Merge branch 'bridgge-mcast'

Nikolay Aleksandrov says:

====================
net: bridge: mcast: dump querier state

This set adds the ability to dump the current multicast querier state.
This is extremely useful when debugging multicast issues, we've had
many cases of unexpected queriers causing strange behaviour and mcast
test failures. The first patch changes the querier struct to record
a port device's ifindex instead of a pointer to the port itself so we
can later retrieve it, I chose this way because it's much simpler
and doesn't require us to do querier port ref counting, it is best
effort anyway. Then patch 02 makes the querier address/port updates
consistent via a combination of multicast_lock and seqcount, so readers
can only use seqcount to get a consistent snapshot of address and port.
Patch 03 is a minor cleanup in preparation for the dump support, it
consolidates IPv4 and IPv6 querier selection paths as they share most of
the logic (except address comparisons of course). Finally the last three
patches add the new querier state dumping support, for the bridge's
global multicast context we embed the BRIDGE_QUERIER_xxx attributes
into IFLA_BR_MCAST_QUERIER_STATE and for the per-vlan global mcast
contexts we embed them into BRIDGE_VLANDB_GOPTS_MCAST_QUERIER_STATE.

The structure is:
  [IFLA_BR_MCAST_QUERIER_STATE / BRIDGE_VLANDB_GOPTS_MCAST_QUERIER_STATE]
  `[BRIDGE_QUERIER_IP_ADDRESS] - ip address of the querier
  `[BRIDGE_QUERIER_IP_PORT]    - bridge port ifindex where the querier was
                                 seen (set only if external querier)
  `[BRIDGE_QUERIER_IP_OTHER_TIMER]   -  other querier timeout
  `[BRIDGE_QUERIER_IPV6_ADDRESS] - ip address of the querier
  `[BRIDGE_QUERIER_IPV6_PORT]    - bridge port ifindex where the querier
                                   was seen (set only if external querier)
  `[BRIDGE_QUERIER_IPV6_OTHER_TIMER]   -  other querier timeout

Later we can also add IGMP version of seen queriers and last seen values
from the queries.
====================

6 files changed, 199 insertions, 44 deletions

diff --git a/include/uapi/linux/if_bridge.h b/include/uapi/linux/if_bridge.h
index 620d86e825b8..f71a81fdbbc6 100644
--- a/include/uapi/linux/if_bridge.h
+++ b/include/uapi/linux/if_bridge.h
@@ -563,6 +563,7 @@ enum {
 	BRIDGE_VLANDB_GOPTS_MCAST_QUERIER,
 	BRIDGE_VLANDB_GOPTS_MCAST_ROUTER,
 	BRIDGE_VLANDB_GOPTS_MCAST_ROUTER_PORTS,
+	BRIDGE_VLANDB_GOPTS_MCAST_QUERIER_STATE,
 	__BRIDGE_VLANDB_GOPTS_MAX
 };
 #define BRIDGE_VLANDB_GOPTS_MAX (__BRIDGE_VLANDB_GOPTS_MAX - 1)
@@ -770,4 +771,17 @@ struct br_boolopt_multi {
 	__u32 optval;
 	__u32 optmask;
 };
+
+enum {
+	BRIDGE_QUERIER_UNSPEC,
+	BRIDGE_QUERIER_IP_ADDRESS,
+	BRIDGE_QUERIER_IP_PORT,
+	BRIDGE_QUERIER_IP_OTHER_TIMER,
+	BRIDGE_QUERIER_PAD,
+	BRIDGE_QUERIER_IPV6_ADDRESS,
+	BRIDGE_QUERIER_IPV6_PORT,
+	BRIDGE_QUERIER_IPV6_OTHER_TIMER,
+	__BRIDGE_QUERIER_MAX
+};
+#define BRIDGE_QUERIER_MAX (__BRIDGE_QUERIER_MAX - 1)
 #endif /* _UAPI_LINUX_IF_BRIDGE_H */
diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index 5310003523ce..8aad65b69054 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h
@@ -479,6 +479,7 @@ enum {
 	IFLA_BR_MCAST_MLD_VERSION,
 	IFLA_BR_VLAN_STATS_PER_PORT,
 	IFLA_BR_MULTI_BOOLOPT,
+	IFLA_BR_MCAST_QUERIER_STATE,
 	__IFLA_BR_MAX,
 };
 
diff --git a/net/bridge/br_multicast.c b/net/bridge/br_multicast.c
index df6bf6a237aa..0e5d6ba03457 100644
--- a/net/bridge/br_multicast.c
+++ b/net/bridge/br_multicast.c
@@ -1659,11 +1659,38 @@ again_under_lmqt:
 	}
 }
 
+static void br_multicast_read_querier(const struct bridge_mcast_querier *querier,
+				      struct bridge_mcast_querier *dest)
+{
+	unsigned int seq;
+
+	memset(dest, 0, sizeof(*dest));
+	do {
+		seq = read_seqcount_begin(&querier->seq);
+		dest->port_ifidx = querier->port_ifidx;
+		memcpy(&dest->addr, &querier->addr, sizeof(struct br_ip));
+	} while (read_seqcount_retry(&querier->seq, seq));
+}
+
+static void br_multicast_update_querier(struct net_bridge_mcast *brmctx,
+					struct bridge_mcast_querier *querier,
+					int ifindex,
+					struct br_ip *saddr)
+{
+	lockdep_assert_held_once(&brmctx->br->multicast_lock);
+
+	write_seqcount_begin(&querier->seq);
+	querier->port_ifidx = ifindex;
+	memcpy(&querier->addr, saddr, sizeof(*saddr));
+	write_seqcount_end(&querier->seq);
+}
+
 static void br_multicast_send_query(struct net_bridge_mcast *brmctx,
 				    struct net_bridge_mcast_port *pmctx,
 				    struct bridge_mcast_own_query *own_query)
 {
 	struct bridge_mcast_other_query *other_query = NULL;
+	struct bridge_mcast_querier *querier;
 	struct br_ip br_group;
 	unsigned long time;
 
@@ -1676,10 +1703,12 @@ static void br_multicast_send_query(struct net_bridge_mcast *brmctx,
 
 	if (pmctx ? (own_query == &pmctx->ip4_own_query) :
 		    (own_query == &brmctx->ip4_own_query)) {
+		querier = &brmctx->ip4_querier;
 		other_query = &brmctx->ip4_other_query;
 		br_group.proto = htons(ETH_P_IP);
 #if IS_ENABLED(CONFIG_IPV6)
 	} else {
+		querier = &brmctx->ip6_querier;
 		other_query = &brmctx->ip6_other_query;
 		br_group.proto = htons(ETH_P_IPV6);
 #endif
@@ -1688,6 +1717,13 @@ static void br_multicast_send_query(struct net_bridge_mcast *brmctx,
 	if (!other_query || timer_pending(&other_query->timer))
 		return;
 
+	/* we're about to select ourselves as querier */
+	if (!pmctx && querier->port_ifidx) {
+		struct br_ip zeroip = {};
+
+		br_multicast_update_querier(brmctx, querier, 0, &zeroip);
+	}
+
 	__br_multicast_send_query(brmctx, pmctx, NULL, NULL, &br_group, false,
 				  0, NULL);
 
@@ -2828,58 +2864,147 @@ unlock_continue:
 }
 #endif
 
-static bool br_ip4_multicast_select_querier(struct net_bridge_mcast *brmctx,
-					    struct net_bridge_mcast_port *pmctx,
-					    __be32 saddr)
+static bool br_multicast_select_querier(struct net_bridge_mcast *brmctx,
+					struct net_bridge_mcast_port *pmctx,
+					struct br_ip *saddr)
 {
-	struct net_bridge_port *port = pmctx ? pmctx->port : NULL;
+	int port_ifidx = pmctx ? pmctx->port->dev->ifindex : 0;
+	struct timer_list *own_timer, *other_timer;
+	struct bridge_mcast_querier *querier;
 
-	if (!timer_pending(&brmctx->ip4_own_query.timer) &&
-	    !timer_pending(&brmctx->ip4_other_query.timer))
-		goto update;
-
-	if (!brmctx->ip4_querier.addr.src.ip4)
-		goto update;
+	switch (saddr->proto) {
+	case htons(ETH_P_IP):
+		querier = &brmctx->ip4_querier;
+		own_timer = &brmctx->ip4_own_query.timer;
+		other_timer = &brmctx->ip4_other_query.timer;
+		if (!querier->addr.src.ip4 ||
+		    ntohl(saddr->src.ip4) <= ntohl(querier->addr.src.ip4))
+			goto update;
+		break;
+#if IS_ENABLED(CONFIG_IPV6)
+	case htons(ETH_P_IPV6):
+		querier = &brmctx->ip6_querier;
+		own_timer = &brmctx->ip6_own_query.timer;
+		other_timer = &brmctx->ip6_other_query.timer;
+		if (ipv6_addr_cmp(&saddr->src.ip6, &querier->addr.src.ip6) <= 0)
+			goto update;
+		break;
+#endif
+	default:
+		return false;
+	}
 
-	if (ntohl(saddr) <= ntohl(brmctx->ip4_querier.addr.src.ip4))
+	if (!timer_pending(own_timer) && !timer_pending(other_timer))
 		goto update;
 
 	return false;
 
 update:
-	brmctx->ip4_querier.addr.src.ip4 = saddr;
-
-	/* update protected by general multicast_lock by caller */
-	rcu_assign_pointer(brmctx->ip4_querier.port, port);
+	br_multicast_update_querier(brmctx, querier, port_ifidx, saddr);
 
 	return true;
 }
 
-#if IS_ENABLED(CONFIG_IPV6)
-static bool br_ip6_multicast_select_querier(struct net_bridge_mcast *brmctx,
-					    struct net_bridge_mcast_port *pmctx,
-					    struct in6_addr *saddr)
+static struct net_bridge_port *
+__br_multicast_get_querier_port(struct net_bridge *br,
+				const struct bridge_mcast_querier *querier)
 {
-	struct net_bridge_port *port = pmctx ? pmctx->port : NULL;
+	int port_ifidx = READ_ONCE(querier->port_ifidx);
+	struct net_bridge_port *p;
+	struct net_device *dev;
 
-	if (!timer_pending(&brmctx->ip6_own_query.timer) &&
-	    !timer_pending(&brmctx->ip6_other_query.timer))
-		goto update;
+	if (port_ifidx == 0)
+		return NULL;
 
-	if (ipv6_addr_cmp(saddr, &brmctx->ip6_querier.addr.src.ip6) <= 0)
-		goto update;
+	dev = dev_get_by_index_rcu(dev_net(br->dev), port_ifidx);
+	if (!dev)
+		return NULL;
+	p = br_port_get_rtnl_rcu(dev);
+	if (!p || p->br != br)
+		return NULL;
 
-	return false;
+	return p;
+}
 
-update:
-	brmctx->ip6_querier.addr.src.ip6 = *saddr;
+size_t br_multicast_querier_state_size(void)
+{
+	return nla_total_size(sizeof(0)) +      /* nest attribute */
+	       nla_total_size(sizeof(__be32)) + /* BRIDGE_QUERIER_IP_ADDRESS */
+	       nla_total_size(sizeof(int)) +    /* BRIDGE_QUERIER_IP_PORT */
+	       nla_total_size_64bit(sizeof(u64)); /* BRIDGE_QUERIER_IP_OTHER_TIMER */
+}
 
-	/* update protected by general multicast_lock by caller */
-	rcu_assign_pointer(brmctx->ip6_querier.port, port);
+/* protected by rtnl or rcu */
+int br_multicast_dump_querier_state(struct sk_buff *skb,
+				    const struct net_bridge_mcast *brmctx,
+				    int nest_attr)
+{
+	struct bridge_mcast_querier querier = {};
+	struct net_bridge_port *p;
+	struct nlattr *nest;
 
-	return true;
-}
+	nest = nla_nest_start(skb, nest_attr);
+	if (!nest)
+		return -EMSGSIZE;
+
+	rcu_read_lock();
+	if (!brmctx->multicast_querier &&
+	    !timer_pending(&brmctx->ip4_other_query.timer))
+		goto out_v6;
+
+	br_multicast_read_querier(&brmctx->ip4_querier, &querier);
+	if (nla_put_in_addr(skb, BRIDGE_QUERIER_IP_ADDRESS,
+			    querier.addr.src.ip4)) {
+		rcu_read_unlock();
+		goto out_err;
+	}
+
+	p = __br_multicast_get_querier_port(brmctx->br, &querier);
+	if (timer_pending(&brmctx->ip4_other_query.timer) &&
+	    (nla_put_u64_64bit(skb, BRIDGE_QUERIER_IP_OTHER_TIMER,
+			       br_timer_value(&brmctx->ip4_other_query.timer),
+			       BRIDGE_QUERIER_PAD) ||
+	     (p && nla_put_u32(skb, BRIDGE_QUERIER_IP_PORT, p->dev->ifindex)))) {
+		rcu_read_unlock();
+		goto out_err;
+	}
+
+out_v6:
+#if IS_ENABLED(CONFIG_IPV6)
+	if (!brmctx->multicast_querier &&
+	    !timer_pending(&brmctx->ip6_other_query.timer))
+		goto out;
+
+	br_multicast_read_querier(&brmctx->ip6_querier, &querier);
+	if (nla_put_in6_addr(skb, BRIDGE_QUERIER_IPV6_ADDRESS,
+			     &querier.addr.src.ip6)) {
+		rcu_read_unlock();
+		goto out_err;
+	}
+
+	p = __br_multicast_get_querier_port(brmctx->br, &querier);
+	if (timer_pending(&brmctx->ip6_other_query.timer) &&
+	    (nla_put_u64_64bit(skb, BRIDGE_QUERIER_IPV6_OTHER_TIMER,
+			       br_timer_value(&brmctx->ip6_other_query.timer),
+			       BRIDGE_QUERIER_PAD) ||
+	     (p && nla_put_u32(skb, BRIDGE_QUERIER_IPV6_PORT,
+			       p->dev->ifindex)))) {
+		rcu_read_unlock();
+		goto out_err;
+	}
+out:
 #endif
+	rcu_read_unlock();
+	nla_nest_end(skb, nest);
+	if (!nla_len(nest))
+		nla_nest_cancel(skb, nest);
+
+	return 0;
+
+out_err:
+	nla_nest_cancel(skb, nest);
+	return -EMSGSIZE;
+}
 
 static void
 br_multicast_update_query_timer(struct net_bridge_mcast *brmctx,
@@ -3082,7 +3207,7 @@ br_ip4_multicast_query_received(struct net_bridge_mcast *brmctx,
 				struct br_ip *saddr,
 				unsigned long max_delay)
 {
-	if (!br_ip4_multicast_select_querier(brmctx, pmctx, saddr->src.ip4))
+	if (!br_multicast_select_querier(brmctx, pmctx, saddr))
 		return;
 
 	br_multicast_update_query_timer(brmctx, query, max_delay);
@@ -3097,7 +3222,7 @@ br_ip6_multicast_query_received(struct net_bridge_mcast *brmctx,
 				struct br_ip *saddr,
 				unsigned long max_delay)
 {
-	if (!br_ip6_multicast_select_querier(brmctx, pmctx, &saddr->src.ip6))
+	if (!br_multicast_select_querier(brmctx, pmctx, saddr))
 		return;
 
 	br_multicast_update_query_timer(brmctx, query, max_delay);
@@ -3117,7 +3242,7 @@ static void br_ip4_multicast_query(struct net_bridge_mcast *brmctx,
 	struct igmpv3_query *ih3;
 	struct net_bridge_port_group *p;
 	struct net_bridge_port_group __rcu **pp;
-	struct br_ip saddr;
+	struct br_ip saddr = {};
 	unsigned long max_delay;
 	unsigned long now = jiffies;
 	__be32 group;
@@ -3197,7 +3322,7 @@ static int br_ip6_multicast_query(struct net_bridge_mcast *brmctx,
 	struct mld2_query *mld2q;
 	struct net_bridge_port_group *p;
 	struct net_bridge_port_group __rcu **pp;
-	struct br_ip saddr;
+	struct br_ip saddr = {};
 	unsigned long max_delay;
 	unsigned long now = jiffies;
 	unsigned int offset = skb_transport_offset(skb);
@@ -3675,7 +3800,6 @@ static void br_multicast_query_expired(struct net_bridge_mcast *brmctx,
 	if (query->startup_sent < brmctx->multicast_startup_query_count)
 		query->startup_sent++;
 
-	RCU_INIT_POINTER(querier->port, NULL);
 	br_multicast_send_query(brmctx, NULL, query);
 out:
 	spin_unlock(&brmctx->br->multicast_lock);
@@ -3732,12 +3856,14 @@ void br_multicast_ctx_init(struct net_bridge *br,
 	brmctx->multicast_membership_interval = 260 * HZ;
 
 	brmctx->ip4_other_query.delay_time = 0;
-	brmctx->ip4_querier.port = NULL;
+	brmctx->ip4_querier.port_ifidx = 0;
+	seqcount_init(&brmctx->ip4_querier.seq);
 	brmctx->multicast_igmp_version = 2;
 #if IS_ENABLED(CONFIG_IPV6)
 	brmctx->multicast_mld_version = 1;
 	brmctx->ip6_other_query.delay_time = 0;
-	brmctx->ip6_querier.port = NULL;
+	brmctx->ip6_querier.port_ifidx = 0;
+	seqcount_init(&brmctx->ip6_querier.seq);
 #endif
 
 	timer_setup(&brmctx->ip4_mc_router_timer,
@@ -4479,6 +4605,7 @@ bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
 	struct net_bridge *br;
 	struct net_bridge_port *port;
 	bool ret = false;
+	int port_ifidx;
 
 	rcu_read_lock();
 	if (!netif_is_bridge_port(dev))
@@ -4493,14 +4620,16 @@ bool br_multicast_has_querier_adjacent(struct net_device *dev, int proto)
 
 	switch (proto) {
 	case ETH_P_IP:
+		port_ifidx = brmctx->ip4_querier.port_ifidx;
 		if (!timer_pending(&brmctx->ip4_other_query.timer) ||
-		    rcu_dereference(brmctx->ip4_querier.port) == port)
+		    port_ifidx == port->dev->ifindex)
 			goto unlock;
 		break;
 #if IS_ENABLED(CONFIG_IPV6)
 	case ETH_P_IPV6:
+		port_ifidx = brmctx->ip6_querier.port_ifidx;
 		if (!timer_pending(&brmctx->ip6_other_query.timer) ||
-		    rcu_dereference(brmctx->ip6_querier.port) == port)
+		    port_ifidx == port->dev->ifindex)
 			goto unlock;
 		break;
 #endif
diff --git a/net/bridge/br_netlink.c b/net/bridge/br_netlink.c
index 8ae026fa2ad7..2f184ad8ae29 100644
--- a/net/bridge/br_netlink.c
+++ b/net/bridge/br_netlink.c
@@ -1501,6 +1501,7 @@ static size_t br_get_size(const struct net_device *brdev)
 	       nla_total_size_64bit(sizeof(u64)) + /* IFLA_BR_MCAST_STARTUP_QUERY_INTVL */
 	       nla_total_size(sizeof(u8)) +	/* IFLA_BR_MCAST_IGMP_VERSION */
 	       nla_total_size(sizeof(u8)) +	/* IFLA_BR_MCAST_MLD_VERSION */
+	       br_multicast_querier_state_size() + /* IFLA_BR_MCAST_QUERIER_STATE */
 #endif
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 	       nla_total_size(sizeof(u8)) +     /* IFLA_BR_NF_CALL_IPTABLES */
@@ -1587,7 +1588,9 @@ static int br_fill_info(struct sk_buff *skb, const struct net_device *brdev)
 	    nla_put_u32(skb, IFLA_BR_MCAST_STARTUP_QUERY_CNT,
 			br->multicast_ctx.multicast_startup_query_count) ||
 	    nla_put_u8(skb, IFLA_BR_MCAST_IGMP_VERSION,
-		       br->multicast_ctx.multicast_igmp_version))
+		       br->multicast_ctx.multicast_igmp_version) ||
+	    br_multicast_dump_querier_state(skb, &br->multicast_ctx,
+					    IFLA_BR_MCAST_QUERIER_STATE))
 		return -EMSGSIZE;
 #if IS_ENABLED(CONFIG_IPV6)
 	if (nla_put_u8(skb, IFLA_BR_MCAST_MLD_VERSION,
diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index c4a8fee990c9..9b1bf98a2c5a 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h
@@ -81,7 +81,8 @@ struct bridge_mcast_other_query {
 /* selected querier */
 struct bridge_mcast_querier {
 	struct br_ip addr;
-	struct net_bridge_port __rcu	*port;
+	int port_ifidx;
+	seqcount_t seq;
 };
 
 /* IGMP/MLD statistics */
@@ -947,6 +948,10 @@ int br_mdb_replay(struct net_device *br_dev, struct net_device *dev,
 		  struct netlink_ext_ack *extack);
 int br_rports_fill_info(struct sk_buff *skb,
 			const struct net_bridge_mcast *brmctx);
+int br_multicast_dump_querier_state(struct sk_buff *skb,
+				    const struct net_bridge_mcast *brmctx,
+				    int nest_attr);
+size_t br_multicast_querier_state_size(void);
 
 static inline bool br_group_is_l2(const struct br_ip *group)
 {
diff --git a/net/bridge/br_vlan_options.c b/net/bridge/br_vlan_options.c
index b4fd5fa441b7..49dec53a4a74 100644
--- a/net/bridge/br_vlan_options.c
+++ b/net/bridge/br_vlan_options.c
@@ -299,7 +299,9 @@ bool br_vlan_global_opts_fill(struct sk_buff *skb, u16 vid, u16 vid_range,
 	    nla_put_u8(skb, BRIDGE_VLANDB_GOPTS_MCAST_QUERIER,
 		       v_opts->br_mcast_ctx.multicast_querier) ||
 	    nla_put_u8(skb, BRIDGE_VLANDB_GOPTS_MCAST_ROUTER,
-		       v_opts->br_mcast_ctx.multicast_router))
+		       v_opts->br_mcast_ctx.multicast_router) ||
+	    br_multicast_dump_querier_state(skb, &v_opts->br_mcast_ctx,
+					    BRIDGE_VLANDB_GOPTS_MCAST_QUERIER_STATE))
 		goto out_err;
 
 	clockval = jiffies_to_clock_t(v_opts->br_mcast_ctx.multicast_last_member_interval);
@@ -379,6 +381,7 @@ static size_t rtnl_vlan_global_opts_nlmsg_size(void)
 		+ nla_total_size(sizeof(u64)) /* BRIDGE_VLANDB_GOPTS_MCAST_STARTUP_QUERY_INTVL */
 		+ nla_total_size(sizeof(u8)) /* BRIDGE_VLANDB_GOPTS_MCAST_QUERIER */
 		+ nla_total_size(sizeof(u8)) /* BRIDGE_VLANDB_GOPTS_MCAST_ROUTER */
+		+ br_multicast_querier_state_size() /* BRIDGE_VLANDB_GOPTS_MCAST_QUERIER_STATE */
 #endif
 		+ nla_total_size(sizeof(u16)); /* BRIDGE_VLANDB_GOPTS_RANGE */
 }