Merge pull request #10775 from opensourcerouting/pim6-mld-pr

pim6d: MLD code

doc/user/pimv6.rst

@@ -218,6 +218,9 @@ vrf is specified then the default vrf is assumed. Finally the special keyword
 'all' allows you to look at all vrfs for the command. Naming a vrf 'all' will
 cause great confusion.
 
+PIM protocol state
+------------------
+
 .. clicmd:: show ipv6 pim [vrf NAME] group-type [json]
 
    Display SSM group ranges.
@@ -289,6 +292,39 @@ cause great confusion.
 
    Display upstream information for S,G's and the RPF data associated with them.
 
+
+MLD state
+---------
+
+.. clicmd:: show ipv6 mld [vrf NAME] interface [IFNAME] [detail|json]
+
+   Display per-interface MLD state, elected querier and related timers.  Use
+   the ``detail`` or ``json`` options for further information (the JSON output
+   always contains all details.)
+
+.. clicmd:: show ipv6 mld [vrf NAME] statistics [interface IFNAME] [json]
+
+   Display packet and error counters for MLD interfaces.  All counters are
+   packet counters (not bytes) and wrap at 64 bit.  In some rare cases,
+   malformed received MLD reports may be partially processed and counted on
+   multiple counters.
+
+.. clicmd:: show ipv6 mld [vrf NAME] joins [{interface IFNAME|groups X:X::X:X/M|sources X:X::X:X/M|detail}] [json]
+
+   Display joined groups tracked by MLD.  ``interface``, ``groups`` and
+   ``sources`` options may be used to limit output to a subset (note ``sources``
+   refers to the multicast traffic sender, not the host that joined to receive
+   the traffic.)
+
+   The ``detail`` option also reports which hosts have joined (subscribed) to
+   particular ``S,G``.  This information is only available for MLDv2 hosts with
+   a MLDv2 querier.  MLDv1 joins are recorded as "untracked" and shown in the
+   ``NonTrkSeen`` output column.
+
+
+General multicast routing state
+-------------------------------
+
 .. clicmd:: show ipv6 multicast
 
    Display various information about the interfaces used in this pim instance.

pimd/pim6_cmd.c

@@ -707,7 +707,7 @@ DEFPY (interface_ipv6_mld_query_max_response_time,
        IPV6_STR
        IFACE_MLD_STR
        IFACE_MLD_QUERY_MAX_RESPONSE_TIME_STR
-       "Query response value in deci-seconds\n")
+       "Query response value in milliseconds\n")
 {
 	return gm_process_query_max_response_time_cmd(vty, qmrt_str);
 }

pimd/pim6_main.c

@@ -38,6 +38,7 @@
 #include "pim_zebra.h"
 #include "pim_nb.h"
 #include "pim6_cmd.h"
+#include "pim6_mld.h"
 
 zebra_capabilities_t _caps_p[] = {
 	ZCAP_SYS_ADMIN,
@@ -133,7 +134,6 @@ FRR_DAEMON_INFO(pim6d, PIM6,
 );
 /* clang-format on */
 
-
 int main(int argc, char **argv, char **envp)
 {
 	static struct option longopts[] = {
@@ -184,6 +184,8 @@ int main(int argc, char **argv, char **envp)
 	 */
 	pim_iface_init();
 
+	gm_cli_init();
+
 	pim_zebra_init();
 #if 0
 	pim_bfd_init();

pimd/pim6_mld.h

@@ -0,0 +1,367 @@
+/*
+ * PIMv6 MLD querier
+ * Copyright (C) 2021-2022  David Lamparter for NetDEF, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef PIM6_MLD_H
+#define PIM6_MLD_H
+
+#include "typesafe.h"
+#include "pim_addr.h"
+
+struct thread;
+struct pim_instance;
+struct gm_packet_sg;
+struct gm_if;
+struct channel_oil;
+
+#define MLD_DEFAULT_VERSION 2
+
+/* see comment below on subs_negative/subs_positive */
+enum gm_sub_sense {
+	/* negative/pruning: S,G in EXCLUDE */
+	GM_SUB_NEG = 0,
+	/* positive/joining: *,G in EXCLUDE and S,G in INCLUDE */
+	GM_SUB_POS = 1,
+};
+
+enum gm_sg_state {
+	GM_SG_NOINFO = 0,
+	GM_SG_JOIN,
+	GM_SG_JOIN_EXPIRING,
+	/* remaining 3 only valid for S,G when *,G in EXCLUDE */
+	GM_SG_PRUNE,
+	GM_SG_NOPRUNE,
+	GM_SG_NOPRUNE_EXPIRING,
+};
+
+static inline bool gm_sg_state_want_join(enum gm_sg_state state)
+{
+	return state != GM_SG_NOINFO && state != GM_SG_PRUNE;
+}
+
+/* MLD (S,G) state (on an interface)
+ *
+ * group is always != ::, src is :: for (*,G) joins.  sort order in RB tree is
+ * such that sources for a particular group can be iterated by starting at the
+ * group.  For INCLUDE, no (*,G) entry exists, only (S,G).
+ */
+
+PREDECL_RBTREE_UNIQ(gm_packet_sg_subs);
+PREDECL_RBTREE_UNIQ(gm_sgs);
+struct gm_sg {
+	pim_sgaddr sgaddr;
+	struct gm_if *iface;
+	struct gm_sgs_item itm;
+
+	enum gm_sg_state state;
+	struct channel_oil *oil;
+	bool tib_joined;
+
+	struct timeval created;
+
+	/* if a group- or group-and-source specific query is running
+	 * (implies we haven't received any report yet, since it's cancelled
+	 * by that)
+	 */
+	struct thread *t_sg_expire;
+
+	/* last-member-left triggered queries (group/group-source specific)
+	 *
+	 * this timer will be running even if we aren't the elected querier,
+	 * in case the election result changes midway through.
+	 */
+	struct thread *t_sg_query;
+
+	/* we must keep sending (QRV) queries even if we get a positive
+	 * response, to make sure other routers are updated.  query_sbit
+	 * will be set in that case, since other routers need the *response*,
+	 * not the *query*
+	 */
+	uint8_t n_query;
+	bool query_sbit;
+
+	/* subs_positive tracks gm_packet_sg resulting in a JOIN, i.e. for
+	 * (*,G) it has *EXCLUDE* items, for (S,G) it has *INCLUDE* items.
+	 *
+	 * subs_negative is always empty for (*,G) and tracks EXCLUDE items
+	 * for (S,G).  This means that an (S,G) entry is active as a PRUNE if
+	 *   len(src->subs_negative) == len(grp->subs_positive)
+	 *   && len(src->subs_positive) == 0
+	 * (i.e. all receivers for the group opted to exclude this S,G and
+	 * noone did an SSM join for the S,G)
+	 */
+	union {
+		struct {
+			struct gm_packet_sg_subs_head subs_negative[1];
+			struct gm_packet_sg_subs_head subs_positive[1];
+		};
+		struct gm_packet_sg_subs_head subs[2];
+	};
+
+	/* If the elected querier is not ourselves, queries and reports might
+	 * get reordered in rare circumstances, i.e. the report could arrive
+	 * just a microsecond before the query kicks off the timer.  This can
+	 * then result in us thinking there are no more receivers since no
+	 * report might be received during the query period.
+	 *
+	 * To avoid this, keep track of the most recent report for this (S,G)
+	 * so we can do a quick check to add just a little bit of slack.
+	 *
+	 * EXCLUDE S,Gs are never in most_recent.
+	 */
+	struct gm_packet_sg *most_recent;
+};
+
+/* host tracking entry.  addr will be one of:
+ *
+ * ::		- used by hosts during address acquisition
+ * ::1		- may show up on some OS for joins by the router itself
+ * link-local	- regular operation by MLDv2 hosts
+ * ffff:..:ffff	- MLDv1 entry (cannot be tracked due to report suppression)
+ *
+ * global scope IPv6 addresses can never show up here
+ */
+PREDECL_HASH(gm_subscribers);
+PREDECL_DLIST(gm_packets);
+struct gm_subscriber {
+	pim_addr addr;
+	struct gm_subscribers_item itm;
+
+	struct gm_if *iface;
+	size_t refcount;
+
+	struct gm_packets_head packets[1];
+
+	struct timeval created;
+};
+
+/*
+ * MLD join state is kept batched by packet.  Since the timers for all items
+ * in a packet are the same, this reduces the number of timers we're keeping
+ * track of.  It also eases tracking for EXCLUDE state groups because the
+ * excluded sources are in the same packet.  (MLD does not support splitting
+ * that if it exceeds MTU, it's always a full replace for exclude.)
+ *
+ * Since packets may be partially superseded by newer packets, the "active"
+ * field is used to track this.
+ */
+
+/* gm_packet_sg is allocated as part of gm_packet_state, note the items[0]
+ * array at the end of that.  gm_packet_sg is NEVER directly allocated with
+ * XMALLOC/XFREE.
+ */
+struct gm_packet_sg {
+	/* non-NULL as long as this gm_packet_sg is the most recent entry
+	 * for (subscriber,S,G).  Cleared to NULL when a newer packet by the
+	 * subscriber replaces this item.
+	 *
+	 * (Old items are kept around so we don't need to realloc/resize
+	 * gm_packet_state, which would mess up a whole lot of pointers)
+	 */
+	struct gm_sg *sg;
+
+	/* gm_sg -> (subscriber, gm_packet_sg)
+	 * only on RB-tree while sg != NULL, i.e. not superseded by newer.
+	 */
+	struct gm_packet_sg_subs_item subs_itm;
+
+	bool is_src : 1; /* := (src != ::) */
+	bool is_excl : 1;
+
+	/* for getting back to struct gm_packet_state, cf.
+	 * gm_packet_sg2state() below
+	 */
+	uint16_t offset;
+
+	/* if this is a group entry in EXCLUDE state, n_exclude counts how
+	 * many sources are on the exclude list here.  They follow immediately
+	 * after.
+	 */
+	uint16_t n_exclude;
+};
+
+#define gm_packet_sg2state(sg)                                                 \
+	container_of(sg, struct gm_packet_state, items[sg->offset])
+
+PREDECL_DLIST(gm_packet_expires);
+struct gm_packet_state {
+	struct gm_if *iface;
+	struct gm_subscriber *subscriber;
+	struct gm_packets_item pkt_itm;
+
+	struct timeval received;
+	struct gm_packet_expires_item exp_itm;
+
+	/* n_active starts equal to n_sg;  whenever active is set to false on
+	 * an item it is decremented.  When n_active == 0, the packet can be
+	 * freed.
+	 */
+	uint16_t n_sg, n_active;
+	struct gm_packet_sg items[0];
+};
+
+/* general queries are rather different from group/S,G specific queries;  it's
+ * not particularly efficient or useful to try to shoehorn them into the S,G
+ * timers.  Instead, we keep a history of recent queries and their implied
+ * expiries.
+ */
+struct gm_general_pending {
+	struct timeval query, expiry;
+};
+
+/* similarly, group queries also age out S,G entries for the group, but in
+ * this case we only keep one query for each group
+ *
+ * why is this not in the *,G gm_sg?  There may not be one (for INCLUDE mode
+ * groups, or groups we don't know about.)  Also, malicious clients could spam
+ * random group-specific queries to trigger resource exhaustion, so it makes
+ * sense to limit these.
+ */
+PREDECL_RBTREE_UNIQ(gm_grp_pends);
+struct gm_grp_pending {
+	struct gm_grp_pends_item itm;
+	struct gm_if *iface;
+	pim_addr grp;
+
+	struct timeval query;
+	struct thread *t_expire;
+};
+
+/* guaranteed MTU for IPv6 is 1280 bytes.  IPv6 header is 40 bytes, MLDv2
+ * query header is 24 bytes, RA option is 8 bytes - leaves 1208 bytes for the
+ * source list, which is 151 IPv6 addresses.  But we may have some more IPv6
+ * extension headers (e.g. IPsec AH), so just cap to 128
+ */
+#define MLD_V2Q_MTU_MAX_SOURCES 128
+
+/* group-and-source-specific queries are bundled together, if some host joins
+ * multiple sources it's likely to drop all at the same time.
+ *
+ * Unlike gm_grp_pending, this is only used for aggregation since the S,G
+ * state is kept directly in the gm_sg structure.
+ */
+PREDECL_HASH(gm_gsq_pends);
+struct gm_gsq_pending {
+	struct gm_gsq_pends_item itm;
+
+	struct gm_if *iface;
+	struct thread *t_send;
+
+	pim_addr grp;
+	bool s_bit;
+
+	size_t n_src;
+	pim_addr srcs[MLD_V2Q_MTU_MAX_SOURCES];
+};
+
+
+/* The size of this history is limited by QRV, i.e. there can't be more than
+ * 8 items here.
+ */
+#define GM_MAX_PENDING 8
+
+enum gm_version {
+	GM_NONE,
+	GM_MLDV1,
+	GM_MLDV2,
+};
+
+struct gm_if_stats {
+	uint64_t rx_drop_csum;
+	uint64_t rx_drop_srcaddr;
+	uint64_t rx_drop_dstaddr;
+	uint64_t rx_drop_ra;
+	uint64_t rx_drop_malformed;
+	uint64_t rx_trunc_report;
+
+	/* since the types are different, this is rx_old_* not of rx_*_old */
+	uint64_t rx_old_report;
+	uint64_t rx_old_leave;
+	uint64_t rx_new_report;
+
+	uint64_t rx_query_new_general;
+	uint64_t rx_query_new_group;
+	uint64_t rx_query_new_groupsrc;
+	uint64_t rx_query_new_sbit;
+	uint64_t rx_query_old_general;
+	uint64_t rx_query_old_group;
+
+	uint64_t tx_query_new_general;
+	uint64_t tx_query_new_group;
+	uint64_t tx_query_new_groupsrc;
+	uint64_t tx_query_old_general;
+	uint64_t tx_query_old_group;
+
+	uint64_t tx_query_fail;
+};
+
+struct gm_if {
+	struct interface *ifp;
+	struct pim_instance *pim;
+	struct thread *t_query, *t_other_querier, *t_expire;
+
+	bool stopping;
+
+	uint8_t n_startup;
+
+	uint8_t cur_qrv;
+	unsigned int cur_query_intv;	  /* ms */
+	unsigned int cur_query_intv_trig; /* ms */
+	unsigned int cur_max_resp;	  /* ms */
+	enum gm_version cur_version;
+
+	/* this value (positive, default 10ms) defines our "timing tolerance":
+	 * - added to deadlines for expiring joins
+	 * - used to look backwards in time for queries, in case a report was
+	 *   reordered before the query
+	 */
+	struct timeval cfg_timing_fuzz;
+
+	/* items in pending[] are sorted by expiry, pending[0] is earliest */
+	struct gm_general_pending pending[GM_MAX_PENDING];
+	uint8_t n_pending;
+	struct gm_grp_pends_head grp_pends[1];
+	struct gm_gsq_pends_head gsq_pends[1];
+
+	pim_addr querier;
+	pim_addr cur_ll_lowest;
+
+	struct gm_sgs_head sgs[1];
+	struct gm_subscribers_head subscribers[1];
+	struct gm_packet_expires_head expires[1];
+
+	struct timeval started;
+	struct gm_if_stats stats;
+};
+
+#if PIM_IPV == 6
+extern void gm_ifp_update(struct interface *ifp);
+extern void gm_ifp_teardown(struct interface *ifp);
+#else
+static inline void gm_ifp_update(struct interface *ifp)
+{
+}
+
+static inline void gm_ifp_teardown(struct interface *ifp)
+{
+}
+#endif
+
+extern void gm_cli_init(void);
+
+#endif /* PIM6_MLD_H */

pimd/pim6_mld_protocol.h

@@ -0,0 +1,125 @@
+/*
+ * MLD protocol definitions
+ * Copyright (C) 2022  David Lamparter for NetDEF, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef _PIM6_MLD_PROTOCOL_H
+#define _PIM6_MLD_PROTOCOL_H
+
+#include <stdalign.h>
+#include <stdint.h>
+
+/* There is a struct icmp6_hdr provided by OS, but it includes 4 bytes of data.
+ * Not helpful for us if we want to put the MLD struct after it.
+ */
+
+struct icmp6_plain_hdr {
+	uint8_t icmp6_type;
+	uint8_t icmp6_code;
+	uint16_t icmp6_cksum;
+};
+static_assert(sizeof(struct icmp6_plain_hdr) == 4, "struct mismatch");
+static_assert(alignof(struct icmp6_plain_hdr) <= 4, "struct mismatch");
+
+/* for MLDv1 query, report and leave all use the same packet format */
+struct mld_v1_pkt {
+	uint16_t max_resp_code;
+	uint16_t rsvd0;
+	struct in6_addr grp;
+};
+static_assert(sizeof(struct mld_v1_pkt) == 20, "struct mismatch");
+static_assert(alignof(struct mld_v1_pkt) <= 4, "struct mismatch");
+
+
+struct mld_v2_query_hdr {
+	uint16_t max_resp_code;
+	uint16_t rsvd0;
+	struct in6_addr grp;
+	uint8_t flags;
+	uint8_t qqic;
+	uint16_t n_src;
+	struct in6_addr srcs[0];
+};
+static_assert(sizeof(struct mld_v2_query_hdr) == 24, "struct mismatch");
+static_assert(alignof(struct mld_v2_query_hdr) <= 4, "struct mismatch");
+
+
+struct mld_v2_report_hdr {
+	uint16_t rsvd;
+	uint16_t n_records;
+};
+static_assert(sizeof(struct mld_v2_report_hdr) == 4, "struct mismatch");
+static_assert(alignof(struct mld_v2_report_hdr) <= 4, "struct mismatch");
+
+
+struct mld_v2_rec_hdr {
+	uint8_t type;
+	uint8_t aux_len;
+	uint16_t n_src;
+	struct in6_addr grp;
+	struct in6_addr srcs[0];
+};
+static_assert(sizeof(struct mld_v2_rec_hdr) == 20, "struct mismatch");
+static_assert(alignof(struct mld_v2_rec_hdr) <= 4, "struct mismatch");
+
+/* clang-format off */
+enum icmp6_mld_type {
+	ICMP6_MLD_QUERY			= 130,
+	ICMP6_MLD_V1_REPORT		= 131,
+	ICMP6_MLD_V1_DONE		= 132,
+	ICMP6_MLD_V2_REPORT		= 143,
+};
+
+enum mld_v2_rec_type {
+	MLD_RECTYPE_IS_INCLUDE		= 1,
+	MLD_RECTYPE_IS_EXCLUDE		= 2,
+	MLD_RECTYPE_CHANGE_TO_INCLUDE	= 3,
+	MLD_RECTYPE_CHANGE_TO_EXCLUDE	= 4,
+	MLD_RECTYPE_ALLOW_NEW_SOURCES	= 5,
+	MLD_RECTYPE_BLOCK_OLD_SOURCES	= 6,
+};
+/* clang-format on */
+
+/* helper functions */
+
+static inline unsigned int mld_max_resp_decode(uint16_t wire)
+{
+	uint16_t code = ntohs(wire);
+	uint8_t exp;
+
+	if (code < 0x8000)
+		return code;
+	exp = (code >> 12) & 0x7;
+	return ((code & 0xfff) | 0x1000) << (exp + 3);
+}
+
+static inline uint16_t mld_max_resp_encode(uint32_t value)
+{
+	uint16_t code;
+	uint8_t exp;
+
+	if (value < 0x8000)
+		code = value;
+	else {
+		exp = 16 - __builtin_clz(value);
+		code = (value >> (exp + 3)) & 0xfff;
+		code |= 0x8000 | (exp << 12);
+	}
+	return htons(code);
+}
+
+#endif /* _PIM6_MLD_PROTOCOL_H */

pimd/pim_iface.c

@@ -50,6 +50,8 @@
 #include "pim_igmp_join.h"
 #include "pim_vxlan.h"
 
+#include "pim6_mld.h"
+
 #if PIM_IPV == 4
 static void pim_if_igmp_join_del_all(struct interface *ifp);
 static int igmp_join_sock(const char *ifname, ifindex_t ifindex,
@@ -127,6 +129,7 @@ struct pim_interface *pim_if_new(struct interface *ifp, bool igmp, bool pim,
 	pim_ifp->mroute_vif_index = -1;
 
 	pim_ifp->igmp_version = IGMP_DEFAULT_VERSION;
+	pim_ifp->mld_version = MLD_DEFAULT_VERSION;
 	pim_ifp->gm_default_robustness_variable =
 		IGMP_DEFAULT_ROBUSTNESS_VARIABLE;
 	pim_ifp->gm_default_query_interval = IGMP_GENERAL_QUERY_INTERVAL;
@@ -651,6 +654,7 @@ void pim_if_addr_add(struct connected *ifc)
 		vxlan_term = pim_vxlan_is_term_dev_cfg(pim_ifp->pim, ifp);
 		pim_if_add_vif(ifp, false, vxlan_term);
 	}
+	gm_ifp_update(ifp);
 	pim_ifchannel_scan_forward_start(ifp);
 }
 
@@ -763,6 +767,8 @@ void pim_if_addr_del(struct connected *ifc, int force_prim_as_any)
 				"%s: removed link-local %pI6, lowest now %pI6, highest %pI6",
 				ifc->ifp->name, &ifc->address->u.prefix6,
 				&pim_ifp->ll_lowest, &pim_ifp->ll_highest);
+
+		gm_ifp_update(ifp);
 	}
 #endif
 
@@ -822,6 +828,7 @@ void pim_if_addr_add_all(struct interface *ifp)
 		vxlan_term = pim_vxlan_is_term_dev_cfg(pim_ifp->pim, ifp);
 		pim_if_add_vif(ifp, false, vxlan_term);
 	}
+	gm_ifp_update(ifp);
 	pim_ifchannel_scan_forward_start(ifp);
 
 	pim_rp_setup(pim_ifp->pim);
@@ -1000,12 +1007,15 @@ int pim_if_add_vif(struct interface *ifp, bool ispimreg, bool is_vxlan_term)
 	}
 
 	ifaddr = pim_ifp->primary_address;
+#if PIM_IPV != 6
+	/* IPv6 API is always by interface index */
 	if (!ispimreg && !is_vxlan_term && pim_addr_is_any(ifaddr)) {
 		zlog_warn(
 			"%s: could not get address for interface %s ifindex=%d",
 			__func__, ifp->name, ifp->ifindex);
 		return -4;
 	}
+#endif
 
 	pim_ifp->mroute_vif_index = pim_iface_next_vif_index(ifp);
 
@@ -1030,9 +1040,10 @@ int pim_if_add_vif(struct interface *ifp, bool ispimreg, bool is_vxlan_term)
 
 	pim_ifp->pim->iface_vif_index[pim_ifp->mroute_vif_index] = 1;
 
+	gm_ifp_update(ifp);
+
 	/* if the device qualifies as pim_vxlan iif/oif update vxlan entries */
 	pim_vxlan_add_vif(ifp);
-
 	return 0;
 }
 
@@ -1050,6 +1061,8 @@ int pim_if_del_vif(struct interface *ifp)
 	/* if the device was a pim_vxlan iif/oif update vxlan mroute entries */
 	pim_vxlan_del_vif(ifp);
 
+	gm_ifp_teardown(ifp);
+
 	pim_mroute_del_vif(ifp);
 
 	/*
@@ -1058,7 +1071,6 @@ int pim_if_del_vif(struct interface *ifp)
 	pim_ifp->pim->iface_vif_index[pim_ifp->mroute_vif_index] = 0;
 
 	pim_ifp->mroute_vif_index = -1;
-
 	return 0;
 }
 

pimd/pim_iface.h

@@ -68,6 +68,8 @@ struct pim_secondary_addr {
 	enum pim_secondary_addr_flags flags;
 };
 
+struct gm_if;
+
 struct pim_interface {
 	bool pim_enable : 1;
 	bool pim_can_disable_join_suppression : 1;
@@ -90,6 +92,7 @@ struct pim_interface {
 					 * address of the interface */
 
 	int igmp_version;		     /* IGMP version */
+	int mld_version;
 	int gm_default_robustness_variable;  /* IGMP or MLD QRV */
 	int gm_default_query_interval;       /* IGMP or MLD secs between general
 						  queries */
@@ -111,6 +114,8 @@ struct pim_interface {
 	struct list *gm_group_list;  /* list of struct IGMP or MLD group */
 	struct hash *gm_group_hash;
 
+	struct gm_if *mld;
+
 	int pim_sock_fd;		/* PIM socket file descriptor */
 	struct thread *t_pim_sock_read; /* thread for reading PIM socket */
 	int64_t pim_sock_creation;      /* timestamp of PIM socket creation */

pimd/pim_instance.c

@@ -115,6 +115,8 @@ static struct pim_instance *pim_instance_init(struct vrf *vrf)
 
 	pim->send_v6_secondary = 1;
 
+	pim->gm_socket = -1;
+
 	pim_rp_init(pim);
 
 	pim_bsm_proc_init(pim);

pimd/pim_instance.h

@@ -167,6 +167,10 @@ struct pim_instance {
 	struct list *ssmpingd_list;
 	pim_addr ssmpingd_group_addr;
 
+	unsigned int gm_socket_if_count;
+	int gm_socket;
+	struct thread *t_gm_recv;
+
 	unsigned int igmp_group_count;
 	unsigned int igmp_watermark_limit;
 	unsigned int keep_alive_time;
@@ -194,6 +198,8 @@ struct pim_instance {
 	int64_t nexthop_lookups;
 	int64_t nexthop_lookups_avoided;
 	int64_t last_route_change_time;
+
+	uint64_t gm_rx_drop_sys;
 };
 
 void pim_vrf_init(void);

pimd/pim_nb_config.c

@@ -35,6 +35,7 @@
 #include "log.h"
 #include "lib_errors.h"
 #include "pim_util.h"
+#include "pim6_mld.h"
 
 #if PIM_IPV == 6
 #define pim6_msdp_err(funcname, argtype)                                       \
@@ -2702,12 +2703,22 @@ int lib_interface_gmp_address_family_igmp_version_destroy(
 int lib_interface_gmp_address_family_mld_version_modify(
 	struct nb_cb_modify_args *args)
 {
+	struct interface *ifp;
+	struct pim_interface *pim_ifp;
+
 	switch (args->event) {
 	case NB_EV_VALIDATE:
 	case NB_EV_PREPARE:
 	case NB_EV_ABORT:
+		break;
 	case NB_EV_APPLY:
-		/* TBD depends on MLD data structure changes */
+		ifp = nb_running_get_entry(args->dnode, NULL, true);
+		pim_ifp = ifp->info;
+		if (!pim_ifp)
+			return NB_ERR_INCONSISTENCY;
+
+		pim_ifp->mld_version = yang_dnode_get_uint8(args->dnode, NULL);
+		gm_ifp_update(ifp);
 		break;
 	}
 
@@ -2717,11 +2728,22 @@ int lib_interface_gmp_address_family_mld_version_modify(
 int lib_interface_gmp_address_family_mld_version_destroy(
 	struct nb_cb_destroy_args *args)
 {
+	struct interface *ifp;
+	struct pim_interface *pim_ifp;
+
 	switch (args->event) {
 	case NB_EV_VALIDATE:
 	case NB_EV_PREPARE:
 	case NB_EV_ABORT:
+		break;
 	case NB_EV_APPLY:
+		ifp = nb_running_get_entry(args->dnode, NULL, true);
+		pim_ifp = ifp->info;
+		if (!pim_ifp)
+			return NB_ERR_INCONSISTENCY;
+
+		pim_ifp->mld_version = 2;
+		gm_ifp_update(ifp);
 		break;
 	}
 
@@ -2734,10 +2756,10 @@ int lib_interface_gmp_address_family_mld_version_destroy(
 int lib_interface_gmp_address_family_query_interval_modify(
 	struct nb_cb_modify_args *args)
 {
-#if PIM_IPV == 4
 	struct interface *ifp;
 	int query_interval;
 
+#if PIM_IPV == 4
 	switch (args->event) {
 	case NB_EV_VALIDATE:
 	case NB_EV_PREPARE:
@@ -2749,7 +2771,23 @@ int lib_interface_gmp_address_family_query_interval_modify(
 		change_query_interval(ifp->info, query_interval);
 	}
 #else
-	/* TBD Depends on MLD data structure changes */
+	struct pim_interface *pim_ifp;
+
+	switch (args->event) {
+	case NB_EV_VALIDATE:
+	case NB_EV_PREPARE:
+	case NB_EV_ABORT:
+		break;
+	case NB_EV_APPLY:
+		ifp = nb_running_get_entry(args->dnode, NULL, true);
+		pim_ifp = ifp->info;
+		if (!pim_ifp)
+			return NB_ERR_INCONSISTENCY;
+
+		query_interval = yang_dnode_get_uint16(args->dnode, NULL);
+		pim_ifp->gm_default_query_interval = query_interval;
+		gm_ifp_update(ifp);
+	}
 #endif
 	return NB_OK;
 }

pimd/pim_vty.c

@@ -40,6 +40,7 @@
 #include "pim_bfd.h"
 #include "pim_bsm.h"
 #include "pim_vxlan.h"
+#include "pim6_mld.h"
 
 int pim_debug_config_write(struct vty *vty)
 {
@@ -291,7 +292,7 @@ int pim_global_config_write_worker(struct pim_instance *pim, struct vty *vty)
 }
 
 #if PIM_IPV == 4
-static int pim_igmp_config_write(struct vty *vty, int writes,
+static int gm_config_write(struct vty *vty, int writes,
 			   struct pim_interface *pim_ifp)
 {
 	/* IF ip igmp */
@@ -360,6 +361,17 @@ static int pim_igmp_config_write(struct vty *vty, int writes,
 
 	return writes;
 }
+#else
+static int gm_config_write(struct vty *vty, int writes,
+			   struct pim_interface *pim_ifp)
+{
+	if (pim_ifp->mld_version != MLD_DEFAULT_VERSION)
+		vty_out(vty, " ipv6 mld version %d\n", pim_ifp->mld_version);
+	if (pim_ifp->gm_default_query_interval != IGMP_GENERAL_QUERY_INTERVAL)
+		vty_out(vty, " ipv6 mld query-interval %d\n",
+			pim_ifp->gm_default_query_interval);
+	return 0;
+}
 #endif
 
 int pim_config_write(struct vty *vty, int writes, struct interface *ifp,
@@ -388,9 +400,7 @@ int pim_config_write(struct vty *vty, int writes, struct interface *ifp,
 		++writes;
 	}
 
-#if PIM_IPV == 4
+	writes += gm_config_write(vty, writes, pim_ifp);
-	writes += pim_igmp_config_write(vty, writes, pim_ifp);
-#endif
 
 	/* update source */
 	if (!pim_addr_is_any(pim_ifp->update_source)) {

pimd/subdir.am

@@ -9,6 +9,7 @@ noinst_PROGRAMS += pimd/test_igmpv3_join
 vtysh_scan += \
 	pimd/pim_cmd.c \
 	pimd/pim6_cmd.c \
+	pimd/pim6_mld.c \
 	#end
 vtysh_daemons += pimd
 vtysh_daemons += pim6d
@@ -89,6 +90,7 @@ nodist_pimd_pimd_SOURCES = \
 pimd_pim6d_SOURCES = \
 	$(pim_common) \
 	pimd/pim6_main.c \
+	pimd/pim6_mld.c \
 	pimd/pim6_stubs.c \
 	pimd/pim6_cmd.c \
 	pimd/pim6_mroute_msg.c \
@@ -155,6 +157,8 @@ noinst_HEADERS += \
 	pimd/pim_vxlan.h \
 	pimd/pim_vxlan_instance.h \
 	pimd/pimd.h \
+	pimd/pim6_mld.h \
+	pimd/pim6_mld_protocol.h \
 	pimd/mtracebis_netlink.h \
 	pimd/mtracebis_routeget.h \
 	pimd/pim6_cmd.h \
@@ -163,6 +167,7 @@ noinst_HEADERS += \
 clippy_scan += \
 	pimd/pim_cmd.c \
 	pimd/pim6_cmd.c \
+	pimd/pim6_mld.c \
 	# end
 
 pimd_pimd_CFLAGS = $(AM_CFLAGS) -DPIM_IPV=4

yang/frr-gmp.yang

@@ -96,6 +96,7 @@ module frr-gmp {
       type uint8 {
         range "1..2";
       }
+      default "2";
       description
         "MLD version.";
     }