 if (mli->mli_version == MLD_VERSION_1) {
2194 if (inm->in6m_state == MLD_G_QUERY_PENDING_MEMBER ||
2195 inm->in6m_state == MLD_SG_QUERY_PENDING_MEMBER)
2196 panic("%s: MLDv2 state reached, not MLDv2 mode",
2199 mld_v1_transmit_report(inm, MLD_LISTENER_DONE);
2200 inm->in6m_state = MLD_NOT_MEMBER;
2201 V_current_state_timers_running6 = 1;
2202 } else if (mli->mli_version == MLD_VERSION_2) {
2204 * Stop group timer and all pending reports.
2205 * Immediately enqueue a state-change report
2206 * TO_IN {} to be sent on the next fast timeout,
2207 * giving us an opportunity to merge reports.
2209 _IF_DRAIN(&inm->in6m_scq);
2210 inm->in6m_timer = 0;
2211 inm->in6m_scrv = mli->mli_rv;
2212 CTR4(KTR_MLD, "%s: Leaving %s/%s with %d "
2213 "pending retransmissions.", __func__,
2214 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2215 if_name(inm->in6m_ifp), inm->in6m_scrv);
2216 if (inm->in6m_scrv == 0) {
2217 inm->in6m_state = MLD_NOT_MEMBER;
2218 inm->in6m_sctimer = 0;
2222 in6m_acquire_locked(inm);
2224 retval = mld_v2_enqueue_group_record(
2225 &inm->in6m_scq, inm, 1, 0, 0,
2226 (mli->mli_flags & MLIF_USEALLOW));
2227 KASSERT(retval != 0,
2228 ("%s: enqueue record = %d", __func__,
2231 inm->in6m_state = MLD_LEAVING_MEMBER;
2232 inm->in6m_sctimer = 1;
2233 V_state_change_timers_running6 = 1;
2239 case MLD_LAZY_MEMBER:
2240 case MLD_SLEEPING_MEMBER:
2241 case MLD_AWAKENING_MEMBER:
2242 /* Our reports are suppressed; do nothing. */
2248 CTR3(KTR_MLD, "%s: T1 -> T0 for %s/%s", __func__,
2249 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2250 if_name(inm->in6m_ifp));
2251 inm->in6m_st[1].iss_fmode = MCAST_UNDEFINED;
2252 CTR3(KTR_MLD, "%s: T1 now MCAST_UNDEFINED for %p/%s",
2253 __func__, &inm->in6m_addr, if_name(inm->in6m_ifp));
2258 * Enqueue an MLDv2 group record to the given output queue.
2260 * If is_state_change is zero, a current-state record is appended.
2261 * If is_state_change is non-zero, a state-change report is appended.
2263 * If is_group_query is non-zero, an mbuf packet chain is allocated.
2264 * If is_group_query is zero, and if there is a packet with free space
2265 * at the tail of the queue, it will be appended to providing there
2266 * is enough free space.
2267 * Otherwise a new mbuf packet chain is allocated.
2269 * If is_source_query is non-zero, each source is checked to see if
2270 * it was recorded for a Group-Source query, and will be omitted if
2271 * it is not both in-mode and recorded.
2273 * If use_block_allow is non-zero, state change reports for initial join
2274 * and final leave, on an inclusive mode group with a source list, will be
2275 * rewritten to use the ALLOW_NEW and BLOCK_OLD record types, respectively.
2277 * The function will attempt to allocate leading space in the packet
2278 * for the IPv6+ICMP headers to be prepended without fragmenting the chain.
2280 * If successful the size of all data appended to the queue is returned,
2281 * otherwise an error code less than zero is returned, or zero if
2282 * no record(s) were appended.
2285 mld_v2_enqueue_group_record(struct ifqueue *ifq, struct in6_multi *inm,
2286 const int is_state_change, const int is_group_query,
2287 const int is_source_query, const int use_block_allow)
2289 struct mldv2_record mr;
2290 struct mldv2_record *pmr;
2292 struct ip6_msource *ims, *nims;
2293 struct mbuf *m0, *m, *md;
2294 int error, is_filter_list_change;
2295 int minrec0len, m0srcs, msrcs, nbytes, off;
2296 int record_has_sources;
2301 char ip6tbuf[INET6_ADDRSTRLEN];
2304 IN6_MULTI_LOCK_ASSERT();
2307 ifp = inm->in6m_ifp;
2308 is_filter_list_change = 0;
2315 record_has_sources = 1;
2317 type = MLD_DO_NOTHING;
2318 mode = inm->in6m_st[1].iss_fmode;
2321 * If we did not transition out of ASM mode during t0->t1,
2322 * and there are no source nodes to process, we can skip
2323 * the generation of source records.
2325 if (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0 &&
2326 inm->in6m_nsrc == 0)
2327 record_has_sources = 0;
2329 if (is_state_change) {
2331 * Queue a state change record.
2332 * If the mode did not change, and there are non-ASM
2333 * listeners or source filters present,
2334 * we potentially need to issue two records for the group.
2335 * If there are ASM listeners, and there was no filter
2336 * mode transition of any kind, do nothing.
2338 * If we are transitioning to MCAST_UNDEFINED, we need
2339 * not send any sources. A transition to/from this state is
2340 * considered inclusive with some special treatment.
2342 * If we are rewriting initial joins/leaves to use
2343 * ALLOW/BLOCK, and the group's membership is inclusive,
2344 * we need to send sources in all cases.
2346 if (mode != inm->in6m_st[0].iss_fmode) {
2347 if (mode == MCAST_EXCLUDE) {
2348 CTR1(KTR_MLD, "%s: change to EXCLUDE",
2350 type = MLD_CHANGE_TO_EXCLUDE_MODE;
2352 CTR1(KTR_MLD, "%s: change to INCLUDE",
2354 if (use_block_allow) {
2357 * Here we're interested in state
2358 * edges either direction between
2359 * MCAST_UNDEFINED and MCAST_INCLUDE.
2360 * Perhaps we should just check
2361 * the group state, rather than
2364 if (mode == MCAST_UNDEFINED) {
2365 type = MLD_BLOCK_OLD_SOURCES;
2367 type = MLD_ALLOW_NEW_SOURCES;
2370 type = MLD_CHANGE_TO_INCLUDE_MODE;
2371 if (mode == MCAST_UNDEFINED)
2372 record_has_sources = 0;
2376 if (record_has_sources) {
2377 is_filter_list_change = 1;
2379 type = MLD_DO_NOTHING;
2384 * Queue a current state record.
2386 if (mode == MCAST_EXCLUDE) {
2387 type = MLD_MODE_IS_EXCLUDE;
2388 } else if (mode == MCAST_INCLUDE) {
2389 type = MLD_MODE_IS_INCLUDE;
2390 KASSERT(inm->in6m_st[1].iss_asm == 0,
2391 ("%s: inm %p is INCLUDE but ASM count is %d",
2392 __func__, inm, inm->in6m_st[1].iss_asm));
2397 * Generate the filter list changes using a separate function.
2399 if (is_filter_list_change)
2400 return (mld_v2_enqueue_filter_change(ifq, inm));
2402 if (type == MLD_DO_NOTHING) {
2403 CTR3(KTR_MLD, "%s: nothing to do for %s/%s",
2404 __func__, ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2405 if_name(inm->in6m_ifp));
2410 * If any sources are present, we must be able to fit at least
2411 * one in the trailing space of the tail packet's mbuf,
2414 minrec0len = sizeof(struct mldv2_record);
2415 if (record_has_sources)
2416 minrec0len += sizeof(struct in6_addr);
2418 CTR4(KTR_MLD, "%s: queueing %s for %s/%s", __func__,
2419 mld_rec_type_to_str(type),
2420 ip6_sprintf(ip6tbuf, &inm->in6m_addr),
2421 if_name(inm->in6m_ifp));
2424 * Check if we have a packet in the tail of the queue for this
2425 * group into which the first group record for this group will fit.
2426 * Otherwise allocate a new packet.
2427 * Always allocate leading space for IP6+RA+ICMPV6+REPORT.
2428 * Note: Group records for G/GSR query responses MUST be sent
2429 * in their own packet.
2432 if (!is_group_query &&
2434 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <= MLD_V2_REPORT_MAXRECS) &&
2435 (m0->m_pkthdr.len + minrec0len) <
2436 (ifp->if_mtu - MLD_MTUSPACE)) {
2437 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2438 sizeof(struct mldv2_record)) /
2439 sizeof(struct in6_addr);
2441 CTR1(KTR_MLD, "%s: use existing packet", __func__);
2443 if (_IF_QFULL(ifq)) {
2444 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2448 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2449 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2450 if (!is_state_change && !is_group_query)
2451 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2453 m = m_gethdr(M_NOWAIT, MT_DATA);
2457 mld_save_context(m, ifp);
2459 CTR1(KTR_MLD, "%s: allocated first packet", __func__);
2463 * Append group record.
2464 * If we have sources, we don't know how many yet.
2469 mr.mr_addr = inm->in6m_addr;
2470 in6_clearscope(&mr.mr_addr);
2471 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2474 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2477 nbytes += sizeof(struct mldv2_record);
2480 * Append as many sources as will fit in the first packet.
2481 * If we are appending to a new packet, the chain allocation
2482 * may potentially use clusters; use m_getptr() in this case.
2483 * If we are appending to an existing packet, we need to obtain
2484 * a pointer to the group record after m_append(), in case a new
2485 * mbuf was allocated.
2487 * Only append sources which are in-mode at t1. If we are
2488 * transitioning to MCAST_UNDEFINED state on the group, and
2489 * use_block_allow is zero, do not include source entries.
2490 * Otherwise, we need to include this source in the report.
2492 * Only report recorded sources in our filter set when responding
2493 * to a group-source query.
2495 if (record_has_sources) {
2498 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2499 md->m_len - nbytes);
2501 md = m_getptr(m, 0, &off);
2502 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) +
2506 RB_FOREACH_SAFE(ims, ip6_msource_tree, &inm->in6m_srcs,
2508 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2509 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2510 now = im6s_get_mode(inm, ims, 1);
2511 CTR2(KTR_MLD, "%s: node is %d", __func__, now);
2512 if ((now != mode) ||
2514 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2515 CTR1(KTR_MLD, "%s: skip node", __func__);
2518 if (is_source_query && ims->im6s_stp == 0) {
2519 CTR1(KTR_MLD, "%s: skip unrecorded node",
2523 CTR1(KTR_MLD, "%s: append node", __func__);
2524 if (!m_append(m, sizeof(struct in6_addr),
2525 (void *)&ims->im6s_addr)) {
2528 CTR1(KTR_MLD, "%s: m_append() failed.",
2532 nbytes += sizeof(struct in6_addr);
2534 if (msrcs == m0srcs)
2537 CTR2(KTR_MLD, "%s: msrcs is %d this packet", __func__,
2539 pmr->mr_numsrc = htons(msrcs);
2540 nbytes += (msrcs * sizeof(struct in6_addr));
2543 if (is_source_query && msrcs == 0) {
2544 CTR1(KTR_MLD, "%s: no recorded sources to report", __func__);
2551 * We are good to go with first packet.
2554 CTR1(KTR_MLD, "%s: enqueueing first packet", __func__);
2555 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2556 _IF_ENQUEUE(ifq, m);
2558 m->m_pkthdr.PH_vt.vt_nrecs++;
2561 * No further work needed if no source list in packet(s).
2563 if (!record_has_sources)
2567 * Whilst sources remain to be announced, we need to allocate
2568 * a new packet and fill out as many sources as will fit.
2569 * Always try for a cluster first.
2571 while (nims != NULL) {
2572 if (_IF_QFULL(ifq)) {
2573 CTR1(KTR_MLD, "%s: outbound queue full", __func__);
2576 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2578 m = m_gethdr(M_NOWAIT, MT_DATA);
2581 mld_save_context(m, ifp);
2582 md = m_getptr(m, 0, &off);
2583 pmr = (struct mldv2_record *)(mtod(md, uint8_t *) + off);
2584 CTR1(KTR_MLD, "%s: allocated next packet", __func__);
2586 if (!m_append(m, sizeof(struct mldv2_record), (void *)&mr)) {
2589 CTR1(KTR_MLD, "%s: m_append() failed.", __func__);
2592 m->m_pkthdr.PH_vt.vt_nrecs = 1;
2593 nbytes += sizeof(struct mldv2_record);
2595 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2596 sizeof(struct mldv2_record)) / sizeof(struct in6_addr);
2599 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2600 CTR2(KTR_MLD, "%s: visit node %s",
2601 __func__, ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2602 now = im6s_get_mode(inm, ims, 1);
2603 if ((now != mode) ||
2605 (!use_block_allow && mode == MCAST_UNDEFINED))) {
2606 CTR1(KTR_MLD, "%s: skip node", __func__);
2609 if (is_source_query && ims->im6s_stp == 0) {
2610 CTR1(KTR_MLD, "%s: skip unrecorded node",
2614 CTR1(KTR_MLD, "%s: append node", __func__);
2615 if (!m_append(m, sizeof(struct in6_addr),
2616 (void *)&ims->im6s_addr)) {
2619 CTR1(KTR_MLD, "%s: m_append() failed.",
2624 if (msrcs == m0srcs)
2627 pmr->mr_numsrc = htons(msrcs);
2628 nbytes += (msrcs * sizeof(struct in6_addr));
2630 CTR1(KTR_MLD, "%s: enqueueing next packet", __func__);
2631 _IF_ENQUEUE(ifq, m);
2638 * Type used to mark record pass completion.
2639 * We exploit the fact we can cast to this easily from the
2640 * current filter modes on each ip_msource node.
2643 REC_NONE = 0x00, /* MCAST_UNDEFINED */
2644 REC_ALLOW = 0x01, /* MCAST_INCLUDE */
2645 REC_BLOCK = 0x02, /* MCAST_EXCLUDE */
2646 REC_FULL = REC_ALLOW | REC_BLOCK
2650 * Enqueue an MLDv2 filter list change to the given output queue.
2652 * Source list filter state is held in an RB-tree. When the filter list
2653 * for a group is changed without changing its mode, we need to compute
2654 * the deltas between T0 and T1 for each source in the filter set,
2655 * and enqueue the appropriate ALLOW_NEW/BLOCK_OLD records.
2657 * As we may potentially queue two record types, and the entire R-B tree
2658 * needs to be walked at once, we break this out into its own function
2659 * so we can generate a tightly packed queue of packets.
2661 * XXX This could be written to only use one tree walk, although that makes
2662 * serializing into the mbuf chains a bit harder. For now we do two walks
2663 * which makes things easier on us, and it may or may not be harder on
2666 * If successful the size of all data appended to the queue is returned,
2667 * otherwise an error code less than zero is returned, or zero if
2668 * no record(s) were appended.
2671 mld_v2_enqueue_filter_change(struct ifqueue *ifq, struct in6_multi *inm)
2673 static const int MINRECLEN =
2674 sizeof(struct mldv2_record) + sizeof(struct in6_addr);
2676 struct mldv2_record mr;
2677 struct mldv2_record *pmr;
2678 struct ip6_msource *ims, *nims;
2679 struct mbuf *m, *m0, *md;
2680 int m0srcs, nbytes, npbytes, off, rsrcs, schanged;
2682 uint8_t mode, now, then;
2683 rectype_t crt, drt, nrt;
2685 char ip6tbuf[INET6_ADDRSTRLEN];
2688 IN6_MULTI_LOCK_ASSERT();
2690 if (inm->in6m_nsrc == 0 ||
2691 (inm->in6m_st[0].iss_asm > 0 && inm->in6m_st[1].iss_asm > 0))
2694 ifp = inm->in6m_ifp; /* interface */
2695 mode = inm->in6m_st[1].iss_fmode; /* filter mode at t1 */
2696 crt = REC_NONE; /* current group record type */
2697 drt = REC_NONE; /* mask of completed group record types */
2698 nrt = REC_NONE; /* record type for current node */
2699 m0srcs = 0; /* # source which will fit in current mbuf chain */
2700 npbytes = 0; /* # of bytes appended this packet */
2701 nbytes = 0; /* # of bytes appended to group's state-change queue */
2702 rsrcs = 0; /* # sources encoded in current record */
2703 schanged = 0; /* # nodes encoded in overall filter change */
2704 nallow = 0; /* # of source entries in ALLOW_NEW */
2705 nblock = 0; /* # of source entries in BLOCK_OLD */
2706 nims = NULL; /* next tree node pointer */
2709 * For each possible filter record mode.
2710 * The first kind of source we encounter tells us which
2711 * is the first kind of record we start appending.
2712 * If a node transitioned to UNDEFINED at t1, its mode is treated
2713 * as the inverse of the group's filter mode.
2715 while (drt != REC_FULL) {
2719 (m0->m_pkthdr.PH_vt.vt_nrecs + 1 <=
2720 MLD_V2_REPORT_MAXRECS) &&
2721 (m0->m_pkthdr.len + MINRECLEN) <
2722 (ifp->if_mtu - MLD_MTUSPACE)) {
2724 m0srcs = (ifp->if_mtu - m0->m_pkthdr.len -
2725 sizeof(struct mldv2_record)) /
2726 sizeof(struct in6_addr);
2728 "%s: use previous packet", __func__);
2730 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2732 m = m_gethdr(M_NOWAIT, MT_DATA);
2735 "%s: m_get*() failed", __func__);
2738 m->m_pkthdr.PH_vt.vt_nrecs = 0;
2739 mld_save_context(m, ifp);
2740 m0srcs = (ifp->if_mtu - MLD_MTUSPACE -
2741 sizeof(struct mldv2_record)) /
2742 sizeof(struct in6_addr);
2745 "%s: allocated new packet", __func__);
2748 * Append the MLD group record header to the
2749 * current packet's data area.
2750 * Recalculate pointer to free space for next
2751 * group record, in case m_append() allocated
2752 * a new mbuf or cluster.
2754 memset(&mr, 0, sizeof(mr));
2755 mr.mr_addr = inm->in6m_addr;
2756 in6_clearscope(&mr.mr_addr);
2757 if (!m_append(m, sizeof(mr), (void *)&mr)) {
2761 "%s: m_append() failed", __func__);
2764 npbytes += sizeof(struct mldv2_record);
2766 /* new packet; offset in chain */
2767 md = m_getptr(m, npbytes -
2768 sizeof(struct mldv2_record), &off);
2769 pmr = (struct mldv2_record *)(mtod(md,
2772 /* current packet; offset from last append */
2774 pmr = (struct mldv2_record *)(mtod(md,
2775 uint8_t *) + md->m_len -
2776 sizeof(struct mldv2_record));
2779 * Begin walking the tree for this record type
2780 * pass, or continue from where we left off
2781 * previously if we had to allocate a new packet.
2782 * Only report deltas in-mode at t1.
2783 * We need not report included sources as allowed
2784 * if we are in inclusive mode on the group,
2785 * however the converse is not true.
2789 nims = RB_MIN(ip6_msource_tree,
2792 RB_FOREACH_FROM(ims, ip6_msource_tree, nims) {
2793 CTR2(KTR_MLD, "%s: visit node %s", __func__,
2794 ip6_sprintf(ip6tbuf, &ims->im6s_addr));
2795 now = im6s_get_mode(inm, ims, 1);
2796 then = im6s_get_mode(inm, ims, 0);
2797 CTR3(KTR_MLD, "%s: mode: t0 %d, t1 %d",
2798 __func__, then, now);
2801 "%s: skip unchanged", __func__);
2804 if (mode == MCAST_EXCLUDE &&
2805 now == MCAST_INCLUDE) {
2807 "%s: skip IN src on EX group",
2811 nrt = (rectype_t)now;
2812 if (nrt == REC_NONE)
2813 nrt = (rectype_t)(~mode & REC_FULL);
2814 if (schanged++ == 0) {
2816 } else if (crt != nrt)
2818 if (!m_append(m, sizeof(struct in6_addr),
2819 (void *)&ims->im6s_addr)) {
2823 "%s: m_append() failed", __func__);
2826 nallow += !!(crt == REC_ALLOW);
2827 nblock += !!(crt == REC_BLOCK);
2828 if (++rsrcs == m0srcs)
2832 * If we did not append any tree nodes on this
2833 * pass, back out of allocations.
2836 npbytes -= sizeof(struct mldv2_record);
2839 "%s: m_free(m)", __func__);
2843 "%s: m_adj(m, -mr)", __func__);
2844 m_adj(m, -((int)sizeof(
2845 struct mldv2_record)));
2849 npbytes += (rsrcs * sizeof(struct in6_addr));
2850 if (crt == REC_ALLOW)
2851 pmr->mr_type = MLD_ALLOW_NEW_SOURCES;
2852 else if (crt == REC_BLOCK)
2853 pmr->mr_type = MLD_BLOCK_OLD_SOURCES;
2854 pmr->mr_numsrc = htons(rsrcs);
2856 * Count the new group record, and enqueue this
2857 * packet if it wasn't already queued.
2859 m->m_pkthdr.PH_vt.vt_nrecs++;
2861 _IF_ENQUEUE(ifq, m);
2863 } while (nims != NULL);
2865 crt = (~crt & REC_FULL);
2868 CTR3(KTR_MLD, "%s: queued %d ALLOW_NEW, %d BLOCK_OLD", __func__,
2875 mld_v2_merge_state_changes(struct in6_multi *inm, struct ifqueue *ifscq)
2878 struct mbuf *m; /* pending state-change */
2879 struct mbuf *m0; /* copy of pending state-change */
2880 struct mbuf *mt; /* last state-change in packet */
2881 int docopy, domerge;
2888 IN6_MULTI_LOCK_ASSERT();
2892 * If there are further pending retransmissions, make a writable
2893 * copy of each queued state-change message before merging.
2895 if (inm->in6m_scrv > 0)
2898 gq = &inm->in6m_scq;
2900 if (gq->ifq_head == NULL) {
2901 CTR2(KTR_MLD, "%s: WARNING: queue for inm %p is empty",
2909 * Only merge the report into the current packet if
2910 * there is sufficient space to do so; an MLDv2 report
2911 * packet may only contain 65,535 group records.
2912 * Always use a simple mbuf chain concatentation to do this,
2913 * as large state changes for single groups may have
2914 * allocated clusters.
2917 mt = ifscq->ifq_tail;
2919 recslen = m_length(m, NULL);
2921 if ((mt->m_pkthdr.PH_vt.vt_nrecs +
2922 m->m_pkthdr.PH_vt.vt_nrecs <=
2923 MLD_V2_REPORT_MAXRECS) &&
2924 (mt->m_pkthdr.len + recslen <=
2925 (inm->in6m_ifp->if_mtu - MLD_MTUSPACE)))
2929 if (!domerge && _IF_QFULL(gq)) {
2931 "%s: outbound queue full, skipping whole packet %p",
2941 CTR2(KTR_MLD, "%s: dequeueing %p", __func__, m);
2942 _IF_DEQUEUE(gq, m0);
2945 CTR2(KTR_MLD, "%s: copying %p", __func__, m);
2946 m0 = m_dup(m, M_NOWAIT);
2949 m0->m_nextpkt = NULL;
2954 CTR3(KTR_MLD, "%s: queueing %p to ifscq %p)",
2955 __func__, m0, ifscq);
2956 _IF_ENQUEUE(ifscq, m0);
2958 struct mbuf *mtl; /* last mbuf of packet mt */
2960 CTR3(KTR_MLD, "%s: merging %p with ifscq tail %p)",
2964 m0->m_flags &= ~M_PKTHDR;
2965 mt->m_pkthdr.len += recslen;
2966 mt->m_pkthdr.PH_vt.vt_nrecs +=
2967 m0->m_pkthdr.PH_vt.vt_nrecs;
2977 * Respond to a pending MLDv2 General Query.
2980 mld_v2_dispatch_general_query(struct mld_ifinfo *mli)
2982 struct ifmultiaddr *ifma;
2984 struct in6_multi *inm;
2987 IN6_MULTI_LOCK_ASSERT();
2990 KASSERT(mli->mli_version == MLD_VERSION_2,
2991 ("%s: called when version %d", __func__, mli->mli_version));
2996 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2997 if (ifma->ifma_addr->sa_family != AF_INET6 ||
2998 ifma->ifma_protospec == NULL)
3001 inm = (struct in6_multi *)ifma->ifma_protospec;
3002 KASSERT(ifp == inm->in6m_ifp,
3003 ("%s: inconsistent ifp", __func__));
3005 switch (inm->in6m_state) {
3006 case MLD_NOT_MEMBER:
3007 case MLD_SILENT_MEMBER:
3009 case MLD_REPORTING_MEMBER:
3010 case MLD_IDLE_MEMBER:
3011 case MLD_LAZY_MEMBER:
3012 case MLD_SLEEPING_MEMBER:
3013 case MLD_AWAKENING_MEMBER:
3014 inm->in6m_state = MLD_REPORTING_MEMBER;
3015 retval = mld_v2_enqueue_group_record(&mli->mli_gq,
3017 CTR2(KTR_MLD, "%s: enqueue record = %d",
3020 case MLD_G_QUERY_PENDING_MEMBER:
3021 case MLD_SG_QUERY_PENDING_MEMBER:
3022 case MLD_LEAVING_MEMBER:
3026 IF_ADDR_RUNLOCK(ifp);
3028 mld_dispatch_queue(&mli->mli_gq, MLD_MAX_RESPONSE_BURST);
3031 * Slew transmission of bursts over 500ms intervals.
3033 if (mli->mli_gq.ifq_head != NULL) {
3034 mli->mli_v2_timer = 1 + MLD_RANDOM_DELAY(
3035 MLD_RESPONSE_BURST_INTERVAL);
3036 V_interface_timers_running6 = 1;
3041 * Transmit the next pending message in the output queue.
3043 * VIMAGE: Needs to store/restore vnet pointer on a per-mbuf-chain basis.
3044 * MRT: Nothing needs to be done, as MLD traffic is always local to
3045 * a link and uses a link-scope multicast address.
3048 mld_dispatch_packet(struct mbuf *m)
3050 struct ip6_moptions im6o;
3055 struct ip6_hdr *ip6;
3056 struct mld_hdr *mld;
3062 CTR2(KTR_MLD, "%s: transmit %p", __func__, m);
3065 * Set VNET image pointer from enqueued mbuf chain
3066 * before doing anything else. Whilst we use interface
3067 * indexes to guard against interface detach, they are
3068 * unique to each VIMAGE and must be retrieved.
3070 ifindex = mld_restore_context(m);
3073 * Check if the ifnet still exists. This limits the scope of
3074 * any race in the absence of a global ifp lock for low cost
3075 * (an array lookup).
3077 ifp = ifnet_byindex(ifindex);
3079 CTR3(KTR_MLD, "%s: dropped %p as ifindex %u went away.",
3080 __func__, m, ifindex);
3082 IP6STAT_INC(ip6s_noroute);
3086 im6o.im6o_multicast_hlim = 1;
3087 im6o.im6o_multicast_loop = (V_ip6_mrouter != NULL);
3088 im6o.im6o_multicast_ifp = ifp;
3090 if (m->m_flags & M_MLDV1) {
3093 m0 = mld_v2_encap_report(ifp, m);
3095 CTR2(KTR_MLD, "%s: dropped %p", __func__, m);
3096 IP6STAT_INC(ip6s_odropped);
3101 mld_scrub_context(m0);
3103 m0->m_pkthdr.rcvif = V_loif;
3105 ip6 = mtod(m0, struct ip6_hdr *);
3107 (void)in6_setscope(&ip6->ip6_dst, ifp, NULL); /* XXX LOR */
3110 * XXX XXX Break some KPI rules to prevent an LOR which would
3111 * occur if we called in6_setscope() at transmission.
3112 * See comments at top of file.
3114 MLD_EMBEDSCOPE(&ip6->ip6_dst, ifp->if_index);
3118 * Retrieve the ICMPv6 type before handoff to ip6_output(),
3119 * so we can bump the stats.
3121 md = m_getptr(m0, sizeof(struct ip6_hdr), &off);
3122 mld = (struct mld_hdr *)(mtod(md, uint8_t *) + off);
3123 type = mld->mld_type;
3125 error = ip6_output(m0, &mld_po, NULL, IPV6_UNSPECSRC, &im6o,
3128 CTR3(KTR_MLD, "%s: ip6_output(%p) = %d", __func__, m0, error);
3131 ICMP6STAT_INC(icp6s_outhist[type]);
3133 icmp6_ifstat_inc(oifp, ifs6_out_msg);
3135 case MLD_LISTENER_REPORT:
3136 case MLDV2_LISTENER_REPORT:
3137 icmp6_ifstat_inc(oifp, ifs6_out_mldreport);
3139 case MLD_LISTENER_DONE:
3140 icmp6_ifstat_inc(oifp, ifs6_out_mlddone);
3149 * Encapsulate an MLDv2 report.
3151 * KAME IPv6 requires that hop-by-hop options be passed separately,
3152 * and that the IPv6 header be prepended in a separate mbuf.
3154 * Returns a pointer to the new mbuf chain head, or NULL if the
3155 * allocation failed.
3157 static struct mbuf *
3158 mld_v2_encap_report(struct ifnet *ifp, struct mbuf *m)
3161 struct mldv2_report *mld;
3162 struct ip6_hdr *ip6;
3163 struct in6_ifaddr *ia;
3166 KASSERT(ifp != NULL, ("%s: null ifp", __func__));
3167 KASSERT((m->m_flags & M_PKTHDR),
3168 ("%s: mbuf chain %p is !M_PKTHDR", __func__, m));
3171 * RFC3590: OK to send as :: or tentative during DAD.
3173 ia = in6ifa_ifpforlinklocal(ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
3175 CTR1(KTR_MLD, "%s: warning: ia is NULL", __func__);
3177 mh = m_gethdr(M_NOWAIT, MT_DATA);
3180 ifa_free(&ia->ia_ifa);
3184 MH_ALIGN(mh, sizeof(struct ip6_hdr) + sizeof(struct mldv2_report));
3186 mldreclen = m_length(m, NULL);
3187 CTR2(KTR_MLD, "%s: mldreclen is %d", __func__, mldreclen);
3189 mh->m_len = sizeof(struct ip6_hdr) + sizeof(struct mldv2_report);
3190 mh->m_pkthdr.len = sizeof(struct ip6_hdr) +
3191 sizeof(struct mldv2_report) + mldreclen;
3193 ip6 = mtod(mh, struct ip6_hdr *);
3195 ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
3196 ip6->ip6_vfc |= IPV6_VERSION;
3197 ip6->ip6_nxt = IPPROTO_ICMPV6;
3198 ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any;
3200 ifa_free(&ia->ia_ifa);
3201 ip6->ip6_dst = in6addr_linklocal_allv2routers;
3202 /* scope ID will be set in netisr */
3204 mld = (struct mldv2_report *)(ip6 + 1);
3205 mld->mld_type = MLDV2_LISTENER_REPORT;
3208 mld->mld_v2_reserved = 0;
3209 mld->mld_v2_numrecs = htons(m->m_pkthdr.PH_vt.vt_nrecs);
3210 m->m_pkthdr.PH_vt.vt_nrecs = 0;
3213 mld->mld_cksum = in6_cksum(mh, IPPROTO_ICMPV6,
3214 sizeof(struct ip6_hdr), sizeof(struct mldv2_report) + mldreclen);
3220 mld_rec_type_to_str(const int type)
3224 case MLD_CHANGE_TO_EXCLUDE_MODE:
3227 case MLD_CHANGE_TO_INCLUDE_MODE:
3230 case MLD_MODE_IS_EXCLUDE:
3233 case MLD_MODE_IS_INCLUDE:
3236 case MLD_ALLOW_NEW_SOURCES:
3239 case MLD_BLOCK_OLD_SOURCES:
3250 mld_init(void *unused __unused)
3253 CTR1(KTR_MLD, "%s: initializing", __func__);
3256 ip6_initpktopts(&mld_po);
3257 mld_po.ip6po_hlim = 1;
3258 mld_po.ip6po_hbh = &mld_ra.hbh;
3259 mld_po.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER;
3260 mld_po.ip6po_flags = IP6PO_DONTFRAG;
3262 SYSINIT(mld_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_init, NULL);
3265 mld_uninit(void *unused __unused)
3268 CTR1(KTR_MLD, "%s: tearing down", __func__);
3271 SYSUNINIT(mld_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, mld_uninit, NULL);
3274 vnet_mld_init(const void *unused __unused)
3277 CTR1(KTR_MLD, "%s: initializing", __func__);
3279 LIST_INIT(&V_mli_head);
3281 VNET_SYSINIT(vnet_mld_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_init,
3285 vnet_mld_uninit(const void *unused __unused)
3288 CTR1(KTR_MLD, "%s: tearing down", __func__);
3290 KASSERT(LIST_EMPTY(&V_mli_head),
3291 ("%s: mli list not empty; ifnets not detached?", __func__));
3293 VNET_SYSUNINIT(vnet_mld_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_mld_uninit,
3297 mld_modevent(module_t mod, int type, void *unused __unused)
3305 return (EOPNOTSUPP);
3310 static moduledata_t mld_mod = {
3315 DECLARE_MODULE(mld, mld_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);