xref: /openbmc/linux/net/tipc/group.c (revision da2ef666)
1 /*
2  * net/tipc/group.c: TIPC group messaging code
3  *
4  * Copyright (c) 2017, Ericsson AB
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions are met:
9  *
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the names of the copyright holders nor the names of its
16  *    contributors may be used to endorse or promote products derived from
17  *    this software without specific prior written permission.
18  *
19  * Alternatively, this software may be distributed under the terms of the
20  * GNU General Public License ("GPL") version 2 as published by the Free
21  * Software Foundation.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
24  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
27  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  */
35 
36 #include "core.h"
37 #include "addr.h"
38 #include "group.h"
39 #include "bcast.h"
40 #include "topsrv.h"
41 #include "msg.h"
42 #include "socket.h"
43 #include "node.h"
44 #include "name_table.h"
45 #include "subscr.h"
46 
47 #define ADV_UNIT (((MAX_MSG_SIZE + MAX_H_SIZE) / FLOWCTL_BLK_SZ) + 1)
48 #define ADV_IDLE ADV_UNIT
49 #define ADV_ACTIVE (ADV_UNIT * 12)
50 
51 enum mbr_state {
52 	MBR_JOINING,
53 	MBR_PUBLISHED,
54 	MBR_JOINED,
55 	MBR_PENDING,
56 	MBR_ACTIVE,
57 	MBR_RECLAIMING,
58 	MBR_REMITTED,
59 	MBR_LEAVING
60 };
61 
62 struct tipc_member {
63 	struct rb_node tree_node;
64 	struct list_head list;
65 	struct list_head small_win;
66 	struct sk_buff_head deferredq;
67 	struct tipc_group *group;
68 	u32 node;
69 	u32 port;
70 	u32 instance;
71 	enum mbr_state state;
72 	u16 advertised;
73 	u16 window;
74 	u16 bc_rcv_nxt;
75 	u16 bc_syncpt;
76 	u16 bc_acked;
77 };
78 
79 struct tipc_group {
80 	struct rb_root members;
81 	struct list_head small_win;
82 	struct list_head pending;
83 	struct list_head active;
84 	struct tipc_nlist dests;
85 	struct net *net;
86 	int subid;
87 	u32 type;
88 	u32 instance;
89 	u32 scope;
90 	u32 portid;
91 	u16 member_cnt;
92 	u16 active_cnt;
93 	u16 max_active;
94 	u16 bc_snd_nxt;
95 	u16 bc_ackers;
96 	bool *open;
97 	bool loopback;
98 	bool events;
99 };
100 
101 static void tipc_group_proto_xmit(struct tipc_group *grp, struct tipc_member *m,
102 				  int mtyp, struct sk_buff_head *xmitq);
103 
104 static void tipc_group_open(struct tipc_member *m, bool *wakeup)
105 {
106 	*wakeup = false;
107 	if (list_empty(&m->small_win))
108 		return;
109 	list_del_init(&m->small_win);
110 	*m->group->open = true;
111 	*wakeup = true;
112 }
113 
114 static void tipc_group_decr_active(struct tipc_group *grp,
115 				   struct tipc_member *m)
116 {
117 	if (m->state == MBR_ACTIVE || m->state == MBR_RECLAIMING ||
118 	    m->state == MBR_REMITTED)
119 		grp->active_cnt--;
120 }
121 
122 static int tipc_group_rcvbuf_limit(struct tipc_group *grp)
123 {
124 	int max_active, active_pool, idle_pool;
125 	int mcnt = grp->member_cnt + 1;
126 
127 	/* Limit simultaneous reception from other members */
128 	max_active = min(mcnt / 8, 64);
129 	max_active = max(max_active, 16);
130 	grp->max_active = max_active;
131 
132 	/* Reserve blocks for active and idle members */
133 	active_pool = max_active * ADV_ACTIVE;
134 	idle_pool = (mcnt - max_active) * ADV_IDLE;
135 
136 	/* Scale to bytes, considering worst-case truesize/msgsize ratio */
137 	return (active_pool + idle_pool) * FLOWCTL_BLK_SZ * 4;
138 }
139 
140 u16 tipc_group_bc_snd_nxt(struct tipc_group *grp)
141 {
142 	return grp->bc_snd_nxt;
143 }
144 
145 static bool tipc_group_is_receiver(struct tipc_member *m)
146 {
147 	return m && m->state != MBR_JOINING && m->state != MBR_LEAVING;
148 }
149 
150 static bool tipc_group_is_sender(struct tipc_member *m)
151 {
152 	return m && m->state != MBR_JOINING && m->state != MBR_PUBLISHED;
153 }
154 
155 u32 tipc_group_exclude(struct tipc_group *grp)
156 {
157 	if (!grp->loopback)
158 		return grp->portid;
159 	return 0;
160 }
161 
162 struct tipc_group *tipc_group_create(struct net *net, u32 portid,
163 				     struct tipc_group_req *mreq,
164 				     bool *group_is_open)
165 {
166 	u32 filter = TIPC_SUB_PORTS | TIPC_SUB_NO_STATUS;
167 	bool global = mreq->scope != TIPC_NODE_SCOPE;
168 	struct tipc_group *grp;
169 	u32 type = mreq->type;
170 
171 	grp = kzalloc(sizeof(*grp), GFP_ATOMIC);
172 	if (!grp)
173 		return NULL;
174 	tipc_nlist_init(&grp->dests, tipc_own_addr(net));
175 	INIT_LIST_HEAD(&grp->small_win);
176 	INIT_LIST_HEAD(&grp->active);
177 	INIT_LIST_HEAD(&grp->pending);
178 	grp->members = RB_ROOT;
179 	grp->net = net;
180 	grp->portid = portid;
181 	grp->type = type;
182 	grp->instance = mreq->instance;
183 	grp->scope = mreq->scope;
184 	grp->loopback = mreq->flags & TIPC_GROUP_LOOPBACK;
185 	grp->events = mreq->flags & TIPC_GROUP_MEMBER_EVTS;
186 	grp->open = group_is_open;
187 	*grp->open = false;
188 	filter |= global ? TIPC_SUB_CLUSTER_SCOPE : TIPC_SUB_NODE_SCOPE;
189 	if (tipc_topsrv_kern_subscr(net, portid, type, 0, ~0,
190 				    filter, &grp->subid))
191 		return grp;
192 	kfree(grp);
193 	return NULL;
194 }
195 
196 void tipc_group_join(struct net *net, struct tipc_group *grp, int *sk_rcvbuf)
197 {
198 	struct rb_root *tree = &grp->members;
199 	struct tipc_member *m, *tmp;
200 	struct sk_buff_head xmitq;
201 
202 	skb_queue_head_init(&xmitq);
203 	rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
204 		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, &xmitq);
205 		tipc_group_update_member(m, 0);
206 	}
207 	tipc_node_distr_xmit(net, &xmitq);
208 	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
209 }
210 
211 void tipc_group_delete(struct net *net, struct tipc_group *grp)
212 {
213 	struct rb_root *tree = &grp->members;
214 	struct tipc_member *m, *tmp;
215 	struct sk_buff_head xmitq;
216 
217 	__skb_queue_head_init(&xmitq);
218 
219 	rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
220 		tipc_group_proto_xmit(grp, m, GRP_LEAVE_MSG, &xmitq);
221 		list_del(&m->list);
222 		kfree(m);
223 	}
224 	tipc_node_distr_xmit(net, &xmitq);
225 	tipc_nlist_purge(&grp->dests);
226 	tipc_topsrv_kern_unsubscr(net, grp->subid);
227 	kfree(grp);
228 }
229 
230 static struct tipc_member *tipc_group_find_member(struct tipc_group *grp,
231 						  u32 node, u32 port)
232 {
233 	struct rb_node *n = grp->members.rb_node;
234 	u64 nkey, key = (u64)node << 32 | port;
235 	struct tipc_member *m;
236 
237 	while (n) {
238 		m = container_of(n, struct tipc_member, tree_node);
239 		nkey = (u64)m->node << 32 | m->port;
240 		if (key < nkey)
241 			n = n->rb_left;
242 		else if (key > nkey)
243 			n = n->rb_right;
244 		else
245 			return m;
246 	}
247 	return NULL;
248 }
249 
250 static struct tipc_member *tipc_group_find_dest(struct tipc_group *grp,
251 						u32 node, u32 port)
252 {
253 	struct tipc_member *m;
254 
255 	m = tipc_group_find_member(grp, node, port);
256 	if (m && tipc_group_is_receiver(m))
257 		return m;
258 	return NULL;
259 }
260 
261 static struct tipc_member *tipc_group_find_node(struct tipc_group *grp,
262 						u32 node)
263 {
264 	struct tipc_member *m;
265 	struct rb_node *n;
266 
267 	for (n = rb_first(&grp->members); n; n = rb_next(n)) {
268 		m = container_of(n, struct tipc_member, tree_node);
269 		if (m->node == node)
270 			return m;
271 	}
272 	return NULL;
273 }
274 
275 static void tipc_group_add_to_tree(struct tipc_group *grp,
276 				   struct tipc_member *m)
277 {
278 	u64 nkey, key = (u64)m->node << 32 | m->port;
279 	struct rb_node **n, *parent = NULL;
280 	struct tipc_member *tmp;
281 
282 	n = &grp->members.rb_node;
283 	while (*n) {
284 		tmp = container_of(*n, struct tipc_member, tree_node);
285 		parent = *n;
286 		tmp = container_of(parent, struct tipc_member, tree_node);
287 		nkey = (u64)tmp->node << 32 | tmp->port;
288 		if (key < nkey)
289 			n = &(*n)->rb_left;
290 		else if (key > nkey)
291 			n = &(*n)->rb_right;
292 		else
293 			return;
294 	}
295 	rb_link_node(&m->tree_node, parent, n);
296 	rb_insert_color(&m->tree_node, &grp->members);
297 }
298 
299 static struct tipc_member *tipc_group_create_member(struct tipc_group *grp,
300 						    u32 node, u32 port,
301 						    u32 instance, int state)
302 {
303 	struct tipc_member *m;
304 
305 	m = kzalloc(sizeof(*m), GFP_ATOMIC);
306 	if (!m)
307 		return NULL;
308 	INIT_LIST_HEAD(&m->list);
309 	INIT_LIST_HEAD(&m->small_win);
310 	__skb_queue_head_init(&m->deferredq);
311 	m->group = grp;
312 	m->node = node;
313 	m->port = port;
314 	m->instance = instance;
315 	m->bc_acked = grp->bc_snd_nxt - 1;
316 	grp->member_cnt++;
317 	tipc_group_add_to_tree(grp, m);
318 	tipc_nlist_add(&grp->dests, m->node);
319 	m->state = state;
320 	return m;
321 }
322 
323 void tipc_group_add_member(struct tipc_group *grp, u32 node,
324 			   u32 port, u32 instance)
325 {
326 	tipc_group_create_member(grp, node, port, instance, MBR_PUBLISHED);
327 }
328 
329 static void tipc_group_delete_member(struct tipc_group *grp,
330 				     struct tipc_member *m)
331 {
332 	rb_erase(&m->tree_node, &grp->members);
333 	grp->member_cnt--;
334 
335 	/* Check if we were waiting for replicast ack from this member */
336 	if (grp->bc_ackers && less(m->bc_acked, grp->bc_snd_nxt - 1))
337 		grp->bc_ackers--;
338 
339 	list_del_init(&m->list);
340 	list_del_init(&m->small_win);
341 	tipc_group_decr_active(grp, m);
342 
343 	/* If last member on a node, remove node from dest list */
344 	if (!tipc_group_find_node(grp, m->node))
345 		tipc_nlist_del(&grp->dests, m->node);
346 
347 	kfree(m);
348 }
349 
350 struct tipc_nlist *tipc_group_dests(struct tipc_group *grp)
351 {
352 	return &grp->dests;
353 }
354 
355 void tipc_group_self(struct tipc_group *grp, struct tipc_name_seq *seq,
356 		     int *scope)
357 {
358 	seq->type = grp->type;
359 	seq->lower = grp->instance;
360 	seq->upper = grp->instance;
361 	*scope = grp->scope;
362 }
363 
364 void tipc_group_update_member(struct tipc_member *m, int len)
365 {
366 	struct tipc_group *grp = m->group;
367 	struct tipc_member *_m, *tmp;
368 
369 	if (!tipc_group_is_receiver(m))
370 		return;
371 
372 	m->window -= len;
373 
374 	if (m->window >= ADV_IDLE)
375 		return;
376 
377 	list_del_init(&m->small_win);
378 
379 	/* Sort member into small_window members' list */
380 	list_for_each_entry_safe(_m, tmp, &grp->small_win, small_win) {
381 		if (_m->window > m->window)
382 			break;
383 	}
384 	list_add_tail(&m->small_win, &_m->small_win);
385 }
386 
387 void tipc_group_update_bc_members(struct tipc_group *grp, int len, bool ack)
388 {
389 	u16 prev = grp->bc_snd_nxt - 1;
390 	struct tipc_member *m;
391 	struct rb_node *n;
392 	u16 ackers = 0;
393 
394 	for (n = rb_first(&grp->members); n; n = rb_next(n)) {
395 		m = container_of(n, struct tipc_member, tree_node);
396 		if (tipc_group_is_receiver(m)) {
397 			tipc_group_update_member(m, len);
398 			m->bc_acked = prev;
399 			ackers++;
400 		}
401 	}
402 
403 	/* Mark number of acknowledges to expect, if any */
404 	if (ack)
405 		grp->bc_ackers = ackers;
406 	grp->bc_snd_nxt++;
407 }
408 
409 bool tipc_group_cong(struct tipc_group *grp, u32 dnode, u32 dport,
410 		     int len, struct tipc_member **mbr)
411 {
412 	struct sk_buff_head xmitq;
413 	struct tipc_member *m;
414 	int adv, state;
415 
416 	m = tipc_group_find_dest(grp, dnode, dport);
417 	if (!tipc_group_is_receiver(m)) {
418 		*mbr = NULL;
419 		return false;
420 	}
421 	*mbr = m;
422 
423 	if (m->window >= len)
424 		return false;
425 
426 	*grp->open = false;
427 
428 	/* If not fully advertised, do it now to prevent mutual blocking */
429 	adv = m->advertised;
430 	state = m->state;
431 	if (state == MBR_JOINED && adv == ADV_IDLE)
432 		return true;
433 	if (state == MBR_ACTIVE && adv == ADV_ACTIVE)
434 		return true;
435 	if (state == MBR_PENDING && adv == ADV_IDLE)
436 		return true;
437 	skb_queue_head_init(&xmitq);
438 	tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, &xmitq);
439 	tipc_node_distr_xmit(grp->net, &xmitq);
440 	return true;
441 }
442 
443 bool tipc_group_bc_cong(struct tipc_group *grp, int len)
444 {
445 	struct tipc_member *m = NULL;
446 
447 	/* If prev bcast was replicast, reject until all receivers have acked */
448 	if (grp->bc_ackers) {
449 		*grp->open = false;
450 		return true;
451 	}
452 	if (list_empty(&grp->small_win))
453 		return false;
454 
455 	m = list_first_entry(&grp->small_win, struct tipc_member, small_win);
456 	if (m->window >= len)
457 		return false;
458 
459 	return tipc_group_cong(grp, m->node, m->port, len, &m);
460 }
461 
462 /* tipc_group_sort_msg() - sort msg into queue by bcast sequence number
463  */
464 static void tipc_group_sort_msg(struct sk_buff *skb, struct sk_buff_head *defq)
465 {
466 	struct tipc_msg *_hdr, *hdr = buf_msg(skb);
467 	u16 bc_seqno = msg_grp_bc_seqno(hdr);
468 	struct sk_buff *_skb, *tmp;
469 	int mtyp = msg_type(hdr);
470 
471 	/* Bcast/mcast may be bypassed by ucast or other bcast, - sort it in */
472 	if (mtyp == TIPC_GRP_BCAST_MSG || mtyp == TIPC_GRP_MCAST_MSG) {
473 		skb_queue_walk_safe(defq, _skb, tmp) {
474 			_hdr = buf_msg(_skb);
475 			if (!less(bc_seqno, msg_grp_bc_seqno(_hdr)))
476 				continue;
477 			__skb_queue_before(defq, _skb, skb);
478 			return;
479 		}
480 		/* Bcast was not bypassed, - add to tail */
481 	}
482 	/* Unicasts are never bypassed, - always add to tail */
483 	__skb_queue_tail(defq, skb);
484 }
485 
486 /* tipc_group_filter_msg() - determine if we should accept arriving message
487  */
488 void tipc_group_filter_msg(struct tipc_group *grp, struct sk_buff_head *inputq,
489 			   struct sk_buff_head *xmitq)
490 {
491 	struct sk_buff *skb = __skb_dequeue(inputq);
492 	bool ack, deliver, update, leave = false;
493 	struct sk_buff_head *defq;
494 	struct tipc_member *m;
495 	struct tipc_msg *hdr;
496 	u32 node, port;
497 	int mtyp, blks;
498 
499 	if (!skb)
500 		return;
501 
502 	hdr = buf_msg(skb);
503 	node =  msg_orignode(hdr);
504 	port = msg_origport(hdr);
505 
506 	if (!msg_in_group(hdr))
507 		goto drop;
508 
509 	m = tipc_group_find_member(grp, node, port);
510 	if (!tipc_group_is_sender(m))
511 		goto drop;
512 
513 	if (less(msg_grp_bc_seqno(hdr), m->bc_rcv_nxt))
514 		goto drop;
515 
516 	TIPC_SKB_CB(skb)->orig_member = m->instance;
517 	defq = &m->deferredq;
518 	tipc_group_sort_msg(skb, defq);
519 
520 	while ((skb = skb_peek(defq))) {
521 		hdr = buf_msg(skb);
522 		mtyp = msg_type(hdr);
523 		blks = msg_blocks(hdr);
524 		deliver = true;
525 		ack = false;
526 		update = false;
527 
528 		if (more(msg_grp_bc_seqno(hdr), m->bc_rcv_nxt))
529 			break;
530 
531 		/* Decide what to do with message */
532 		switch (mtyp) {
533 		case TIPC_GRP_MCAST_MSG:
534 			if (msg_nameinst(hdr) != grp->instance) {
535 				update = true;
536 				deliver = false;
537 			}
538 			/* Fall thru */
539 		case TIPC_GRP_BCAST_MSG:
540 			m->bc_rcv_nxt++;
541 			ack = msg_grp_bc_ack_req(hdr);
542 			break;
543 		case TIPC_GRP_UCAST_MSG:
544 			break;
545 		case TIPC_GRP_MEMBER_EVT:
546 			if (m->state == MBR_LEAVING)
547 				leave = true;
548 			if (!grp->events)
549 				deliver = false;
550 			break;
551 		default:
552 			break;
553 		}
554 
555 		/* Execute decisions */
556 		__skb_dequeue(defq);
557 		if (deliver)
558 			__skb_queue_tail(inputq, skb);
559 		else
560 			kfree_skb(skb);
561 
562 		if (ack)
563 			tipc_group_proto_xmit(grp, m, GRP_ACK_MSG, xmitq);
564 
565 		if (leave) {
566 			__skb_queue_purge(defq);
567 			tipc_group_delete_member(grp, m);
568 			break;
569 		}
570 		if (!update)
571 			continue;
572 
573 		tipc_group_update_rcv_win(grp, blks, node, port, xmitq);
574 	}
575 	return;
576 drop:
577 	kfree_skb(skb);
578 }
579 
580 void tipc_group_update_rcv_win(struct tipc_group *grp, int blks, u32 node,
581 			       u32 port, struct sk_buff_head *xmitq)
582 {
583 	struct list_head *active = &grp->active;
584 	int max_active = grp->max_active;
585 	int reclaim_limit = max_active * 3 / 4;
586 	int active_cnt = grp->active_cnt;
587 	struct tipc_member *m, *rm, *pm;
588 
589 	m = tipc_group_find_member(grp, node, port);
590 	if (!m)
591 		return;
592 
593 	m->advertised -= blks;
594 
595 	switch (m->state) {
596 	case MBR_JOINED:
597 		/* First, decide if member can go active */
598 		if (active_cnt <= max_active) {
599 			m->state = MBR_ACTIVE;
600 			list_add_tail(&m->list, active);
601 			grp->active_cnt++;
602 			tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
603 		} else {
604 			m->state = MBR_PENDING;
605 			list_add_tail(&m->list, &grp->pending);
606 		}
607 
608 		if (active_cnt < reclaim_limit)
609 			break;
610 
611 		/* Reclaim from oldest active member, if possible */
612 		if (!list_empty(active)) {
613 			rm = list_first_entry(active, struct tipc_member, list);
614 			rm->state = MBR_RECLAIMING;
615 			list_del_init(&rm->list);
616 			tipc_group_proto_xmit(grp, rm, GRP_RECLAIM_MSG, xmitq);
617 			break;
618 		}
619 		/* Nobody to reclaim from; - revert oldest pending to JOINED */
620 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
621 		list_del_init(&pm->list);
622 		pm->state = MBR_JOINED;
623 		tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
624 		break;
625 	case MBR_ACTIVE:
626 		if (!list_is_last(&m->list, &grp->active))
627 			list_move_tail(&m->list, &grp->active);
628 		if (m->advertised > (ADV_ACTIVE * 3 / 4))
629 			break;
630 		tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
631 		break;
632 	case MBR_REMITTED:
633 		if (m->advertised > ADV_IDLE)
634 			break;
635 		m->state = MBR_JOINED;
636 		grp->active_cnt--;
637 		if (m->advertised < ADV_IDLE) {
638 			pr_warn_ratelimited("Rcv unexpected msg after REMIT\n");
639 			tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
640 		}
641 
642 		if (list_empty(&grp->pending))
643 			return;
644 
645 		/* Set oldest pending member to active and advertise */
646 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
647 		pm->state = MBR_ACTIVE;
648 		list_move_tail(&pm->list, &grp->active);
649 		grp->active_cnt++;
650 		tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
651 		break;
652 	case MBR_RECLAIMING:
653 	case MBR_JOINING:
654 	case MBR_LEAVING:
655 	default:
656 		break;
657 	}
658 }
659 
660 static void tipc_group_create_event(struct tipc_group *grp,
661 				    struct tipc_member *m,
662 				    u32 event, u16 seqno,
663 				    struct sk_buff_head *inputq)
664 {	u32 dnode = tipc_own_addr(grp->net);
665 	struct tipc_event evt;
666 	struct sk_buff *skb;
667 	struct tipc_msg *hdr;
668 
669 	evt.event = event;
670 	evt.found_lower = m->instance;
671 	evt.found_upper = m->instance;
672 	evt.port.ref = m->port;
673 	evt.port.node = m->node;
674 	evt.s.seq.type = grp->type;
675 	evt.s.seq.lower = m->instance;
676 	evt.s.seq.upper = m->instance;
677 
678 	skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_GRP_MEMBER_EVT,
679 			      GROUP_H_SIZE, sizeof(evt), dnode, m->node,
680 			      grp->portid, m->port, 0);
681 	if (!skb)
682 		return;
683 
684 	hdr = buf_msg(skb);
685 	msg_set_nametype(hdr, grp->type);
686 	msg_set_grp_evt(hdr, event);
687 	msg_set_dest_droppable(hdr, true);
688 	msg_set_grp_bc_seqno(hdr, seqno);
689 	memcpy(msg_data(hdr), &evt, sizeof(evt));
690 	TIPC_SKB_CB(skb)->orig_member = m->instance;
691 	__skb_queue_tail(inputq, skb);
692 }
693 
694 static void tipc_group_proto_xmit(struct tipc_group *grp, struct tipc_member *m,
695 				  int mtyp, struct sk_buff_head *xmitq)
696 {
697 	struct tipc_msg *hdr;
698 	struct sk_buff *skb;
699 	int adv = 0;
700 
701 	skb = tipc_msg_create(GROUP_PROTOCOL, mtyp, INT_H_SIZE, 0,
702 			      m->node, tipc_own_addr(grp->net),
703 			      m->port, grp->portid, 0);
704 	if (!skb)
705 		return;
706 
707 	if (m->state == MBR_ACTIVE)
708 		adv = ADV_ACTIVE - m->advertised;
709 	else if (m->state == MBR_JOINED || m->state == MBR_PENDING)
710 		adv = ADV_IDLE - m->advertised;
711 
712 	hdr = buf_msg(skb);
713 
714 	if (mtyp == GRP_JOIN_MSG) {
715 		msg_set_grp_bc_syncpt(hdr, grp->bc_snd_nxt);
716 		msg_set_adv_win(hdr, adv);
717 		m->advertised += adv;
718 	} else if (mtyp == GRP_LEAVE_MSG) {
719 		msg_set_grp_bc_syncpt(hdr, grp->bc_snd_nxt);
720 	} else if (mtyp == GRP_ADV_MSG) {
721 		msg_set_adv_win(hdr, adv);
722 		m->advertised += adv;
723 	} else if (mtyp == GRP_ACK_MSG) {
724 		msg_set_grp_bc_acked(hdr, m->bc_rcv_nxt);
725 	} else if (mtyp == GRP_REMIT_MSG) {
726 		msg_set_grp_remitted(hdr, m->window);
727 	}
728 	msg_set_dest_droppable(hdr, true);
729 	__skb_queue_tail(xmitq, skb);
730 }
731 
732 void tipc_group_proto_rcv(struct tipc_group *grp, bool *usr_wakeup,
733 			  struct tipc_msg *hdr, struct sk_buff_head *inputq,
734 			  struct sk_buff_head *xmitq)
735 {
736 	u32 node = msg_orignode(hdr);
737 	u32 port = msg_origport(hdr);
738 	struct tipc_member *m, *pm;
739 	u16 remitted, in_flight;
740 
741 	if (!grp)
742 		return;
743 
744 	if (grp->scope == TIPC_NODE_SCOPE && node != tipc_own_addr(grp->net))
745 		return;
746 
747 	m = tipc_group_find_member(grp, node, port);
748 
749 	switch (msg_type(hdr)) {
750 	case GRP_JOIN_MSG:
751 		if (!m)
752 			m = tipc_group_create_member(grp, node, port,
753 						     0, MBR_JOINING);
754 		if (!m)
755 			return;
756 		m->bc_syncpt = msg_grp_bc_syncpt(hdr);
757 		m->bc_rcv_nxt = m->bc_syncpt;
758 		m->window += msg_adv_win(hdr);
759 
760 		/* Wait until PUBLISH event is received if necessary */
761 		if (m->state != MBR_PUBLISHED)
762 			return;
763 
764 		/* Member can be taken into service */
765 		m->state = MBR_JOINED;
766 		tipc_group_open(m, usr_wakeup);
767 		tipc_group_update_member(m, 0);
768 		tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
769 		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
770 					m->bc_syncpt, inputq);
771 		return;
772 	case GRP_LEAVE_MSG:
773 		if (!m)
774 			return;
775 		m->bc_syncpt = msg_grp_bc_syncpt(hdr);
776 		list_del_init(&m->list);
777 		tipc_group_open(m, usr_wakeup);
778 		tipc_group_decr_active(grp, m);
779 		m->state = MBR_LEAVING;
780 		tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
781 					m->bc_syncpt, inputq);
782 		return;
783 	case GRP_ADV_MSG:
784 		if (!m)
785 			return;
786 		m->window += msg_adv_win(hdr);
787 		tipc_group_open(m, usr_wakeup);
788 		return;
789 	case GRP_ACK_MSG:
790 		if (!m)
791 			return;
792 		m->bc_acked = msg_grp_bc_acked(hdr);
793 		if (--grp->bc_ackers)
794 			return;
795 		list_del_init(&m->small_win);
796 		*m->group->open = true;
797 		*usr_wakeup = true;
798 		tipc_group_update_member(m, 0);
799 		return;
800 	case GRP_RECLAIM_MSG:
801 		if (!m)
802 			return;
803 		tipc_group_proto_xmit(grp, m, GRP_REMIT_MSG, xmitq);
804 		m->window = ADV_IDLE;
805 		tipc_group_open(m, usr_wakeup);
806 		return;
807 	case GRP_REMIT_MSG:
808 		if (!m || m->state != MBR_RECLAIMING)
809 			return;
810 
811 		remitted = msg_grp_remitted(hdr);
812 
813 		/* Messages preceding the REMIT still in receive queue */
814 		if (m->advertised > remitted) {
815 			m->state = MBR_REMITTED;
816 			in_flight = m->advertised - remitted;
817 			m->advertised = ADV_IDLE + in_flight;
818 			return;
819 		}
820 		/* This should never happen */
821 		if (m->advertised < remitted)
822 			pr_warn_ratelimited("Unexpected REMIT msg\n");
823 
824 		/* All messages preceding the REMIT have been read */
825 		m->state = MBR_JOINED;
826 		grp->active_cnt--;
827 		m->advertised = ADV_IDLE;
828 
829 		/* Set oldest pending member to active and advertise */
830 		if (list_empty(&grp->pending))
831 			return;
832 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
833 		pm->state = MBR_ACTIVE;
834 		list_move_tail(&pm->list, &grp->active);
835 		grp->active_cnt++;
836 		if (pm->advertised <= (ADV_ACTIVE * 3 / 4))
837 			tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
838 		return;
839 	default:
840 		pr_warn("Received unknown GROUP_PROTO message\n");
841 	}
842 }
843 
844 /* tipc_group_member_evt() - receive and handle a member up/down event
845  */
846 void tipc_group_member_evt(struct tipc_group *grp,
847 			   bool *usr_wakeup,
848 			   int *sk_rcvbuf,
849 			   struct tipc_msg *hdr,
850 			   struct sk_buff_head *inputq,
851 			   struct sk_buff_head *xmitq)
852 {
853 	struct tipc_event *evt = (void *)msg_data(hdr);
854 	u32 instance = evt->found_lower;
855 	u32 node = evt->port.node;
856 	u32 port = evt->port.ref;
857 	int event = evt->event;
858 	struct tipc_member *m;
859 	struct net *net;
860 	u32 self;
861 
862 	if (!grp)
863 		return;
864 
865 	net = grp->net;
866 	self = tipc_own_addr(net);
867 	if (!grp->loopback && node == self && port == grp->portid)
868 		return;
869 
870 	m = tipc_group_find_member(grp, node, port);
871 
872 	switch (event) {
873 	case TIPC_PUBLISHED:
874 		/* Send and wait for arrival of JOIN message if necessary */
875 		if (!m) {
876 			m = tipc_group_create_member(grp, node, port, instance,
877 						     MBR_PUBLISHED);
878 			if (!m)
879 				break;
880 			tipc_group_update_member(m, 0);
881 			tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
882 			break;
883 		}
884 
885 		if (m->state != MBR_JOINING)
886 			break;
887 
888 		/* Member can be taken into service */
889 		m->instance = instance;
890 		m->state = MBR_JOINED;
891 		tipc_group_open(m, usr_wakeup);
892 		tipc_group_update_member(m, 0);
893 		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
894 		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
895 					m->bc_syncpt, inputq);
896 		break;
897 	case TIPC_WITHDRAWN:
898 		if (!m)
899 			break;
900 
901 		tipc_group_decr_active(grp, m);
902 		m->state = MBR_LEAVING;
903 		list_del_init(&m->list);
904 		tipc_group_open(m, usr_wakeup);
905 
906 		/* Only send event if no LEAVE message can be expected */
907 		if (!tipc_node_is_up(net, node))
908 			tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
909 						m->bc_rcv_nxt, inputq);
910 		break;
911 	default:
912 		break;
913 	}
914 	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
915 }
916 
917 int tipc_group_fill_sock_diag(struct tipc_group *grp, struct sk_buff *skb)
918 {
919 	struct nlattr *group = nla_nest_start(skb, TIPC_NLA_SOCK_GROUP);
920 
921 	if (nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_ID,
922 			grp->type) ||
923 	    nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_INSTANCE,
924 			grp->instance) ||
925 	    nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_BC_SEND_NEXT,
926 			grp->bc_snd_nxt))
927 		goto group_msg_cancel;
928 
929 	if (grp->scope == TIPC_NODE_SCOPE)
930 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_NODE_SCOPE))
931 			goto group_msg_cancel;
932 
933 	if (grp->scope == TIPC_CLUSTER_SCOPE)
934 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_CLUSTER_SCOPE))
935 			goto group_msg_cancel;
936 
937 	if (*grp->open)
938 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_OPEN))
939 			goto group_msg_cancel;
940 
941 	nla_nest_end(skb, group);
942 	return 0;
943 
944 group_msg_cancel:
945 	nla_nest_cancel(skb, group);
946 	return -1;
947 }
948