Linux Audio

Check our new training course

Loading...
v6.8
  1/*
  2 * net/tipc/monitor.c
  3 *
  4 * Copyright (c) 2016, 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 <net/genetlink.h>
 37#include "core.h"
 38#include "addr.h"
 39#include "monitor.h"
 40#include "bearer.h"
 41
 42#define MAX_MON_DOMAIN       64
 43#define MON_TIMEOUT          120000
 44#define MAX_PEER_DOWN_EVENTS 4
 45
 46/* struct tipc_mon_domain: domain record to be transferred between peers
 47 * @len: actual size of domain record
 48 * @gen: current generation of sender's domain
 49 * @ack_gen: most recent generation of self's domain acked by peer
 50 * @member_cnt: number of domain member nodes described in this record
 51 * @up_map: bit map indicating which of the members the sender considers up
 52 * @members: identity of the domain members
 53 */
 54struct tipc_mon_domain {
 55	u16 len;
 56	u16 gen;
 57	u16 ack_gen;
 58	u16 member_cnt;
 59	u64 up_map;
 60	u32 members[MAX_MON_DOMAIN];
 61};
 62
 63/* struct tipc_peer: state of a peer node and its domain
 64 * @addr: tipc node identity of peer
 65 * @head_map: shows which other nodes currently consider peer 'up'
 66 * @domain: most recent domain record from peer
 67 * @hash: position in hashed lookup list
 68 * @list: position in linked list, in circular ascending order by 'addr'
 69 * @applied: number of reported domain members applied on this monitor list
 70 * @is_up: peer is up as seen from this node
 71 * @is_head: peer is assigned domain head as seen from this node
 72 * @is_local: peer is in local domain and should be continuously monitored
 73 * @down_cnt: - numbers of other peers which have reported this on lost
 74 */
 75struct tipc_peer {
 76	u32 addr;
 77	struct tipc_mon_domain *domain;
 78	struct hlist_node hash;
 79	struct list_head list;
 80	u8 applied;
 81	u8 down_cnt;
 82	bool is_up;
 83	bool is_head;
 84	bool is_local;
 85};
 86
 87struct tipc_monitor {
 88	struct hlist_head peers[NODE_HTABLE_SIZE];
 89	int peer_cnt;
 90	struct tipc_peer *self;
 91	rwlock_t lock;
 92	struct tipc_mon_domain cache;
 93	u16 list_gen;
 94	u16 dom_gen;
 95	struct net *net;
 96	struct timer_list timer;
 97	unsigned long timer_intv;
 98};
 99
100static struct tipc_monitor *tipc_monitor(struct net *net, int bearer_id)
101{
102	return tipc_net(net)->monitors[bearer_id];
103}
104
105const int tipc_max_domain_size = sizeof(struct tipc_mon_domain);
106
107static inline u16 mon_cpu_to_le16(u16 val)
108{
109	return (__force __u16)htons(val);
110}
111
112static inline u32 mon_cpu_to_le32(u32 val)
113{
114	return (__force __u32)htonl(val);
115}
116
117static inline u64 mon_cpu_to_le64(u64 val)
118{
119	return (__force __u64)cpu_to_be64(val);
120}
121
122static inline u16 mon_le16_to_cpu(u16 val)
123{
124	return ntohs((__force __be16)val);
125}
126
127static inline u32 mon_le32_to_cpu(u32 val)
128{
129	return ntohl((__force __be32)val);
130}
131
132static inline u64 mon_le64_to_cpu(u64 val)
133{
134	return be64_to_cpu((__force __be64)val);
135}
136
137/* dom_rec_len(): actual length of domain record for transport
138 */
139static int dom_rec_len(struct tipc_mon_domain *dom, u16 mcnt)
140{
141	return (offsetof(struct tipc_mon_domain, members)) + (mcnt * sizeof(u32));
142}
143
144/* dom_size() : calculate size of own domain based on number of peers
145 */
146static int dom_size(int peers)
147{
148	int i = 0;
149
150	while ((i * i) < peers)
151		i++;
152	return i < MAX_MON_DOMAIN ? i : MAX_MON_DOMAIN;
153}
154
155static void map_set(u64 *up_map, int i, unsigned int v)
156{
157	*up_map &= ~(1ULL << i);
158	*up_map |= ((u64)v << i);
159}
160
161static int map_get(u64 up_map, int i)
162{
163	return (up_map & (1ULL << i)) >> i;
164}
165
166static struct tipc_peer *peer_prev(struct tipc_peer *peer)
167{
168	return list_last_entry(&peer->list, struct tipc_peer, list);
169}
170
171static struct tipc_peer *peer_nxt(struct tipc_peer *peer)
172{
173	return list_first_entry(&peer->list, struct tipc_peer, list);
174}
175
176static struct tipc_peer *peer_head(struct tipc_peer *peer)
177{
178	while (!peer->is_head)
179		peer = peer_prev(peer);
180	return peer;
181}
182
183static struct tipc_peer *get_peer(struct tipc_monitor *mon, u32 addr)
184{
185	struct tipc_peer *peer;
186	unsigned int thash = tipc_hashfn(addr);
187
188	hlist_for_each_entry(peer, &mon->peers[thash], hash) {
189		if (peer->addr == addr)
190			return peer;
191	}
192	return NULL;
193}
194
195static struct tipc_peer *get_self(struct net *net, int bearer_id)
196{
197	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
198
199	return mon->self;
200}
201
202static inline bool tipc_mon_is_active(struct net *net, struct tipc_monitor *mon)
203{
204	struct tipc_net *tn = tipc_net(net);
205
206	return mon->peer_cnt > tn->mon_threshold;
207}
208
209/* mon_identify_lost_members() : - identify amd mark potentially lost members
210 */
211static void mon_identify_lost_members(struct tipc_peer *peer,
212				      struct tipc_mon_domain *dom_bef,
213				      int applied_bef)
214{
215	struct tipc_peer *member = peer;
216	struct tipc_mon_domain *dom_aft = peer->domain;
217	int applied_aft = peer->applied;
218	int i;
219
220	for (i = 0; i < applied_bef; i++) {
221		member = peer_nxt(member);
222
223		/* Do nothing if self or peer already see member as down */
224		if (!member->is_up || !map_get(dom_bef->up_map, i))
225			continue;
226
227		/* Loss of local node must be detected by active probing */
228		if (member->is_local)
229			continue;
230
231		/* Start probing if member was removed from applied domain */
232		if (!applied_aft || (applied_aft < i)) {
233			member->down_cnt = 1;
234			continue;
235		}
236
237		/* Member loss is confirmed if it is still in applied domain */
238		if (!map_get(dom_aft->up_map, i))
239			member->down_cnt++;
240	}
241}
242
243/* mon_apply_domain() : match a peer's domain record against monitor list
244 */
245static void mon_apply_domain(struct tipc_monitor *mon,
246			     struct tipc_peer *peer)
247{
248	struct tipc_mon_domain *dom = peer->domain;
249	struct tipc_peer *member;
250	u32 addr;
251	int i;
252
253	if (!dom || !peer->is_up)
254		return;
255
256	/* Scan across domain members and match against monitor list */
257	peer->applied = 0;
258	member = peer_nxt(peer);
259	for (i = 0; i < dom->member_cnt; i++) {
260		addr = dom->members[i];
261		if (addr != member->addr)
262			return;
263		peer->applied++;
264		member = peer_nxt(member);
265	}
266}
267
268/* mon_update_local_domain() : update after peer addition/removal/up/down
269 */
270static void mon_update_local_domain(struct tipc_monitor *mon)
271{
272	struct tipc_peer *self = mon->self;
273	struct tipc_mon_domain *cache = &mon->cache;
274	struct tipc_mon_domain *dom = self->domain;
275	struct tipc_peer *peer = self;
276	u64 prev_up_map = dom->up_map;
277	u16 member_cnt, i;
278	bool diff;
279
280	/* Update local domain size based on current size of cluster */
281	member_cnt = dom_size(mon->peer_cnt) - 1;
282	self->applied = member_cnt;
283
284	/* Update native and cached outgoing local domain records */
285	dom->len = dom_rec_len(dom, member_cnt);
286	diff = dom->member_cnt != member_cnt;
287	dom->member_cnt = member_cnt;
288	for (i = 0; i < member_cnt; i++) {
289		peer = peer_nxt(peer);
290		diff |= dom->members[i] != peer->addr;
291		dom->members[i] = peer->addr;
292		map_set(&dom->up_map, i, peer->is_up);
293		cache->members[i] = mon_cpu_to_le32(peer->addr);
294	}
295	diff |= dom->up_map != prev_up_map;
296	if (!diff)
297		return;
298	dom->gen = ++mon->dom_gen;
299	cache->len = mon_cpu_to_le16(dom->len);
300	cache->gen = mon_cpu_to_le16(dom->gen);
301	cache->member_cnt = mon_cpu_to_le16(member_cnt);
302	cache->up_map = mon_cpu_to_le64(dom->up_map);
303	mon_apply_domain(mon, self);
304}
305
306/* mon_update_neighbors() : update preceding neighbors of added/removed peer
307 */
308static void mon_update_neighbors(struct tipc_monitor *mon,
309				 struct tipc_peer *peer)
310{
311	int dz, i;
312
313	dz = dom_size(mon->peer_cnt);
314	for (i = 0; i < dz; i++) {
315		mon_apply_domain(mon, peer);
316		peer = peer_prev(peer);
317	}
318}
319
320/* mon_assign_roles() : reassign peer roles after a network change
321 * The monitor list is consistent at this stage; i.e., each peer is monitoring
322 * a set of domain members as matched between domain record and the monitor list
323 */
324static void mon_assign_roles(struct tipc_monitor *mon, struct tipc_peer *head)
325{
326	struct tipc_peer *peer = peer_nxt(head);
327	struct tipc_peer *self = mon->self;
328	int i = 0;
329
330	for (; peer != self; peer = peer_nxt(peer)) {
331		peer->is_local = false;
332
333		/* Update domain member */
334		if (i++ < head->applied) {
335			peer->is_head = false;
336			if (head == self)
337				peer->is_local = true;
338			continue;
339		}
340		/* Assign next domain head */
341		if (!peer->is_up)
342			continue;
343		if (peer->is_head)
344			break;
345		head = peer;
346		head->is_head = true;
347		i = 0;
348	}
349	mon->list_gen++;
350}
351
352void tipc_mon_remove_peer(struct net *net, u32 addr, int bearer_id)
353{
354	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
355	struct tipc_peer *self;
356	struct tipc_peer *peer, *prev, *head;
357
358	if (!mon)
359		return;
360
361	self = get_self(net, bearer_id);
362	write_lock_bh(&mon->lock);
363	peer = get_peer(mon, addr);
364	if (!peer)
365		goto exit;
366	prev = peer_prev(peer);
367	list_del(&peer->list);
368	hlist_del(&peer->hash);
369	kfree(peer->domain);
370	kfree(peer);
371	mon->peer_cnt--;
372	head = peer_head(prev);
373	if (head == self)
374		mon_update_local_domain(mon);
375	mon_update_neighbors(mon, prev);
376
377	/* Revert to full-mesh monitoring if we reach threshold */
378	if (!tipc_mon_is_active(net, mon)) {
379		list_for_each_entry(peer, &self->list, list) {
380			kfree(peer->domain);
381			peer->domain = NULL;
382			peer->applied = 0;
383		}
384	}
385	mon_assign_roles(mon, head);
386exit:
387	write_unlock_bh(&mon->lock);
388}
389
390static bool tipc_mon_add_peer(struct tipc_monitor *mon, u32 addr,
391			      struct tipc_peer **peer)
392{
393	struct tipc_peer *self = mon->self;
394	struct tipc_peer *cur, *prev, *p;
395
396	p = kzalloc(sizeof(*p), GFP_ATOMIC);
397	*peer = p;
398	if (!p)
399		return false;
400	p->addr = addr;
401
402	/* Add new peer to lookup list */
403	INIT_LIST_HEAD(&p->list);
404	hlist_add_head(&p->hash, &mon->peers[tipc_hashfn(addr)]);
405
406	/* Sort new peer into iterator list, in ascending circular order */
407	prev = self;
408	list_for_each_entry(cur, &self->list, list) {
409		if ((addr > prev->addr) && (addr < cur->addr))
410			break;
411		if (((addr < cur->addr) || (addr > prev->addr)) &&
412		    (prev->addr > cur->addr))
413			break;
414		prev = cur;
415	}
416	list_add_tail(&p->list, &cur->list);
417	mon->peer_cnt++;
418	mon_update_neighbors(mon, p);
419	return true;
420}
421
422void tipc_mon_peer_up(struct net *net, u32 addr, int bearer_id)
423{
424	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
425	struct tipc_peer *self = get_self(net, bearer_id);
426	struct tipc_peer *peer, *head;
427
428	write_lock_bh(&mon->lock);
429	peer = get_peer(mon, addr);
430	if (!peer && !tipc_mon_add_peer(mon, addr, &peer))
431		goto exit;
432	peer->is_up = true;
433	head = peer_head(peer);
434	if (head == self)
435		mon_update_local_domain(mon);
436	mon_assign_roles(mon, head);
437exit:
438	write_unlock_bh(&mon->lock);
439}
440
441void tipc_mon_peer_down(struct net *net, u32 addr, int bearer_id)
442{
443	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
444	struct tipc_peer *self;
445	struct tipc_peer *peer, *head;
446	struct tipc_mon_domain *dom;
447	int applied;
448
449	if (!mon)
450		return;
451
452	self = get_self(net, bearer_id);
453	write_lock_bh(&mon->lock);
454	peer = get_peer(mon, addr);
455	if (!peer) {
456		pr_warn("Mon: unknown link %x/%u DOWN\n", addr, bearer_id);
457		goto exit;
458	}
459	applied = peer->applied;
460	peer->applied = 0;
461	dom = peer->domain;
462	peer->domain = NULL;
463	if (peer->is_head)
464		mon_identify_lost_members(peer, dom, applied);
465	kfree(dom);
466	peer->is_up = false;
467	peer->is_head = false;
468	peer->is_local = false;
469	peer->down_cnt = 0;
470	head = peer_head(peer);
471	if (head == self)
472		mon_update_local_domain(mon);
473	mon_assign_roles(mon, head);
474exit:
475	write_unlock_bh(&mon->lock);
476}
477
478/* tipc_mon_rcv - process monitor domain event message
479 */
480void tipc_mon_rcv(struct net *net, void *data, u16 dlen, u32 addr,
481		  struct tipc_mon_state *state, int bearer_id)
482{
483	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
484	struct tipc_mon_domain *arrv_dom = data;
485	struct tipc_mon_domain dom_bef;
486	struct tipc_mon_domain *dom;
487	struct tipc_peer *peer;
488	u16 new_member_cnt = mon_le16_to_cpu(arrv_dom->member_cnt);
489	int new_dlen = dom_rec_len(arrv_dom, new_member_cnt);
490	u16 new_gen = mon_le16_to_cpu(arrv_dom->gen);
491	u16 acked_gen = mon_le16_to_cpu(arrv_dom->ack_gen);
492	u16 arrv_dlen = mon_le16_to_cpu(arrv_dom->len);
493	bool probing = state->probing;
494	int i, applied_bef;
495
496	state->probing = false;
497
498	/* Sanity check received domain record */
499	if (new_member_cnt > MAX_MON_DOMAIN)
500		return;
501	if (dlen < dom_rec_len(arrv_dom, 0))
502		return;
503	if (dlen != dom_rec_len(arrv_dom, new_member_cnt))
504		return;
505	if (dlen < new_dlen || arrv_dlen != new_dlen)
506		return;
507
508	/* Synch generation numbers with peer if link just came up */
509	if (!state->synched) {
510		state->peer_gen = new_gen - 1;
511		state->acked_gen = acked_gen;
512		state->synched = true;
513	}
514
515	if (more(acked_gen, state->acked_gen))
516		state->acked_gen = acked_gen;
517
518	/* Drop duplicate unless we are waiting for a probe response */
519	if (!more(new_gen, state->peer_gen) && !probing)
520		return;
521
522	write_lock_bh(&mon->lock);
523	peer = get_peer(mon, addr);
524	if (!peer || !peer->is_up)
525		goto exit;
526
527	/* Peer is confirmed, stop any ongoing probing */
528	peer->down_cnt = 0;
529
530	/* Task is done for duplicate record */
531	if (!more(new_gen, state->peer_gen))
532		goto exit;
533
534	state->peer_gen = new_gen;
535
536	/* Cache current domain record for later use */
537	dom_bef.member_cnt = 0;
538	dom = peer->domain;
539	if (dom)
540		memcpy(&dom_bef, dom, dom->len);
541
542	/* Transform and store received domain record */
543	if (!dom || (dom->len < new_dlen)) {
544		kfree(dom);
545		dom = kmalloc(new_dlen, GFP_ATOMIC);
546		peer->domain = dom;
547		if (!dom)
548			goto exit;
549	}
550	dom->len = new_dlen;
551	dom->gen = new_gen;
552	dom->member_cnt = new_member_cnt;
553	dom->up_map = mon_le64_to_cpu(arrv_dom->up_map);
554	for (i = 0; i < new_member_cnt; i++)
555		dom->members[i] = mon_le32_to_cpu(arrv_dom->members[i]);
556
557	/* Update peers affected by this domain record */
558	applied_bef = peer->applied;
559	mon_apply_domain(mon, peer);
560	mon_identify_lost_members(peer, &dom_bef, applied_bef);
561	mon_assign_roles(mon, peer_head(peer));
562exit:
563	write_unlock_bh(&mon->lock);
564}
565
566void tipc_mon_prep(struct net *net, void *data, int *dlen,
567		   struct tipc_mon_state *state, int bearer_id)
568{
569	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
570	struct tipc_mon_domain *dom = data;
571	u16 gen = mon->dom_gen;
572	u16 len;
573
574	/* Send invalid record if not active */
575	if (!tipc_mon_is_active(net, mon)) {
576		dom->len = 0;
577		return;
578	}
579
580	/* Send only a dummy record with ack if peer has acked our last sent */
581	if (likely(state->acked_gen == gen)) {
582		len = dom_rec_len(dom, 0);
583		*dlen = len;
584		dom->len = mon_cpu_to_le16(len);
585		dom->gen = mon_cpu_to_le16(gen);
586		dom->ack_gen = mon_cpu_to_le16(state->peer_gen);
587		dom->member_cnt = 0;
588		return;
589	}
590	/* Send the full record */
591	read_lock_bh(&mon->lock);
592	len = mon_le16_to_cpu(mon->cache.len);
593	*dlen = len;
594	memcpy(data, &mon->cache, len);
595	read_unlock_bh(&mon->lock);
596	dom->ack_gen = mon_cpu_to_le16(state->peer_gen);
597}
598
599void tipc_mon_get_state(struct net *net, u32 addr,
600			struct tipc_mon_state *state,
601			int bearer_id)
602{
603	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
604	struct tipc_peer *peer;
605
606	if (!tipc_mon_is_active(net, mon)) {
607		state->probing = false;
608		state->monitoring = true;
609		return;
610	}
611
612	/* Used cached state if table has not changed */
613	if (!state->probing &&
614	    (state->list_gen == mon->list_gen) &&
615	    (state->acked_gen == mon->dom_gen))
616		return;
617
618	read_lock_bh(&mon->lock);
619	peer = get_peer(mon, addr);
620	if (peer) {
621		state->probing = state->acked_gen != mon->dom_gen;
622		state->probing |= peer->down_cnt;
623		state->reset |= peer->down_cnt >= MAX_PEER_DOWN_EVENTS;
624		state->monitoring = peer->is_local;
625		state->monitoring |= peer->is_head;
626		state->list_gen = mon->list_gen;
627	}
628	read_unlock_bh(&mon->lock);
629}
630
631static void mon_timeout(struct timer_list *t)
632{
633	struct tipc_monitor *mon = from_timer(mon, t, timer);
634	struct tipc_peer *self;
635	int best_member_cnt = dom_size(mon->peer_cnt) - 1;
636
637	write_lock_bh(&mon->lock);
638	self = mon->self;
639	if (self && (best_member_cnt != self->applied)) {
640		mon_update_local_domain(mon);
641		mon_assign_roles(mon, self);
642	}
643	write_unlock_bh(&mon->lock);
644	mod_timer(&mon->timer, jiffies + mon->timer_intv);
645}
646
647int tipc_mon_create(struct net *net, int bearer_id)
648{
649	struct tipc_net *tn = tipc_net(net);
650	struct tipc_monitor *mon;
651	struct tipc_peer *self;
652	struct tipc_mon_domain *dom;
653
654	if (tn->monitors[bearer_id])
655		return 0;
656
657	mon = kzalloc(sizeof(*mon), GFP_ATOMIC);
658	self = kzalloc(sizeof(*self), GFP_ATOMIC);
659	dom = kzalloc(sizeof(*dom), GFP_ATOMIC);
660	if (!mon || !self || !dom) {
661		kfree(mon);
662		kfree(self);
663		kfree(dom);
664		return -ENOMEM;
665	}
666	tn->monitors[bearer_id] = mon;
667	rwlock_init(&mon->lock);
668	mon->net = net;
669	mon->peer_cnt = 1;
670	mon->self = self;
671	self->domain = dom;
672	self->addr = tipc_own_addr(net);
673	self->is_up = true;
674	self->is_head = true;
675	INIT_LIST_HEAD(&self->list);
676	timer_setup(&mon->timer, mon_timeout, 0);
677	mon->timer_intv = msecs_to_jiffies(MON_TIMEOUT + (tn->random & 0xffff));
678	mod_timer(&mon->timer, jiffies + mon->timer_intv);
679	return 0;
680}
681
682void tipc_mon_delete(struct net *net, int bearer_id)
683{
684	struct tipc_net *tn = tipc_net(net);
685	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
686	struct tipc_peer *self;
687	struct tipc_peer *peer, *tmp;
688
689	if (!mon)
690		return;
691
692	self = get_self(net, bearer_id);
693	write_lock_bh(&mon->lock);
694	tn->monitors[bearer_id] = NULL;
695	list_for_each_entry_safe(peer, tmp, &self->list, list) {
696		list_del(&peer->list);
697		hlist_del(&peer->hash);
698		kfree(peer->domain);
699		kfree(peer);
700	}
701	mon->self = NULL;
702	write_unlock_bh(&mon->lock);
703	timer_shutdown_sync(&mon->timer);
704	kfree(self->domain);
705	kfree(self);
706	kfree(mon);
707}
708
709void tipc_mon_reinit_self(struct net *net)
710{
711	struct tipc_monitor *mon;
712	int bearer_id;
713
714	for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
715		mon = tipc_monitor(net, bearer_id);
716		if (!mon)
717			continue;
718		write_lock_bh(&mon->lock);
719		mon->self->addr = tipc_own_addr(net);
720		write_unlock_bh(&mon->lock);
721	}
722}
723
724int tipc_nl_monitor_set_threshold(struct net *net, u32 cluster_size)
725{
726	struct tipc_net *tn = tipc_net(net);
727
728	if (cluster_size > TIPC_CLUSTER_SIZE)
729		return -EINVAL;
730
731	tn->mon_threshold = cluster_size;
732
733	return 0;
734}
735
736int tipc_nl_monitor_get_threshold(struct net *net)
737{
738	struct tipc_net *tn = tipc_net(net);
739
740	return tn->mon_threshold;
741}
742
743static int __tipc_nl_add_monitor_peer(struct tipc_peer *peer,
744				      struct tipc_nl_msg *msg)
745{
746	struct tipc_mon_domain *dom = peer->domain;
747	struct nlattr *attrs;
748	void *hdr;
749
750	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
751			  NLM_F_MULTI, TIPC_NL_MON_PEER_GET);
752	if (!hdr)
753		return -EMSGSIZE;
754
755	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_MON_PEER);
756	if (!attrs)
757		goto msg_full;
758
759	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_ADDR, peer->addr))
760		goto attr_msg_full;
761	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_APPLIED, peer->applied))
762		goto attr_msg_full;
763
764	if (peer->is_up)
765		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_UP))
766			goto attr_msg_full;
767	if (peer->is_local)
768		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_LOCAL))
769			goto attr_msg_full;
770	if (peer->is_head)
771		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_HEAD))
772			goto attr_msg_full;
773
774	if (dom) {
775		if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_DOMGEN, dom->gen))
776			goto attr_msg_full;
777		if (nla_put_u64_64bit(msg->skb, TIPC_NLA_MON_PEER_UPMAP,
778				      dom->up_map, TIPC_NLA_MON_PEER_PAD))
779			goto attr_msg_full;
780		if (nla_put(msg->skb, TIPC_NLA_MON_PEER_MEMBERS,
781			    dom->member_cnt * sizeof(u32), &dom->members))
782			goto attr_msg_full;
783	}
784
785	nla_nest_end(msg->skb, attrs);
786	genlmsg_end(msg->skb, hdr);
787	return 0;
788
789attr_msg_full:
790	nla_nest_cancel(msg->skb, attrs);
791msg_full:
792	genlmsg_cancel(msg->skb, hdr);
793
794	return -EMSGSIZE;
795}
796
797int tipc_nl_add_monitor_peer(struct net *net, struct tipc_nl_msg *msg,
798			     u32 bearer_id, u32 *prev_node)
799{
800	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
801	struct tipc_peer *peer;
802
803	if (!mon)
804		return -EINVAL;
805
806	read_lock_bh(&mon->lock);
807	peer = mon->self;
808	do {
809		if (*prev_node) {
810			if (peer->addr == *prev_node)
811				*prev_node = 0;
812			else
813				continue;
814		}
815		if (__tipc_nl_add_monitor_peer(peer, msg)) {
816			*prev_node = peer->addr;
817			read_unlock_bh(&mon->lock);
818			return -EMSGSIZE;
819		}
820	} while ((peer = peer_nxt(peer)) != mon->self);
821	read_unlock_bh(&mon->lock);
822
823	return 0;
824}
825
826int __tipc_nl_add_monitor(struct net *net, struct tipc_nl_msg *msg,
827			  u32 bearer_id)
828{
829	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
830	char bearer_name[TIPC_MAX_BEARER_NAME];
831	struct nlattr *attrs;
832	void *hdr;
833	int ret;
834
835	ret = tipc_bearer_get_name(net, bearer_name, bearer_id);
836	if (ret || !mon)
837		return 0;
838
839	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
840			  NLM_F_MULTI, TIPC_NL_MON_GET);
841	if (!hdr)
842		return -EMSGSIZE;
843
844	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_MON);
845	if (!attrs)
846		goto msg_full;
847
848	read_lock_bh(&mon->lock);
849	if (nla_put_u32(msg->skb, TIPC_NLA_MON_REF, bearer_id))
850		goto attr_msg_full;
851	if (tipc_mon_is_active(net, mon))
852		if (nla_put_flag(msg->skb, TIPC_NLA_MON_ACTIVE))
853			goto attr_msg_full;
854	if (nla_put_string(msg->skb, TIPC_NLA_MON_BEARER_NAME, bearer_name))
855		goto attr_msg_full;
856	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEERCNT, mon->peer_cnt))
857		goto attr_msg_full;
858	if (nla_put_u32(msg->skb, TIPC_NLA_MON_LISTGEN, mon->list_gen))
859		goto attr_msg_full;
860
861	read_unlock_bh(&mon->lock);
862	nla_nest_end(msg->skb, attrs);
863	genlmsg_end(msg->skb, hdr);
864
865	return 0;
866
867attr_msg_full:
868	read_unlock_bh(&mon->lock);
869	nla_nest_cancel(msg->skb, attrs);
870msg_full:
871	genlmsg_cancel(msg->skb, hdr);
872
873	return -EMSGSIZE;
874}
v5.9
  1/*
  2 * net/tipc/monitor.c
  3 *
  4 * Copyright (c) 2016, 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 <net/genetlink.h>
 37#include "core.h"
 38#include "addr.h"
 39#include "monitor.h"
 40#include "bearer.h"
 41
 42#define MAX_MON_DOMAIN       64
 43#define MON_TIMEOUT          120000
 44#define MAX_PEER_DOWN_EVENTS 4
 45
 46/* struct tipc_mon_domain: domain record to be transferred between peers
 47 * @len: actual size of domain record
 48 * @gen: current generation of sender's domain
 49 * @ack_gen: most recent generation of self's domain acked by peer
 50 * @member_cnt: number of domain member nodes described in this record
 51 * @up_map: bit map indicating which of the members the sender considers up
 52 * @members: identity of the domain members
 53 */
 54struct tipc_mon_domain {
 55	u16 len;
 56	u16 gen;
 57	u16 ack_gen;
 58	u16 member_cnt;
 59	u64 up_map;
 60	u32 members[MAX_MON_DOMAIN];
 61};
 62
 63/* struct tipc_peer: state of a peer node and its domain
 64 * @addr: tipc node identity of peer
 65 * @head_map: shows which other nodes currently consider peer 'up'
 66 * @domain: most recent domain record from peer
 67 * @hash: position in hashed lookup list
 68 * @list: position in linked list, in circular ascending order by 'addr'
 69 * @applied: number of reported domain members applied on this monitor list
 70 * @is_up: peer is up as seen from this node
 71 * @is_head: peer is assigned domain head as seen from this node
 72 * @is_local: peer is in local domain and should be continuously monitored
 73 * @down_cnt: - numbers of other peers which have reported this on lost
 74 */
 75struct tipc_peer {
 76	u32 addr;
 77	struct tipc_mon_domain *domain;
 78	struct hlist_node hash;
 79	struct list_head list;
 80	u8 applied;
 81	u8 down_cnt;
 82	bool is_up;
 83	bool is_head;
 84	bool is_local;
 85};
 86
 87struct tipc_monitor {
 88	struct hlist_head peers[NODE_HTABLE_SIZE];
 89	int peer_cnt;
 90	struct tipc_peer *self;
 91	rwlock_t lock;
 92	struct tipc_mon_domain cache;
 93	u16 list_gen;
 94	u16 dom_gen;
 95	struct net *net;
 96	struct timer_list timer;
 97	unsigned long timer_intv;
 98};
 99
100static struct tipc_monitor *tipc_monitor(struct net *net, int bearer_id)
101{
102	return tipc_net(net)->monitors[bearer_id];
103}
104
105const int tipc_max_domain_size = sizeof(struct tipc_mon_domain);
106
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
107/* dom_rec_len(): actual length of domain record for transport
108 */
109static int dom_rec_len(struct tipc_mon_domain *dom, u16 mcnt)
110{
111	return ((void *)&dom->members - (void *)dom) + (mcnt * sizeof(u32));
112}
113
114/* dom_size() : calculate size of own domain based on number of peers
115 */
116static int dom_size(int peers)
117{
118	int i = 0;
119
120	while ((i * i) < peers)
121		i++;
122	return i < MAX_MON_DOMAIN ? i : MAX_MON_DOMAIN;
123}
124
125static void map_set(u64 *up_map, int i, unsigned int v)
126{
127	*up_map &= ~(1ULL << i);
128	*up_map |= ((u64)v << i);
129}
130
131static int map_get(u64 up_map, int i)
132{
133	return (up_map & (1 << i)) >> i;
134}
135
136static struct tipc_peer *peer_prev(struct tipc_peer *peer)
137{
138	return list_last_entry(&peer->list, struct tipc_peer, list);
139}
140
141static struct tipc_peer *peer_nxt(struct tipc_peer *peer)
142{
143	return list_first_entry(&peer->list, struct tipc_peer, list);
144}
145
146static struct tipc_peer *peer_head(struct tipc_peer *peer)
147{
148	while (!peer->is_head)
149		peer = peer_prev(peer);
150	return peer;
151}
152
153static struct tipc_peer *get_peer(struct tipc_monitor *mon, u32 addr)
154{
155	struct tipc_peer *peer;
156	unsigned int thash = tipc_hashfn(addr);
157
158	hlist_for_each_entry(peer, &mon->peers[thash], hash) {
159		if (peer->addr == addr)
160			return peer;
161	}
162	return NULL;
163}
164
165static struct tipc_peer *get_self(struct net *net, int bearer_id)
166{
167	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
168
169	return mon->self;
170}
171
172static inline bool tipc_mon_is_active(struct net *net, struct tipc_monitor *mon)
173{
174	struct tipc_net *tn = tipc_net(net);
175
176	return mon->peer_cnt > tn->mon_threshold;
177}
178
179/* mon_identify_lost_members() : - identify amd mark potentially lost members
180 */
181static void mon_identify_lost_members(struct tipc_peer *peer,
182				      struct tipc_mon_domain *dom_bef,
183				      int applied_bef)
184{
185	struct tipc_peer *member = peer;
186	struct tipc_mon_domain *dom_aft = peer->domain;
187	int applied_aft = peer->applied;
188	int i;
189
190	for (i = 0; i < applied_bef; i++) {
191		member = peer_nxt(member);
192
193		/* Do nothing if self or peer already see member as down */
194		if (!member->is_up || !map_get(dom_bef->up_map, i))
195			continue;
196
197		/* Loss of local node must be detected by active probing */
198		if (member->is_local)
199			continue;
200
201		/* Start probing if member was removed from applied domain */
202		if (!applied_aft || (applied_aft < i)) {
203			member->down_cnt = 1;
204			continue;
205		}
206
207		/* Member loss is confirmed if it is still in applied domain */
208		if (!map_get(dom_aft->up_map, i))
209			member->down_cnt++;
210	}
211}
212
213/* mon_apply_domain() : match a peer's domain record against monitor list
214 */
215static void mon_apply_domain(struct tipc_monitor *mon,
216			     struct tipc_peer *peer)
217{
218	struct tipc_mon_domain *dom = peer->domain;
219	struct tipc_peer *member;
220	u32 addr;
221	int i;
222
223	if (!dom || !peer->is_up)
224		return;
225
226	/* Scan across domain members and match against monitor list */
227	peer->applied = 0;
228	member = peer_nxt(peer);
229	for (i = 0; i < dom->member_cnt; i++) {
230		addr = dom->members[i];
231		if (addr != member->addr)
232			return;
233		peer->applied++;
234		member = peer_nxt(member);
235	}
236}
237
238/* mon_update_local_domain() : update after peer addition/removal/up/down
239 */
240static void mon_update_local_domain(struct tipc_monitor *mon)
241{
242	struct tipc_peer *self = mon->self;
243	struct tipc_mon_domain *cache = &mon->cache;
244	struct tipc_mon_domain *dom = self->domain;
245	struct tipc_peer *peer = self;
246	u64 prev_up_map = dom->up_map;
247	u16 member_cnt, i;
248	bool diff;
249
250	/* Update local domain size based on current size of cluster */
251	member_cnt = dom_size(mon->peer_cnt) - 1;
252	self->applied = member_cnt;
253
254	/* Update native and cached outgoing local domain records */
255	dom->len = dom_rec_len(dom, member_cnt);
256	diff = dom->member_cnt != member_cnt;
257	dom->member_cnt = member_cnt;
258	for (i = 0; i < member_cnt; i++) {
259		peer = peer_nxt(peer);
260		diff |= dom->members[i] != peer->addr;
261		dom->members[i] = peer->addr;
262		map_set(&dom->up_map, i, peer->is_up);
263		cache->members[i] = htonl(peer->addr);
264	}
265	diff |= dom->up_map != prev_up_map;
266	if (!diff)
267		return;
268	dom->gen = ++mon->dom_gen;
269	cache->len = htons(dom->len);
270	cache->gen = htons(dom->gen);
271	cache->member_cnt = htons(member_cnt);
272	cache->up_map = cpu_to_be64(dom->up_map);
273	mon_apply_domain(mon, self);
274}
275
276/* mon_update_neighbors() : update preceding neighbors of added/removed peer
277 */
278static void mon_update_neighbors(struct tipc_monitor *mon,
279				 struct tipc_peer *peer)
280{
281	int dz, i;
282
283	dz = dom_size(mon->peer_cnt);
284	for (i = 0; i < dz; i++) {
285		mon_apply_domain(mon, peer);
286		peer = peer_prev(peer);
287	}
288}
289
290/* mon_assign_roles() : reassign peer roles after a network change
291 * The monitor list is consistent at this stage; i.e., each peer is monitoring
292 * a set of domain members as matched between domain record and the monitor list
293 */
294static void mon_assign_roles(struct tipc_monitor *mon, struct tipc_peer *head)
295{
296	struct tipc_peer *peer = peer_nxt(head);
297	struct tipc_peer *self = mon->self;
298	int i = 0;
299
300	for (; peer != self; peer = peer_nxt(peer)) {
301		peer->is_local = false;
302
303		/* Update domain member */
304		if (i++ < head->applied) {
305			peer->is_head = false;
306			if (head == self)
307				peer->is_local = true;
308			continue;
309		}
310		/* Assign next domain head */
311		if (!peer->is_up)
312			continue;
313		if (peer->is_head)
314			break;
315		head = peer;
316		head->is_head = true;
317		i = 0;
318	}
319	mon->list_gen++;
320}
321
322void tipc_mon_remove_peer(struct net *net, u32 addr, int bearer_id)
323{
324	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
325	struct tipc_peer *self;
326	struct tipc_peer *peer, *prev, *head;
327
328	if (!mon)
329		return;
330
331	self = get_self(net, bearer_id);
332	write_lock_bh(&mon->lock);
333	peer = get_peer(mon, addr);
334	if (!peer)
335		goto exit;
336	prev = peer_prev(peer);
337	list_del(&peer->list);
338	hlist_del(&peer->hash);
339	kfree(peer->domain);
340	kfree(peer);
341	mon->peer_cnt--;
342	head = peer_head(prev);
343	if (head == self)
344		mon_update_local_domain(mon);
345	mon_update_neighbors(mon, prev);
346
347	/* Revert to full-mesh monitoring if we reach threshold */
348	if (!tipc_mon_is_active(net, mon)) {
349		list_for_each_entry(peer, &self->list, list) {
350			kfree(peer->domain);
351			peer->domain = NULL;
352			peer->applied = 0;
353		}
354	}
355	mon_assign_roles(mon, head);
356exit:
357	write_unlock_bh(&mon->lock);
358}
359
360static bool tipc_mon_add_peer(struct tipc_monitor *mon, u32 addr,
361			      struct tipc_peer **peer)
362{
363	struct tipc_peer *self = mon->self;
364	struct tipc_peer *cur, *prev, *p;
365
366	p = kzalloc(sizeof(*p), GFP_ATOMIC);
367	*peer = p;
368	if (!p)
369		return false;
370	p->addr = addr;
371
372	/* Add new peer to lookup list */
373	INIT_LIST_HEAD(&p->list);
374	hlist_add_head(&p->hash, &mon->peers[tipc_hashfn(addr)]);
375
376	/* Sort new peer into iterator list, in ascending circular order */
377	prev = self;
378	list_for_each_entry(cur, &self->list, list) {
379		if ((addr > prev->addr) && (addr < cur->addr))
380			break;
381		if (((addr < cur->addr) || (addr > prev->addr)) &&
382		    (prev->addr > cur->addr))
383			break;
384		prev = cur;
385	}
386	list_add_tail(&p->list, &cur->list);
387	mon->peer_cnt++;
388	mon_update_neighbors(mon, p);
389	return true;
390}
391
392void tipc_mon_peer_up(struct net *net, u32 addr, int bearer_id)
393{
394	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
395	struct tipc_peer *self = get_self(net, bearer_id);
396	struct tipc_peer *peer, *head;
397
398	write_lock_bh(&mon->lock);
399	peer = get_peer(mon, addr);
400	if (!peer && !tipc_mon_add_peer(mon, addr, &peer))
401		goto exit;
402	peer->is_up = true;
403	head = peer_head(peer);
404	if (head == self)
405		mon_update_local_domain(mon);
406	mon_assign_roles(mon, head);
407exit:
408	write_unlock_bh(&mon->lock);
409}
410
411void tipc_mon_peer_down(struct net *net, u32 addr, int bearer_id)
412{
413	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
414	struct tipc_peer *self;
415	struct tipc_peer *peer, *head;
416	struct tipc_mon_domain *dom;
417	int applied;
418
419	if (!mon)
420		return;
421
422	self = get_self(net, bearer_id);
423	write_lock_bh(&mon->lock);
424	peer = get_peer(mon, addr);
425	if (!peer) {
426		pr_warn("Mon: unknown link %x/%u DOWN\n", addr, bearer_id);
427		goto exit;
428	}
429	applied = peer->applied;
430	peer->applied = 0;
431	dom = peer->domain;
432	peer->domain = NULL;
433	if (peer->is_head)
434		mon_identify_lost_members(peer, dom, applied);
435	kfree(dom);
436	peer->is_up = false;
437	peer->is_head = false;
438	peer->is_local = false;
439	peer->down_cnt = 0;
440	head = peer_head(peer);
441	if (head == self)
442		mon_update_local_domain(mon);
443	mon_assign_roles(mon, head);
444exit:
445	write_unlock_bh(&mon->lock);
446}
447
448/* tipc_mon_rcv - process monitor domain event message
449 */
450void tipc_mon_rcv(struct net *net, void *data, u16 dlen, u32 addr,
451		  struct tipc_mon_state *state, int bearer_id)
452{
453	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
454	struct tipc_mon_domain *arrv_dom = data;
455	struct tipc_mon_domain dom_bef;
456	struct tipc_mon_domain *dom;
457	struct tipc_peer *peer;
458	u16 new_member_cnt = ntohs(arrv_dom->member_cnt);
459	int new_dlen = dom_rec_len(arrv_dom, new_member_cnt);
460	u16 new_gen = ntohs(arrv_dom->gen);
461	u16 acked_gen = ntohs(arrv_dom->ack_gen);
 
462	bool probing = state->probing;
463	int i, applied_bef;
464
465	state->probing = false;
466
467	/* Sanity check received domain record */
 
 
468	if (dlen < dom_rec_len(arrv_dom, 0))
469		return;
470	if (dlen != dom_rec_len(arrv_dom, new_member_cnt))
471		return;
472	if ((dlen < new_dlen) || ntohs(arrv_dom->len) != new_dlen)
473		return;
474
475	/* Synch generation numbers with peer if link just came up */
476	if (!state->synched) {
477		state->peer_gen = new_gen - 1;
478		state->acked_gen = acked_gen;
479		state->synched = true;
480	}
481
482	if (more(acked_gen, state->acked_gen))
483		state->acked_gen = acked_gen;
484
485	/* Drop duplicate unless we are waiting for a probe response */
486	if (!more(new_gen, state->peer_gen) && !probing)
487		return;
488
489	write_lock_bh(&mon->lock);
490	peer = get_peer(mon, addr);
491	if (!peer || !peer->is_up)
492		goto exit;
493
494	/* Peer is confirmed, stop any ongoing probing */
495	peer->down_cnt = 0;
496
497	/* Task is done for duplicate record */
498	if (!more(new_gen, state->peer_gen))
499		goto exit;
500
501	state->peer_gen = new_gen;
502
503	/* Cache current domain record for later use */
504	dom_bef.member_cnt = 0;
505	dom = peer->domain;
506	if (dom)
507		memcpy(&dom_bef, dom, dom->len);
508
509	/* Transform and store received domain record */
510	if (!dom || (dom->len < new_dlen)) {
511		kfree(dom);
512		dom = kmalloc(new_dlen, GFP_ATOMIC);
513		peer->domain = dom;
514		if (!dom)
515			goto exit;
516	}
517	dom->len = new_dlen;
518	dom->gen = new_gen;
519	dom->member_cnt = new_member_cnt;
520	dom->up_map = be64_to_cpu(arrv_dom->up_map);
521	for (i = 0; i < new_member_cnt; i++)
522		dom->members[i] = ntohl(arrv_dom->members[i]);
523
524	/* Update peers affected by this domain record */
525	applied_bef = peer->applied;
526	mon_apply_domain(mon, peer);
527	mon_identify_lost_members(peer, &dom_bef, applied_bef);
528	mon_assign_roles(mon, peer_head(peer));
529exit:
530	write_unlock_bh(&mon->lock);
531}
532
533void tipc_mon_prep(struct net *net, void *data, int *dlen,
534		   struct tipc_mon_state *state, int bearer_id)
535{
536	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
537	struct tipc_mon_domain *dom = data;
538	u16 gen = mon->dom_gen;
539	u16 len;
540
541	/* Send invalid record if not active */
542	if (!tipc_mon_is_active(net, mon)) {
543		dom->len = 0;
544		return;
545	}
546
547	/* Send only a dummy record with ack if peer has acked our last sent */
548	if (likely(state->acked_gen == gen)) {
549		len = dom_rec_len(dom, 0);
550		*dlen = len;
551		dom->len = htons(len);
552		dom->gen = htons(gen);
553		dom->ack_gen = htons(state->peer_gen);
554		dom->member_cnt = 0;
555		return;
556	}
557	/* Send the full record */
558	read_lock_bh(&mon->lock);
559	len = ntohs(mon->cache.len);
560	*dlen = len;
561	memcpy(data, &mon->cache, len);
562	read_unlock_bh(&mon->lock);
563	dom->ack_gen = htons(state->peer_gen);
564}
565
566void tipc_mon_get_state(struct net *net, u32 addr,
567			struct tipc_mon_state *state,
568			int bearer_id)
569{
570	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
571	struct tipc_peer *peer;
572
573	if (!tipc_mon_is_active(net, mon)) {
574		state->probing = false;
575		state->monitoring = true;
576		return;
577	}
578
579	/* Used cached state if table has not changed */
580	if (!state->probing &&
581	    (state->list_gen == mon->list_gen) &&
582	    (state->acked_gen == mon->dom_gen))
583		return;
584
585	read_lock_bh(&mon->lock);
586	peer = get_peer(mon, addr);
587	if (peer) {
588		state->probing = state->acked_gen != mon->dom_gen;
589		state->probing |= peer->down_cnt;
590		state->reset |= peer->down_cnt >= MAX_PEER_DOWN_EVENTS;
591		state->monitoring = peer->is_local;
592		state->monitoring |= peer->is_head;
593		state->list_gen = mon->list_gen;
594	}
595	read_unlock_bh(&mon->lock);
596}
597
598static void mon_timeout(struct timer_list *t)
599{
600	struct tipc_monitor *mon = from_timer(mon, t, timer);
601	struct tipc_peer *self;
602	int best_member_cnt = dom_size(mon->peer_cnt) - 1;
603
604	write_lock_bh(&mon->lock);
605	self = mon->self;
606	if (self && (best_member_cnt != self->applied)) {
607		mon_update_local_domain(mon);
608		mon_assign_roles(mon, self);
609	}
610	write_unlock_bh(&mon->lock);
611	mod_timer(&mon->timer, jiffies + mon->timer_intv);
612}
613
614int tipc_mon_create(struct net *net, int bearer_id)
615{
616	struct tipc_net *tn = tipc_net(net);
617	struct tipc_monitor *mon;
618	struct tipc_peer *self;
619	struct tipc_mon_domain *dom;
620
621	if (tn->monitors[bearer_id])
622		return 0;
623
624	mon = kzalloc(sizeof(*mon), GFP_ATOMIC);
625	self = kzalloc(sizeof(*self), GFP_ATOMIC);
626	dom = kzalloc(sizeof(*dom), GFP_ATOMIC);
627	if (!mon || !self || !dom) {
628		kfree(mon);
629		kfree(self);
630		kfree(dom);
631		return -ENOMEM;
632	}
633	tn->monitors[bearer_id] = mon;
634	rwlock_init(&mon->lock);
635	mon->net = net;
636	mon->peer_cnt = 1;
637	mon->self = self;
638	self->domain = dom;
639	self->addr = tipc_own_addr(net);
640	self->is_up = true;
641	self->is_head = true;
642	INIT_LIST_HEAD(&self->list);
643	timer_setup(&mon->timer, mon_timeout, 0);
644	mon->timer_intv = msecs_to_jiffies(MON_TIMEOUT + (tn->random & 0xffff));
645	mod_timer(&mon->timer, jiffies + mon->timer_intv);
646	return 0;
647}
648
649void tipc_mon_delete(struct net *net, int bearer_id)
650{
651	struct tipc_net *tn = tipc_net(net);
652	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
653	struct tipc_peer *self;
654	struct tipc_peer *peer, *tmp;
655
656	if (!mon)
657		return;
658
659	self = get_self(net, bearer_id);
660	write_lock_bh(&mon->lock);
661	tn->monitors[bearer_id] = NULL;
662	list_for_each_entry_safe(peer, tmp, &self->list, list) {
663		list_del(&peer->list);
664		hlist_del(&peer->hash);
665		kfree(peer->domain);
666		kfree(peer);
667	}
668	mon->self = NULL;
669	write_unlock_bh(&mon->lock);
670	del_timer_sync(&mon->timer);
671	kfree(self->domain);
672	kfree(self);
673	kfree(mon);
674}
675
676void tipc_mon_reinit_self(struct net *net)
677{
678	struct tipc_monitor *mon;
679	int bearer_id;
680
681	for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
682		mon = tipc_monitor(net, bearer_id);
683		if (!mon)
684			continue;
685		write_lock_bh(&mon->lock);
686		mon->self->addr = tipc_own_addr(net);
687		write_unlock_bh(&mon->lock);
688	}
689}
690
691int tipc_nl_monitor_set_threshold(struct net *net, u32 cluster_size)
692{
693	struct tipc_net *tn = tipc_net(net);
694
695	if (cluster_size > TIPC_CLUSTER_SIZE)
696		return -EINVAL;
697
698	tn->mon_threshold = cluster_size;
699
700	return 0;
701}
702
703int tipc_nl_monitor_get_threshold(struct net *net)
704{
705	struct tipc_net *tn = tipc_net(net);
706
707	return tn->mon_threshold;
708}
709
710static int __tipc_nl_add_monitor_peer(struct tipc_peer *peer,
711				      struct tipc_nl_msg *msg)
712{
713	struct tipc_mon_domain *dom = peer->domain;
714	struct nlattr *attrs;
715	void *hdr;
716
717	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
718			  NLM_F_MULTI, TIPC_NL_MON_PEER_GET);
719	if (!hdr)
720		return -EMSGSIZE;
721
722	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_MON_PEER);
723	if (!attrs)
724		goto msg_full;
725
726	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_ADDR, peer->addr))
727		goto attr_msg_full;
728	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_APPLIED, peer->applied))
729		goto attr_msg_full;
730
731	if (peer->is_up)
732		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_UP))
733			goto attr_msg_full;
734	if (peer->is_local)
735		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_LOCAL))
736			goto attr_msg_full;
737	if (peer->is_head)
738		if (nla_put_flag(msg->skb, TIPC_NLA_MON_PEER_HEAD))
739			goto attr_msg_full;
740
741	if (dom) {
742		if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEER_DOMGEN, dom->gen))
743			goto attr_msg_full;
744		if (nla_put_u64_64bit(msg->skb, TIPC_NLA_MON_PEER_UPMAP,
745				      dom->up_map, TIPC_NLA_MON_PEER_PAD))
746			goto attr_msg_full;
747		if (nla_put(msg->skb, TIPC_NLA_MON_PEER_MEMBERS,
748			    dom->member_cnt * sizeof(u32), &dom->members))
749			goto attr_msg_full;
750	}
751
752	nla_nest_end(msg->skb, attrs);
753	genlmsg_end(msg->skb, hdr);
754	return 0;
755
756attr_msg_full:
757	nla_nest_cancel(msg->skb, attrs);
758msg_full:
759	genlmsg_cancel(msg->skb, hdr);
760
761	return -EMSGSIZE;
762}
763
764int tipc_nl_add_monitor_peer(struct net *net, struct tipc_nl_msg *msg,
765			     u32 bearer_id, u32 *prev_node)
766{
767	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
768	struct tipc_peer *peer;
769
770	if (!mon)
771		return -EINVAL;
772
773	read_lock_bh(&mon->lock);
774	peer = mon->self;
775	do {
776		if (*prev_node) {
777			if (peer->addr == *prev_node)
778				*prev_node = 0;
779			else
780				continue;
781		}
782		if (__tipc_nl_add_monitor_peer(peer, msg)) {
783			*prev_node = peer->addr;
784			read_unlock_bh(&mon->lock);
785			return -EMSGSIZE;
786		}
787	} while ((peer = peer_nxt(peer)) != mon->self);
788	read_unlock_bh(&mon->lock);
789
790	return 0;
791}
792
793int __tipc_nl_add_monitor(struct net *net, struct tipc_nl_msg *msg,
794			  u32 bearer_id)
795{
796	struct tipc_monitor *mon = tipc_monitor(net, bearer_id);
797	char bearer_name[TIPC_MAX_BEARER_NAME];
798	struct nlattr *attrs;
799	void *hdr;
800	int ret;
801
802	ret = tipc_bearer_get_name(net, bearer_name, bearer_id);
803	if (ret || !mon)
804		return 0;
805
806	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
807			  NLM_F_MULTI, TIPC_NL_MON_GET);
808	if (!hdr)
809		return -EMSGSIZE;
810
811	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_MON);
812	if (!attrs)
813		goto msg_full;
814
815	read_lock_bh(&mon->lock);
816	if (nla_put_u32(msg->skb, TIPC_NLA_MON_REF, bearer_id))
817		goto attr_msg_full;
818	if (tipc_mon_is_active(net, mon))
819		if (nla_put_flag(msg->skb, TIPC_NLA_MON_ACTIVE))
820			goto attr_msg_full;
821	if (nla_put_string(msg->skb, TIPC_NLA_MON_BEARER_NAME, bearer_name))
822		goto attr_msg_full;
823	if (nla_put_u32(msg->skb, TIPC_NLA_MON_PEERCNT, mon->peer_cnt))
824		goto attr_msg_full;
825	if (nla_put_u32(msg->skb, TIPC_NLA_MON_LISTGEN, mon->list_gen))
826		goto attr_msg_full;
827
828	read_unlock_bh(&mon->lock);
829	nla_nest_end(msg->skb, attrs);
830	genlmsg_end(msg->skb, hdr);
831
832	return 0;
833
834attr_msg_full:
835	read_unlock_bh(&mon->lock);
836	nla_nest_cancel(msg->skb, attrs);
837msg_full:
838	genlmsg_cancel(msg->skb, hdr);
839
840	return -EMSGSIZE;
841}