• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * net/tipc/link.c: TIPC link code
3  *
4  * Copyright (c) 1996-2007, 2012-2016, Ericsson AB
5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
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 "subscr.h"
39 #include "link.h"
40 #include "bcast.h"
41 #include "socket.h"
42 #include "name_distr.h"
43 #include "discover.h"
44 #include "netlink.h"
45 #include "monitor.h"
46 #include "trace.h"
47 #include "crypto.h"
48 
49 #include <linux/pkt_sched.h>
50 
51 struct tipc_stats {
52 	u32 sent_pkts;
53 	u32 recv_pkts;
54 	u32 sent_states;
55 	u32 recv_states;
56 	u32 sent_probes;
57 	u32 recv_probes;
58 	u32 sent_nacks;
59 	u32 recv_nacks;
60 	u32 sent_acks;
61 	u32 sent_bundled;
62 	u32 sent_bundles;
63 	u32 recv_bundled;
64 	u32 recv_bundles;
65 	u32 retransmitted;
66 	u32 sent_fragmented;
67 	u32 sent_fragments;
68 	u32 recv_fragmented;
69 	u32 recv_fragments;
70 	u32 link_congs;		/* # port sends blocked by congestion */
71 	u32 deferred_recv;
72 	u32 duplicates;
73 	u32 max_queue_sz;	/* send queue size high water mark */
74 	u32 accu_queue_sz;	/* used for send queue size profiling */
75 	u32 queue_sz_counts;	/* used for send queue size profiling */
76 	u32 msg_length_counts;	/* used for message length profiling */
77 	u32 msg_lengths_total;	/* used for message length profiling */
78 	u32 msg_length_profile[7]; /* used for msg. length profiling */
79 };
80 
81 /**
82  * struct tipc_link - TIPC link data structure
83  * @addr: network address of link's peer node
84  * @name: link name character string
85  * @media_addr: media address to use when sending messages over link
86  * @timer: link timer
87  * @net: pointer to namespace struct
88  * @refcnt: reference counter for permanent references (owner node & timer)
89  * @peer_session: link session # being used by peer end of link
90  * @peer_bearer_id: bearer id used by link's peer endpoint
91  * @bearer_id: local bearer id used by link
92  * @tolerance: minimum link continuity loss needed to reset link [in ms]
93  * @abort_limit: # of unacknowledged continuity probes needed to reset link
94  * @state: current state of link FSM
95  * @peer_caps: bitmap describing capabilities of peer node
96  * @silent_intv_cnt: # of timer intervals without any reception from peer
97  * @proto_msg: template for control messages generated by link
98  * @pmsg: convenience pointer to "proto_msg" field
99  * @priority: current link priority
100  * @net_plane: current link network plane ('A' through 'H')
101  * @mon_state: cookie with information needed by link monitor
102  * @backlog_limit: backlog queue congestion thresholds (indexed by importance)
103  * @exp_msg_count: # of tunnelled messages expected during link changeover
104  * @reset_rcv_checkpt: seq # of last acknowledged message at time of link reset
105  * @mtu: current maximum packet size for this link
106  * @advertised_mtu: advertised own mtu when link is being established
107  * @transmitq: queue for sent, non-acked messages
108  * @backlogq: queue for messages waiting to be sent
109  * @snt_nxt: next sequence number to use for outbound messages
110  * @ackers: # of peers that needs to ack each packet before it can be released
111  * @acked: # last packet acked by a certain peer. Used for broadcast.
112  * @rcv_nxt: next sequence number to expect for inbound messages
113  * @deferred_queue: deferred queue saved OOS b'cast message received from node
114  * @unacked_window: # of inbound messages rx'd without ack'ing back to peer
115  * @inputq: buffer queue for messages to be delivered upwards
116  * @namedq: buffer queue for name table messages to be delivered upwards
117  * @next_out: ptr to first unsent outbound message in queue
118  * @wakeupq: linked list of wakeup msgs waiting for link congestion to abate
119  * @long_msg_seq_no: next identifier to use for outbound fragmented messages
120  * @reasm_buf: head of partially reassembled inbound message fragments
121  * @bc_rcvr: marks that this is a broadcast receiver link
122  * @stats: collects statistics regarding link activity
123  */
124 struct tipc_link {
125 	u32 addr;
126 	char name[TIPC_MAX_LINK_NAME];
127 	struct net *net;
128 
129 	/* Management and link supervision data */
130 	u16 peer_session;
131 	u16 session;
132 	u16 snd_nxt_state;
133 	u16 rcv_nxt_state;
134 	u32 peer_bearer_id;
135 	u32 bearer_id;
136 	u32 tolerance;
137 	u32 abort_limit;
138 	u32 state;
139 	u16 peer_caps;
140 	bool in_session;
141 	bool active;
142 	u32 silent_intv_cnt;
143 	char if_name[TIPC_MAX_IF_NAME];
144 	u32 priority;
145 	char net_plane;
146 	struct tipc_mon_state mon_state;
147 	u16 rst_cnt;
148 
149 	/* Failover/synch */
150 	u16 drop_point;
151 	struct sk_buff *failover_reasm_skb;
152 	struct sk_buff_head failover_deferdq;
153 
154 	/* Max packet negotiation */
155 	u16 mtu;
156 	u16 advertised_mtu;
157 
158 	/* Sending */
159 	struct sk_buff_head transmq;
160 	struct sk_buff_head backlogq;
161 	struct {
162 		u16 len;
163 		u16 limit;
164 		struct sk_buff *target_bskb;
165 	} backlog[5];
166 	u16 snd_nxt;
167 
168 	/* Reception */
169 	u16 rcv_nxt;
170 	u32 rcv_unacked;
171 	struct sk_buff_head deferdq;
172 	struct sk_buff_head *inputq;
173 	struct sk_buff_head *namedq;
174 
175 	/* Congestion handling */
176 	struct sk_buff_head wakeupq;
177 	u16 window;
178 	u16 min_win;
179 	u16 ssthresh;
180 	u16 max_win;
181 	u16 cong_acks;
182 	u16 checkpoint;
183 
184 	/* Fragmentation/reassembly */
185 	struct sk_buff *reasm_buf;
186 	struct sk_buff *reasm_tnlmsg;
187 
188 	/* Broadcast */
189 	u16 ackers;
190 	u16 acked;
191 	u16 last_gap;
192 	struct tipc_gap_ack_blks *last_ga;
193 	struct tipc_link *bc_rcvlink;
194 	struct tipc_link *bc_sndlink;
195 	u8 nack_state;
196 	bool bc_peer_is_up;
197 
198 	/* Statistics */
199 	struct tipc_stats stats;
200 };
201 
202 /*
203  * Error message prefixes
204  */
205 static const char *link_co_err = "Link tunneling error, ";
206 static const char *link_rst_msg = "Resetting link ";
207 
208 /* Send states for broadcast NACKs
209  */
210 enum {
211 	BC_NACK_SND_CONDITIONAL,
212 	BC_NACK_SND_UNCONDITIONAL,
213 	BC_NACK_SND_SUPPRESS,
214 };
215 
216 #define TIPC_BC_RETR_LIM  (jiffies + msecs_to_jiffies(10))
217 #define TIPC_UC_RETR_TIME (jiffies + msecs_to_jiffies(1))
218 
219 /* Link FSM states:
220  */
221 enum {
222 	LINK_ESTABLISHED     = 0xe,
223 	LINK_ESTABLISHING    = 0xe  << 4,
224 	LINK_RESET           = 0x1  << 8,
225 	LINK_RESETTING       = 0x2  << 12,
226 	LINK_PEER_RESET      = 0xd  << 16,
227 	LINK_FAILINGOVER     = 0xf  << 20,
228 	LINK_SYNCHING        = 0xc  << 24
229 };
230 
231 /* Link FSM state checking routines
232  */
link_is_up(struct tipc_link * l)233 static int link_is_up(struct tipc_link *l)
234 {
235 	return l->state & (LINK_ESTABLISHED | LINK_SYNCHING);
236 }
237 
238 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
239 			       struct sk_buff_head *xmitq);
240 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
241 				      bool probe_reply, u16 rcvgap,
242 				      int tolerance, int priority,
243 				      struct sk_buff_head *xmitq);
244 static void link_print(struct tipc_link *l, const char *str);
245 static int tipc_link_build_nack_msg(struct tipc_link *l,
246 				    struct sk_buff_head *xmitq);
247 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
248 					struct sk_buff_head *xmitq);
249 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
250 				    struct tipc_link *l, u8 start_index);
251 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr);
252 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
253 				     u16 acked, u16 gap,
254 				     struct tipc_gap_ack_blks *ga,
255 				     struct sk_buff_head *xmitq,
256 				     bool *retransmitted, int *rc);
257 static void tipc_link_update_cwin(struct tipc_link *l, int released,
258 				  bool retransmitted);
259 /*
260  *  Simple non-static link routines (i.e. referenced outside this file)
261  */
tipc_link_is_up(struct tipc_link * l)262 bool tipc_link_is_up(struct tipc_link *l)
263 {
264 	return link_is_up(l);
265 }
266 
tipc_link_peer_is_down(struct tipc_link * l)267 bool tipc_link_peer_is_down(struct tipc_link *l)
268 {
269 	return l->state == LINK_PEER_RESET;
270 }
271 
tipc_link_is_reset(struct tipc_link * l)272 bool tipc_link_is_reset(struct tipc_link *l)
273 {
274 	return l->state & (LINK_RESET | LINK_FAILINGOVER | LINK_ESTABLISHING);
275 }
276 
tipc_link_is_establishing(struct tipc_link * l)277 bool tipc_link_is_establishing(struct tipc_link *l)
278 {
279 	return l->state == LINK_ESTABLISHING;
280 }
281 
tipc_link_is_synching(struct tipc_link * l)282 bool tipc_link_is_synching(struct tipc_link *l)
283 {
284 	return l->state == LINK_SYNCHING;
285 }
286 
tipc_link_is_failingover(struct tipc_link * l)287 bool tipc_link_is_failingover(struct tipc_link *l)
288 {
289 	return l->state == LINK_FAILINGOVER;
290 }
291 
tipc_link_is_blocked(struct tipc_link * l)292 bool tipc_link_is_blocked(struct tipc_link *l)
293 {
294 	return l->state & (LINK_RESETTING | LINK_PEER_RESET | LINK_FAILINGOVER);
295 }
296 
link_is_bc_sndlink(struct tipc_link * l)297 static bool link_is_bc_sndlink(struct tipc_link *l)
298 {
299 	return !l->bc_sndlink;
300 }
301 
link_is_bc_rcvlink(struct tipc_link * l)302 static bool link_is_bc_rcvlink(struct tipc_link *l)
303 {
304 	return ((l->bc_rcvlink == l) && !link_is_bc_sndlink(l));
305 }
306 
tipc_link_set_active(struct tipc_link * l,bool active)307 void tipc_link_set_active(struct tipc_link *l, bool active)
308 {
309 	l->active = active;
310 }
311 
tipc_link_id(struct tipc_link * l)312 u32 tipc_link_id(struct tipc_link *l)
313 {
314 	return l->peer_bearer_id << 16 | l->bearer_id;
315 }
316 
tipc_link_min_win(struct tipc_link * l)317 int tipc_link_min_win(struct tipc_link *l)
318 {
319 	return l->min_win;
320 }
321 
tipc_link_max_win(struct tipc_link * l)322 int tipc_link_max_win(struct tipc_link *l)
323 {
324 	return l->max_win;
325 }
326 
tipc_link_prio(struct tipc_link * l)327 int tipc_link_prio(struct tipc_link *l)
328 {
329 	return l->priority;
330 }
331 
tipc_link_tolerance(struct tipc_link * l)332 unsigned long tipc_link_tolerance(struct tipc_link *l)
333 {
334 	return l->tolerance;
335 }
336 
tipc_link_inputq(struct tipc_link * l)337 struct sk_buff_head *tipc_link_inputq(struct tipc_link *l)
338 {
339 	return l->inputq;
340 }
341 
tipc_link_plane(struct tipc_link * l)342 char tipc_link_plane(struct tipc_link *l)
343 {
344 	return l->net_plane;
345 }
346 
tipc_link_update_caps(struct tipc_link * l,u16 capabilities)347 void tipc_link_update_caps(struct tipc_link *l, u16 capabilities)
348 {
349 	l->peer_caps = capabilities;
350 }
351 
tipc_link_add_bc_peer(struct tipc_link * snd_l,struct tipc_link * uc_l,struct sk_buff_head * xmitq)352 void tipc_link_add_bc_peer(struct tipc_link *snd_l,
353 			   struct tipc_link *uc_l,
354 			   struct sk_buff_head *xmitq)
355 {
356 	struct tipc_link *rcv_l = uc_l->bc_rcvlink;
357 
358 	snd_l->ackers++;
359 	rcv_l->acked = snd_l->snd_nxt - 1;
360 	snd_l->state = LINK_ESTABLISHED;
361 	tipc_link_build_bc_init_msg(uc_l, xmitq);
362 }
363 
tipc_link_remove_bc_peer(struct tipc_link * snd_l,struct tipc_link * rcv_l,struct sk_buff_head * xmitq)364 void tipc_link_remove_bc_peer(struct tipc_link *snd_l,
365 			      struct tipc_link *rcv_l,
366 			      struct sk_buff_head *xmitq)
367 {
368 	u16 ack = snd_l->snd_nxt - 1;
369 
370 	snd_l->ackers--;
371 	rcv_l->bc_peer_is_up = true;
372 	rcv_l->state = LINK_ESTABLISHED;
373 	tipc_link_bc_ack_rcv(rcv_l, ack, 0, NULL, xmitq, NULL);
374 	trace_tipc_link_reset(rcv_l, TIPC_DUMP_ALL, "bclink removed!");
375 	tipc_link_reset(rcv_l);
376 	rcv_l->state = LINK_RESET;
377 	if (!snd_l->ackers) {
378 		trace_tipc_link_reset(snd_l, TIPC_DUMP_ALL, "zero ackers!");
379 		tipc_link_reset(snd_l);
380 		snd_l->state = LINK_RESET;
381 		__skb_queue_purge(xmitq);
382 	}
383 }
384 
tipc_link_bc_peers(struct tipc_link * l)385 int tipc_link_bc_peers(struct tipc_link *l)
386 {
387 	return l->ackers;
388 }
389 
link_bc_rcv_gap(struct tipc_link * l)390 static u16 link_bc_rcv_gap(struct tipc_link *l)
391 {
392 	struct sk_buff *skb = skb_peek(&l->deferdq);
393 	u16 gap = 0;
394 
395 	if (more(l->snd_nxt, l->rcv_nxt))
396 		gap = l->snd_nxt - l->rcv_nxt;
397 	if (skb)
398 		gap = buf_seqno(skb) - l->rcv_nxt;
399 	return gap;
400 }
401 
tipc_link_set_mtu(struct tipc_link * l,int mtu)402 void tipc_link_set_mtu(struct tipc_link *l, int mtu)
403 {
404 	l->mtu = mtu;
405 }
406 
tipc_link_mtu(struct tipc_link * l)407 int tipc_link_mtu(struct tipc_link *l)
408 {
409 	return l->mtu;
410 }
411 
tipc_link_mss(struct tipc_link * l)412 int tipc_link_mss(struct tipc_link *l)
413 {
414 #ifdef CONFIG_TIPC_CRYPTO
415 	return l->mtu - INT_H_SIZE - EMSG_OVERHEAD;
416 #else
417 	return l->mtu - INT_H_SIZE;
418 #endif
419 }
420 
tipc_link_rcv_nxt(struct tipc_link * l)421 u16 tipc_link_rcv_nxt(struct tipc_link *l)
422 {
423 	return l->rcv_nxt;
424 }
425 
tipc_link_acked(struct tipc_link * l)426 u16 tipc_link_acked(struct tipc_link *l)
427 {
428 	return l->acked;
429 }
430 
tipc_link_name(struct tipc_link * l)431 char *tipc_link_name(struct tipc_link *l)
432 {
433 	return l->name;
434 }
435 
tipc_link_state(struct tipc_link * l)436 u32 tipc_link_state(struct tipc_link *l)
437 {
438 	return l->state;
439 }
440 
441 /**
442  * tipc_link_create - create a new link
443  * @net: pointer to associated network namespace
444  * @if_name: associated interface name
445  * @bearer_id: id (index) of associated bearer
446  * @tolerance: link tolerance to be used by link
447  * @net_plane: network plane (A,B,c..) this link belongs to
448  * @mtu: mtu to be advertised by link
449  * @priority: priority to be used by link
450  * @min_win: minimal send window to be used by link
451  * @max_win: maximal send window to be used by link
452  * @session: session to be used by link
453  * @ownnode: identity of own node
454  * @peer: node id of peer node
455  * @peer_caps: bitmap describing peer node capabilities
456  * @bc_sndlink: the namespace global link used for broadcast sending
457  * @bc_rcvlink: the peer specific link used for broadcast reception
458  * @inputq: queue to put messages ready for delivery
459  * @namedq: queue to put binding table update messages ready for delivery
460  * @link: return value, pointer to put the created link
461  *
462  * Returns true if link was created, otherwise false
463  */
tipc_link_create(struct net * net,char * if_name,int bearer_id,int tolerance,char net_plane,u32 mtu,int priority,u32 min_win,u32 max_win,u32 session,u32 self,u32 peer,u8 * peer_id,u16 peer_caps,struct tipc_link * bc_sndlink,struct tipc_link * bc_rcvlink,struct sk_buff_head * inputq,struct sk_buff_head * namedq,struct tipc_link ** link)464 bool tipc_link_create(struct net *net, char *if_name, int bearer_id,
465 		      int tolerance, char net_plane, u32 mtu, int priority,
466 		      u32 min_win, u32 max_win, u32 session, u32 self,
467 		      u32 peer, u8 *peer_id, u16 peer_caps,
468 		      struct tipc_link *bc_sndlink,
469 		      struct tipc_link *bc_rcvlink,
470 		      struct sk_buff_head *inputq,
471 		      struct sk_buff_head *namedq,
472 		      struct tipc_link **link)
473 {
474 	char peer_str[NODE_ID_STR_LEN] = {0,};
475 	char self_str[NODE_ID_STR_LEN] = {0,};
476 	struct tipc_link *l;
477 
478 	l = kzalloc(sizeof(*l), GFP_ATOMIC);
479 	if (!l)
480 		return false;
481 	*link = l;
482 	l->session = session;
483 
484 	/* Set link name for unicast links only */
485 	if (peer_id) {
486 		tipc_nodeid2string(self_str, tipc_own_id(net));
487 		if (strlen(self_str) > 16)
488 			sprintf(self_str, "%x", self);
489 		tipc_nodeid2string(peer_str, peer_id);
490 		if (strlen(peer_str) > 16)
491 			sprintf(peer_str, "%x", peer);
492 	}
493 	/* Peer i/f name will be completed by reset/activate message */
494 	snprintf(l->name, sizeof(l->name), "%s:%s-%s:unknown",
495 		 self_str, if_name, peer_str);
496 
497 	strcpy(l->if_name, if_name);
498 	l->addr = peer;
499 	l->peer_caps = peer_caps;
500 	l->net = net;
501 	l->in_session = false;
502 	l->bearer_id = bearer_id;
503 	l->tolerance = tolerance;
504 	if (bc_rcvlink)
505 		bc_rcvlink->tolerance = tolerance;
506 	l->net_plane = net_plane;
507 	l->advertised_mtu = mtu;
508 	l->mtu = mtu;
509 	l->priority = priority;
510 	tipc_link_set_queue_limits(l, min_win, max_win);
511 	l->ackers = 1;
512 	l->bc_sndlink = bc_sndlink;
513 	l->bc_rcvlink = bc_rcvlink;
514 	l->inputq = inputq;
515 	l->namedq = namedq;
516 	l->state = LINK_RESETTING;
517 	__skb_queue_head_init(&l->transmq);
518 	__skb_queue_head_init(&l->backlogq);
519 	__skb_queue_head_init(&l->deferdq);
520 	__skb_queue_head_init(&l->failover_deferdq);
521 	skb_queue_head_init(&l->wakeupq);
522 	skb_queue_head_init(l->inputq);
523 	return true;
524 }
525 
526 /**
527  * tipc_link_bc_create - create new link to be used for broadcast
528  * @net: pointer to associated network namespace
529  * @mtu: mtu to be used initially if no peers
530  * @min_win: minimal send window to be used by link
531  * @max_win: maximal send window to be used by link
532  * @inputq: queue to put messages ready for delivery
533  * @namedq: queue to put binding table update messages ready for delivery
534  * @link: return value, pointer to put the created link
535  *
536  * Returns true if link was created, otherwise false
537  */
tipc_link_bc_create(struct net * net,u32 ownnode,u32 peer,u8 * peer_id,int mtu,u32 min_win,u32 max_win,u16 peer_caps,struct sk_buff_head * inputq,struct sk_buff_head * namedq,struct tipc_link * bc_sndlink,struct tipc_link ** link)538 bool tipc_link_bc_create(struct net *net, u32 ownnode, u32 peer, u8 *peer_id,
539 			 int mtu, u32 min_win, u32 max_win, u16 peer_caps,
540 			 struct sk_buff_head *inputq,
541 			 struct sk_buff_head *namedq,
542 			 struct tipc_link *bc_sndlink,
543 			 struct tipc_link **link)
544 {
545 	struct tipc_link *l;
546 
547 	if (!tipc_link_create(net, "", MAX_BEARERS, 0, 'Z', mtu, 0, min_win,
548 			      max_win, 0, ownnode, peer, NULL, peer_caps,
549 			      bc_sndlink, NULL, inputq, namedq, link))
550 		return false;
551 
552 	l = *link;
553 	if (peer_id) {
554 		char peer_str[NODE_ID_STR_LEN] = {0,};
555 
556 		tipc_nodeid2string(peer_str, peer_id);
557 		if (strlen(peer_str) > 16)
558 			sprintf(peer_str, "%x", peer);
559 		/* Broadcast receiver link name: "broadcast-link:<peer>" */
560 		snprintf(l->name, sizeof(l->name), "%s:%s", tipc_bclink_name,
561 			 peer_str);
562 	} else {
563 		strcpy(l->name, tipc_bclink_name);
564 	}
565 	trace_tipc_link_reset(l, TIPC_DUMP_ALL, "bclink created!");
566 	tipc_link_reset(l);
567 	l->state = LINK_RESET;
568 	l->ackers = 0;
569 	l->bc_rcvlink = l;
570 
571 	/* Broadcast send link is always up */
572 	if (link_is_bc_sndlink(l))
573 		l->state = LINK_ESTABLISHED;
574 
575 	/* Disable replicast if even a single peer doesn't support it */
576 	if (link_is_bc_rcvlink(l) && !(peer_caps & TIPC_BCAST_RCAST))
577 		tipc_bcast_toggle_rcast(net, false);
578 
579 	return true;
580 }
581 
582 /**
583  * tipc_link_fsm_evt - link finite state machine
584  * @l: pointer to link
585  * @evt: state machine event to be processed
586  */
tipc_link_fsm_evt(struct tipc_link * l,int evt)587 int tipc_link_fsm_evt(struct tipc_link *l, int evt)
588 {
589 	int rc = 0;
590 	int old_state = l->state;
591 
592 	switch (l->state) {
593 	case LINK_RESETTING:
594 		switch (evt) {
595 		case LINK_PEER_RESET_EVT:
596 			l->state = LINK_PEER_RESET;
597 			break;
598 		case LINK_RESET_EVT:
599 			l->state = LINK_RESET;
600 			break;
601 		case LINK_FAILURE_EVT:
602 		case LINK_FAILOVER_BEGIN_EVT:
603 		case LINK_ESTABLISH_EVT:
604 		case LINK_FAILOVER_END_EVT:
605 		case LINK_SYNCH_BEGIN_EVT:
606 		case LINK_SYNCH_END_EVT:
607 		default:
608 			goto illegal_evt;
609 		}
610 		break;
611 	case LINK_RESET:
612 		switch (evt) {
613 		case LINK_PEER_RESET_EVT:
614 			l->state = LINK_ESTABLISHING;
615 			break;
616 		case LINK_FAILOVER_BEGIN_EVT:
617 			l->state = LINK_FAILINGOVER;
618 		case LINK_FAILURE_EVT:
619 		case LINK_RESET_EVT:
620 		case LINK_ESTABLISH_EVT:
621 		case LINK_FAILOVER_END_EVT:
622 			break;
623 		case LINK_SYNCH_BEGIN_EVT:
624 		case LINK_SYNCH_END_EVT:
625 		default:
626 			goto illegal_evt;
627 		}
628 		break;
629 	case LINK_PEER_RESET:
630 		switch (evt) {
631 		case LINK_RESET_EVT:
632 			l->state = LINK_ESTABLISHING;
633 			break;
634 		case LINK_PEER_RESET_EVT:
635 		case LINK_ESTABLISH_EVT:
636 		case LINK_FAILURE_EVT:
637 			break;
638 		case LINK_SYNCH_BEGIN_EVT:
639 		case LINK_SYNCH_END_EVT:
640 		case LINK_FAILOVER_BEGIN_EVT:
641 		case LINK_FAILOVER_END_EVT:
642 		default:
643 			goto illegal_evt;
644 		}
645 		break;
646 	case LINK_FAILINGOVER:
647 		switch (evt) {
648 		case LINK_FAILOVER_END_EVT:
649 			l->state = LINK_RESET;
650 			break;
651 		case LINK_PEER_RESET_EVT:
652 		case LINK_RESET_EVT:
653 		case LINK_ESTABLISH_EVT:
654 		case LINK_FAILURE_EVT:
655 			break;
656 		case LINK_FAILOVER_BEGIN_EVT:
657 		case LINK_SYNCH_BEGIN_EVT:
658 		case LINK_SYNCH_END_EVT:
659 		default:
660 			goto illegal_evt;
661 		}
662 		break;
663 	case LINK_ESTABLISHING:
664 		switch (evt) {
665 		case LINK_ESTABLISH_EVT:
666 			l->state = LINK_ESTABLISHED;
667 			break;
668 		case LINK_FAILOVER_BEGIN_EVT:
669 			l->state = LINK_FAILINGOVER;
670 			break;
671 		case LINK_RESET_EVT:
672 			l->state = LINK_RESET;
673 			break;
674 		case LINK_FAILURE_EVT:
675 		case LINK_PEER_RESET_EVT:
676 		case LINK_SYNCH_BEGIN_EVT:
677 		case LINK_FAILOVER_END_EVT:
678 			break;
679 		case LINK_SYNCH_END_EVT:
680 		default:
681 			goto illegal_evt;
682 		}
683 		break;
684 	case LINK_ESTABLISHED:
685 		switch (evt) {
686 		case LINK_PEER_RESET_EVT:
687 			l->state = LINK_PEER_RESET;
688 			rc |= TIPC_LINK_DOWN_EVT;
689 			break;
690 		case LINK_FAILURE_EVT:
691 			l->state = LINK_RESETTING;
692 			rc |= TIPC_LINK_DOWN_EVT;
693 			break;
694 		case LINK_RESET_EVT:
695 			l->state = LINK_RESET;
696 			break;
697 		case LINK_ESTABLISH_EVT:
698 		case LINK_SYNCH_END_EVT:
699 			break;
700 		case LINK_SYNCH_BEGIN_EVT:
701 			l->state = LINK_SYNCHING;
702 			break;
703 		case LINK_FAILOVER_BEGIN_EVT:
704 		case LINK_FAILOVER_END_EVT:
705 		default:
706 			goto illegal_evt;
707 		}
708 		break;
709 	case LINK_SYNCHING:
710 		switch (evt) {
711 		case LINK_PEER_RESET_EVT:
712 			l->state = LINK_PEER_RESET;
713 			rc |= TIPC_LINK_DOWN_EVT;
714 			break;
715 		case LINK_FAILURE_EVT:
716 			l->state = LINK_RESETTING;
717 			rc |= TIPC_LINK_DOWN_EVT;
718 			break;
719 		case LINK_RESET_EVT:
720 			l->state = LINK_RESET;
721 			break;
722 		case LINK_ESTABLISH_EVT:
723 		case LINK_SYNCH_BEGIN_EVT:
724 			break;
725 		case LINK_SYNCH_END_EVT:
726 			l->state = LINK_ESTABLISHED;
727 			break;
728 		case LINK_FAILOVER_BEGIN_EVT:
729 		case LINK_FAILOVER_END_EVT:
730 		default:
731 			goto illegal_evt;
732 		}
733 		break;
734 	default:
735 		pr_err("Unknown FSM state %x in %s\n", l->state, l->name);
736 	}
737 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
738 	return rc;
739 illegal_evt:
740 	pr_err("Illegal FSM event %x in state %x on link %s\n",
741 	       evt, l->state, l->name);
742 	trace_tipc_link_fsm(l->name, old_state, l->state, evt);
743 	return rc;
744 }
745 
746 /* link_profile_stats - update statistical profiling of traffic
747  */
link_profile_stats(struct tipc_link * l)748 static void link_profile_stats(struct tipc_link *l)
749 {
750 	struct sk_buff *skb;
751 	struct tipc_msg *msg;
752 	int length;
753 
754 	/* Update counters used in statistical profiling of send traffic */
755 	l->stats.accu_queue_sz += skb_queue_len(&l->transmq);
756 	l->stats.queue_sz_counts++;
757 
758 	skb = skb_peek(&l->transmq);
759 	if (!skb)
760 		return;
761 	msg = buf_msg(skb);
762 	length = msg_size(msg);
763 
764 	if (msg_user(msg) == MSG_FRAGMENTER) {
765 		if (msg_type(msg) != FIRST_FRAGMENT)
766 			return;
767 		length = msg_size(msg_inner_hdr(msg));
768 	}
769 	l->stats.msg_lengths_total += length;
770 	l->stats.msg_length_counts++;
771 	if (length <= 64)
772 		l->stats.msg_length_profile[0]++;
773 	else if (length <= 256)
774 		l->stats.msg_length_profile[1]++;
775 	else if (length <= 1024)
776 		l->stats.msg_length_profile[2]++;
777 	else if (length <= 4096)
778 		l->stats.msg_length_profile[3]++;
779 	else if (length <= 16384)
780 		l->stats.msg_length_profile[4]++;
781 	else if (length <= 32768)
782 		l->stats.msg_length_profile[5]++;
783 	else
784 		l->stats.msg_length_profile[6]++;
785 }
786 
787 /**
788  * tipc_link_too_silent - check if link is "too silent"
789  * @l: tipc link to be checked
790  *
791  * Returns true if the link 'silent_intv_cnt' is about to reach the
792  * 'abort_limit' value, otherwise false
793  */
tipc_link_too_silent(struct tipc_link * l)794 bool tipc_link_too_silent(struct tipc_link *l)
795 {
796 	return (l->silent_intv_cnt + 2 > l->abort_limit);
797 }
798 
799 /* tipc_link_timeout - perform periodic task as instructed from node timeout
800  */
tipc_link_timeout(struct tipc_link * l,struct sk_buff_head * xmitq)801 int tipc_link_timeout(struct tipc_link *l, struct sk_buff_head *xmitq)
802 {
803 	int mtyp = 0;
804 	int rc = 0;
805 	bool state = false;
806 	bool probe = false;
807 	bool setup = false;
808 	u16 bc_snt = l->bc_sndlink->snd_nxt - 1;
809 	u16 bc_acked = l->bc_rcvlink->acked;
810 	struct tipc_mon_state *mstate = &l->mon_state;
811 
812 	trace_tipc_link_timeout(l, TIPC_DUMP_NONE, " ");
813 	trace_tipc_link_too_silent(l, TIPC_DUMP_ALL, " ");
814 	switch (l->state) {
815 	case LINK_ESTABLISHED:
816 	case LINK_SYNCHING:
817 		mtyp = STATE_MSG;
818 		link_profile_stats(l);
819 		tipc_mon_get_state(l->net, l->addr, mstate, l->bearer_id);
820 		if (mstate->reset || (l->silent_intv_cnt > l->abort_limit))
821 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
822 		state = bc_acked != bc_snt;
823 		state |= l->bc_rcvlink->rcv_unacked;
824 		state |= l->rcv_unacked;
825 		state |= !skb_queue_empty(&l->transmq);
826 		probe = mstate->probing;
827 		probe |= l->silent_intv_cnt;
828 		if (probe || mstate->monitoring)
829 			l->silent_intv_cnt++;
830 		probe |= !skb_queue_empty(&l->deferdq);
831 		if (l->snd_nxt == l->checkpoint) {
832 			tipc_link_update_cwin(l, 0, 0);
833 			probe = true;
834 		}
835 		l->checkpoint = l->snd_nxt;
836 		break;
837 	case LINK_RESET:
838 		setup = l->rst_cnt++ <= 4;
839 		setup |= !(l->rst_cnt % 16);
840 		mtyp = RESET_MSG;
841 		break;
842 	case LINK_ESTABLISHING:
843 		setup = true;
844 		mtyp = ACTIVATE_MSG;
845 		break;
846 	case LINK_PEER_RESET:
847 	case LINK_RESETTING:
848 	case LINK_FAILINGOVER:
849 		break;
850 	default:
851 		break;
852 	}
853 
854 	if (state || probe || setup)
855 		tipc_link_build_proto_msg(l, mtyp, probe, 0, 0, 0, 0, xmitq);
856 
857 	return rc;
858 }
859 
860 /**
861  * link_schedule_user - schedule a message sender for wakeup after congestion
862  * @l: congested link
863  * @hdr: header of message that is being sent
864  * Create pseudo msg to send back to user when congestion abates
865  */
link_schedule_user(struct tipc_link * l,struct tipc_msg * hdr)866 static int link_schedule_user(struct tipc_link *l, struct tipc_msg *hdr)
867 {
868 	u32 dnode = tipc_own_addr(l->net);
869 	u32 dport = msg_origport(hdr);
870 	struct sk_buff *skb;
871 
872 	/* Create and schedule wakeup pseudo message */
873 	skb = tipc_msg_create(SOCK_WAKEUP, 0, INT_H_SIZE, 0,
874 			      dnode, l->addr, dport, 0, 0);
875 	if (!skb)
876 		return -ENOBUFS;
877 	msg_set_dest_droppable(buf_msg(skb), true);
878 	TIPC_SKB_CB(skb)->chain_imp = msg_importance(hdr);
879 	skb_queue_tail(&l->wakeupq, skb);
880 	l->stats.link_congs++;
881 	trace_tipc_link_conges(l, TIPC_DUMP_ALL, "wakeup scheduled!");
882 	return -ELINKCONG;
883 }
884 
885 /**
886  * link_prepare_wakeup - prepare users for wakeup after congestion
887  * @l: congested link
888  * Wake up a number of waiting users, as permitted by available space
889  * in the send queue
890  */
link_prepare_wakeup(struct tipc_link * l)891 static void link_prepare_wakeup(struct tipc_link *l)
892 {
893 	struct sk_buff_head *wakeupq = &l->wakeupq;
894 	struct sk_buff_head *inputq = l->inputq;
895 	struct sk_buff *skb, *tmp;
896 	struct sk_buff_head tmpq;
897 	int avail[5] = {0,};
898 	int imp = 0;
899 
900 	__skb_queue_head_init(&tmpq);
901 
902 	for (; imp <= TIPC_SYSTEM_IMPORTANCE; imp++)
903 		avail[imp] = l->backlog[imp].limit - l->backlog[imp].len;
904 
905 	skb_queue_walk_safe(wakeupq, skb, tmp) {
906 		imp = TIPC_SKB_CB(skb)->chain_imp;
907 		if (avail[imp] <= 0)
908 			continue;
909 		avail[imp]--;
910 		__skb_unlink(skb, wakeupq);
911 		__skb_queue_tail(&tmpq, skb);
912 	}
913 
914 	spin_lock_bh(&inputq->lock);
915 	skb_queue_splice_tail(&tmpq, inputq);
916 	spin_unlock_bh(&inputq->lock);
917 
918 }
919 
920 /**
921  * tipc_link_set_skb_retransmit_time - set the time at which retransmission of
922  *                                     the given skb should be next attempted
923  * @skb: skb to set a future retransmission time for
924  * @l: link the skb will be transmitted on
925  */
tipc_link_set_skb_retransmit_time(struct sk_buff * skb,struct tipc_link * l)926 static void tipc_link_set_skb_retransmit_time(struct sk_buff *skb,
927 					      struct tipc_link *l)
928 {
929 	if (link_is_bc_sndlink(l))
930 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_BC_RETR_LIM;
931 	else
932 		TIPC_SKB_CB(skb)->nxt_retr = TIPC_UC_RETR_TIME;
933 }
934 
tipc_link_reset(struct tipc_link * l)935 void tipc_link_reset(struct tipc_link *l)
936 {
937 	struct sk_buff_head list;
938 	u32 imp;
939 
940 	__skb_queue_head_init(&list);
941 
942 	l->in_session = false;
943 	/* Force re-synch of peer session number before establishing */
944 	l->peer_session--;
945 	l->session++;
946 	l->mtu = l->advertised_mtu;
947 
948 	spin_lock_bh(&l->wakeupq.lock);
949 	skb_queue_splice_init(&l->wakeupq, &list);
950 	spin_unlock_bh(&l->wakeupq.lock);
951 
952 	spin_lock_bh(&l->inputq->lock);
953 	skb_queue_splice_init(&list, l->inputq);
954 	spin_unlock_bh(&l->inputq->lock);
955 
956 	__skb_queue_purge(&l->transmq);
957 	__skb_queue_purge(&l->deferdq);
958 	__skb_queue_purge(&l->backlogq);
959 	__skb_queue_purge(&l->failover_deferdq);
960 	for (imp = 0; imp <= TIPC_SYSTEM_IMPORTANCE; imp++) {
961 		l->backlog[imp].len = 0;
962 		l->backlog[imp].target_bskb = NULL;
963 	}
964 	kfree_skb(l->reasm_buf);
965 	kfree_skb(l->reasm_tnlmsg);
966 	kfree_skb(l->failover_reasm_skb);
967 	l->reasm_buf = NULL;
968 	l->reasm_tnlmsg = NULL;
969 	l->failover_reasm_skb = NULL;
970 	l->rcv_unacked = 0;
971 	l->snd_nxt = 1;
972 	l->rcv_nxt = 1;
973 	l->snd_nxt_state = 1;
974 	l->rcv_nxt_state = 1;
975 	l->acked = 0;
976 	l->last_gap = 0;
977 	kfree(l->last_ga);
978 	l->last_ga = NULL;
979 	l->silent_intv_cnt = 0;
980 	l->rst_cnt = 0;
981 	l->bc_peer_is_up = false;
982 	memset(&l->mon_state, 0, sizeof(l->mon_state));
983 	tipc_link_reset_stats(l);
984 }
985 
986 /**
987  * tipc_link_xmit(): enqueue buffer list according to queue situation
988  * @l: link to use
989  * @list: chain of buffers containing message
990  * @xmitq: returned list of packets to be sent by caller
991  *
992  * Consumes the buffer chain.
993  * Returns 0 if success, or errno: -ELINKCONG, -EMSGSIZE or -ENOBUFS
994  * Messages at TIPC_SYSTEM_IMPORTANCE are always accepted
995  */
tipc_link_xmit(struct tipc_link * l,struct sk_buff_head * list,struct sk_buff_head * xmitq)996 int tipc_link_xmit(struct tipc_link *l, struct sk_buff_head *list,
997 		   struct sk_buff_head *xmitq)
998 {
999 	struct sk_buff_head *backlogq = &l->backlogq;
1000 	struct sk_buff_head *transmq = &l->transmq;
1001 	struct sk_buff *skb, *_skb;
1002 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1003 	u16 ack = l->rcv_nxt - 1;
1004 	u16 seqno = l->snd_nxt;
1005 	int pkt_cnt = skb_queue_len(list);
1006 	unsigned int mss = tipc_link_mss(l);
1007 	unsigned int cwin = l->window;
1008 	unsigned int mtu = l->mtu;
1009 	struct tipc_msg *hdr;
1010 	bool new_bundle;
1011 	int rc = 0;
1012 	int imp;
1013 
1014 	if (pkt_cnt <= 0)
1015 		return 0;
1016 
1017 	hdr = buf_msg(skb_peek(list));
1018 	if (unlikely(msg_size(hdr) > mtu)) {
1019 		pr_warn("Too large msg, purging xmit list %d %d %d %d %d!\n",
1020 			skb_queue_len(list), msg_user(hdr),
1021 			msg_type(hdr), msg_size(hdr), mtu);
1022 		__skb_queue_purge(list);
1023 		return -EMSGSIZE;
1024 	}
1025 
1026 	imp = msg_importance(hdr);
1027 	/* Allow oversubscription of one data msg per source at congestion */
1028 	if (unlikely(l->backlog[imp].len >= l->backlog[imp].limit)) {
1029 		if (imp == TIPC_SYSTEM_IMPORTANCE) {
1030 			pr_warn("%s<%s>, link overflow", link_rst_msg, l->name);
1031 			return -ENOBUFS;
1032 		}
1033 		rc = link_schedule_user(l, hdr);
1034 	}
1035 
1036 	if (pkt_cnt > 1) {
1037 		l->stats.sent_fragmented++;
1038 		l->stats.sent_fragments += pkt_cnt;
1039 	}
1040 
1041 	/* Prepare each packet for sending, and add to relevant queue: */
1042 	while ((skb = __skb_dequeue(list))) {
1043 		if (likely(skb_queue_len(transmq) < cwin)) {
1044 			hdr = buf_msg(skb);
1045 			msg_set_seqno(hdr, seqno);
1046 			msg_set_ack(hdr, ack);
1047 			msg_set_bcast_ack(hdr, bc_ack);
1048 			_skb = skb_clone(skb, GFP_ATOMIC);
1049 			if (!_skb) {
1050 				kfree_skb(skb);
1051 				__skb_queue_purge(list);
1052 				return -ENOBUFS;
1053 			}
1054 			__skb_queue_tail(transmq, skb);
1055 			tipc_link_set_skb_retransmit_time(skb, l);
1056 			__skb_queue_tail(xmitq, _skb);
1057 			TIPC_SKB_CB(skb)->ackers = l->ackers;
1058 			l->rcv_unacked = 0;
1059 			l->stats.sent_pkts++;
1060 			seqno++;
1061 			continue;
1062 		}
1063 		if (tipc_msg_try_bundle(l->backlog[imp].target_bskb, &skb,
1064 					mss, l->addr, &new_bundle)) {
1065 			if (skb) {
1066 				/* Keep a ref. to the skb for next try */
1067 				l->backlog[imp].target_bskb = skb;
1068 				l->backlog[imp].len++;
1069 				__skb_queue_tail(backlogq, skb);
1070 			} else {
1071 				if (new_bundle) {
1072 					l->stats.sent_bundles++;
1073 					l->stats.sent_bundled++;
1074 				}
1075 				l->stats.sent_bundled++;
1076 			}
1077 			continue;
1078 		}
1079 		l->backlog[imp].target_bskb = NULL;
1080 		l->backlog[imp].len += (1 + skb_queue_len(list));
1081 		__skb_queue_tail(backlogq, skb);
1082 		skb_queue_splice_tail_init(list, backlogq);
1083 	}
1084 	l->snd_nxt = seqno;
1085 	return rc;
1086 }
1087 
tipc_link_update_cwin(struct tipc_link * l,int released,bool retransmitted)1088 static void tipc_link_update_cwin(struct tipc_link *l, int released,
1089 				  bool retransmitted)
1090 {
1091 	int bklog_len = skb_queue_len(&l->backlogq);
1092 	struct sk_buff_head *txq = &l->transmq;
1093 	int txq_len = skb_queue_len(txq);
1094 	u16 cwin = l->window;
1095 
1096 	/* Enter fast recovery */
1097 	if (unlikely(retransmitted)) {
1098 		l->ssthresh = max_t(u16, l->window / 2, 300);
1099 		l->window = min_t(u16, l->ssthresh, l->window);
1100 		return;
1101 	}
1102 	/* Enter slow start */
1103 	if (unlikely(!released)) {
1104 		l->ssthresh = max_t(u16, l->window / 2, 300);
1105 		l->window = l->min_win;
1106 		return;
1107 	}
1108 	/* Don't increase window if no pressure on the transmit queue */
1109 	if (txq_len + bklog_len < cwin)
1110 		return;
1111 
1112 	/* Don't increase window if there are holes the transmit queue */
1113 	if (txq_len && l->snd_nxt - buf_seqno(skb_peek(txq)) != txq_len)
1114 		return;
1115 
1116 	l->cong_acks += released;
1117 
1118 	/* Slow start  */
1119 	if (cwin <= l->ssthresh) {
1120 		l->window = min_t(u16, cwin + released, l->max_win);
1121 		return;
1122 	}
1123 	/* Congestion avoidance */
1124 	if (l->cong_acks < cwin)
1125 		return;
1126 	l->window = min_t(u16, ++cwin, l->max_win);
1127 	l->cong_acks = 0;
1128 }
1129 
tipc_link_advance_backlog(struct tipc_link * l,struct sk_buff_head * xmitq)1130 static void tipc_link_advance_backlog(struct tipc_link *l,
1131 				      struct sk_buff_head *xmitq)
1132 {
1133 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1134 	struct sk_buff_head *txq = &l->transmq;
1135 	struct sk_buff *skb, *_skb;
1136 	u16 ack = l->rcv_nxt - 1;
1137 	u16 seqno = l->snd_nxt;
1138 	struct tipc_msg *hdr;
1139 	u16 cwin = l->window;
1140 	u32 imp;
1141 
1142 	while (skb_queue_len(txq) < cwin) {
1143 		skb = skb_peek(&l->backlogq);
1144 		if (!skb)
1145 			break;
1146 		_skb = skb_clone(skb, GFP_ATOMIC);
1147 		if (!_skb)
1148 			break;
1149 		__skb_dequeue(&l->backlogq);
1150 		hdr = buf_msg(skb);
1151 		imp = msg_importance(hdr);
1152 		l->backlog[imp].len--;
1153 		if (unlikely(skb == l->backlog[imp].target_bskb))
1154 			l->backlog[imp].target_bskb = NULL;
1155 		__skb_queue_tail(&l->transmq, skb);
1156 		tipc_link_set_skb_retransmit_time(skb, l);
1157 
1158 		__skb_queue_tail(xmitq, _skb);
1159 		TIPC_SKB_CB(skb)->ackers = l->ackers;
1160 		msg_set_seqno(hdr, seqno);
1161 		msg_set_ack(hdr, ack);
1162 		msg_set_bcast_ack(hdr, bc_ack);
1163 		l->rcv_unacked = 0;
1164 		l->stats.sent_pkts++;
1165 		seqno++;
1166 	}
1167 	l->snd_nxt = seqno;
1168 }
1169 
1170 /**
1171  * link_retransmit_failure() - Detect repeated retransmit failures
1172  * @l: tipc link sender
1173  * @r: tipc link receiver (= l in case of unicast)
1174  * @rc: returned code
1175  *
1176  * Return: true if the repeated retransmit failures happens, otherwise
1177  * false
1178  */
link_retransmit_failure(struct tipc_link * l,struct tipc_link * r,int * rc)1179 static bool link_retransmit_failure(struct tipc_link *l, struct tipc_link *r,
1180 				    int *rc)
1181 {
1182 	struct sk_buff *skb = skb_peek(&l->transmq);
1183 	struct tipc_msg *hdr;
1184 
1185 	if (!skb)
1186 		return false;
1187 
1188 	if (!TIPC_SKB_CB(skb)->retr_cnt)
1189 		return false;
1190 
1191 	if (!time_after(jiffies, TIPC_SKB_CB(skb)->retr_stamp +
1192 			msecs_to_jiffies(r->tolerance * 10)))
1193 		return false;
1194 
1195 	hdr = buf_msg(skb);
1196 	if (link_is_bc_sndlink(l) && !less(r->acked, msg_seqno(hdr)))
1197 		return false;
1198 
1199 	pr_warn("Retransmission failure on link <%s>\n", l->name);
1200 	link_print(l, "State of link ");
1201 	pr_info("Failed msg: usr %u, typ %u, len %u, err %u\n",
1202 		msg_user(hdr), msg_type(hdr), msg_size(hdr), msg_errcode(hdr));
1203 	pr_info("sqno %u, prev: %x, dest: %x\n",
1204 		msg_seqno(hdr), msg_prevnode(hdr), msg_destnode(hdr));
1205 	pr_info("retr_stamp %d, retr_cnt %d\n",
1206 		jiffies_to_msecs(TIPC_SKB_CB(skb)->retr_stamp),
1207 		TIPC_SKB_CB(skb)->retr_cnt);
1208 
1209 	trace_tipc_list_dump(&l->transmq, true, "retrans failure!");
1210 	trace_tipc_link_dump(l, TIPC_DUMP_NONE, "retrans failure!");
1211 	trace_tipc_link_dump(r, TIPC_DUMP_NONE, "retrans failure!");
1212 
1213 	if (link_is_bc_sndlink(l)) {
1214 		r->state = LINK_RESET;
1215 		*rc |= TIPC_LINK_DOWN_EVT;
1216 	} else {
1217 		*rc |= tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1218 	}
1219 
1220 	return true;
1221 }
1222 
1223 /* tipc_data_input - deliver data and name distr msgs to upper layer
1224  *
1225  * Consumes buffer if message is of right type
1226  * Node lock must be held
1227  */
tipc_data_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1228 static bool tipc_data_input(struct tipc_link *l, struct sk_buff *skb,
1229 			    struct sk_buff_head *inputq)
1230 {
1231 	struct sk_buff_head *mc_inputq = l->bc_rcvlink->inputq;
1232 	struct tipc_msg *hdr = buf_msg(skb);
1233 
1234 	switch (msg_user(hdr)) {
1235 	case TIPC_LOW_IMPORTANCE:
1236 	case TIPC_MEDIUM_IMPORTANCE:
1237 	case TIPC_HIGH_IMPORTANCE:
1238 	case TIPC_CRITICAL_IMPORTANCE:
1239 		if (unlikely(msg_in_group(hdr) || msg_mcast(hdr))) {
1240 			skb_queue_tail(mc_inputq, skb);
1241 			return true;
1242 		}
1243 		fallthrough;
1244 	case CONN_MANAGER:
1245 		skb_queue_tail(inputq, skb);
1246 		return true;
1247 	case GROUP_PROTOCOL:
1248 		skb_queue_tail(mc_inputq, skb);
1249 		return true;
1250 	case NAME_DISTRIBUTOR:
1251 		l->bc_rcvlink->state = LINK_ESTABLISHED;
1252 		skb_queue_tail(l->namedq, skb);
1253 		return true;
1254 	case MSG_BUNDLER:
1255 	case TUNNEL_PROTOCOL:
1256 	case MSG_FRAGMENTER:
1257 	case BCAST_PROTOCOL:
1258 		return false;
1259 #ifdef CONFIG_TIPC_CRYPTO
1260 	case MSG_CRYPTO:
1261 		tipc_crypto_msg_rcv(l->net, skb);
1262 		return true;
1263 #endif
1264 	default:
1265 		pr_warn("Dropping received illegal msg type\n");
1266 		kfree_skb(skb);
1267 		return true;
1268 	};
1269 }
1270 
1271 /* tipc_link_input - process packet that has passed link protocol check
1272  *
1273  * Consumes buffer
1274  */
tipc_link_input(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq,struct sk_buff ** reasm_skb)1275 static int tipc_link_input(struct tipc_link *l, struct sk_buff *skb,
1276 			   struct sk_buff_head *inputq,
1277 			   struct sk_buff **reasm_skb)
1278 {
1279 	struct tipc_msg *hdr = buf_msg(skb);
1280 	struct sk_buff *iskb;
1281 	struct sk_buff_head tmpq;
1282 	int usr = msg_user(hdr);
1283 	int pos = 0;
1284 
1285 	if (usr == MSG_BUNDLER) {
1286 		skb_queue_head_init(&tmpq);
1287 		l->stats.recv_bundles++;
1288 		l->stats.recv_bundled += msg_msgcnt(hdr);
1289 		while (tipc_msg_extract(skb, &iskb, &pos))
1290 			tipc_data_input(l, iskb, &tmpq);
1291 		tipc_skb_queue_splice_tail(&tmpq, inputq);
1292 		return 0;
1293 	} else if (usr == MSG_FRAGMENTER) {
1294 		l->stats.recv_fragments++;
1295 		if (tipc_buf_append(reasm_skb, &skb)) {
1296 			l->stats.recv_fragmented++;
1297 			tipc_data_input(l, skb, inputq);
1298 		} else if (!*reasm_skb && !link_is_bc_rcvlink(l)) {
1299 			pr_warn_ratelimited("Unable to build fragment list\n");
1300 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1301 		}
1302 		return 0;
1303 	} else if (usr == BCAST_PROTOCOL) {
1304 		tipc_bcast_lock(l->net);
1305 		tipc_link_bc_init_rcv(l->bc_rcvlink, hdr);
1306 		tipc_bcast_unlock(l->net);
1307 	}
1308 
1309 	kfree_skb(skb);
1310 	return 0;
1311 }
1312 
1313 /* tipc_link_tnl_rcv() - receive TUNNEL_PROTOCOL message, drop or process the
1314  *			 inner message along with the ones in the old link's
1315  *			 deferdq
1316  * @l: tunnel link
1317  * @skb: TUNNEL_PROTOCOL message
1318  * @inputq: queue to put messages ready for delivery
1319  */
tipc_link_tnl_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * inputq)1320 static int tipc_link_tnl_rcv(struct tipc_link *l, struct sk_buff *skb,
1321 			     struct sk_buff_head *inputq)
1322 {
1323 	struct sk_buff **reasm_skb = &l->failover_reasm_skb;
1324 	struct sk_buff **reasm_tnlmsg = &l->reasm_tnlmsg;
1325 	struct sk_buff_head *fdefq = &l->failover_deferdq;
1326 	struct tipc_msg *hdr = buf_msg(skb);
1327 	struct sk_buff *iskb;
1328 	int ipos = 0;
1329 	int rc = 0;
1330 	u16 seqno;
1331 
1332 	if (msg_type(hdr) == SYNCH_MSG) {
1333 		kfree_skb(skb);
1334 		return 0;
1335 	}
1336 
1337 	/* Not a fragment? */
1338 	if (likely(!msg_nof_fragms(hdr))) {
1339 		if (unlikely(!tipc_msg_extract(skb, &iskb, &ipos))) {
1340 			pr_warn_ratelimited("Unable to extract msg, defq: %d\n",
1341 					    skb_queue_len(fdefq));
1342 			return 0;
1343 		}
1344 		kfree_skb(skb);
1345 	} else {
1346 		/* Set fragment type for buf_append */
1347 		if (msg_fragm_no(hdr) == 1)
1348 			msg_set_type(hdr, FIRST_FRAGMENT);
1349 		else if (msg_fragm_no(hdr) < msg_nof_fragms(hdr))
1350 			msg_set_type(hdr, FRAGMENT);
1351 		else
1352 			msg_set_type(hdr, LAST_FRAGMENT);
1353 
1354 		if (!tipc_buf_append(reasm_tnlmsg, &skb)) {
1355 			/* Successful but non-complete reassembly? */
1356 			if (*reasm_tnlmsg || link_is_bc_rcvlink(l))
1357 				return 0;
1358 			pr_warn_ratelimited("Unable to reassemble tunnel msg\n");
1359 			return tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
1360 		}
1361 		iskb = skb;
1362 	}
1363 
1364 	do {
1365 		seqno = buf_seqno(iskb);
1366 		if (unlikely(less(seqno, l->drop_point))) {
1367 			kfree_skb(iskb);
1368 			continue;
1369 		}
1370 		if (unlikely(seqno != l->drop_point)) {
1371 			__tipc_skb_queue_sorted(fdefq, seqno, iskb);
1372 			continue;
1373 		}
1374 
1375 		l->drop_point++;
1376 		if (!tipc_data_input(l, iskb, inputq))
1377 			rc |= tipc_link_input(l, iskb, inputq, reasm_skb);
1378 		if (unlikely(rc))
1379 			break;
1380 	} while ((iskb = __tipc_skb_dequeue(fdefq, l->drop_point)));
1381 
1382 	return rc;
1383 }
1384 
1385 /**
1386  * tipc_get_gap_ack_blks - get Gap ACK blocks from PROTOCOL/STATE_MSG
1387  * @ga: returned pointer to the Gap ACK blocks if any
1388  * @l: the tipc link
1389  * @hdr: the PROTOCOL/STATE_MSG header
1390  * @uc: desired Gap ACK blocks type, i.e. unicast (= 1) or broadcast (= 0)
1391  *
1392  * Return: the total Gap ACK blocks size
1393  */
tipc_get_gap_ack_blks(struct tipc_gap_ack_blks ** ga,struct tipc_link * l,struct tipc_msg * hdr,bool uc)1394 u16 tipc_get_gap_ack_blks(struct tipc_gap_ack_blks **ga, struct tipc_link *l,
1395 			  struct tipc_msg *hdr, bool uc)
1396 {
1397 	struct tipc_gap_ack_blks *p;
1398 	u16 sz = 0;
1399 
1400 	/* Does peer support the Gap ACK blocks feature? */
1401 	if (l->peer_caps & TIPC_GAP_ACK_BLOCK) {
1402 		p = (struct tipc_gap_ack_blks *)msg_data(hdr);
1403 		sz = ntohs(p->len);
1404 		/* Sanity check */
1405 		if (sz == struct_size(p, gacks, p->ugack_cnt + p->bgack_cnt)) {
1406 			/* Good, check if the desired type exists */
1407 			if ((uc && p->ugack_cnt) || (!uc && p->bgack_cnt))
1408 				goto ok;
1409 		/* Backward compatible: peer might not support bc, but uc? */
1410 		} else if (uc && sz == struct_size(p, gacks, p->ugack_cnt)) {
1411 			if (p->ugack_cnt) {
1412 				p->bgack_cnt = 0;
1413 				goto ok;
1414 			}
1415 		}
1416 	}
1417 	/* Other cases: ignore! */
1418 	p = NULL;
1419 
1420 ok:
1421 	*ga = p;
1422 	return sz;
1423 }
1424 
__tipc_build_gap_ack_blks(struct tipc_gap_ack_blks * ga,struct tipc_link * l,u8 start_index)1425 static u8 __tipc_build_gap_ack_blks(struct tipc_gap_ack_blks *ga,
1426 				    struct tipc_link *l, u8 start_index)
1427 {
1428 	struct tipc_gap_ack *gacks = &ga->gacks[start_index];
1429 	struct sk_buff *skb = skb_peek(&l->deferdq);
1430 	u16 expect, seqno = 0;
1431 	u8 n = 0;
1432 
1433 	if (!skb)
1434 		return 0;
1435 
1436 	expect = buf_seqno(skb);
1437 	skb_queue_walk(&l->deferdq, skb) {
1438 		seqno = buf_seqno(skb);
1439 		if (unlikely(more(seqno, expect))) {
1440 			gacks[n].ack = htons(expect - 1);
1441 			gacks[n].gap = htons(seqno - expect);
1442 			if (++n >= MAX_GAP_ACK_BLKS / 2) {
1443 				pr_info_ratelimited("Gacks on %s: %d, ql: %d!\n",
1444 						    l->name, n,
1445 						    skb_queue_len(&l->deferdq));
1446 				return n;
1447 			}
1448 		} else if (unlikely(less(seqno, expect))) {
1449 			pr_warn("Unexpected skb in deferdq!\n");
1450 			continue;
1451 		}
1452 		expect = seqno + 1;
1453 	}
1454 
1455 	/* last block */
1456 	gacks[n].ack = htons(seqno);
1457 	gacks[n].gap = 0;
1458 	n++;
1459 	return n;
1460 }
1461 
1462 /* tipc_build_gap_ack_blks - build Gap ACK blocks
1463  * @l: tipc unicast link
1464  * @hdr: the tipc message buffer to store the Gap ACK blocks after built
1465  *
1466  * The function builds Gap ACK blocks for both the unicast & broadcast receiver
1467  * links of a certain peer, the buffer after built has the network data format
1468  * as found at the struct tipc_gap_ack_blks definition.
1469  *
1470  * returns the actual allocated memory size
1471  */
tipc_build_gap_ack_blks(struct tipc_link * l,struct tipc_msg * hdr)1472 static u16 tipc_build_gap_ack_blks(struct tipc_link *l, struct tipc_msg *hdr)
1473 {
1474 	struct tipc_link *bcl = l->bc_rcvlink;
1475 	struct tipc_gap_ack_blks *ga;
1476 	u16 len;
1477 
1478 	ga = (struct tipc_gap_ack_blks *)msg_data(hdr);
1479 
1480 	/* Start with broadcast link first */
1481 	tipc_bcast_lock(bcl->net);
1482 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1483 	msg_set_bc_gap(hdr, link_bc_rcv_gap(bcl));
1484 	ga->bgack_cnt = __tipc_build_gap_ack_blks(ga, bcl, 0);
1485 	tipc_bcast_unlock(bcl->net);
1486 
1487 	/* Now for unicast link, but an explicit NACK only (???) */
1488 	ga->ugack_cnt = (msg_seq_gap(hdr)) ?
1489 			__tipc_build_gap_ack_blks(ga, l, ga->bgack_cnt) : 0;
1490 
1491 	/* Total len */
1492 	len = struct_size(ga, gacks, ga->bgack_cnt + ga->ugack_cnt);
1493 	ga->len = htons(len);
1494 	return len;
1495 }
1496 
1497 /* tipc_link_advance_transmq - advance TIPC link transmq queue by releasing
1498  *			       acked packets, also doing retransmissions if
1499  *			       gaps found
1500  * @l: tipc link with transmq queue to be advanced
1501  * @r: tipc link "receiver" i.e. in case of broadcast (= "l" if unicast)
1502  * @acked: seqno of last packet acked by peer without any gaps before
1503  * @gap: # of gap packets
1504  * @ga: buffer pointer to Gap ACK blocks from peer
1505  * @xmitq: queue for accumulating the retransmitted packets if any
1506  * @retransmitted: returned boolean value if a retransmission is really issued
1507  * @rc: returned code e.g. TIPC_LINK_DOWN_EVT if a repeated retransmit failures
1508  *      happens (- unlikely case)
1509  *
1510  * Return: the number of packets released from the link transmq
1511  */
tipc_link_advance_transmq(struct tipc_link * l,struct tipc_link * r,u16 acked,u16 gap,struct tipc_gap_ack_blks * ga,struct sk_buff_head * xmitq,bool * retransmitted,int * rc)1512 static int tipc_link_advance_transmq(struct tipc_link *l, struct tipc_link *r,
1513 				     u16 acked, u16 gap,
1514 				     struct tipc_gap_ack_blks *ga,
1515 				     struct sk_buff_head *xmitq,
1516 				     bool *retransmitted, int *rc)
1517 {
1518 	struct tipc_gap_ack_blks *last_ga = r->last_ga, *this_ga = NULL;
1519 	struct tipc_gap_ack *gacks = NULL;
1520 	struct sk_buff *skb, *_skb, *tmp;
1521 	struct tipc_msg *hdr;
1522 	u32 qlen = skb_queue_len(&l->transmq);
1523 	u16 nacked = acked, ngap = gap, gack_cnt = 0;
1524 	u16 bc_ack = l->bc_rcvlink->rcv_nxt - 1;
1525 	u16 ack = l->rcv_nxt - 1;
1526 	u16 seqno, n = 0;
1527 	u16 end = r->acked, start = end, offset = r->last_gap;
1528 	u16 si = (last_ga) ? last_ga->start_index : 0;
1529 	bool is_uc = !link_is_bc_sndlink(l);
1530 	bool bc_has_acked = false;
1531 
1532 	trace_tipc_link_retrans(r, acked + 1, acked + gap, &l->transmq);
1533 
1534 	/* Determine Gap ACK blocks if any for the particular link */
1535 	if (ga && is_uc) {
1536 		/* Get the Gap ACKs, uc part */
1537 		gack_cnt = ga->ugack_cnt;
1538 		gacks = &ga->gacks[ga->bgack_cnt];
1539 	} else if (ga) {
1540 		/* Copy the Gap ACKs, bc part, for later renewal if needed */
1541 		this_ga = kmemdup(ga, struct_size(ga, gacks, ga->bgack_cnt),
1542 				  GFP_ATOMIC);
1543 		if (likely(this_ga)) {
1544 			this_ga->start_index = 0;
1545 			/* Start with the bc Gap ACKs */
1546 			gack_cnt = this_ga->bgack_cnt;
1547 			gacks = &this_ga->gacks[0];
1548 		} else {
1549 			/* Hmm, we can get in trouble..., simply ignore it */
1550 			pr_warn_ratelimited("Ignoring bc Gap ACKs, no memory\n");
1551 		}
1552 	}
1553 
1554 	/* Advance the link transmq */
1555 	skb_queue_walk_safe(&l->transmq, skb, tmp) {
1556 		seqno = buf_seqno(skb);
1557 
1558 next_gap_ack:
1559 		if (less_eq(seqno, nacked)) {
1560 			if (is_uc)
1561 				goto release;
1562 			/* Skip packets peer has already acked */
1563 			if (!more(seqno, r->acked))
1564 				continue;
1565 			/* Get the next of last Gap ACK blocks */
1566 			while (more(seqno, end)) {
1567 				if (!last_ga || si >= last_ga->bgack_cnt)
1568 					break;
1569 				start = end + offset + 1;
1570 				end = ntohs(last_ga->gacks[si].ack);
1571 				offset = ntohs(last_ga->gacks[si].gap);
1572 				si++;
1573 				WARN_ONCE(more(start, end) ||
1574 					  (!offset &&
1575 					   si < last_ga->bgack_cnt) ||
1576 					  si > MAX_GAP_ACK_BLKS,
1577 					  "Corrupted Gap ACK: %d %d %d %d %d\n",
1578 					  start, end, offset, si,
1579 					  last_ga->bgack_cnt);
1580 			}
1581 			/* Check against the last Gap ACK block */
1582 			if (in_range(seqno, start, end))
1583 				continue;
1584 			/* Update/release the packet peer is acking */
1585 			bc_has_acked = true;
1586 			if (--TIPC_SKB_CB(skb)->ackers)
1587 				continue;
1588 release:
1589 			/* release skb */
1590 			__skb_unlink(skb, &l->transmq);
1591 			kfree_skb(skb);
1592 		} else if (less_eq(seqno, nacked + ngap)) {
1593 			/* First gap: check if repeated retrans failures? */
1594 			if (unlikely(seqno == acked + 1 &&
1595 				     link_retransmit_failure(l, r, rc))) {
1596 				/* Ignore this bc Gap ACKs if any */
1597 				kfree(this_ga);
1598 				this_ga = NULL;
1599 				break;
1600 			}
1601 			/* retransmit skb if unrestricted*/
1602 			if (time_before(jiffies, TIPC_SKB_CB(skb)->nxt_retr))
1603 				continue;
1604 			tipc_link_set_skb_retransmit_time(skb, l);
1605 			_skb = pskb_copy(skb, GFP_ATOMIC);
1606 			if (!_skb)
1607 				continue;
1608 			hdr = buf_msg(_skb);
1609 			msg_set_ack(hdr, ack);
1610 			msg_set_bcast_ack(hdr, bc_ack);
1611 			_skb->priority = TC_PRIO_CONTROL;
1612 			__skb_queue_tail(xmitq, _skb);
1613 			l->stats.retransmitted++;
1614 			if (!is_uc)
1615 				r->stats.retransmitted++;
1616 			*retransmitted = true;
1617 			/* Increase actual retrans counter & mark first time */
1618 			if (!TIPC_SKB_CB(skb)->retr_cnt++)
1619 				TIPC_SKB_CB(skb)->retr_stamp = jiffies;
1620 		} else {
1621 			/* retry with Gap ACK blocks if any */
1622 			if (n >= gack_cnt)
1623 				break;
1624 			nacked = ntohs(gacks[n].ack);
1625 			ngap = ntohs(gacks[n].gap);
1626 			n++;
1627 			goto next_gap_ack;
1628 		}
1629 	}
1630 
1631 	/* Renew last Gap ACK blocks for bc if needed */
1632 	if (bc_has_acked) {
1633 		if (this_ga) {
1634 			kfree(last_ga);
1635 			r->last_ga = this_ga;
1636 			r->last_gap = gap;
1637 		} else if (last_ga) {
1638 			if (less(acked, start)) {
1639 				si--;
1640 				offset = start - acked - 1;
1641 			} else if (less(acked, end)) {
1642 				acked = end;
1643 			}
1644 			if (si < last_ga->bgack_cnt) {
1645 				last_ga->start_index = si;
1646 				r->last_gap = offset;
1647 			} else {
1648 				kfree(last_ga);
1649 				r->last_ga = NULL;
1650 				r->last_gap = 0;
1651 			}
1652 		} else {
1653 			r->last_gap = 0;
1654 		}
1655 		r->acked = acked;
1656 	} else {
1657 		kfree(this_ga);
1658 	}
1659 
1660 	return qlen - skb_queue_len(&l->transmq);
1661 }
1662 
1663 /* tipc_link_build_state_msg: prepare link state message for transmission
1664  *
1665  * Note that sending of broadcast ack is coordinated among nodes, to reduce
1666  * risk of ack storms towards the sender
1667  */
tipc_link_build_state_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1668 int tipc_link_build_state_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1669 {
1670 	if (!l)
1671 		return 0;
1672 
1673 	/* Broadcast ACK must be sent via a unicast link => defer to caller */
1674 	if (link_is_bc_rcvlink(l)) {
1675 		if (((l->rcv_nxt ^ tipc_own_addr(l->net)) & 0xf) != 0xf)
1676 			return 0;
1677 		l->rcv_unacked = 0;
1678 
1679 		/* Use snd_nxt to store peer's snd_nxt in broadcast rcv link */
1680 		l->snd_nxt = l->rcv_nxt;
1681 		return TIPC_LINK_SND_STATE;
1682 	}
1683 	/* Unicast ACK */
1684 	l->rcv_unacked = 0;
1685 	l->stats.sent_acks++;
1686 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, 0, xmitq);
1687 	return 0;
1688 }
1689 
1690 /* tipc_link_build_reset_msg: prepare link RESET or ACTIVATE message
1691  */
tipc_link_build_reset_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1692 void tipc_link_build_reset_msg(struct tipc_link *l, struct sk_buff_head *xmitq)
1693 {
1694 	int mtyp = RESET_MSG;
1695 	struct sk_buff *skb;
1696 
1697 	if (l->state == LINK_ESTABLISHING)
1698 		mtyp = ACTIVATE_MSG;
1699 
1700 	tipc_link_build_proto_msg(l, mtyp, 0, 0, 0, 0, 0, xmitq);
1701 
1702 	/* Inform peer that this endpoint is going down if applicable */
1703 	skb = skb_peek_tail(xmitq);
1704 	if (skb && (l->state == LINK_RESET))
1705 		msg_set_peer_stopping(buf_msg(skb), 1);
1706 }
1707 
1708 /* tipc_link_build_nack_msg: prepare link nack message for transmission
1709  * Note that sending of broadcast NACK is coordinated among nodes, to
1710  * reduce the risk of NACK storms towards the sender
1711  */
tipc_link_build_nack_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1712 static int tipc_link_build_nack_msg(struct tipc_link *l,
1713 				    struct sk_buff_head *xmitq)
1714 {
1715 	u32 def_cnt = ++l->stats.deferred_recv;
1716 	struct sk_buff_head *dfq = &l->deferdq;
1717 	u32 defq_len = skb_queue_len(dfq);
1718 	int match1, match2;
1719 
1720 	if (link_is_bc_rcvlink(l)) {
1721 		match1 = def_cnt & 0xf;
1722 		match2 = tipc_own_addr(l->net) & 0xf;
1723 		if (match1 == match2)
1724 			return TIPC_LINK_SND_STATE;
1725 		return 0;
1726 	}
1727 
1728 	if (defq_len >= 3 && !((defq_len - 3) % 16)) {
1729 		u16 rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1730 
1731 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0,
1732 					  rcvgap, 0, 0, xmitq);
1733 	}
1734 	return 0;
1735 }
1736 
1737 /* tipc_link_rcv - process TIPC packets/messages arriving from off-node
1738  * @l: the link that should handle the message
1739  * @skb: TIPC packet
1740  * @xmitq: queue to place packets to be sent after this call
1741  */
tipc_link_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)1742 int tipc_link_rcv(struct tipc_link *l, struct sk_buff *skb,
1743 		  struct sk_buff_head *xmitq)
1744 {
1745 	struct sk_buff_head *defq = &l->deferdq;
1746 	struct tipc_msg *hdr = buf_msg(skb);
1747 	u16 seqno, rcv_nxt, win_lim;
1748 	int released = 0;
1749 	int rc = 0;
1750 
1751 	/* Verify and update link state */
1752 	if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1753 		return tipc_link_proto_rcv(l, skb, xmitq);
1754 
1755 	/* Don't send probe at next timeout expiration */
1756 	l->silent_intv_cnt = 0;
1757 
1758 	do {
1759 		hdr = buf_msg(skb);
1760 		seqno = msg_seqno(hdr);
1761 		rcv_nxt = l->rcv_nxt;
1762 		win_lim = rcv_nxt + TIPC_MAX_LINK_WIN;
1763 
1764 		if (unlikely(!link_is_up(l))) {
1765 			if (l->state == LINK_ESTABLISHING)
1766 				rc = TIPC_LINK_UP_EVT;
1767 			kfree_skb(skb);
1768 			break;
1769 		}
1770 
1771 		/* Drop if outside receive window */
1772 		if (unlikely(less(seqno, rcv_nxt) || more(seqno, win_lim))) {
1773 			l->stats.duplicates++;
1774 			kfree_skb(skb);
1775 			break;
1776 		}
1777 		released += tipc_link_advance_transmq(l, l, msg_ack(hdr), 0,
1778 						      NULL, NULL, NULL, NULL);
1779 
1780 		/* Defer delivery if sequence gap */
1781 		if (unlikely(seqno != rcv_nxt)) {
1782 			if (!__tipc_skb_queue_sorted(defq, seqno, skb))
1783 				l->stats.duplicates++;
1784 			rc |= tipc_link_build_nack_msg(l, xmitq);
1785 			break;
1786 		}
1787 
1788 		/* Deliver packet */
1789 		l->rcv_nxt++;
1790 		l->stats.recv_pkts++;
1791 
1792 		if (unlikely(msg_user(hdr) == TUNNEL_PROTOCOL))
1793 			rc |= tipc_link_tnl_rcv(l, skb, l->inputq);
1794 		else if (!tipc_data_input(l, skb, l->inputq))
1795 			rc |= tipc_link_input(l, skb, l->inputq, &l->reasm_buf);
1796 		if (unlikely(++l->rcv_unacked >= TIPC_MIN_LINK_WIN))
1797 			rc |= tipc_link_build_state_msg(l, xmitq);
1798 		if (unlikely(rc & ~TIPC_LINK_SND_STATE))
1799 			break;
1800 	} while ((skb = __tipc_skb_dequeue(defq, l->rcv_nxt)));
1801 
1802 	/* Forward queues and wake up waiting users */
1803 	if (released) {
1804 		tipc_link_update_cwin(l, released, 0);
1805 		tipc_link_advance_backlog(l, xmitq);
1806 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
1807 			link_prepare_wakeup(l);
1808 	}
1809 	return rc;
1810 }
1811 
tipc_link_build_proto_msg(struct tipc_link * l,int mtyp,bool probe,bool probe_reply,u16 rcvgap,int tolerance,int priority,struct sk_buff_head * xmitq)1812 static void tipc_link_build_proto_msg(struct tipc_link *l, int mtyp, bool probe,
1813 				      bool probe_reply, u16 rcvgap,
1814 				      int tolerance, int priority,
1815 				      struct sk_buff_head *xmitq)
1816 {
1817 	struct tipc_mon_state *mstate = &l->mon_state;
1818 	struct sk_buff_head *dfq = &l->deferdq;
1819 	struct tipc_link *bcl = l->bc_rcvlink;
1820 	struct tipc_msg *hdr;
1821 	struct sk_buff *skb;
1822 	bool node_up = link_is_up(bcl);
1823 	u16 glen = 0, bc_rcvgap = 0;
1824 	int dlen = 0;
1825 	void *data;
1826 
1827 	/* Don't send protocol message during reset or link failover */
1828 	if (tipc_link_is_blocked(l))
1829 		return;
1830 
1831 	if (!tipc_link_is_up(l) && (mtyp == STATE_MSG))
1832 		return;
1833 
1834 	if ((probe || probe_reply) && !skb_queue_empty(dfq))
1835 		rcvgap = buf_seqno(skb_peek(dfq)) - l->rcv_nxt;
1836 
1837 	skb = tipc_msg_create(LINK_PROTOCOL, mtyp, INT_H_SIZE,
1838 			      tipc_max_domain_size + MAX_GAP_ACK_BLKS_SZ,
1839 			      l->addr, tipc_own_addr(l->net), 0, 0, 0);
1840 	if (!skb)
1841 		return;
1842 
1843 	hdr = buf_msg(skb);
1844 	data = msg_data(hdr);
1845 	msg_set_session(hdr, l->session);
1846 	msg_set_bearer_id(hdr, l->bearer_id);
1847 	msg_set_net_plane(hdr, l->net_plane);
1848 	msg_set_next_sent(hdr, l->snd_nxt);
1849 	msg_set_ack(hdr, l->rcv_nxt - 1);
1850 	msg_set_bcast_ack(hdr, bcl->rcv_nxt - 1);
1851 	msg_set_bc_ack_invalid(hdr, !node_up);
1852 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
1853 	msg_set_link_tolerance(hdr, tolerance);
1854 	msg_set_linkprio(hdr, priority);
1855 	msg_set_redundant_link(hdr, node_up);
1856 	msg_set_seq_gap(hdr, 0);
1857 	msg_set_seqno(hdr, l->snd_nxt + U16_MAX / 2);
1858 
1859 	if (mtyp == STATE_MSG) {
1860 		if (l->peer_caps & TIPC_LINK_PROTO_SEQNO)
1861 			msg_set_seqno(hdr, l->snd_nxt_state++);
1862 		msg_set_seq_gap(hdr, rcvgap);
1863 		bc_rcvgap = link_bc_rcv_gap(bcl);
1864 		msg_set_bc_gap(hdr, bc_rcvgap);
1865 		msg_set_probe(hdr, probe);
1866 		msg_set_is_keepalive(hdr, probe || probe_reply);
1867 		if (l->peer_caps & TIPC_GAP_ACK_BLOCK)
1868 			glen = tipc_build_gap_ack_blks(l, hdr);
1869 		tipc_mon_prep(l->net, data + glen, &dlen, mstate, l->bearer_id);
1870 		msg_set_size(hdr, INT_H_SIZE + glen + dlen);
1871 		skb_trim(skb, INT_H_SIZE + glen + dlen);
1872 		l->stats.sent_states++;
1873 		l->rcv_unacked = 0;
1874 	} else {
1875 		/* RESET_MSG or ACTIVATE_MSG */
1876 		if (mtyp == ACTIVATE_MSG) {
1877 			msg_set_dest_session_valid(hdr, 1);
1878 			msg_set_dest_session(hdr, l->peer_session);
1879 		}
1880 		msg_set_max_pkt(hdr, l->advertised_mtu);
1881 		strcpy(data, l->if_name);
1882 		msg_set_size(hdr, INT_H_SIZE + TIPC_MAX_IF_NAME);
1883 		skb_trim(skb, INT_H_SIZE + TIPC_MAX_IF_NAME);
1884 	}
1885 	if (probe)
1886 		l->stats.sent_probes++;
1887 	if (rcvgap)
1888 		l->stats.sent_nacks++;
1889 	if (bc_rcvgap)
1890 		bcl->stats.sent_nacks++;
1891 	skb->priority = TC_PRIO_CONTROL;
1892 	__skb_queue_tail(xmitq, skb);
1893 	trace_tipc_proto_build(skb, false, l->name);
1894 }
1895 
tipc_link_create_dummy_tnl_msg(struct tipc_link * l,struct sk_buff_head * xmitq)1896 void tipc_link_create_dummy_tnl_msg(struct tipc_link *l,
1897 				    struct sk_buff_head *xmitq)
1898 {
1899 	u32 onode = tipc_own_addr(l->net);
1900 	struct tipc_msg *hdr, *ihdr;
1901 	struct sk_buff_head tnlq;
1902 	struct sk_buff *skb;
1903 	u32 dnode = l->addr;
1904 
1905 	__skb_queue_head_init(&tnlq);
1906 	skb = tipc_msg_create(TUNNEL_PROTOCOL, FAILOVER_MSG,
1907 			      INT_H_SIZE, BASIC_H_SIZE,
1908 			      dnode, onode, 0, 0, 0);
1909 	if (!skb) {
1910 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1911 		return;
1912 	}
1913 
1914 	hdr = buf_msg(skb);
1915 	msg_set_msgcnt(hdr, 1);
1916 	msg_set_bearer_id(hdr, l->peer_bearer_id);
1917 
1918 	ihdr = (struct tipc_msg *)msg_data(hdr);
1919 	tipc_msg_init(onode, ihdr, TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1920 		      BASIC_H_SIZE, dnode);
1921 	msg_set_errcode(ihdr, TIPC_ERR_NO_PORT);
1922 	__skb_queue_tail(&tnlq, skb);
1923 	tipc_link_xmit(l, &tnlq, xmitq);
1924 }
1925 
1926 /* tipc_link_tnl_prepare(): prepare and return a list of tunnel packets
1927  * with contents of the link's transmit and backlog queues.
1928  */
tipc_link_tnl_prepare(struct tipc_link * l,struct tipc_link * tnl,int mtyp,struct sk_buff_head * xmitq)1929 void tipc_link_tnl_prepare(struct tipc_link *l, struct tipc_link *tnl,
1930 			   int mtyp, struct sk_buff_head *xmitq)
1931 {
1932 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
1933 	struct sk_buff *skb, *tnlskb;
1934 	struct tipc_msg *hdr, tnlhdr;
1935 	struct sk_buff_head *queue = &l->transmq;
1936 	struct sk_buff_head tmpxq, tnlq, frags;
1937 	u16 pktlen, pktcnt, seqno = l->snd_nxt;
1938 	bool pktcnt_need_update = false;
1939 	u16 syncpt;
1940 	int rc;
1941 
1942 	if (!tnl)
1943 		return;
1944 
1945 	__skb_queue_head_init(&tnlq);
1946 	/* Link Synching:
1947 	 * From now on, send only one single ("dummy") SYNCH message
1948 	 * to peer. The SYNCH message does not contain any data, just
1949 	 * a header conveying the synch point to the peer.
1950 	 */
1951 	if (mtyp == SYNCH_MSG && (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
1952 		tnlskb = tipc_msg_create(TUNNEL_PROTOCOL, SYNCH_MSG,
1953 					 INT_H_SIZE, 0, l->addr,
1954 					 tipc_own_addr(l->net),
1955 					 0, 0, 0);
1956 		if (!tnlskb) {
1957 			pr_warn("%sunable to create dummy SYNCH_MSG\n",
1958 				link_co_err);
1959 			return;
1960 		}
1961 
1962 		hdr = buf_msg(tnlskb);
1963 		syncpt = l->snd_nxt + skb_queue_len(&l->backlogq) - 1;
1964 		msg_set_syncpt(hdr, syncpt);
1965 		msg_set_bearer_id(hdr, l->peer_bearer_id);
1966 		__skb_queue_tail(&tnlq, tnlskb);
1967 		tipc_link_xmit(tnl, &tnlq, xmitq);
1968 		return;
1969 	}
1970 
1971 	__skb_queue_head_init(&tmpxq);
1972 	__skb_queue_head_init(&frags);
1973 	/* At least one packet required for safe algorithm => add dummy */
1974 	skb = tipc_msg_create(TIPC_LOW_IMPORTANCE, TIPC_DIRECT_MSG,
1975 			      BASIC_H_SIZE, 0, l->addr, tipc_own_addr(l->net),
1976 			      0, 0, TIPC_ERR_NO_PORT);
1977 	if (!skb) {
1978 		pr_warn("%sunable to create tunnel packet\n", link_co_err);
1979 		return;
1980 	}
1981 	__skb_queue_tail(&tnlq, skb);
1982 	tipc_link_xmit(l, &tnlq, &tmpxq);
1983 	__skb_queue_purge(&tmpxq);
1984 
1985 	/* Initialize reusable tunnel packet header */
1986 	tipc_msg_init(tipc_own_addr(l->net), &tnlhdr, TUNNEL_PROTOCOL,
1987 		      mtyp, INT_H_SIZE, l->addr);
1988 	if (mtyp == SYNCH_MSG)
1989 		pktcnt = l->snd_nxt - buf_seqno(skb_peek(&l->transmq));
1990 	else
1991 		pktcnt = skb_queue_len(&l->transmq);
1992 	pktcnt += skb_queue_len(&l->backlogq);
1993 	msg_set_msgcnt(&tnlhdr, pktcnt);
1994 	msg_set_bearer_id(&tnlhdr, l->peer_bearer_id);
1995 tnl:
1996 	/* Wrap each packet into a tunnel packet */
1997 	skb_queue_walk(queue, skb) {
1998 		hdr = buf_msg(skb);
1999 		if (queue == &l->backlogq)
2000 			msg_set_seqno(hdr, seqno++);
2001 		pktlen = msg_size(hdr);
2002 
2003 		/* Tunnel link MTU is not large enough? This could be
2004 		 * due to:
2005 		 * 1) Link MTU has just changed or set differently;
2006 		 * 2) Or FAILOVER on the top of a SYNCH message
2007 		 *
2008 		 * The 2nd case should not happen if peer supports
2009 		 * TIPC_TUNNEL_ENHANCED
2010 		 */
2011 		if (pktlen > tnl->mtu - INT_H_SIZE) {
2012 			if (mtyp == FAILOVER_MSG &&
2013 			    (tnl->peer_caps & TIPC_TUNNEL_ENHANCED)) {
2014 				rc = tipc_msg_fragment(skb, &tnlhdr, tnl->mtu,
2015 						       &frags);
2016 				if (rc) {
2017 					pr_warn("%sunable to frag msg: rc %d\n",
2018 						link_co_err, rc);
2019 					return;
2020 				}
2021 				pktcnt += skb_queue_len(&frags) - 1;
2022 				pktcnt_need_update = true;
2023 				skb_queue_splice_tail_init(&frags, &tnlq);
2024 				continue;
2025 			}
2026 			/* Unluckily, peer doesn't have TIPC_TUNNEL_ENHANCED
2027 			 * => Just warn it and return!
2028 			 */
2029 			pr_warn_ratelimited("%stoo large msg <%d, %d>: %d!\n",
2030 					    link_co_err, msg_user(hdr),
2031 					    msg_type(hdr), msg_size(hdr));
2032 			return;
2033 		}
2034 
2035 		msg_set_size(&tnlhdr, pktlen + INT_H_SIZE);
2036 		tnlskb = tipc_buf_acquire(pktlen + INT_H_SIZE, GFP_ATOMIC);
2037 		if (!tnlskb) {
2038 			pr_warn("%sunable to send packet\n", link_co_err);
2039 			return;
2040 		}
2041 		skb_copy_to_linear_data(tnlskb, &tnlhdr, INT_H_SIZE);
2042 		skb_copy_to_linear_data_offset(tnlskb, INT_H_SIZE, hdr, pktlen);
2043 		__skb_queue_tail(&tnlq, tnlskb);
2044 	}
2045 	if (queue != &l->backlogq) {
2046 		queue = &l->backlogq;
2047 		goto tnl;
2048 	}
2049 
2050 	if (pktcnt_need_update)
2051 		skb_queue_walk(&tnlq, skb) {
2052 			hdr = buf_msg(skb);
2053 			msg_set_msgcnt(hdr, pktcnt);
2054 		}
2055 
2056 	tipc_link_xmit(tnl, &tnlq, xmitq);
2057 
2058 	if (mtyp == FAILOVER_MSG) {
2059 		tnl->drop_point = l->rcv_nxt;
2060 		tnl->failover_reasm_skb = l->reasm_buf;
2061 		l->reasm_buf = NULL;
2062 
2063 		/* Failover the link's deferdq */
2064 		if (unlikely(!skb_queue_empty(fdefq))) {
2065 			pr_warn("Link failover deferdq not empty: %d!\n",
2066 				skb_queue_len(fdefq));
2067 			__skb_queue_purge(fdefq);
2068 		}
2069 		skb_queue_splice_init(&l->deferdq, fdefq);
2070 	}
2071 }
2072 
2073 /**
2074  * tipc_link_failover_prepare() - prepare tnl for link failover
2075  *
2076  * This is a special version of the precursor - tipc_link_tnl_prepare(),
2077  * see the tipc_node_link_failover() for details
2078  *
2079  * @l: failover link
2080  * @tnl: tunnel link
2081  * @xmitq: queue for messages to be xmited
2082  */
tipc_link_failover_prepare(struct tipc_link * l,struct tipc_link * tnl,struct sk_buff_head * xmitq)2083 void tipc_link_failover_prepare(struct tipc_link *l, struct tipc_link *tnl,
2084 				struct sk_buff_head *xmitq)
2085 {
2086 	struct sk_buff_head *fdefq = &tnl->failover_deferdq;
2087 
2088 	tipc_link_create_dummy_tnl_msg(tnl, xmitq);
2089 
2090 	/* This failover link endpoint was never established before,
2091 	 * so it has not received anything from peer.
2092 	 * Otherwise, it must be a normal failover situation or the
2093 	 * node has entered SELF_DOWN_PEER_LEAVING and both peer nodes
2094 	 * would have to start over from scratch instead.
2095 	 */
2096 	tnl->drop_point = 1;
2097 	tnl->failover_reasm_skb = NULL;
2098 
2099 	/* Initiate the link's failover deferdq */
2100 	if (unlikely(!skb_queue_empty(fdefq))) {
2101 		pr_warn("Link failover deferdq not empty: %d!\n",
2102 			skb_queue_len(fdefq));
2103 		__skb_queue_purge(fdefq);
2104 	}
2105 }
2106 
2107 /* tipc_link_validate_msg(): validate message against current link state
2108  * Returns true if message should be accepted, otherwise false
2109  */
tipc_link_validate_msg(struct tipc_link * l,struct tipc_msg * hdr)2110 bool tipc_link_validate_msg(struct tipc_link *l, struct tipc_msg *hdr)
2111 {
2112 	u16 curr_session = l->peer_session;
2113 	u16 session = msg_session(hdr);
2114 	int mtyp = msg_type(hdr);
2115 
2116 	if (msg_user(hdr) != LINK_PROTOCOL)
2117 		return true;
2118 
2119 	switch (mtyp) {
2120 	case RESET_MSG:
2121 		if (!l->in_session)
2122 			return true;
2123 		/* Accept only RESET with new session number */
2124 		return more(session, curr_session);
2125 	case ACTIVATE_MSG:
2126 		if (!l->in_session)
2127 			return true;
2128 		/* Accept only ACTIVATE with new or current session number */
2129 		return !less(session, curr_session);
2130 	case STATE_MSG:
2131 		/* Accept only STATE with current session number */
2132 		if (!l->in_session)
2133 			return false;
2134 		if (session != curr_session)
2135 			return false;
2136 		/* Extra sanity check */
2137 		if (!link_is_up(l) && msg_ack(hdr))
2138 			return false;
2139 		if (!(l->peer_caps & TIPC_LINK_PROTO_SEQNO))
2140 			return true;
2141 		/* Accept only STATE with new sequence number */
2142 		return !less(msg_seqno(hdr), l->rcv_nxt_state);
2143 	default:
2144 		return false;
2145 	}
2146 }
2147 
2148 /* tipc_link_proto_rcv(): receive link level protocol message :
2149  * Note that network plane id propagates through the network, and may
2150  * change at any time. The node with lowest numerical id determines
2151  * network plane
2152  */
tipc_link_proto_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2153 static int tipc_link_proto_rcv(struct tipc_link *l, struct sk_buff *skb,
2154 			       struct sk_buff_head *xmitq)
2155 {
2156 	struct tipc_msg *hdr = buf_msg(skb);
2157 	struct tipc_gap_ack_blks *ga = NULL;
2158 	bool reply = msg_probe(hdr), retransmitted = false;
2159 	u32 dlen = msg_data_sz(hdr), glen = 0;
2160 	u16 peers_snd_nxt =  msg_next_sent(hdr);
2161 	u16 peers_tol = msg_link_tolerance(hdr);
2162 	u16 peers_prio = msg_linkprio(hdr);
2163 	u16 gap = msg_seq_gap(hdr);
2164 	u16 ack = msg_ack(hdr);
2165 	u16 rcv_nxt = l->rcv_nxt;
2166 	u16 rcvgap = 0;
2167 	int mtyp = msg_type(hdr);
2168 	int rc = 0, released;
2169 	char *if_name;
2170 	void *data;
2171 
2172 	trace_tipc_proto_rcv(skb, false, l->name);
2173 
2174 	if (dlen > U16_MAX)
2175 		goto exit;
2176 
2177 	if (tipc_link_is_blocked(l) || !xmitq)
2178 		goto exit;
2179 
2180 	if (tipc_own_addr(l->net) > msg_prevnode(hdr))
2181 		l->net_plane = msg_net_plane(hdr);
2182 
2183 	skb_linearize(skb);
2184 	hdr = buf_msg(skb);
2185 	data = msg_data(hdr);
2186 
2187 	if (!tipc_link_validate_msg(l, hdr)) {
2188 		trace_tipc_skb_dump(skb, false, "PROTO invalid (1)!");
2189 		trace_tipc_link_dump(l, TIPC_DUMP_NONE, "PROTO invalid (1)!");
2190 		goto exit;
2191 	}
2192 
2193 	switch (mtyp) {
2194 	case RESET_MSG:
2195 	case ACTIVATE_MSG:
2196 		/* Complete own link name with peer's interface name */
2197 		if_name =  strrchr(l->name, ':') + 1;
2198 		if (sizeof(l->name) - (if_name - l->name) <= TIPC_MAX_IF_NAME)
2199 			break;
2200 		if (msg_data_sz(hdr) < TIPC_MAX_IF_NAME)
2201 			break;
2202 		strncpy(if_name, data, TIPC_MAX_IF_NAME);
2203 
2204 		/* Update own tolerance if peer indicates a non-zero value */
2205 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2206 			l->tolerance = peers_tol;
2207 			l->bc_rcvlink->tolerance = peers_tol;
2208 		}
2209 		/* Update own priority if peer's priority is higher */
2210 		if (in_range(peers_prio, l->priority + 1, TIPC_MAX_LINK_PRI))
2211 			l->priority = peers_prio;
2212 
2213 		/* If peer is going down we want full re-establish cycle */
2214 		if (msg_peer_stopping(hdr)) {
2215 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2216 			break;
2217 		}
2218 
2219 		/* If this endpoint was re-created while peer was ESTABLISHING
2220 		 * it doesn't know current session number. Force re-synch.
2221 		 */
2222 		if (mtyp == ACTIVATE_MSG && msg_dest_session_valid(hdr) &&
2223 		    l->session != msg_dest_session(hdr)) {
2224 			if (less(l->session, msg_dest_session(hdr)))
2225 				l->session = msg_dest_session(hdr) + 1;
2226 			break;
2227 		}
2228 
2229 		/* ACTIVATE_MSG serves as PEER_RESET if link is already down */
2230 		if (mtyp == RESET_MSG || !link_is_up(l))
2231 			rc = tipc_link_fsm_evt(l, LINK_PEER_RESET_EVT);
2232 
2233 		/* ACTIVATE_MSG takes up link if it was already locally reset */
2234 		if (mtyp == ACTIVATE_MSG && l->state == LINK_ESTABLISHING)
2235 			rc = TIPC_LINK_UP_EVT;
2236 
2237 		l->peer_session = msg_session(hdr);
2238 		l->in_session = true;
2239 		l->peer_bearer_id = msg_bearer_id(hdr);
2240 		if (l->mtu > msg_max_pkt(hdr))
2241 			l->mtu = msg_max_pkt(hdr);
2242 		break;
2243 
2244 	case STATE_MSG:
2245 		l->rcv_nxt_state = msg_seqno(hdr) + 1;
2246 
2247 		/* Update own tolerance if peer indicates a non-zero value */
2248 		if (in_range(peers_tol, TIPC_MIN_LINK_TOL, TIPC_MAX_LINK_TOL)) {
2249 			l->tolerance = peers_tol;
2250 			l->bc_rcvlink->tolerance = peers_tol;
2251 		}
2252 		/* Update own prio if peer indicates a different value */
2253 		if ((peers_prio != l->priority) &&
2254 		    in_range(peers_prio, 1, TIPC_MAX_LINK_PRI)) {
2255 			l->priority = peers_prio;
2256 			rc = tipc_link_fsm_evt(l, LINK_FAILURE_EVT);
2257 		}
2258 
2259 		l->silent_intv_cnt = 0;
2260 		l->stats.recv_states++;
2261 		if (msg_probe(hdr))
2262 			l->stats.recv_probes++;
2263 
2264 		if (!link_is_up(l)) {
2265 			if (l->state == LINK_ESTABLISHING)
2266 				rc = TIPC_LINK_UP_EVT;
2267 			break;
2268 		}
2269 
2270 		/* Receive Gap ACK blocks from peer if any */
2271 		glen = tipc_get_gap_ack_blks(&ga, l, hdr, true);
2272 		if(glen > dlen)
2273 			break;
2274 		tipc_mon_rcv(l->net, data + glen, dlen - glen, l->addr,
2275 			     &l->mon_state, l->bearer_id);
2276 
2277 		/* Send NACK if peer has sent pkts we haven't received yet */
2278 		if ((reply || msg_is_keepalive(hdr)) &&
2279 		    more(peers_snd_nxt, rcv_nxt) &&
2280 		    !tipc_link_is_synching(l) &&
2281 		    skb_queue_empty(&l->deferdq))
2282 			rcvgap = peers_snd_nxt - l->rcv_nxt;
2283 		if (rcvgap || reply)
2284 			tipc_link_build_proto_msg(l, STATE_MSG, 0, reply,
2285 						  rcvgap, 0, 0, xmitq);
2286 
2287 		released = tipc_link_advance_transmq(l, l, ack, gap, ga, xmitq,
2288 						     &retransmitted, &rc);
2289 		if (gap)
2290 			l->stats.recv_nacks++;
2291 		if (released || retransmitted)
2292 			tipc_link_update_cwin(l, released, retransmitted);
2293 		if (released)
2294 			tipc_link_advance_backlog(l, xmitq);
2295 		if (unlikely(!skb_queue_empty(&l->wakeupq)))
2296 			link_prepare_wakeup(l);
2297 	}
2298 exit:
2299 	kfree_skb(skb);
2300 	return rc;
2301 }
2302 
2303 /* tipc_link_build_bc_proto_msg() - create broadcast protocol message
2304  */
tipc_link_build_bc_proto_msg(struct tipc_link * l,bool bcast,u16 peers_snd_nxt,struct sk_buff_head * xmitq)2305 static bool tipc_link_build_bc_proto_msg(struct tipc_link *l, bool bcast,
2306 					 u16 peers_snd_nxt,
2307 					 struct sk_buff_head *xmitq)
2308 {
2309 	struct sk_buff *skb;
2310 	struct tipc_msg *hdr;
2311 	struct sk_buff *dfrd_skb = skb_peek(&l->deferdq);
2312 	u16 ack = l->rcv_nxt - 1;
2313 	u16 gap_to = peers_snd_nxt - 1;
2314 
2315 	skb = tipc_msg_create(BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE,
2316 			      0, l->addr, tipc_own_addr(l->net), 0, 0, 0);
2317 	if (!skb)
2318 		return false;
2319 	hdr = buf_msg(skb);
2320 	msg_set_last_bcast(hdr, l->bc_sndlink->snd_nxt - 1);
2321 	msg_set_bcast_ack(hdr, ack);
2322 	msg_set_bcgap_after(hdr, ack);
2323 	if (dfrd_skb)
2324 		gap_to = buf_seqno(dfrd_skb) - 1;
2325 	msg_set_bcgap_to(hdr, gap_to);
2326 	msg_set_non_seq(hdr, bcast);
2327 	__skb_queue_tail(xmitq, skb);
2328 	return true;
2329 }
2330 
2331 /* tipc_link_build_bc_init_msg() - synchronize broadcast link endpoints.
2332  *
2333  * Give a newly added peer node the sequence number where it should
2334  * start receiving and acking broadcast packets.
2335  */
tipc_link_build_bc_init_msg(struct tipc_link * l,struct sk_buff_head * xmitq)2336 static void tipc_link_build_bc_init_msg(struct tipc_link *l,
2337 					struct sk_buff_head *xmitq)
2338 {
2339 	struct sk_buff_head list;
2340 
2341 	__skb_queue_head_init(&list);
2342 	if (!tipc_link_build_bc_proto_msg(l->bc_rcvlink, false, 0, &list))
2343 		return;
2344 	msg_set_bc_ack_invalid(buf_msg(skb_peek(&list)), true);
2345 	tipc_link_xmit(l, &list, xmitq);
2346 }
2347 
2348 /* tipc_link_bc_init_rcv - receive initial broadcast synch data from peer
2349  */
tipc_link_bc_init_rcv(struct tipc_link * l,struct tipc_msg * hdr)2350 void tipc_link_bc_init_rcv(struct tipc_link *l, struct tipc_msg *hdr)
2351 {
2352 	int mtyp = msg_type(hdr);
2353 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2354 
2355 	if (link_is_up(l))
2356 		return;
2357 
2358 	if (msg_user(hdr) == BCAST_PROTOCOL) {
2359 		l->rcv_nxt = peers_snd_nxt;
2360 		l->state = LINK_ESTABLISHED;
2361 		return;
2362 	}
2363 
2364 	if (l->peer_caps & TIPC_BCAST_SYNCH)
2365 		return;
2366 
2367 	if (msg_peer_node_is_up(hdr))
2368 		return;
2369 
2370 	/* Compatibility: accept older, less safe initial synch data */
2371 	if ((mtyp == RESET_MSG) || (mtyp == ACTIVATE_MSG))
2372 		l->rcv_nxt = peers_snd_nxt;
2373 }
2374 
2375 /* tipc_link_bc_sync_rcv - update rcv link according to peer's send state
2376  */
tipc_link_bc_sync_rcv(struct tipc_link * l,struct tipc_msg * hdr,struct sk_buff_head * xmitq)2377 int tipc_link_bc_sync_rcv(struct tipc_link *l, struct tipc_msg *hdr,
2378 			  struct sk_buff_head *xmitq)
2379 {
2380 	u16 peers_snd_nxt = msg_bc_snd_nxt(hdr);
2381 	int rc = 0;
2382 
2383 	if (!link_is_up(l))
2384 		return rc;
2385 
2386 	if (!msg_peer_node_is_up(hdr))
2387 		return rc;
2388 
2389 	/* Open when peer ackowledges our bcast init msg (pkt #1) */
2390 	if (msg_ack(hdr))
2391 		l->bc_peer_is_up = true;
2392 
2393 	if (!l->bc_peer_is_up)
2394 		return rc;
2395 
2396 	/* Ignore if peers_snd_nxt goes beyond receive window */
2397 	if (more(peers_snd_nxt, l->rcv_nxt + l->window))
2398 		return rc;
2399 
2400 	l->snd_nxt = peers_snd_nxt;
2401 	if (link_bc_rcv_gap(l))
2402 		rc |= TIPC_LINK_SND_STATE;
2403 
2404 	/* Return now if sender supports nack via STATE messages */
2405 	if (l->peer_caps & TIPC_BCAST_STATE_NACK)
2406 		return rc;
2407 
2408 	/* Otherwise, be backwards compatible */
2409 
2410 	if (!more(peers_snd_nxt, l->rcv_nxt)) {
2411 		l->nack_state = BC_NACK_SND_CONDITIONAL;
2412 		return 0;
2413 	}
2414 
2415 	/* Don't NACK if one was recently sent or peeked */
2416 	if (l->nack_state == BC_NACK_SND_SUPPRESS) {
2417 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2418 		return 0;
2419 	}
2420 
2421 	/* Conditionally delay NACK sending until next synch rcv */
2422 	if (l->nack_state == BC_NACK_SND_CONDITIONAL) {
2423 		l->nack_state = BC_NACK_SND_UNCONDITIONAL;
2424 		if ((peers_snd_nxt - l->rcv_nxt) < TIPC_MIN_LINK_WIN)
2425 			return 0;
2426 	}
2427 
2428 	/* Send NACK now but suppress next one */
2429 	tipc_link_build_bc_proto_msg(l, true, peers_snd_nxt, xmitq);
2430 	l->nack_state = BC_NACK_SND_SUPPRESS;
2431 	return 0;
2432 }
2433 
tipc_link_bc_ack_rcv(struct tipc_link * r,u16 acked,u16 gap,struct tipc_gap_ack_blks * ga,struct sk_buff_head * xmitq,struct sk_buff_head * retrq)2434 int tipc_link_bc_ack_rcv(struct tipc_link *r, u16 acked, u16 gap,
2435 			 struct tipc_gap_ack_blks *ga,
2436 			 struct sk_buff_head *xmitq,
2437 			 struct sk_buff_head *retrq)
2438 {
2439 	struct tipc_link *l = r->bc_sndlink;
2440 	bool unused = false;
2441 	int rc = 0;
2442 
2443 	if (!link_is_up(r) || !r->bc_peer_is_up)
2444 		return 0;
2445 
2446 	if (gap) {
2447 		l->stats.recv_nacks++;
2448 		r->stats.recv_nacks++;
2449 	}
2450 
2451 	if (less(acked, r->acked) || (acked == r->acked && !gap && !ga))
2452 		return 0;
2453 
2454 	trace_tipc_link_bc_ack(r, acked, gap, &l->transmq);
2455 	tipc_link_advance_transmq(l, r, acked, gap, ga, retrq, &unused, &rc);
2456 
2457 	tipc_link_advance_backlog(l, xmitq);
2458 	if (unlikely(!skb_queue_empty(&l->wakeupq)))
2459 		link_prepare_wakeup(l);
2460 
2461 	return rc;
2462 }
2463 
2464 /* tipc_link_bc_nack_rcv(): receive broadcast nack message
2465  * This function is here for backwards compatibility, since
2466  * no BCAST_PROTOCOL/STATE messages occur from TIPC v2.5.
2467  */
tipc_link_bc_nack_rcv(struct tipc_link * l,struct sk_buff * skb,struct sk_buff_head * xmitq)2468 int tipc_link_bc_nack_rcv(struct tipc_link *l, struct sk_buff *skb,
2469 			  struct sk_buff_head *xmitq)
2470 {
2471 	struct tipc_msg *hdr = buf_msg(skb);
2472 	u32 dnode = msg_destnode(hdr);
2473 	int mtyp = msg_type(hdr);
2474 	u16 acked = msg_bcast_ack(hdr);
2475 	u16 from = acked + 1;
2476 	u16 to = msg_bcgap_to(hdr);
2477 	u16 peers_snd_nxt = to + 1;
2478 	int rc = 0;
2479 
2480 	kfree_skb(skb);
2481 
2482 	if (!tipc_link_is_up(l) || !l->bc_peer_is_up)
2483 		return 0;
2484 
2485 	if (mtyp != STATE_MSG)
2486 		return 0;
2487 
2488 	if (dnode == tipc_own_addr(l->net)) {
2489 		rc = tipc_link_bc_ack_rcv(l, acked, to - acked, NULL, xmitq,
2490 					  xmitq);
2491 		l->stats.recv_nacks++;
2492 		return rc;
2493 	}
2494 
2495 	/* Msg for other node => suppress own NACK at next sync if applicable */
2496 	if (more(peers_snd_nxt, l->rcv_nxt) && !less(l->rcv_nxt, from))
2497 		l->nack_state = BC_NACK_SND_SUPPRESS;
2498 
2499 	return 0;
2500 }
2501 
tipc_link_set_queue_limits(struct tipc_link * l,u32 min_win,u32 max_win)2502 void tipc_link_set_queue_limits(struct tipc_link *l, u32 min_win, u32 max_win)
2503 {
2504 	int max_bulk = TIPC_MAX_PUBL / (l->mtu / ITEM_SIZE);
2505 
2506 	l->min_win = min_win;
2507 	l->ssthresh = max_win;
2508 	l->max_win = max_win;
2509 	l->window = min_win;
2510 	l->backlog[TIPC_LOW_IMPORTANCE].limit      = min_win * 2;
2511 	l->backlog[TIPC_MEDIUM_IMPORTANCE].limit   = min_win * 4;
2512 	l->backlog[TIPC_HIGH_IMPORTANCE].limit     = min_win * 6;
2513 	l->backlog[TIPC_CRITICAL_IMPORTANCE].limit = min_win * 8;
2514 	l->backlog[TIPC_SYSTEM_IMPORTANCE].limit   = max_bulk;
2515 }
2516 
2517 /**
2518  * link_reset_stats - reset link statistics
2519  * @l: pointer to link
2520  */
tipc_link_reset_stats(struct tipc_link * l)2521 void tipc_link_reset_stats(struct tipc_link *l)
2522 {
2523 	memset(&l->stats, 0, sizeof(l->stats));
2524 }
2525 
link_print(struct tipc_link * l,const char * str)2526 static void link_print(struct tipc_link *l, const char *str)
2527 {
2528 	struct sk_buff *hskb = skb_peek(&l->transmq);
2529 	u16 head = hskb ? msg_seqno(buf_msg(hskb)) : l->snd_nxt - 1;
2530 	u16 tail = l->snd_nxt - 1;
2531 
2532 	pr_info("%s Link <%s> state %x\n", str, l->name, l->state);
2533 	pr_info("XMTQ: %u [%u-%u], BKLGQ: %u, SNDNX: %u, RCVNX: %u\n",
2534 		skb_queue_len(&l->transmq), head, tail,
2535 		skb_queue_len(&l->backlogq), l->snd_nxt, l->rcv_nxt);
2536 }
2537 
2538 /* Parse and validate nested (link) properties valid for media, bearer and link
2539  */
tipc_nl_parse_link_prop(struct nlattr * prop,struct nlattr * props[])2540 int tipc_nl_parse_link_prop(struct nlattr *prop, struct nlattr *props[])
2541 {
2542 	int err;
2543 
2544 	err = nla_parse_nested_deprecated(props, TIPC_NLA_PROP_MAX, prop,
2545 					  tipc_nl_prop_policy, NULL);
2546 	if (err)
2547 		return err;
2548 
2549 	if (props[TIPC_NLA_PROP_PRIO]) {
2550 		u32 prio;
2551 
2552 		prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
2553 		if (prio > TIPC_MAX_LINK_PRI)
2554 			return -EINVAL;
2555 	}
2556 
2557 	if (props[TIPC_NLA_PROP_TOL]) {
2558 		u32 tol;
2559 
2560 		tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
2561 		if ((tol < TIPC_MIN_LINK_TOL) || (tol > TIPC_MAX_LINK_TOL))
2562 			return -EINVAL;
2563 	}
2564 
2565 	if (props[TIPC_NLA_PROP_WIN]) {
2566 		u32 max_win;
2567 
2568 		max_win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
2569 		if (max_win < TIPC_DEF_LINK_WIN || max_win > TIPC_MAX_LINK_WIN)
2570 			return -EINVAL;
2571 	}
2572 
2573 	return 0;
2574 }
2575 
__tipc_nl_add_stats(struct sk_buff * skb,struct tipc_stats * s)2576 static int __tipc_nl_add_stats(struct sk_buff *skb, struct tipc_stats *s)
2577 {
2578 	int i;
2579 	struct nlattr *stats;
2580 
2581 	struct nla_map {
2582 		u32 key;
2583 		u32 val;
2584 	};
2585 
2586 	struct nla_map map[] = {
2587 		{TIPC_NLA_STATS_RX_INFO, 0},
2588 		{TIPC_NLA_STATS_RX_FRAGMENTS, s->recv_fragments},
2589 		{TIPC_NLA_STATS_RX_FRAGMENTED, s->recv_fragmented},
2590 		{TIPC_NLA_STATS_RX_BUNDLES, s->recv_bundles},
2591 		{TIPC_NLA_STATS_RX_BUNDLED, s->recv_bundled},
2592 		{TIPC_NLA_STATS_TX_INFO, 0},
2593 		{TIPC_NLA_STATS_TX_FRAGMENTS, s->sent_fragments},
2594 		{TIPC_NLA_STATS_TX_FRAGMENTED, s->sent_fragmented},
2595 		{TIPC_NLA_STATS_TX_BUNDLES, s->sent_bundles},
2596 		{TIPC_NLA_STATS_TX_BUNDLED, s->sent_bundled},
2597 		{TIPC_NLA_STATS_MSG_PROF_TOT, (s->msg_length_counts) ?
2598 			s->msg_length_counts : 1},
2599 		{TIPC_NLA_STATS_MSG_LEN_CNT, s->msg_length_counts},
2600 		{TIPC_NLA_STATS_MSG_LEN_TOT, s->msg_lengths_total},
2601 		{TIPC_NLA_STATS_MSG_LEN_P0, s->msg_length_profile[0]},
2602 		{TIPC_NLA_STATS_MSG_LEN_P1, s->msg_length_profile[1]},
2603 		{TIPC_NLA_STATS_MSG_LEN_P2, s->msg_length_profile[2]},
2604 		{TIPC_NLA_STATS_MSG_LEN_P3, s->msg_length_profile[3]},
2605 		{TIPC_NLA_STATS_MSG_LEN_P4, s->msg_length_profile[4]},
2606 		{TIPC_NLA_STATS_MSG_LEN_P5, s->msg_length_profile[5]},
2607 		{TIPC_NLA_STATS_MSG_LEN_P6, s->msg_length_profile[6]},
2608 		{TIPC_NLA_STATS_RX_STATES, s->recv_states},
2609 		{TIPC_NLA_STATS_RX_PROBES, s->recv_probes},
2610 		{TIPC_NLA_STATS_RX_NACKS, s->recv_nacks},
2611 		{TIPC_NLA_STATS_RX_DEFERRED, s->deferred_recv},
2612 		{TIPC_NLA_STATS_TX_STATES, s->sent_states},
2613 		{TIPC_NLA_STATS_TX_PROBES, s->sent_probes},
2614 		{TIPC_NLA_STATS_TX_NACKS, s->sent_nacks},
2615 		{TIPC_NLA_STATS_TX_ACKS, s->sent_acks},
2616 		{TIPC_NLA_STATS_RETRANSMITTED, s->retransmitted},
2617 		{TIPC_NLA_STATS_DUPLICATES, s->duplicates},
2618 		{TIPC_NLA_STATS_LINK_CONGS, s->link_congs},
2619 		{TIPC_NLA_STATS_MAX_QUEUE, s->max_queue_sz},
2620 		{TIPC_NLA_STATS_AVG_QUEUE, s->queue_sz_counts ?
2621 			(s->accu_queue_sz / s->queue_sz_counts) : 0}
2622 	};
2623 
2624 	stats = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2625 	if (!stats)
2626 		return -EMSGSIZE;
2627 
2628 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2629 		if (nla_put_u32(skb, map[i].key, map[i].val))
2630 			goto msg_full;
2631 
2632 	nla_nest_end(skb, stats);
2633 
2634 	return 0;
2635 msg_full:
2636 	nla_nest_cancel(skb, stats);
2637 
2638 	return -EMSGSIZE;
2639 }
2640 
2641 /* Caller should hold appropriate locks to protect the link */
__tipc_nl_add_link(struct net * net,struct tipc_nl_msg * msg,struct tipc_link * link,int nlflags)2642 int __tipc_nl_add_link(struct net *net, struct tipc_nl_msg *msg,
2643 		       struct tipc_link *link, int nlflags)
2644 {
2645 	u32 self = tipc_own_addr(net);
2646 	struct nlattr *attrs;
2647 	struct nlattr *prop;
2648 	void *hdr;
2649 	int err;
2650 
2651 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2652 			  nlflags, TIPC_NL_LINK_GET);
2653 	if (!hdr)
2654 		return -EMSGSIZE;
2655 
2656 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2657 	if (!attrs)
2658 		goto msg_full;
2659 
2660 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, link->name))
2661 		goto attr_msg_full;
2662 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_DEST, tipc_cluster_mask(self)))
2663 		goto attr_msg_full;
2664 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_MTU, link->mtu))
2665 		goto attr_msg_full;
2666 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, link->stats.recv_pkts))
2667 		goto attr_msg_full;
2668 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, link->stats.sent_pkts))
2669 		goto attr_msg_full;
2670 
2671 	if (tipc_link_is_up(link))
2672 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2673 			goto attr_msg_full;
2674 	if (link->active)
2675 		if (nla_put_flag(msg->skb, TIPC_NLA_LINK_ACTIVE))
2676 			goto attr_msg_full;
2677 
2678 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2679 	if (!prop)
2680 		goto attr_msg_full;
2681 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2682 		goto prop_msg_full;
2683 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_TOL, link->tolerance))
2684 		goto prop_msg_full;
2685 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN,
2686 			link->window))
2687 		goto prop_msg_full;
2688 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_PRIO, link->priority))
2689 		goto prop_msg_full;
2690 	nla_nest_end(msg->skb, prop);
2691 
2692 	err = __tipc_nl_add_stats(msg->skb, &link->stats);
2693 	if (err)
2694 		goto attr_msg_full;
2695 
2696 	nla_nest_end(msg->skb, attrs);
2697 	genlmsg_end(msg->skb, hdr);
2698 
2699 	return 0;
2700 
2701 prop_msg_full:
2702 	nla_nest_cancel(msg->skb, prop);
2703 attr_msg_full:
2704 	nla_nest_cancel(msg->skb, attrs);
2705 msg_full:
2706 	genlmsg_cancel(msg->skb, hdr);
2707 
2708 	return -EMSGSIZE;
2709 }
2710 
__tipc_nl_add_bc_link_stat(struct sk_buff * skb,struct tipc_stats * stats)2711 static int __tipc_nl_add_bc_link_stat(struct sk_buff *skb,
2712 				      struct tipc_stats *stats)
2713 {
2714 	int i;
2715 	struct nlattr *nest;
2716 
2717 	struct nla_map {
2718 		__u32 key;
2719 		__u32 val;
2720 	};
2721 
2722 	struct nla_map map[] = {
2723 		{TIPC_NLA_STATS_RX_INFO, stats->recv_pkts},
2724 		{TIPC_NLA_STATS_RX_FRAGMENTS, stats->recv_fragments},
2725 		{TIPC_NLA_STATS_RX_FRAGMENTED, stats->recv_fragmented},
2726 		{TIPC_NLA_STATS_RX_BUNDLES, stats->recv_bundles},
2727 		{TIPC_NLA_STATS_RX_BUNDLED, stats->recv_bundled},
2728 		{TIPC_NLA_STATS_TX_INFO, stats->sent_pkts},
2729 		{TIPC_NLA_STATS_TX_FRAGMENTS, stats->sent_fragments},
2730 		{TIPC_NLA_STATS_TX_FRAGMENTED, stats->sent_fragmented},
2731 		{TIPC_NLA_STATS_TX_BUNDLES, stats->sent_bundles},
2732 		{TIPC_NLA_STATS_TX_BUNDLED, stats->sent_bundled},
2733 		{TIPC_NLA_STATS_RX_NACKS, stats->recv_nacks},
2734 		{TIPC_NLA_STATS_RX_DEFERRED, stats->deferred_recv},
2735 		{TIPC_NLA_STATS_TX_NACKS, stats->sent_nacks},
2736 		{TIPC_NLA_STATS_TX_ACKS, stats->sent_acks},
2737 		{TIPC_NLA_STATS_RETRANSMITTED, stats->retransmitted},
2738 		{TIPC_NLA_STATS_DUPLICATES, stats->duplicates},
2739 		{TIPC_NLA_STATS_LINK_CONGS, stats->link_congs},
2740 		{TIPC_NLA_STATS_MAX_QUEUE, stats->max_queue_sz},
2741 		{TIPC_NLA_STATS_AVG_QUEUE, stats->queue_sz_counts ?
2742 			(stats->accu_queue_sz / stats->queue_sz_counts) : 0}
2743 	};
2744 
2745 	nest = nla_nest_start_noflag(skb, TIPC_NLA_LINK_STATS);
2746 	if (!nest)
2747 		return -EMSGSIZE;
2748 
2749 	for (i = 0; i <  ARRAY_SIZE(map); i++)
2750 		if (nla_put_u32(skb, map[i].key, map[i].val))
2751 			goto msg_full;
2752 
2753 	nla_nest_end(skb, nest);
2754 
2755 	return 0;
2756 msg_full:
2757 	nla_nest_cancel(skb, nest);
2758 
2759 	return -EMSGSIZE;
2760 }
2761 
tipc_nl_add_bc_link(struct net * net,struct tipc_nl_msg * msg,struct tipc_link * bcl)2762 int tipc_nl_add_bc_link(struct net *net, struct tipc_nl_msg *msg,
2763 			struct tipc_link *bcl)
2764 {
2765 	int err;
2766 	void *hdr;
2767 	struct nlattr *attrs;
2768 	struct nlattr *prop;
2769 	u32 bc_mode = tipc_bcast_get_mode(net);
2770 	u32 bc_ratio = tipc_bcast_get_broadcast_ratio(net);
2771 
2772 	if (!bcl)
2773 		return 0;
2774 
2775 	tipc_bcast_lock(net);
2776 
2777 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
2778 			  NLM_F_MULTI, TIPC_NL_LINK_GET);
2779 	if (!hdr) {
2780 		tipc_bcast_unlock(net);
2781 		return -EMSGSIZE;
2782 	}
2783 
2784 	attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK);
2785 	if (!attrs)
2786 		goto msg_full;
2787 
2788 	/* The broadcast link is always up */
2789 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_UP))
2790 		goto attr_msg_full;
2791 
2792 	if (nla_put_flag(msg->skb, TIPC_NLA_LINK_BROADCAST))
2793 		goto attr_msg_full;
2794 	if (nla_put_string(msg->skb, TIPC_NLA_LINK_NAME, bcl->name))
2795 		goto attr_msg_full;
2796 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_RX, 0))
2797 		goto attr_msg_full;
2798 	if (nla_put_u32(msg->skb, TIPC_NLA_LINK_TX, 0))
2799 		goto attr_msg_full;
2800 
2801 	prop = nla_nest_start_noflag(msg->skb, TIPC_NLA_LINK_PROP);
2802 	if (!prop)
2803 		goto attr_msg_full;
2804 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_WIN, bcl->max_win))
2805 		goto prop_msg_full;
2806 	if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST, bc_mode))
2807 		goto prop_msg_full;
2808 	if (bc_mode & BCLINK_MODE_SEL)
2809 		if (nla_put_u32(msg->skb, TIPC_NLA_PROP_BROADCAST_RATIO,
2810 				bc_ratio))
2811 			goto prop_msg_full;
2812 	nla_nest_end(msg->skb, prop);
2813 
2814 	err = __tipc_nl_add_bc_link_stat(msg->skb, &bcl->stats);
2815 	if (err)
2816 		goto attr_msg_full;
2817 
2818 	tipc_bcast_unlock(net);
2819 	nla_nest_end(msg->skb, attrs);
2820 	genlmsg_end(msg->skb, hdr);
2821 
2822 	return 0;
2823 
2824 prop_msg_full:
2825 	nla_nest_cancel(msg->skb, prop);
2826 attr_msg_full:
2827 	nla_nest_cancel(msg->skb, attrs);
2828 msg_full:
2829 	tipc_bcast_unlock(net);
2830 	genlmsg_cancel(msg->skb, hdr);
2831 
2832 	return -EMSGSIZE;
2833 }
2834 
tipc_link_set_tolerance(struct tipc_link * l,u32 tol,struct sk_buff_head * xmitq)2835 void tipc_link_set_tolerance(struct tipc_link *l, u32 tol,
2836 			     struct sk_buff_head *xmitq)
2837 {
2838 	l->tolerance = tol;
2839 	if (l->bc_rcvlink)
2840 		l->bc_rcvlink->tolerance = tol;
2841 	if (link_is_up(l))
2842 		tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, tol, 0, xmitq);
2843 }
2844 
tipc_link_set_prio(struct tipc_link * l,u32 prio,struct sk_buff_head * xmitq)2845 void tipc_link_set_prio(struct tipc_link *l, u32 prio,
2846 			struct sk_buff_head *xmitq)
2847 {
2848 	l->priority = prio;
2849 	tipc_link_build_proto_msg(l, STATE_MSG, 0, 0, 0, 0, prio, xmitq);
2850 }
2851 
tipc_link_set_abort_limit(struct tipc_link * l,u32 limit)2852 void tipc_link_set_abort_limit(struct tipc_link *l, u32 limit)
2853 {
2854 	l->abort_limit = limit;
2855 }
2856 
2857 /**
2858  * tipc_link_dump - dump TIPC link data
2859  * @l: tipc link to be dumped
2860  * @dqueues: bitmask to decide if any link queue to be dumped?
2861  *           - TIPC_DUMP_NONE: don't dump link queues
2862  *           - TIPC_DUMP_TRANSMQ: dump link transmq queue
2863  *           - TIPC_DUMP_BACKLOGQ: dump link backlog queue
2864  *           - TIPC_DUMP_DEFERDQ: dump link deferd queue
2865  *           - TIPC_DUMP_INPUTQ: dump link input queue
2866  *           - TIPC_DUMP_WAKEUP: dump link wakeup queue
2867  *           - TIPC_DUMP_ALL: dump all the link queues above
2868  * @buf: returned buffer of dump data in format
2869  */
tipc_link_dump(struct tipc_link * l,u16 dqueues,char * buf)2870 int tipc_link_dump(struct tipc_link *l, u16 dqueues, char *buf)
2871 {
2872 	int i = 0;
2873 	size_t sz = (dqueues) ? LINK_LMAX : LINK_LMIN;
2874 	struct sk_buff_head *list;
2875 	struct sk_buff *hskb, *tskb;
2876 	u32 len;
2877 
2878 	if (!l) {
2879 		i += scnprintf(buf, sz, "link data: (null)\n");
2880 		return i;
2881 	}
2882 
2883 	i += scnprintf(buf, sz, "link data: %x", l->addr);
2884 	i += scnprintf(buf + i, sz - i, " %x", l->state);
2885 	i += scnprintf(buf + i, sz - i, " %u", l->in_session);
2886 	i += scnprintf(buf + i, sz - i, " %u", l->session);
2887 	i += scnprintf(buf + i, sz - i, " %u", l->peer_session);
2888 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt);
2889 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt);
2890 	i += scnprintf(buf + i, sz - i, " %u", l->snd_nxt_state);
2891 	i += scnprintf(buf + i, sz - i, " %u", l->rcv_nxt_state);
2892 	i += scnprintf(buf + i, sz - i, " %x", l->peer_caps);
2893 	i += scnprintf(buf + i, sz - i, " %u", l->silent_intv_cnt);
2894 	i += scnprintf(buf + i, sz - i, " %u", l->rst_cnt);
2895 	i += scnprintf(buf + i, sz - i, " %u", 0);
2896 	i += scnprintf(buf + i, sz - i, " %u", 0);
2897 	i += scnprintf(buf + i, sz - i, " %u", l->acked);
2898 
2899 	list = &l->transmq;
2900 	len = skb_queue_len(list);
2901 	hskb = skb_peek(list);
2902 	tskb = skb_peek_tail(list);
2903 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2904 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2905 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2906 
2907 	list = &l->deferdq;
2908 	len = skb_queue_len(list);
2909 	hskb = skb_peek(list);
2910 	tskb = skb_peek_tail(list);
2911 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2912 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2913 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2914 
2915 	list = &l->backlogq;
2916 	len = skb_queue_len(list);
2917 	hskb = skb_peek(list);
2918 	tskb = skb_peek_tail(list);
2919 	i += scnprintf(buf + i, sz - i, " | %u %u %u", len,
2920 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2921 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2922 
2923 	list = l->inputq;
2924 	len = skb_queue_len(list);
2925 	hskb = skb_peek(list);
2926 	tskb = skb_peek_tail(list);
2927 	i += scnprintf(buf + i, sz - i, " | %u %u %u\n", len,
2928 		       (hskb) ? msg_seqno(buf_msg(hskb)) : 0,
2929 		       (tskb) ? msg_seqno(buf_msg(tskb)) : 0);
2930 
2931 	if (dqueues & TIPC_DUMP_TRANSMQ) {
2932 		i += scnprintf(buf + i, sz - i, "transmq: ");
2933 		i += tipc_list_dump(&l->transmq, false, buf + i);
2934 	}
2935 	if (dqueues & TIPC_DUMP_BACKLOGQ) {
2936 		i += scnprintf(buf + i, sz - i,
2937 			       "backlogq: <%u %u %u %u %u>, ",
2938 			       l->backlog[TIPC_LOW_IMPORTANCE].len,
2939 			       l->backlog[TIPC_MEDIUM_IMPORTANCE].len,
2940 			       l->backlog[TIPC_HIGH_IMPORTANCE].len,
2941 			       l->backlog[TIPC_CRITICAL_IMPORTANCE].len,
2942 			       l->backlog[TIPC_SYSTEM_IMPORTANCE].len);
2943 		i += tipc_list_dump(&l->backlogq, false, buf + i);
2944 	}
2945 	if (dqueues & TIPC_DUMP_DEFERDQ) {
2946 		i += scnprintf(buf + i, sz - i, "deferdq: ");
2947 		i += tipc_list_dump(&l->deferdq, false, buf + i);
2948 	}
2949 	if (dqueues & TIPC_DUMP_INPUTQ) {
2950 		i += scnprintf(buf + i, sz - i, "inputq: ");
2951 		i += tipc_list_dump(l->inputq, false, buf + i);
2952 	}
2953 	if (dqueues & TIPC_DUMP_WAKEUP) {
2954 		i += scnprintf(buf + i, sz - i, "wakeup: ");
2955 		i += tipc_list_dump(&l->wakeupq, false, buf + i);
2956 	}
2957 
2958 	return i;
2959 }
2960