tipc: send out join messages as soon as new member is discovered

When a socket is joining a group, we look up in the binding table to find if there are already other members of the group present. This is used for being able to return EAGAIN instead of EHOSTUNREACH if the user proceeds directly to a send attempt. However, the information in the binding table can be used to directly set the created member in state MBR_PUBLISHED and send a JOIN message to the peer, instead of waiting for a topology PUBLISH event to do this. When there are many members in a group, the propagation time for such events can be significant, and we can save time during the join operation if we use the initial lookup result fully. In this commit, we eliminate the member state MBR_DISCOVERED which has been the result of the initial lookup, and do instead go directly to MBR_PUBLISHED, which initiates the setup. After this change, the tipc_member FSM looks as follows: +-----------+ ---->| PUBLISHED |-----------------------------------------------+ PUB- +-----------+ LEAVE/WITHRAW | LISH |JOIN | | +-------------------------------------------+ | | | LEAVE/WITHDRAW | | | | +------------+ | | | | +----------->| PENDING |---------+ | | | | |msg/maxactv +-+---+------+ LEAVE/ | | | | | | | | WITHDRAW | | | | | | +----------+ | | | | | | | |revert/maxactv| | | | | | | V V V V V | +----------+ msg +------------+ +-----------+ +-->| JOINED |------>| ACTIVE |------>| LEAVING |---> | +----------+ +--- -+------+ LEAVE/+-----------+DOWN | A A | WITHDRAW A A A EVT | | | |RECLAIM | | | | | |REMIT V | | | | | |== adv +------------+ | | | | | +---------| RECLAIMING |--------+ | | | | +-----+------+ LEAVE/ | | | | |REMIT WITHDRAW | | | | |< adv | | | |msg/ V LEAVE/ | | | |adv==ADV_IDLE+------------+ WITHDRAW | | | +-------------| REMITTED |------------+ | | +------------+ | |PUBLISH | JOIN +-----------+ LEAVE/WITHDRAW | ---->| JOINING |-----------------------------------------------+ +-----------+ Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>

tipc: send out join messages as soon as new member is discovered
When a socket is joining a group, we look up in the binding table to find if there are already other members of the group present. This is used for being able to return EAGAIN instead of EHOSTUNREACH if the user proceeds directly to a send attempt. However, the information in the binding table can be used to directly set the created member in state MBR_PUBLISHED and send a JOIN message to the peer, instead of waiting for a topology PUBLISH event to do this. When there are many members in a group, the propagation time for such events can be significant, and we can save time during the join operation if we use the initial lookup result fully. In this commit, we eliminate the member state MBR_DISCOVERED which has been the result of the initial lookup, and do instead go directly to MBR_PUBLISHED, which initiates the setup. After this change, the tipc_member FSM looks as follows: +-----------+ ---->| PUBLISHED |-----------------------------------------------+ PUB- +-----------+ LEAVE/WITHRAW | LISH |JOIN | | +-------------------------------------------+ | | | LEAVE/WITHDRAW | | | | +------------+ | | | | +----------->| PENDING |---------+ | | | | |msg/maxactv +-+---+------+ LEAVE/ | | | | | | | | WITHDRAW | | | | | | +----------+ | | | | | | | |revert/maxactv| | | | | | | V V V V V | +----------+ msg +------------+ +-----------+ +-->| JOINED |------>| ACTIVE |------>| LEAVING |---> | +----------+ +--- -+------+ LEAVE/+-----------+DOWN | A A | WITHDRAW A A A EVT | | | |RECLAIM | | | | | |REMIT V | | | | | |== adv +------------+ | | | | | +---------| RECLAIMING |--------+ | | | | +-----+------+ LEAVE/ | | | | |REMIT WITHDRAW | | | | |< adv | | | |msg/ V LEAVE/ | | | |adv==ADV_IDLE+------------+ WITHDRAW | | | +-------------| REMITTED |------------+ | | +------------+ | |PUBLISH | JOIN +-----------+ LEAVE/WITHDRAW | ---->| JOINING |-----------------------------------------------+ +-----------+ Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Jon Maloy · David S. Miller
1 parent c2b22bcf2e
Showing 4 changed files with 68 additions and 42 deletions Side-by-side Diff
net/tipc/group.c
net/tipc/group.h
net/tipc/name_table.c
net/tipc/socket.c
@@ -49,7 +49,6 @@
 #define ADV_ACTIVE (ADV_UNIT * 12)
  
 enum mbr_state {
-	MBR_DISCOVERED,
 	MBR_JOINING,
 	MBR_PUBLISHED,
 	MBR_JOINED,
@@ -141,7 +140,7 @@
  
 static bool tipc_group_is_sender(struct tipc_member *m)
 {
-	return m && m->state >= MBR_JOINED;
+	return m && m->state != MBR_JOINING && m->state != MBR_PUBLISHED;
 }
  
 u32 tipc_group_exclude(struct tipc_group *grp)
@@ -184,6 +183,21 @@
 	return NULL;
 }
  
+void tipc_group_join(struct net *net, struct tipc_group *grp, int *sk_rcvbuf)
+{
+	struct rb_root *tree = &grp->members;
+	struct tipc_member *m, *tmp;
+	struct sk_buff_head xmitq;
+
+	skb_queue_head_init(&xmitq);
+	rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
+		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, &xmitq);
+		tipc_group_update_member(m, 0);
+	}
+	tipc_node_distr_xmit(net, &xmitq);
+	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
+}
+
 void tipc_group_delete(struct net *net, struct tipc_group *grp)
 {
 	struct rb_root *tree = &grp->members;
@@ -274,7 +288,7 @@
  
 static struct tipc_member *tipc_group_create_member(struct tipc_group *grp,
 						    u32 node, u32 port,
-						    int state)
+						    u32 instance, int state)
 {
 	struct tipc_member *m;
  
@@ -287,6 +301,7 @@
 	m->group = grp;
 	m->node = node;
 	m->port = port;
+	m->instance = instance;
 	m->bc_acked = grp->bc_snd_nxt - 1;
 	grp->member_cnt++;
 	tipc_group_add_to_tree(grp, m);
  
@@ -295,9 +310,10 @@
 	return m;
 }
  
-void tipc_group_add_member(struct tipc_group *grp, u32 node, u32 port)
+void tipc_group_add_member(struct tipc_group *grp, u32 node,
+			   u32 port, u32 instance)
 {
-	tipc_group_create_member(grp, node, port, MBR_DISCOVERED);
+	tipc_group_create_member(grp, node, port, instance, MBR_PUBLISHED);
 }
  
 static void tipc_group_delete_member(struct tipc_group *grp,
@@ -623,7 +639,6 @@
 		tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
 		break;
 	case MBR_RECLAIMING:
-	case MBR_DISCOVERED:
 	case MBR_JOINING:
 	case MBR_LEAVING:
 	default:
  
  
@@ -721,26 +736,26 @@
 	case GRP_JOIN_MSG:
 		if (!m)
 			m = tipc_group_create_member(grp, node, port,
-						     MBR_JOINING);
+						     0, MBR_JOINING);
 		if (!m)
 			return;
 		m->bc_syncpt = msg_grp_bc_syncpt(hdr);
 		m->bc_rcv_nxt = m->bc_syncpt;
 		m->window += msg_adv_win(hdr);
  
-		/* Wait until PUBLISH event is received */
-		if (m->state == MBR_DISCOVERED) {
-			m->state = MBR_JOINING;
-		} else if (m->state == MBR_PUBLISHED) {
-			m->state = MBR_JOINED;
-			*usr_wakeup = true;
-			m->usr_pending = false;
-			tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
-			tipc_group_create_event(grp, m, TIPC_PUBLISHED,
-						m->bc_syncpt, inputq);
-		}
+		/* Wait until PUBLISH event is received if necessary */
+		if (m->state != MBR_PUBLISHED)
+			return;
+
+		/* Member can be taken into service */
+		m->state = MBR_JOINED;
+		*usr_wakeup = true;
+		m->usr_pending = false;
 		list_del_init(&m->small_win);
 		tipc_group_update_member(m, 0);
+		tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
+		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
+					m->bc_syncpt, inputq);
 		return;
 	case GRP_LEAVE_MSG:
 		if (!m)
  
  
  
  
@@ -844,30 +859,36 @@
  
 	m = tipc_group_find_member(grp, node, port);
  
-	if (event == TIPC_PUBLISHED) {
-		if (!m)
-			m = tipc_group_create_member(grp, node, port,
-						     MBR_DISCOVERED);
-		if (!m)
-			return;
+	switch (event) {
+	case TIPC_PUBLISHED:
+		/* Send and wait for arrival of JOIN message if necessary */
+		if (!m) {
+			m = tipc_group_create_member(grp, node, port, instance,
+						     MBR_PUBLISHED);
+			if (!m)
+				break;
+			tipc_group_update_member(m, 0);
+			tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
+			break;
+		}
  
-		m->instance = instance;
+		if (m->state != MBR_JOINING)
+			break;
  
-		/* Hold back event if JOIN message not yet received */
-		if (m->state == MBR_DISCOVERED) {
-			m->state = MBR_PUBLISHED;
-		} else {
-			tipc_group_create_event(grp, m, TIPC_PUBLISHED,
-						m->bc_syncpt, inputq);
-			m->state = MBR_JOINED;
-			*usr_wakeup = true;
-			m->usr_pending = false;
-		}
-		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
+		/* Member can be taken into service */
+		m->instance = instance;
+		m->state = MBR_JOINED;
+		*usr_wakeup = true;
+		m->usr_pending = false;
+		list_del_init(&m->small_win);
 		tipc_group_update_member(m, 0);
-	} else if (event == TIPC_WITHDRAWN) {
+		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
+		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
+					m->bc_syncpt, inputq);
+		break;
+	case TIPC_WITHDRAWN:
 		if (!m)
-			return;
+			break;
  
 		*usr_wakeup = true;
 		m->usr_pending = false;
@@ -880,6 +901,9 @@
 		if (!tipc_node_is_up(net, node))
 			tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
 						m->bc_rcv_nxt, inputq);
+		break;
+	default:
+		break;
 	}
 	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
 }
@@ -44,8 +44,10 @@
  
 struct tipc_group *tipc_group_create(struct net *net, u32 portid,
 				     struct tipc_group_req *mreq);
+void tipc_group_join(struct net *net, struct tipc_group *grp, int *sk_rcv_buf);
 void tipc_group_delete(struct net *net, struct tipc_group *grp);
-void tipc_group_add_member(struct tipc_group *grp, u32 node, u32 port);
+void tipc_group_add_member(struct tipc_group *grp, u32 node,
+			   u32 port, u32 instance);
 struct tipc_nlist *tipc_group_dests(struct tipc_group *grp);
 void tipc_group_self(struct tipc_group *grp, struct tipc_name_seq *seq,
 		     int *scope);
@@ -732,7 +732,7 @@
 		list_for_each_entry(p, &info->zone_list, zone_list) {
 			if (!tipc_in_scope(domain, p->node))
 				continue;
-			tipc_group_add_member(grp, p->node, p->ref);
+			tipc_group_add_member(grp, p->node, p->ref, p->lower);
 		}
 	}
 	spin_unlock_bh(&seq->lock);
@@ -2757,10 +2757,10 @@
 		tipc_group_delete(net, grp);
 		tsk->group = NULL;
 	}
-
-	/* Eliminate any risk that a broadcast overtakes the sent JOIN */
+	/* Eliminate any risk that a broadcast overtakes sent JOINs */
 	tsk->mc_method.rcast = true;
 	tsk->mc_method.mandatory = true;
+	tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
 	return rc;
 }
...	...	@@ -49,7 +49,6 @@
49	49	#define ADV_ACTIVE (ADV_UNIT * 12)
50	50
51	51	enum mbr_state {
52		- MBR_DISCOVERED,
53	52	MBR_JOINING,
54	53	MBR_PUBLISHED,
55	54	MBR_JOINED,
...	...	@@ -141,7 +140,7 @@
141	140
142	141	static bool tipc_group_is_sender(struct tipc_member *m)
143	142	{
144		- return m && m->state >= MBR_JOINED;
	143	+ return m && m->state != MBR_JOINING && m->state != MBR_PUBLISHED;
145	144	}
146	145
147	146	u32 tipc_group_exclude(struct tipc_group *grp)
...	...	@@ -184,6 +183,21 @@
184	183	return NULL;
185	184	}
186	185
	186	+void tipc_group_join(struct net net, struct tipc_group grp, int *sk_rcvbuf)
	187	+{
	188	+ struct rb_root *tree = &grp->members;
	189	+ struct tipc_member m, tmp;
	190	+ struct sk_buff_head xmitq;
	191	+
	192	+ skb_queue_head_init(&xmitq);
	193	+ rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
	194	+ tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, &xmitq);
	195	+ tipc_group_update_member(m, 0);
	196	+ }
	197	+ tipc_node_distr_xmit(net, &xmitq);
	198	+ *sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
	199	+}
	200	+
187	201	void tipc_group_delete(struct net net, struct tipc_group grp)
188	202	{
189	203	struct rb_root *tree = &grp->members;
...	...	@@ -274,7 +288,7 @@
274	288
275	289	static struct tipc_member tipc_group_create_member(struct tipc_group grp,
276	290	u32 node, u32 port,
277		- int state)
	291	+ u32 instance, int state)
278	292	{
279	293	struct tipc_member *m;
280	294
...	...	@@ -287,6 +301,7 @@
287	301	m->group = grp;
288	302	m->node = node;
289	303	m->port = port;
	304	+ m->instance = instance;
290	305	m->bc_acked = grp->bc_snd_nxt - 1;
291	306	grp->member_cnt++;
292	307	tipc_group_add_to_tree(grp, m);
293	308
...	...	@@ -295,9 +310,10 @@
295	310	return m;
296	311	}
297	312
298		-void tipc_group_add_member(struct tipc_group *grp, u32 node, u32 port)
	313	+void tipc_group_add_member(struct tipc_group *grp, u32 node,
	314	+ u32 port, u32 instance)
299	315	{
300		- tipc_group_create_member(grp, node, port, MBR_DISCOVERED);
	316	+ tipc_group_create_member(grp, node, port, instance, MBR_PUBLISHED);
301	317	}
302	318
303	319	static void tipc_group_delete_member(struct tipc_group *grp,
...	...	@@ -623,7 +639,6 @@
623	639	tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
624	640	break;
625	641	case MBR_RECLAIMING:
626		- case MBR_DISCOVERED:
627	642	case MBR_JOINING:
628	643	case MBR_LEAVING:
629	644	default:
630	645
631	646
...	...	@@ -721,26 +736,26 @@
721	736	case GRP_JOIN_MSG:
722	737	if (!m)
723	738	m = tipc_group_create_member(grp, node, port,
724		- MBR_JOINING);
	739	+ 0, MBR_JOINING);
725	740	if (!m)
726	741	return;
727	742	m->bc_syncpt = msg_grp_bc_syncpt(hdr);
728	743	m->bc_rcv_nxt = m->bc_syncpt;
729	744	m->window += msg_adv_win(hdr);
730	745
731		- /* Wait until PUBLISH event is received */
732		- if (m->state == MBR_DISCOVERED) {
733		- m->state = MBR_JOINING;
734		- } else if (m->state == MBR_PUBLISHED) {
735		- m->state = MBR_JOINED;
736		- *usr_wakeup = true;
737		- m->usr_pending = false;
738		- tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
739		- tipc_group_create_event(grp, m, TIPC_PUBLISHED,
740		- m->bc_syncpt, inputq);
741		- }
	746	+ /* Wait until PUBLISH event is received if necessary */
	747	+ if (m->state != MBR_PUBLISHED)
	748	+ return;
	749	+
	750	+ /* Member can be taken into service */
	751	+ m->state = MBR_JOINED;
	752	+ *usr_wakeup = true;
	753	+ m->usr_pending = false;
742	754	list_del_init(&m->small_win);
743	755	tipc_group_update_member(m, 0);
	756	+ tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
	757	+ tipc_group_create_event(grp, m, TIPC_PUBLISHED,
	758	+ m->bc_syncpt, inputq);
744	759	return;
745	760	case GRP_LEAVE_MSG:
746	761	if (!m)
747	762
748	763
749	764
750	765
...	...	@@ -844,30 +859,36 @@
844	859
845	860	m = tipc_group_find_member(grp, node, port);
846	861
847		- if (event == TIPC_PUBLISHED) {
848		- if (!m)
849		- m = tipc_group_create_member(grp, node, port,
850		- MBR_DISCOVERED);
851		- if (!m)
852		- return;
	862	+ switch (event) {
	863	+ case TIPC_PUBLISHED:
	864	+ /* Send and wait for arrival of JOIN message if necessary */
	865	+ if (!m) {
	866	+ m = tipc_group_create_member(grp, node, port, instance,
	867	+ MBR_PUBLISHED);
	868	+ if (!m)
	869	+ break;
	870	+ tipc_group_update_member(m, 0);
	871	+ tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
	872	+ break;
	873	+ }
853	874
854		- m->instance = instance;
	875	+ if (m->state != MBR_JOINING)
	876	+ break;
855	877
856		- /* Hold back event if JOIN message not yet received */
857		- if (m->state == MBR_DISCOVERED) {
858		- m->state = MBR_PUBLISHED;
859		- } else {
860		- tipc_group_create_event(grp, m, TIPC_PUBLISHED,
861		- m->bc_syncpt, inputq);
862		- m->state = MBR_JOINED;
863		- *usr_wakeup = true;
864		- m->usr_pending = false;
865		- }
866		- tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
	878	+ /* Member can be taken into service */
	879	+ m->instance = instance;
	880	+ m->state = MBR_JOINED;
	881	+ *usr_wakeup = true;
	882	+ m->usr_pending = false;
	883	+ list_del_init(&m->small_win);
867	884	tipc_group_update_member(m, 0);
868		- } else if (event == TIPC_WITHDRAWN) {
	885	+ tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
	886	+ tipc_group_create_event(grp, m, TIPC_PUBLISHED,
	887	+ m->bc_syncpt, inputq);
	888	+ break;
	889	+ case TIPC_WITHDRAWN:
869	890	if (!m)
870		- return;
	891	+ break;
871	892
872	893	*usr_wakeup = true;
873	894	m->usr_pending = false;
...	...	@@ -880,6 +901,9 @@
880	901	if (!tipc_node_is_up(net, node))
881	902	tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
882	903	m->bc_rcv_nxt, inputq);
	904	+ break;
	905	+ default:
	906	+ break;
883	907	}
884	908	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
885	909	}
...	...	@@ -44,8 +44,10 @@
44	44
45	45	struct tipc_group tipc_group_create(struct net net, u32 portid,
46	46	struct tipc_group_req *mreq);
	47	+void tipc_group_join(struct net net, struct tipc_group grp, int *sk_rcv_buf);
47	48	void tipc_group_delete(struct net net, struct tipc_group grp);
48		-void tipc_group_add_member(struct tipc_group *grp, u32 node, u32 port);
	49	+void tipc_group_add_member(struct tipc_group *grp, u32 node,
	50	+ u32 port, u32 instance);
49	51	struct tipc_nlist tipc_group_dests(struct tipc_group grp);
50	52	void tipc_group_self(struct tipc_group grp, struct tipc_name_seq seq,
51	53	int *scope);
...	...	@@ -732,7 +732,7 @@
732	732	list_for_each_entry(p, &info->zone_list, zone_list) {
733	733	if (!tipc_in_scope(domain, p->node))
734	734	continue;
735		- tipc_group_add_member(grp, p->node, p->ref);
	735	+ tipc_group_add_member(grp, p->node, p->ref, p->lower);
736	736	}
737	737	}
738	738	spin_unlock_bh(&seq->lock);
...	...	@@ -2757,10 +2757,10 @@
2757	2757	tipc_group_delete(net, grp);
2758	2758	tsk->group = NULL;
2759	2759	}
2760		-
2761		- /* Eliminate any risk that a broadcast overtakes the sent JOIN */
	2760	+ /* Eliminate any risk that a broadcast overtakes sent JOINs */
2762	2761	tsk->mc_method.rcast = true;
2763	2762	tsk->mc_method.mandatory = true;
	2763	+ tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
2764	2764	return rc;
2765	2765	}
2766	2766