xref: /linux/net/tipc/group.c (revision 24bce201d79807b668bf9d9e0aca801c5c0d5f78)
1 /*
2  * net/tipc/group.c: TIPC group messaging code
3  *
4  * Copyright (c) 2017, Ericsson AB
5  * Copyright (c) 2020, Red Hat Inc
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions are met:
10  *
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the names of the copyright holders nor the names of its
17  *    contributors may be used to endorse or promote products derived from
18  *    this software without specific prior written permission.
19  *
20  * Alternatively, this software may be distributed under the terms of the
21  * GNU General Public License ("GPL") version 2 as published by the Free
22  * Software Foundation.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  */
36 
37 #include "core.h"
38 #include "addr.h"
39 #include "group.h"
40 #include "bcast.h"
41 #include "topsrv.h"
42 #include "msg.h"
43 #include "socket.h"
44 #include "node.h"
45 #include "name_table.h"
46 #include "subscr.h"
47 
48 #define ADV_UNIT (((MAX_MSG_SIZE + MAX_H_SIZE) / FLOWCTL_BLK_SZ) + 1)
49 #define ADV_IDLE ADV_UNIT
50 #define ADV_ACTIVE (ADV_UNIT * 12)
51 
52 enum mbr_state {
53 	MBR_JOINING,
54 	MBR_PUBLISHED,
55 	MBR_JOINED,
56 	MBR_PENDING,
57 	MBR_ACTIVE,
58 	MBR_RECLAIMING,
59 	MBR_REMITTED,
60 	MBR_LEAVING
61 };
62 
63 struct tipc_member {
64 	struct rb_node tree_node;
65 	struct list_head list;
66 	struct list_head small_win;
67 	struct sk_buff_head deferredq;
68 	struct tipc_group *group;
69 	u32 node;
70 	u32 port;
71 	u32 instance;
72 	enum mbr_state state;
73 	u16 advertised;
74 	u16 window;
75 	u16 bc_rcv_nxt;
76 	u16 bc_syncpt;
77 	u16 bc_acked;
78 };
79 
80 struct tipc_group {
81 	struct rb_root members;
82 	struct list_head small_win;
83 	struct list_head pending;
84 	struct list_head active;
85 	struct tipc_nlist dests;
86 	struct net *net;
87 	int subid;
88 	u32 type;
89 	u32 instance;
90 	u32 scope;
91 	u32 portid;
92 	u16 member_cnt;
93 	u16 active_cnt;
94 	u16 max_active;
95 	u16 bc_snd_nxt;
96 	u16 bc_ackers;
97 	bool *open;
98 	bool loopback;
99 	bool events;
100 };
101 
102 static void tipc_group_proto_xmit(struct tipc_group *grp, struct tipc_member *m,
103 				  int mtyp, struct sk_buff_head *xmitq);
104 
105 static void tipc_group_open(struct tipc_member *m, bool *wakeup)
106 {
107 	*wakeup = false;
108 	if (list_empty(&m->small_win))
109 		return;
110 	list_del_init(&m->small_win);
111 	*m->group->open = true;
112 	*wakeup = true;
113 }
114 
115 static void tipc_group_decr_active(struct tipc_group *grp,
116 				   struct tipc_member *m)
117 {
118 	if (m->state == MBR_ACTIVE || m->state == MBR_RECLAIMING ||
119 	    m->state == MBR_REMITTED)
120 		grp->active_cnt--;
121 }
122 
123 static int tipc_group_rcvbuf_limit(struct tipc_group *grp)
124 {
125 	int max_active, active_pool, idle_pool;
126 	int mcnt = grp->member_cnt + 1;
127 
128 	/* Limit simultaneous reception from other members */
129 	max_active = min(mcnt / 8, 64);
130 	max_active = max(max_active, 16);
131 	grp->max_active = max_active;
132 
133 	/* Reserve blocks for active and idle members */
134 	active_pool = max_active * ADV_ACTIVE;
135 	idle_pool = (mcnt - max_active) * ADV_IDLE;
136 
137 	/* Scale to bytes, considering worst-case truesize/msgsize ratio */
138 	return (active_pool + idle_pool) * FLOWCTL_BLK_SZ * 4;
139 }
140 
141 u16 tipc_group_bc_snd_nxt(struct tipc_group *grp)
142 {
143 	return grp->bc_snd_nxt;
144 }
145 
146 static bool tipc_group_is_receiver(struct tipc_member *m)
147 {
148 	return m && m->state != MBR_JOINING && m->state != MBR_LEAVING;
149 }
150 
151 static bool tipc_group_is_sender(struct tipc_member *m)
152 {
153 	return m && m->state != MBR_JOINING && m->state != MBR_PUBLISHED;
154 }
155 
156 u32 tipc_group_exclude(struct tipc_group *grp)
157 {
158 	if (!grp->loopback)
159 		return grp->portid;
160 	return 0;
161 }
162 
163 struct tipc_group *tipc_group_create(struct net *net, u32 portid,
164 				     struct tipc_group_req *mreq,
165 				     bool *group_is_open)
166 {
167 	u32 filter = TIPC_SUB_PORTS | TIPC_SUB_NO_STATUS;
168 	bool global = mreq->scope != TIPC_NODE_SCOPE;
169 	struct tipc_group *grp;
170 	u32 type = mreq->type;
171 
172 	grp = kzalloc(sizeof(*grp), GFP_ATOMIC);
173 	if (!grp)
174 		return NULL;
175 	tipc_nlist_init(&grp->dests, tipc_own_addr(net));
176 	INIT_LIST_HEAD(&grp->small_win);
177 	INIT_LIST_HEAD(&grp->active);
178 	INIT_LIST_HEAD(&grp->pending);
179 	grp->members = RB_ROOT;
180 	grp->net = net;
181 	grp->portid = portid;
182 	grp->type = type;
183 	grp->instance = mreq->instance;
184 	grp->scope = mreq->scope;
185 	grp->loopback = mreq->flags & TIPC_GROUP_LOOPBACK;
186 	grp->events = mreq->flags & TIPC_GROUP_MEMBER_EVTS;
187 	grp->open = group_is_open;
188 	*grp->open = false;
189 	filter |= global ? TIPC_SUB_CLUSTER_SCOPE : TIPC_SUB_NODE_SCOPE;
190 	if (tipc_topsrv_kern_subscr(net, portid, type, 0, ~0,
191 				    filter, &grp->subid))
192 		return grp;
193 	kfree(grp);
194 	return NULL;
195 }
196 
197 void tipc_group_join(struct net *net, struct tipc_group *grp, int *sk_rcvbuf)
198 {
199 	struct rb_root *tree = &grp->members;
200 	struct tipc_member *m, *tmp;
201 	struct sk_buff_head xmitq;
202 
203 	__skb_queue_head_init(&xmitq);
204 	rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
205 		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, &xmitq);
206 		tipc_group_update_member(m, 0);
207 	}
208 	tipc_node_distr_xmit(net, &xmitq);
209 	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
210 }
211 
212 void tipc_group_delete(struct net *net, struct tipc_group *grp)
213 {
214 	struct rb_root *tree = &grp->members;
215 	struct tipc_member *m, *tmp;
216 	struct sk_buff_head xmitq;
217 
218 	__skb_queue_head_init(&xmitq);
219 
220 	rbtree_postorder_for_each_entry_safe(m, tmp, tree, tree_node) {
221 		tipc_group_proto_xmit(grp, m, GRP_LEAVE_MSG, &xmitq);
222 		__skb_queue_purge(&m->deferredq);
223 		list_del(&m->list);
224 		kfree(m);
225 	}
226 	tipc_node_distr_xmit(net, &xmitq);
227 	tipc_nlist_purge(&grp->dests);
228 	tipc_topsrv_kern_unsubscr(net, grp->subid);
229 	kfree(grp);
230 }
231 
232 static struct tipc_member *tipc_group_find_member(struct tipc_group *grp,
233 						  u32 node, u32 port)
234 {
235 	struct rb_node *n = grp->members.rb_node;
236 	u64 nkey, key = (u64)node << 32 | port;
237 	struct tipc_member *m;
238 
239 	while (n) {
240 		m = container_of(n, struct tipc_member, tree_node);
241 		nkey = (u64)m->node << 32 | m->port;
242 		if (key < nkey)
243 			n = n->rb_left;
244 		else if (key > nkey)
245 			n = n->rb_right;
246 		else
247 			return m;
248 	}
249 	return NULL;
250 }
251 
252 static struct tipc_member *tipc_group_find_dest(struct tipc_group *grp,
253 						u32 node, u32 port)
254 {
255 	struct tipc_member *m;
256 
257 	m = tipc_group_find_member(grp, node, port);
258 	if (m && tipc_group_is_receiver(m))
259 		return m;
260 	return NULL;
261 }
262 
263 static struct tipc_member *tipc_group_find_node(struct tipc_group *grp,
264 						u32 node)
265 {
266 	struct tipc_member *m;
267 	struct rb_node *n;
268 
269 	for (n = rb_first(&grp->members); n; n = rb_next(n)) {
270 		m = container_of(n, struct tipc_member, tree_node);
271 		if (m->node == node)
272 			return m;
273 	}
274 	return NULL;
275 }
276 
277 static int tipc_group_add_to_tree(struct tipc_group *grp,
278 				  struct tipc_member *m)
279 {
280 	u64 nkey, key = (u64)m->node << 32 | m->port;
281 	struct rb_node **n, *parent = NULL;
282 	struct tipc_member *tmp;
283 
284 	n = &grp->members.rb_node;
285 	while (*n) {
286 		tmp = container_of(*n, struct tipc_member, tree_node);
287 		parent = *n;
288 		tmp = container_of(parent, struct tipc_member, tree_node);
289 		nkey = (u64)tmp->node << 32 | tmp->port;
290 		if (key < nkey)
291 			n = &(*n)->rb_left;
292 		else if (key > nkey)
293 			n = &(*n)->rb_right;
294 		else
295 			return -EEXIST;
296 	}
297 	rb_link_node(&m->tree_node, parent, n);
298 	rb_insert_color(&m->tree_node, &grp->members);
299 	return 0;
300 }
301 
302 static struct tipc_member *tipc_group_create_member(struct tipc_group *grp,
303 						    u32 node, u32 port,
304 						    u32 instance, int state)
305 {
306 	struct tipc_member *m;
307 	int ret;
308 
309 	m = kzalloc(sizeof(*m), GFP_ATOMIC);
310 	if (!m)
311 		return NULL;
312 	INIT_LIST_HEAD(&m->list);
313 	INIT_LIST_HEAD(&m->small_win);
314 	__skb_queue_head_init(&m->deferredq);
315 	m->group = grp;
316 	m->node = node;
317 	m->port = port;
318 	m->instance = instance;
319 	m->bc_acked = grp->bc_snd_nxt - 1;
320 	ret = tipc_group_add_to_tree(grp, m);
321 	if (ret < 0) {
322 		kfree(m);
323 		return NULL;
324 	}
325 	grp->member_cnt++;
326 	tipc_nlist_add(&grp->dests, m->node);
327 	m->state = state;
328 	return m;
329 }
330 
331 void tipc_group_add_member(struct tipc_group *grp, u32 node,
332 			   u32 port, u32 instance)
333 {
334 	tipc_group_create_member(grp, node, port, instance, MBR_PUBLISHED);
335 }
336 
337 static void tipc_group_delete_member(struct tipc_group *grp,
338 				     struct tipc_member *m)
339 {
340 	rb_erase(&m->tree_node, &grp->members);
341 	grp->member_cnt--;
342 
343 	/* Check if we were waiting for replicast ack from this member */
344 	if (grp->bc_ackers && less(m->bc_acked, grp->bc_snd_nxt - 1))
345 		grp->bc_ackers--;
346 
347 	list_del_init(&m->list);
348 	list_del_init(&m->small_win);
349 	tipc_group_decr_active(grp, m);
350 
351 	/* If last member on a node, remove node from dest list */
352 	if (!tipc_group_find_node(grp, m->node))
353 		tipc_nlist_del(&grp->dests, m->node);
354 
355 	kfree(m);
356 }
357 
358 struct tipc_nlist *tipc_group_dests(struct tipc_group *grp)
359 {
360 	return &grp->dests;
361 }
362 
363 void tipc_group_self(struct tipc_group *grp, struct tipc_service_range *seq,
364 		     int *scope)
365 {
366 	seq->type = grp->type;
367 	seq->lower = grp->instance;
368 	seq->upper = grp->instance;
369 	*scope = grp->scope;
370 }
371 
372 void tipc_group_update_member(struct tipc_member *m, int len)
373 {
374 	struct tipc_group *grp = m->group;
375 	struct tipc_member *_m, *tmp;
376 
377 	if (!tipc_group_is_receiver(m))
378 		return;
379 
380 	m->window -= len;
381 
382 	if (m->window >= ADV_IDLE)
383 		return;
384 
385 	list_del_init(&m->small_win);
386 
387 	/* Sort member into small_window members' list */
388 	list_for_each_entry_safe(_m, tmp, &grp->small_win, small_win) {
389 		if (_m->window > m->window)
390 			break;
391 	}
392 	list_add_tail(&m->small_win, &_m->small_win);
393 }
394 
395 void tipc_group_update_bc_members(struct tipc_group *grp, int len, bool ack)
396 {
397 	u16 prev = grp->bc_snd_nxt - 1;
398 	struct tipc_member *m;
399 	struct rb_node *n;
400 	u16 ackers = 0;
401 
402 	for (n = rb_first(&grp->members); n; n = rb_next(n)) {
403 		m = container_of(n, struct tipc_member, tree_node);
404 		if (tipc_group_is_receiver(m)) {
405 			tipc_group_update_member(m, len);
406 			m->bc_acked = prev;
407 			ackers++;
408 		}
409 	}
410 
411 	/* Mark number of acknowledges to expect, if any */
412 	if (ack)
413 		grp->bc_ackers = ackers;
414 	grp->bc_snd_nxt++;
415 }
416 
417 bool tipc_group_cong(struct tipc_group *grp, u32 dnode, u32 dport,
418 		     int len, struct tipc_member **mbr)
419 {
420 	struct sk_buff_head xmitq;
421 	struct tipc_member *m;
422 	int adv, state;
423 
424 	m = tipc_group_find_dest(grp, dnode, dport);
425 	if (!tipc_group_is_receiver(m)) {
426 		*mbr = NULL;
427 		return false;
428 	}
429 	*mbr = m;
430 
431 	if (m->window >= len)
432 		return false;
433 
434 	*grp->open = false;
435 
436 	/* If not fully advertised, do it now to prevent mutual blocking */
437 	adv = m->advertised;
438 	state = m->state;
439 	if (state == MBR_JOINED && adv == ADV_IDLE)
440 		return true;
441 	if (state == MBR_ACTIVE && adv == ADV_ACTIVE)
442 		return true;
443 	if (state == MBR_PENDING && adv == ADV_IDLE)
444 		return true;
445 	__skb_queue_head_init(&xmitq);
446 	tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, &xmitq);
447 	tipc_node_distr_xmit(grp->net, &xmitq);
448 	return true;
449 }
450 
451 bool tipc_group_bc_cong(struct tipc_group *grp, int len)
452 {
453 	struct tipc_member *m = NULL;
454 
455 	/* If prev bcast was replicast, reject until all receivers have acked */
456 	if (grp->bc_ackers) {
457 		*grp->open = false;
458 		return true;
459 	}
460 	if (list_empty(&grp->small_win))
461 		return false;
462 
463 	m = list_first_entry(&grp->small_win, struct tipc_member, small_win);
464 	if (m->window >= len)
465 		return false;
466 
467 	return tipc_group_cong(grp, m->node, m->port, len, &m);
468 }
469 
470 /* tipc_group_sort_msg() - sort msg into queue by bcast sequence number
471  */
472 static void tipc_group_sort_msg(struct sk_buff *skb, struct sk_buff_head *defq)
473 {
474 	struct tipc_msg *_hdr, *hdr = buf_msg(skb);
475 	u16 bc_seqno = msg_grp_bc_seqno(hdr);
476 	struct sk_buff *_skb, *tmp;
477 	int mtyp = msg_type(hdr);
478 
479 	/* Bcast/mcast may be bypassed by ucast or other bcast, - sort it in */
480 	if (mtyp == TIPC_GRP_BCAST_MSG || mtyp == TIPC_GRP_MCAST_MSG) {
481 		skb_queue_walk_safe(defq, _skb, tmp) {
482 			_hdr = buf_msg(_skb);
483 			if (!less(bc_seqno, msg_grp_bc_seqno(_hdr)))
484 				continue;
485 			__skb_queue_before(defq, _skb, skb);
486 			return;
487 		}
488 		/* Bcast was not bypassed, - add to tail */
489 	}
490 	/* Unicasts are never bypassed, - always add to tail */
491 	__skb_queue_tail(defq, skb);
492 }
493 
494 /* tipc_group_filter_msg() - determine if we should accept arriving message
495  */
496 void tipc_group_filter_msg(struct tipc_group *grp, struct sk_buff_head *inputq,
497 			   struct sk_buff_head *xmitq)
498 {
499 	struct sk_buff *skb = __skb_dequeue(inputq);
500 	bool ack, deliver, update, leave = false;
501 	struct sk_buff_head *defq;
502 	struct tipc_member *m;
503 	struct tipc_msg *hdr;
504 	u32 node, port;
505 	int mtyp, blks;
506 
507 	if (!skb)
508 		return;
509 
510 	hdr = buf_msg(skb);
511 	node =  msg_orignode(hdr);
512 	port = msg_origport(hdr);
513 
514 	if (!msg_in_group(hdr))
515 		goto drop;
516 
517 	m = tipc_group_find_member(grp, node, port);
518 	if (!tipc_group_is_sender(m))
519 		goto drop;
520 
521 	if (less(msg_grp_bc_seqno(hdr), m->bc_rcv_nxt))
522 		goto drop;
523 
524 	TIPC_SKB_CB(skb)->orig_member = m->instance;
525 	defq = &m->deferredq;
526 	tipc_group_sort_msg(skb, defq);
527 
528 	while ((skb = skb_peek(defq))) {
529 		hdr = buf_msg(skb);
530 		mtyp = msg_type(hdr);
531 		blks = msg_blocks(hdr);
532 		deliver = true;
533 		ack = false;
534 		update = false;
535 
536 		if (more(msg_grp_bc_seqno(hdr), m->bc_rcv_nxt))
537 			break;
538 
539 		/* Decide what to do with message */
540 		switch (mtyp) {
541 		case TIPC_GRP_MCAST_MSG:
542 			if (msg_nameinst(hdr) != grp->instance) {
543 				update = true;
544 				deliver = false;
545 			}
546 			fallthrough;
547 		case TIPC_GRP_BCAST_MSG:
548 			m->bc_rcv_nxt++;
549 			ack = msg_grp_bc_ack_req(hdr);
550 			break;
551 		case TIPC_GRP_UCAST_MSG:
552 			break;
553 		case TIPC_GRP_MEMBER_EVT:
554 			if (m->state == MBR_LEAVING)
555 				leave = true;
556 			if (!grp->events)
557 				deliver = false;
558 			break;
559 		default:
560 			break;
561 		}
562 
563 		/* Execute decisions */
564 		__skb_dequeue(defq);
565 		if (deliver)
566 			__skb_queue_tail(inputq, skb);
567 		else
568 			kfree_skb(skb);
569 
570 		if (ack)
571 			tipc_group_proto_xmit(grp, m, GRP_ACK_MSG, xmitq);
572 
573 		if (leave) {
574 			__skb_queue_purge(defq);
575 			tipc_group_delete_member(grp, m);
576 			break;
577 		}
578 		if (!update)
579 			continue;
580 
581 		tipc_group_update_rcv_win(grp, blks, node, port, xmitq);
582 	}
583 	return;
584 drop:
585 	kfree_skb(skb);
586 }
587 
588 void tipc_group_update_rcv_win(struct tipc_group *grp, int blks, u32 node,
589 			       u32 port, struct sk_buff_head *xmitq)
590 {
591 	struct list_head *active = &grp->active;
592 	int max_active = grp->max_active;
593 	int reclaim_limit = max_active * 3 / 4;
594 	int active_cnt = grp->active_cnt;
595 	struct tipc_member *m, *rm, *pm;
596 
597 	m = tipc_group_find_member(grp, node, port);
598 	if (!m)
599 		return;
600 
601 	m->advertised -= blks;
602 
603 	switch (m->state) {
604 	case MBR_JOINED:
605 		/* First, decide if member can go active */
606 		if (active_cnt <= max_active) {
607 			m->state = MBR_ACTIVE;
608 			list_add_tail(&m->list, active);
609 			grp->active_cnt++;
610 			tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
611 		} else {
612 			m->state = MBR_PENDING;
613 			list_add_tail(&m->list, &grp->pending);
614 		}
615 
616 		if (active_cnt < reclaim_limit)
617 			break;
618 
619 		/* Reclaim from oldest active member, if possible */
620 		if (!list_empty(active)) {
621 			rm = list_first_entry(active, struct tipc_member, list);
622 			rm->state = MBR_RECLAIMING;
623 			list_del_init(&rm->list);
624 			tipc_group_proto_xmit(grp, rm, GRP_RECLAIM_MSG, xmitq);
625 			break;
626 		}
627 		/* Nobody to reclaim from; - revert oldest pending to JOINED */
628 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
629 		list_del_init(&pm->list);
630 		pm->state = MBR_JOINED;
631 		tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
632 		break;
633 	case MBR_ACTIVE:
634 		if (!list_is_last(&m->list, &grp->active))
635 			list_move_tail(&m->list, &grp->active);
636 		if (m->advertised > (ADV_ACTIVE * 3 / 4))
637 			break;
638 		tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
639 		break;
640 	case MBR_REMITTED:
641 		if (m->advertised > ADV_IDLE)
642 			break;
643 		m->state = MBR_JOINED;
644 		grp->active_cnt--;
645 		if (m->advertised < ADV_IDLE) {
646 			pr_warn_ratelimited("Rcv unexpected msg after REMIT\n");
647 			tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
648 		}
649 
650 		if (list_empty(&grp->pending))
651 			return;
652 
653 		/* Set oldest pending member to active and advertise */
654 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
655 		pm->state = MBR_ACTIVE;
656 		list_move_tail(&pm->list, &grp->active);
657 		grp->active_cnt++;
658 		tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
659 		break;
660 	case MBR_RECLAIMING:
661 	case MBR_JOINING:
662 	case MBR_LEAVING:
663 	default:
664 		break;
665 	}
666 }
667 
668 static void tipc_group_create_event(struct tipc_group *grp,
669 				    struct tipc_member *m,
670 				    u32 event, u16 seqno,
671 				    struct sk_buff_head *inputq)
672 {	u32 dnode = tipc_own_addr(grp->net);
673 	struct tipc_event evt;
674 	struct sk_buff *skb;
675 	struct tipc_msg *hdr;
676 
677 	memset(&evt, 0, sizeof(evt));
678 	evt.event = event;
679 	evt.found_lower = m->instance;
680 	evt.found_upper = m->instance;
681 	evt.port.ref = m->port;
682 	evt.port.node = m->node;
683 	evt.s.seq.type = grp->type;
684 	evt.s.seq.lower = m->instance;
685 	evt.s.seq.upper = m->instance;
686 
687 	skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_GRP_MEMBER_EVT,
688 			      GROUP_H_SIZE, sizeof(evt), dnode, m->node,
689 			      grp->portid, m->port, 0);
690 	if (!skb)
691 		return;
692 
693 	hdr = buf_msg(skb);
694 	msg_set_nametype(hdr, grp->type);
695 	msg_set_grp_evt(hdr, event);
696 	msg_set_dest_droppable(hdr, true);
697 	msg_set_grp_bc_seqno(hdr, seqno);
698 	memcpy(msg_data(hdr), &evt, sizeof(evt));
699 	TIPC_SKB_CB(skb)->orig_member = m->instance;
700 	__skb_queue_tail(inputq, skb);
701 }
702 
703 static void tipc_group_proto_xmit(struct tipc_group *grp, struct tipc_member *m,
704 				  int mtyp, struct sk_buff_head *xmitq)
705 {
706 	struct tipc_msg *hdr;
707 	struct sk_buff *skb;
708 	int adv = 0;
709 
710 	skb = tipc_msg_create(GROUP_PROTOCOL, mtyp, INT_H_SIZE, 0,
711 			      m->node, tipc_own_addr(grp->net),
712 			      m->port, grp->portid, 0);
713 	if (!skb)
714 		return;
715 
716 	if (m->state == MBR_ACTIVE)
717 		adv = ADV_ACTIVE - m->advertised;
718 	else if (m->state == MBR_JOINED || m->state == MBR_PENDING)
719 		adv = ADV_IDLE - m->advertised;
720 
721 	hdr = buf_msg(skb);
722 
723 	if (mtyp == GRP_JOIN_MSG) {
724 		msg_set_grp_bc_syncpt(hdr, grp->bc_snd_nxt);
725 		msg_set_adv_win(hdr, adv);
726 		m->advertised += adv;
727 	} else if (mtyp == GRP_LEAVE_MSG) {
728 		msg_set_grp_bc_syncpt(hdr, grp->bc_snd_nxt);
729 	} else if (mtyp == GRP_ADV_MSG) {
730 		msg_set_adv_win(hdr, adv);
731 		m->advertised += adv;
732 	} else if (mtyp == GRP_ACK_MSG) {
733 		msg_set_grp_bc_acked(hdr, m->bc_rcv_nxt);
734 	} else if (mtyp == GRP_REMIT_MSG) {
735 		msg_set_grp_remitted(hdr, m->window);
736 	}
737 	msg_set_dest_droppable(hdr, true);
738 	__skb_queue_tail(xmitq, skb);
739 }
740 
741 void tipc_group_proto_rcv(struct tipc_group *grp, bool *usr_wakeup,
742 			  struct tipc_msg *hdr, struct sk_buff_head *inputq,
743 			  struct sk_buff_head *xmitq)
744 {
745 	u32 node = msg_orignode(hdr);
746 	u32 port = msg_origport(hdr);
747 	struct tipc_member *m, *pm;
748 	u16 remitted, in_flight;
749 
750 	if (!grp)
751 		return;
752 
753 	if (grp->scope == TIPC_NODE_SCOPE && node != tipc_own_addr(grp->net))
754 		return;
755 
756 	m = tipc_group_find_member(grp, node, port);
757 
758 	switch (msg_type(hdr)) {
759 	case GRP_JOIN_MSG:
760 		if (!m)
761 			m = tipc_group_create_member(grp, node, port,
762 						     0, MBR_JOINING);
763 		if (!m)
764 			return;
765 		m->bc_syncpt = msg_grp_bc_syncpt(hdr);
766 		m->bc_rcv_nxt = m->bc_syncpt;
767 		m->window += msg_adv_win(hdr);
768 
769 		/* Wait until PUBLISH event is received if necessary */
770 		if (m->state != MBR_PUBLISHED)
771 			return;
772 
773 		/* Member can be taken into service */
774 		m->state = MBR_JOINED;
775 		tipc_group_open(m, usr_wakeup);
776 		tipc_group_update_member(m, 0);
777 		tipc_group_proto_xmit(grp, m, GRP_ADV_MSG, xmitq);
778 		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
779 					m->bc_syncpt, inputq);
780 		return;
781 	case GRP_LEAVE_MSG:
782 		if (!m)
783 			return;
784 		m->bc_syncpt = msg_grp_bc_syncpt(hdr);
785 		list_del_init(&m->list);
786 		tipc_group_open(m, usr_wakeup);
787 		tipc_group_decr_active(grp, m);
788 		m->state = MBR_LEAVING;
789 		tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
790 					m->bc_syncpt, inputq);
791 		return;
792 	case GRP_ADV_MSG:
793 		if (!m)
794 			return;
795 		m->window += msg_adv_win(hdr);
796 		tipc_group_open(m, usr_wakeup);
797 		return;
798 	case GRP_ACK_MSG:
799 		if (!m)
800 			return;
801 		m->bc_acked = msg_grp_bc_acked(hdr);
802 		if (--grp->bc_ackers)
803 			return;
804 		list_del_init(&m->small_win);
805 		*m->group->open = true;
806 		*usr_wakeup = true;
807 		tipc_group_update_member(m, 0);
808 		return;
809 	case GRP_RECLAIM_MSG:
810 		if (!m)
811 			return;
812 		tipc_group_proto_xmit(grp, m, GRP_REMIT_MSG, xmitq);
813 		m->window = ADV_IDLE;
814 		tipc_group_open(m, usr_wakeup);
815 		return;
816 	case GRP_REMIT_MSG:
817 		if (!m || m->state != MBR_RECLAIMING)
818 			return;
819 
820 		remitted = msg_grp_remitted(hdr);
821 
822 		/* Messages preceding the REMIT still in receive queue */
823 		if (m->advertised > remitted) {
824 			m->state = MBR_REMITTED;
825 			in_flight = m->advertised - remitted;
826 			m->advertised = ADV_IDLE + in_flight;
827 			return;
828 		}
829 		/* This should never happen */
830 		if (m->advertised < remitted)
831 			pr_warn_ratelimited("Unexpected REMIT msg\n");
832 
833 		/* All messages preceding the REMIT have been read */
834 		m->state = MBR_JOINED;
835 		grp->active_cnt--;
836 		m->advertised = ADV_IDLE;
837 
838 		/* Set oldest pending member to active and advertise */
839 		if (list_empty(&grp->pending))
840 			return;
841 		pm = list_first_entry(&grp->pending, struct tipc_member, list);
842 		pm->state = MBR_ACTIVE;
843 		list_move_tail(&pm->list, &grp->active);
844 		grp->active_cnt++;
845 		if (pm->advertised <= (ADV_ACTIVE * 3 / 4))
846 			tipc_group_proto_xmit(grp, pm, GRP_ADV_MSG, xmitq);
847 		return;
848 	default:
849 		pr_warn("Received unknown GROUP_PROTO message\n");
850 	}
851 }
852 
853 /* tipc_group_member_evt() - receive and handle a member up/down event
854  */
855 void tipc_group_member_evt(struct tipc_group *grp,
856 			   bool *usr_wakeup,
857 			   int *sk_rcvbuf,
858 			   struct tipc_msg *hdr,
859 			   struct sk_buff_head *inputq,
860 			   struct sk_buff_head *xmitq)
861 {
862 	struct tipc_event *evt = (void *)msg_data(hdr);
863 	u32 instance = evt->found_lower;
864 	u32 node = evt->port.node;
865 	u32 port = evt->port.ref;
866 	int event = evt->event;
867 	struct tipc_member *m;
868 	struct net *net;
869 	u32 self;
870 
871 	if (!grp)
872 		return;
873 
874 	net = grp->net;
875 	self = tipc_own_addr(net);
876 	if (!grp->loopback && node == self && port == grp->portid)
877 		return;
878 
879 	m = tipc_group_find_member(grp, node, port);
880 
881 	switch (event) {
882 	case TIPC_PUBLISHED:
883 		/* Send and wait for arrival of JOIN message if necessary */
884 		if (!m) {
885 			m = tipc_group_create_member(grp, node, port, instance,
886 						     MBR_PUBLISHED);
887 			if (!m)
888 				break;
889 			tipc_group_update_member(m, 0);
890 			tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
891 			break;
892 		}
893 
894 		if (m->state != MBR_JOINING)
895 			break;
896 
897 		/* Member can be taken into service */
898 		m->instance = instance;
899 		m->state = MBR_JOINED;
900 		tipc_group_open(m, usr_wakeup);
901 		tipc_group_update_member(m, 0);
902 		tipc_group_proto_xmit(grp, m, GRP_JOIN_MSG, xmitq);
903 		tipc_group_create_event(grp, m, TIPC_PUBLISHED,
904 					m->bc_syncpt, inputq);
905 		break;
906 	case TIPC_WITHDRAWN:
907 		if (!m)
908 			break;
909 
910 		tipc_group_decr_active(grp, m);
911 		m->state = MBR_LEAVING;
912 		list_del_init(&m->list);
913 		tipc_group_open(m, usr_wakeup);
914 
915 		/* Only send event if no LEAVE message can be expected */
916 		if (!tipc_node_is_up(net, node))
917 			tipc_group_create_event(grp, m, TIPC_WITHDRAWN,
918 						m->bc_rcv_nxt, inputq);
919 		break;
920 	default:
921 		break;
922 	}
923 	*sk_rcvbuf = tipc_group_rcvbuf_limit(grp);
924 }
925 
926 int tipc_group_fill_sock_diag(struct tipc_group *grp, struct sk_buff *skb)
927 {
928 	struct nlattr *group = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_GROUP);
929 
930 	if (!group)
931 		return -EMSGSIZE;
932 
933 	if (nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_ID,
934 			grp->type) ||
935 	    nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_INSTANCE,
936 			grp->instance) ||
937 	    nla_put_u32(skb, TIPC_NLA_SOCK_GROUP_BC_SEND_NEXT,
938 			grp->bc_snd_nxt))
939 		goto group_msg_cancel;
940 
941 	if (grp->scope == TIPC_NODE_SCOPE)
942 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_NODE_SCOPE))
943 			goto group_msg_cancel;
944 
945 	if (grp->scope == TIPC_CLUSTER_SCOPE)
946 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_CLUSTER_SCOPE))
947 			goto group_msg_cancel;
948 
949 	if (*grp->open)
950 		if (nla_put_flag(skb, TIPC_NLA_SOCK_GROUP_OPEN))
951 			goto group_msg_cancel;
952 
953 	nla_nest_end(skb, group);
954 	return 0;
955 
956 group_msg_cancel:
957 	nla_nest_cancel(skb, group);
958 	return -1;
959 }
960